International Journal of Computer Science Issues

IJCSI

International Journal of Computer Science Issues

Volume 7, Issue 3, No 6, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814

© IJCSI PUBLICATION www.IJCSI.org

IJCSI proceedings are currently indexed by:

© IJCSI PUBLICATION 2010 www.IJCSI.org

IJCSI Publicity Board 2010

Dr. Borislav D Dimitrov Department of General Practice, Royal College of Surgeons in Ireland Dublin, Ireland

Dr. Vishal Goyal Department of Computer Science, Punjabi University Patiala, India

Mr. Nehinbe Joshua University of Essex Colchester, Essex, UK

Mr. Vassilis Papataxiarhis Department of Informatics and Telecommunications National and Kapodistrian University of Athens, Athens, Greece

EDITORIAL In this third edition of 2010, we bring forward issues from various dynamic computer science areas ranging from system performance, computer vision, artificial intelligence, software engineering, multimedia , pattern recognition, information retrieval, databases, security and networking among others. As always we thank all our reviewers for providing constructive comments on papers sent to them for review. This helps enormously in improving the quality of papers published in this issue. IJCSI will maintain its policy of sending print copies of the journal to all corresponding authors worldwide free of charge. Apart from availability of the full-texts from the journal website, all published papers are deposited in open-access repositories to make access easier and ensure continuous availability of its proceedings. The transition from the 2nd issue to the 3rd one has been marked with an agreement signed between IJCSI and ProQuest and EBSCOHOST, two leading directories to help in the dissemination of our published papers. We believe further indexing and more dissemination will definitely lead to further citations of our authors’ articles. We are pleased to present IJCSI Volume 7, Issue 3, May 2010, split in eleven numbers (IJCSI Vol. 7, Issue 3, No. 6). The acceptance rate for this issue is 37.88%. We wish you a happy reading!

IJCSI Editorial Board May 2010 Issue ISSN (Print): 1694-0814 ISSN (Online): 1694-0784 © IJCSI Publications www.IJCSI.org

IJCSI Editorial Board 2010

Dr Tristan Vanrullen Chief Editor LPL, Laboratoire Parole et Langage - CNRS - Aix en Provence, France LABRI, Laboratoire Bordelais de Recherche en Informatique - INRIA - Bordeaux, France LEEE, Laboratoire d'Esthétique et Expérimentations de l'Espace - Université d'Auvergne, France

Dr Constantino Malagôn Associate Professor Nebrija University Spain

Dr Lamia Fourati Chaari Associate Professor Multimedia and Informatics Higher Institute in SFAX Tunisia

Dr Mokhtar Beldjehem Professor Sainte-Anne University Halifax, NS, Canada

Dr Pascal Chatonnay Assistant Professor MaÎtre de Conférences Laboratoire d'Informatique de l'Université de Franche-Comté Université de Franche-Comté France

Dr Yee-Ming Chen Professor Department of Industrial Engineering and Management Yuan Ze University Taiwan

Dr Vishal Goyal Assistant Professor Department of Computer Science Punjabi University Patiala, India

Dr Natarajan Meghanathan Assistant Professor REU Program Director Department of Computer Science Jackson State University Jackson, USA

Dr Deepak Laxmi Narasimha Department of Software Engineering, Faculty of Computer Science and Information Technology, University of Malaya, Kuala Lumpur, Malaysia

Dr Navneet Agrawal Assistant Professor Department of ECE, College of Technology & Engineering, MPUAT, Udaipur 313001 Rajasthan, India

Prof N. Jaisankar Assistant Professor School of Computing Sciences, VIT University Vellore, Tamilnadu, India

IJCSI Reviewers Committee 2010  Mr. Markus Schatten, University of Zagreb, Faculty of Organization and Informatics, Croatia  Mr. Vassilis Papataxiarhis, Department of Informatics and Telecommunications, National and Kapodistrian University of Athens, Athens, Greece  Dr Modestos Stavrakis, University of the Aegean, Greece  Dr Fadi KHALIL, LAAS -- CNRS Laboratory, France  Dr Dimitar Trajanov, Faculty of Electrical Engineering and Information technologies, ss. Cyril and Methodius Univesity - Skopje, Macedonia  Dr Jinping Yuan, College of Information System and Management,National Univ. of Defense Tech., China  Dr Alexis Lazanas, Ministry of Education, Greece  Dr Stavroula Mougiakakou, University of Bern, ARTORG Center for Biomedical Engineering Research, Switzerland  Dr Cyril de Runz, CReSTIC-SIC, IUT de Reims, University of Reims, France  Mr. Pramodkumar P. Gupta, Dept of Bioinformatics, Dr D Y Patil University, India  Dr Alireza Fereidunian, School of ECE, University of Tehran, Iran  Mr. Fred Viezens, Otto-Von-Guericke-University Magdeburg, Germany  Dr. Richard G. Bush, Lawrence Technological University, United States  Dr. Ola Osunkoya, Information Security Architect, USA  Mr. Kotsokostas N.Antonios, TEI Piraeus, Hellas  Prof Steven Totosy de Zepetnek, U of Halle-Wittenberg & Purdue U & National Sun Yat-sen U, Germany, USA, Taiwan  Mr. M Arif Siddiqui, Najran University, Saudi Arabia  Ms. Ilknur Icke, The Graduate Center, City University of New York, USA  Prof Miroslav Baca, Faculty of Organization and Informatics, University of Zagreb, Croatia  Dr. Elvia Ruiz Beltrán, Instituto Tecnológico de Aguascalientes, Mexico  Mr. Moustafa Banbouk, Engineer du Telecom, UAE  Mr. Kevin P. Monaghan, Wayne State University, Detroit, Michigan, USA  Ms. Moira Stephens, University of Sydney, Australia  Ms. Maryam Feily, National Advanced IPv6 Centre of Excellence (NAV6) , Universiti Sains Malaysia (USM), Malaysia  Dr. Constantine YIALOURIS, Informatics Laboratory Agricultural University of Athens, Greece  Mrs. Angeles Abella, U. de Montreal, Canada  Dr. Patrizio Arrigo, CNR ISMAC, italy  Mr. Anirban Mukhopadhyay, B.P.Poddar Institute of Management & Technology, India  Mr. Dinesh Kumar, DAV Institute of Engineering & Technology, India  Mr. Jorge L. Hernandez-Ardieta, INDRA SISTEMAS / University Carlos III of Madrid, Spain  Mr. AliReza Shahrestani, University of Malaya (UM), National Advanced IPv6 Centre of Excellence (NAv6), Malaysia  Mr. Blagoj Ristevski, Faculty of Administration and Information Systems Management - Bitola, Republic of Macedonia  Mr. Mauricio Egidio Cantão, Department of Computer Science / University of São Paulo, Brazil  Mr. Jules Ruis, Fractal Consultancy, The Netherlands

 Mr. Mohammad Iftekhar Husain, University at Buffalo, USA  Dr. Deepak Laxmi Narasimha, Department of Software Engineering, Faculty of Computer Science and Information Technology, University of Malaya, Malaysia  Dr. Paola Di Maio, DMEM University of Strathclyde, UK  Dr. Bhanu Pratap Singh, Institute of Instrumentation Engineering, Kurukshetra University Kurukshetra, India  Mr. Sana Ullah, Inha University, South Korea  Mr. Cornelis Pieter Pieters, Condast, The Netherlands  Dr. Amogh Kavimandan, The MathWorks Inc., USA  Dr. Zhinan Zhou, Samsung Telecommunications America, USA  Mr. Alberto de Santos Sierra, Universidad Politécnica de Madrid, Spain  Dr. Md. Atiqur Rahman Ahad, Department of Applied Physics, Electronics & Communication Engineering (APECE), University of Dhaka, Bangladesh  Dr. Charalampos Bratsas, Lab of Medical Informatics, Medical Faculty, Aristotle University, Thessaloniki, Greece  Ms. Alexia Dini Kounoudes, Cyprus University of Technology, Cyprus  Mr. Anthony Gesase, University of Dar es salaam Computing Centre, Tanzania  Dr. Jorge A. Ruiz-Vanoye, Universidad Juárez Autónoma de Tabasco, Mexico  Dr. Alejandro Fuentes Penna, Universidad Popular Autónoma del Estado de Puebla, México  Dr. Ocotlán Díaz-Parra, Universidad Juárez Autónoma de Tabasco, México  Mrs. Nantia Iakovidou, Aristotle University of Thessaloniki, Greece  Mr. Vinay Chopra, DAV Institute of Engineering & Technology, Jalandhar  Ms. Carmen Lastres, Universidad Politécnica de Madrid - Centre for Smart Environments, Spain  Dr. Sanja Lazarova-Molnar, United Arab Emirates University, UAE  Mr. Srikrishna Nudurumati, Imaging & Printing Group R&D Hub, Hewlett-Packard, India  Dr. Olivier Nocent, CReSTIC/SIC, University of Reims, France  Mr. Burak Cizmeci, Isik University, Turkey  Dr. Carlos Jaime Barrios Hernandez, LIG (Laboratory Of Informatics of Grenoble), France  Mr. Md. Rabiul Islam, Rajshahi university of Engineering & Technology (RUET), Bangladesh  Dr. LAKHOUA Mohamed Najeh, ISSAT - Laboratory of Analysis and Control of Systems, Tunisia  Dr. Alessandro Lavacchi, Department of Chemistry - University of Firenze, Italy  Mr. Mungwe, University of Oldenburg, Germany  Mr. Somnath Tagore, Dr D Y Patil University, India  Ms. Xueqin Wang, ATCS, USA  Dr. Borislav D Dimitrov, Department of General Practice, Royal College of Surgeons in Ireland, Dublin, Ireland  Dr. Fondjo Fotou Franklin, Langston University, USA  Dr. Vishal Goyal, Department of Computer Science, Punjabi University, Patiala, India  Mr. Thomas J. Clancy, ACM, United States  Dr. Ahmed Nabih Zaki Rashed, Dr. in Electronic Engineering, Faculty of Electronic Engineering, menouf 32951, Electronics and Electrical Communication Engineering Department, Menoufia university, EGYPT, EGYPT  Dr. Rushed Kanawati, LIPN, France  Mr. Koteshwar Rao, K G Reddy College Of ENGG.&TECH,CHILKUR, RR DIST.,AP, India

 Mr. M. Nagesh Kumar, Department of Electronics and Communication, J.S.S. research foundation, Mysore University, Mysore-6, India  Dr. Ibrahim Noha, Grenoble Informatics Laboratory, France  Mr. Muhammad Yasir Qadri, University of Essex, UK  Mr. Annadurai .P, KMCPGS, Lawspet, Pondicherry, India, (Aff. Pondicherry Univeristy, India  Mr. E Munivel , CEDTI (Govt. of India), India  Dr. Chitra Ganesh Desai, University of Pune, India  Mr. Syed, Analytical Services & Materials, Inc., USA  Dr. Mashud Kabir, Department of Computer Science, University of Tuebingen, Germany  Mrs. Payal N. Raj, Veer South Gujarat University, India  Mrs. Priti Maheshwary, Maulana Azad National Institute of Technology, Bhopal, India  Mr. Mahesh Goyani, S.P. University, India, India  Mr. Vinay Verma, Defence Avionics Research Establishment, DRDO, India  Dr. George A. Papakostas, Democritus University of Thrace, Greece  Mr. Abhijit Sanjiv Kulkarni, DARE, DRDO, India  Mr. Kavi Kumar Khedo, University of Mauritius, Mauritius  Dr. B. Sivaselvan, Indian Institute of Information Technology, Design & Manufacturing, Kancheepuram, IIT Madras Campus, India  Dr. Partha Pratim Bhattacharya, Greater Kolkata College of Engineering and Management, West Bengal University of Technology, India  Mr. Manish Maheshwari, Makhanlal C University of Journalism & Communication, India  Dr. Siddhartha Kumar Khaitan, Iowa State University, USA  Dr. Mandhapati Raju, General Motors Inc, USA  Dr. M.Iqbal Saripan, Universiti Putra Malaysia, Malaysia  Mr. Ahmad Shukri Mohd Noor, University Malaysia Terengganu, Malaysia  Mr. Selvakuberan K, TATA Consultancy Services, India  Dr. Smita Rajpal, Institute of Technology and Management, Gurgaon, India  Mr. Rakesh Kachroo, Tata Consultancy Services, India  Mr. Raman Kumar, National Institute of Technology, Jalandhar, Punjab., India  Mr. Nitesh Sureja, S.P.University, India  Dr. M. Emre Celebi, Louisiana State University, Shreveport, USA  Dr. Aung Kyaw Oo, Defence Services Academy, Myanmar  Mr. Sanjay P. Patel, Sankalchand Patel College of Engineering, Visnagar, Gujarat, India  Dr. Pascal Fallavollita, Queens University, Canada  Mr. Jitendra Agrawal, Rajiv Gandhi Technological University, Bhopal, MP, India  Mr. Ismael Rafael Ponce Medellín, Cenidet (Centro Nacional de Investigación y Desarrollo Tecnológico), Mexico  Mr. Supheakmungkol SARIN, Waseda University, Japan  Mr. Shoukat Ullah, Govt. Post Graduate College Bannu, Pakistan  Dr. Vivian Augustine, Telecom Zimbabwe, Zimbabwe  Mrs. Mutalli Vatila, Offshore Business Philipines, Philipines  Dr. Emanuele Goldoni, University of Pavia, Dept. of Electronics, TLC & Networking Lab, Italy  Mr. Pankaj Kumar, SAMA, India  Dr. Himanshu Aggarwal, Punjabi University,Patiala, India  Dr. Vauvert Guillaume, Europages, France

 Prof Yee Ming Chen, Department of Industrial Engineering and Management, Yuan Ze University, Taiwan  Dr. Constantino Malagón, Nebrija University, Spain  Prof Kanwalvir Singh Dhindsa, B.B.S.B.Engg.College, Fatehgarh Sahib (Punjab), India  Mr. Angkoon Phinyomark, Prince of Singkla University, Thailand  Ms. Nital H. Mistry, Veer Narmad South Gujarat University, Surat, India  Dr. M.R.Sumalatha, Anna University, India  Mr. Somesh Kumar Dewangan, Disha Institute of Management and Technology, India  Mr. Raman Maini, Punjabi University, Patiala(Punjab)-147002, India  Dr. Abdelkader Outtagarts, Alcatel-Lucent Bell-Labs, France  Prof Dr. Abdul Wahid, AKG Engg. College, Ghaziabad, India  Mr. Prabu Mohandas, Anna University/Adhiyamaan College of Engineering, india  Dr. Manish Kumar Jindal, Panjab University Regional Centre, Muktsar, India  Prof Mydhili K Nair, M S Ramaiah Institute of Technnology, Bangalore, India  Dr. C. Suresh Gnana Dhas, VelTech MultiTech Dr.Rangarajan Dr.Sagunthala Engineering College,Chennai,Tamilnadu, India  Prof Akash Rajak, Krishna Institute of Engineering and Technology, Ghaziabad, India  Mr. Ajay Kumar Shrivastava, Krishna Institute of Engineering & Technology, Ghaziabad, India  Mr. Deo Prakash, SMVD University, Kakryal(J&K), India  Dr. Vu Thanh Nguyen, University of Information Technology HoChiMinh City, VietNam  Prof Deo Prakash, SMVD University (A Technical University open on I.I.T. Pattern) Kakryal (J&K), India  Dr. Navneet Agrawal, Dept. of ECE, College of Technology & Engineering, MPUAT, Udaipur 313001 Rajasthan, India  Mr. Sufal Das, Sikkim Manipal Institute of Technology, India  Mr. Anil Kumar, Sikkim Manipal Institute of Technology, India  Dr. B. Prasanalakshmi, King Saud University, Saudi Arabia.  Dr. K D Verma, S.V. (P.G.) College, Aligarh, India  Mr. Mohd Nazri Ismail, System and Networking Department, University of Kuala Lumpur (UniKL), Malaysia  Dr. Nguyen Tuan Dang, University of Information Technology, Vietnam National University Ho Chi Minh city, Vietnam  Dr. Abdul Aziz, University of Central Punjab, Pakistan  Dr. P. Vasudeva Reddy, Andhra University, India  Mrs. Savvas A. Chatzichristofis, Democritus University of Thrace, Greece  Mr. Marcio Dorn, Federal University of Rio Grande do Sul - UFRGS Institute of Informatics, Brazil  Mr. Luca Mazzola, University of Lugano, Switzerland  Mr. Nadeem Mahmood, Department of Computer Science, University of Karachi, Pakistan  Mr. Hafeez Ullah Amin, Kohat University of Science & Technology, Pakistan  Dr. Professor Vikram Singh, Ch. Devi Lal University, Sirsa (Haryana), India  Mr. M. Azath, Calicut/Mets School of Enginerring, India  Dr. J. Hanumanthappa, DoS in CS, University of Mysore, India  Dr. Shahanawaj Ahamad, Department of Computer Science, King Saud University, Saudi Arabia  Dr. K. Duraiswamy, K. S. Rangasamy College of Technology, India  Prof. Dr Mazlina Esa, Universiti Teknologi Malaysia, Malaysia

 Dr. P. Vasant, Power Control Optimization (Global), Malaysia  Dr. Taner Tuncer, Firat University, Turkey  Dr. Norrozila Sulaiman, University Malaysia Pahang, Malaysia  Prof. S K Gupta, BCET, Guradspur, India  Dr. Latha Parameswaran, Amrita Vishwa Vidyapeetham, India  Mr. M. Azath, Anna University, India  Dr. P. Suresh Varma, Adikavi Nannaya University, India  Prof. V. N. Kamalesh, JSS Academy of Technical Education, India  Dr. D Gunaseelan, Ibri College of Technology, Oman  Mr. Sanjay Kumar Anand, CDAC, India  Mr. Akshat Verma, CDAC, India  Mrs. Fazeela Tunnisa, Najran University, Kingdom of Saudi Arabia  Mr. Hasan Asil, Islamic Azad University Tabriz Branch (Azarshahr), Iran  Prof. Dr Sajal Kabiraj, Fr. C Rodrigues Institute of Management Studies (Affiliated to University of Mumbai, India), India  Mr. Syed Fawad Mustafa, GAC Center, Shandong University, China  Dr. Natarajan Meghanathan, Jackson State University, Jackson, MS, USA  Prof. Selvakani Kandeeban, Francis Xavier Engineering College, India  Mr. Tohid Sedghi, Urmia University, Iran  Dr. S. Sasikumar, PSNA College of Engg and Tech, Dindigul, India  Dr. Anupam Shukla, Indian Institute of Information Technology and Management Gwalior, India  Mr. Rahul Kala, Indian Institute of Inforamtion Technology and Management Gwalior, India  Dr. A V Nikolov, National University of Lesotho, Lesotho  Mr. Kamal Sarkar, Department of Computer Science and Engineering, Jadavpur University, India  Dr. Mokhled S. AlTarawneh, Computer Engineering Dept., Faculty of Engineering, Mutah University, Jordan, Jordan  Prof. Sattar J Aboud, Iraqi Council of Representatives, Iraq-Baghdad  Dr. Prasant Kumar Pattnaik, Department of CSE, KIST, India  Dr. Mohammed Amoon, King Saud University, Saudi Arabia  Dr. Tsvetanka Georgieva, Department of Information Technologies, St. Cyril and St. Methodius University of Veliko Tarnovo, Bulgaria  Dr. Eva Volna, University of Ostrava, Czech Republic  Mr. Ujjal Marjit, University of Kalyani, West-Bengal, India  Dr. Prasant Kumar Pattnaik, KIST,Bhubaneswar,India, India  Dr. Guezouri Mustapha, Department of Electronics, Faculty of Electrical Engineering, University of Science and Technology (USTO), Oran, Algeria  Mr. Maniyar Shiraz Ahmed, Najran University, Najran, Saudi Arabia  Dr. Sreedhar Reddy, JNTU, SSIETW, Hyderabad, India  Mr. Bala Dhandayuthapani Veerasamy, Mekelle University, Ethiopa  Mr. Arash Habibi Lashkari, University of Malaya (UM), Malaysia  Mr. Rajesh Prasad, LDC Institute of Technical Studies, Allahabad, India  Ms. Habib Izadkhah, Tabriz University, Iran  Dr. Lokesh Kumar Sharma, Chhattisgarh Swami Vivekanand Technical University Bhilai, India  Mr. Kuldeep Yadav, IIIT Delhi, India  Dr. Naoufel Kraiem, Institut Superieur d'Informatique, Tunisia

 Prof. Frank Ortmeier, Otto-von-Guericke-Universitaet Magdeburg, Germany  Mr. Ashraf Aljammal, USM, Malaysia  Mrs. Amandeep Kaur, Department of Computer Science, Punjabi University, Patiala, Punjab, India  Mr. Babak Basharirad, University Technology of Malaysia, Malaysia  Mr. Avinash singh, Kiet Ghaziabad, India  Dr. Miguel Vargas-Lombardo, Technological University of Panama, Panama  Dr. Tuncay Sevindik, Firat University, Turkey  Ms. Pavai Kandavelu, Anna University Chennai, India  Mr. Ravish Khichar, Global Institute of Technology, India  Mr Aos Alaa Zaidan Ansaef, Multimedia University, Cyberjaya, Malaysia  Dr. Awadhesh Kumar Sharma, Dept. of CSE, MMM Engg College, Gorakhpur-273010, UP, India  Mr. Qasim Siddique, FUIEMS, Pakistan  Dr. Le Hoang Thai, University of Science, Vietnam National University - Ho Chi Minh City, Vietnam  Dr. Saravanan C, NIT, Durgapur, India  Dr. Vijay Kumar Mago, DAV College, Jalandhar, India  Dr. Do Van Nhon, University of Information Technology, Vietnam  Mr. Georgios Kioumourtzis, University of Patras, Greece  Mr. Amol D.Potgantwar, SITRC Nasik, India  Mr. Lesedi Melton Masisi, Council for Scientific and Industrial Research, South Africa  Dr. Karthik.S, Department of Computer Science & Engineering, SNS College of Technology, India  Mr. Nafiz Imtiaz Bin Hamid, Department of Electrical and Electronic Engineering, Islamic University of Technology (IUT), Bangladesh  Mr. Muhammad Imran Khan, Universiti Teknologi PETRONAS, Malaysia  Dr. Abdul Kareem M. Radhi, Information Engineering - Nahrin University, Iraq  Dr. Mohd Nazri Ismail, University of Kuala Lumpur, Malaysia  Dr. Manuj Darbari, BBDNITM, Institute of Technology, A-649, Indira Nagar, Lucknow 226016, India  Ms. Izerrouken, INP-IRIT, France  Mr. Nitin Ashokrao Naik, Dept. of Computer Science, Yeshwant Mahavidyalaya, Nanded, India  Mr. Nikhil Raj, National Institute of Technology, Kurukshetra, India  Prof. Maher Ben Jemaa, National School of Engineers of Sfax, Tunisia  Prof. Rajeshwar Singh, BRCM College of Engineering and Technology, Bahal Bhiwani, Haryana, India  Mr. Gaurav Kumar, Department of Computer Applications, Chitkara Institute of Engineering and Technology, Rajpura, Punjab, India  Mr. Ajeet Kumar Pandey, Indian Institute of Technology, Kharagpur, India  Mr. Rajiv Phougat, IBM Corporation, USA  Mrs. Aysha V, College of Applied Science Pattuvam affiliated with Kannur University, India  Dr. Debotosh Bhattacharjee, Department of Computer Science and Engineering, Jadavpur University, Kolkata-700032, India  Dr. Neelam Srivastava, Institute of engineering & Technology, Lucknow, India  Prof. Sweta Verma, Galgotia's College of Engineering & Technology, Greater Noida, India  Mr. Harminder Singh BIndra, MIMIT, INDIA  Dr. Lokesh Kumar Sharma, Chhattisgarh Swami Vivekanand Technical University, Bhilai, India  Mr. Tarun Kumar, U.P. Technical University/Radha Govinend Engg. College, India  Mr. Tirthraj Rai, Jawahar Lal Nehru University, New Delhi, India

 Mr. Akhilesh Tiwari, Madhav Institute of Technology & Science, India  Mr. Dakshina Ranjan Kisku, Dr. B. C. Roy Engineering College, WBUT, India  Ms. Anu Suneja, Maharshi Markandeshwar University, Mullana, Haryana, India  Mr. Munish Kumar Jindal, Punjabi University Regional Centre, Jaito (Faridkot), India  Dr. Ashraf Bany Mohammed, Management Information Systems Department, Faculty of Administrative and Financial Sciences, Petra University, Jordan  Mrs. Jyoti Jain, R.G.P.V. Bhopal, India  Dr. Lamia Chaari, SFAX University, Tunisia  Mr. Akhter Raza Syed, Department of Computer Science, University of Karachi, Pakistan  Prof. Khubaib Ahmed Qureshi, Information Technology Department, HIMS, Hamdard University, Pakistan  Prof. Boubker Sbihi, Ecole des Sciences de L'Information, Morocco  Dr. S. M. Riazul Islam, Inha University, South Korea  Prof. Lokhande S.N., S.R.T.M.University, Nanded (MH), India  Dr. Vijay H Mankar, Dept. of Electronics, Govt. Polytechnic, Nagpur, India  Dr. M. Sreedhar Reddy, JNTU, Hyderabad, SSIETW, India  Mr. Ojesanmi Olusegun, Ajayi Crowther University, Oyo, Nigeria  Ms. Mamta Juneja, RBIEBT, PTU, India  Dr. Ekta Walia Bhullar, Maharishi Markandeshwar University, Mullana Ambala (Haryana), India  Prof. Chandra Mohan, John Bosco Engineering College, India  Mr. Nitin A. Naik, Yeshwant Mahavidyalaya, Nanded, India  Mr. Sunil Kashibarao Nayak, Bahirji Smarak Mahavidyalaya, Basmathnagar Dist-Hingoli., India  Prof. Rakesh.L, Vijetha Institute of Technology, Bangalore, India  Mr B. M. Patil, Indian Institute of Technology, Roorkee, Uttarakhand, India  Mr. Thipendra Pal Singh, Sharda University, K.P. III, Greater Noida, Uttar Pradesh, India  Prof. Chandra Mohan, John Bosco Engg College, India  Mr. Hadi Saboohi, University of Malaya - Faculty of Computer Science and Information Technology, Malaysia  Dr. R. Baskaran, Anna University, India  Dr. Wichian Sittiprapaporn, Mahasarakham University College of Music, Thailand  Mr. Lai Khin Wee, Universiti Teknologi Malaysia, Malaysia  Dr. Kamaljit I. Lakhtaria, Atmiya Institute of Technology, India  Mrs. Inderpreet Kaur, PTU, Jalandhar, India  Mr. Iqbaldeep Kaur, PTU / RBIEBT, India  Mrs. Vasudha Bahl, Maharaja Agrasen Institute of Technology, Delhi, India  Prof. Vinay Uttamrao Kale, P.R.M. Institute of Technology & Research, Badnera, Amravati, Maharashtra, India  Mr. Suhas J Manangi, Microsoft, India  Ms. Anna Kuzio, Adam Mickiewicz University, School of English, Poland  Dr. Debojyoti Mitra, Sir Padampat Singhania University, India  Prof. Rachit Garg, Department of Computer Science, L K College, India  Mrs. Manjula K A, Kannur University, India  Mr. Rakesh Kumar, Indian Institute of Technology Roorkee, India

TABLE OF CONTENTS

Domain-Independent Genotype to Phenotype Mapping through XML Rules Daniela Xhemali, Christopher J. Hinde, Roger G. Stone

Pg 1-9

Specifying Data Bases Management Systems by Using RM-ODP Engineering Language Jalal Laassiri, Said Elhajji, Mohamed Bouhdadi, Ghizlane Orhanou, Youssef Balouki

Pg 10-17

Performance Comparison of SVM and ANN for Handwritten Devnagari Character Recognition Sandhya Arora, Debotosh Bhattacharjee, Mita Nasipuri, Latesh Malik, Mahantapas Kundu, Dipak Kumar Basu

Pg 18-26

Improving Network Performance using ACO Based Redundant Link Avoidance Algorithm B. Chandra Mohan, R. Baskaran

Pg 27-34

5. An Investigation about Performance Comparison of Multi-Hop Wireless Ad-Hoc Network Routing Protocols in MANET S. Karthik, Kannan.S, V. P. Arunachalam, T. Ravichandran, M. L. Valarmathi

Pg 35-41

6. Latency Optimized Data Aggregation Timing Model for Wireless Sensor Networks A. Sivagami, K. Pavai, D. Sridharan

Pg 42-48

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814

1

Domain-Independent Genotype to Phenotype Mapping through XML Rules Daniela Xhemali1, Christopher J. Hinde2 and Roger G. Stone3 1

Computer Science, Loughborough University, Loughborough, LE11 3TU, UK

2


3


Abstract This paper discusses an innovative approach to mapping Genotypes to Phenotypes through XML rules. Specifically, it concentrates on the mapping process using two very different domains – Regular Expressions (REs) and Software Program Statements. The paper shows that our Genotype-Phenotype system can be applied to any domain that requires the use of REs and it can be adapted to work for any other domain with minimum effort. Keywords: Genetic Evolution, Genotypes, Phenotypes, XML Mapping, Regular Expressions, Complete Software Structures.

1. Introduction GP research has attracted attention in various fields such as: game strategies [14], military defence [13], plant biology [7], electronics [20], railway platform allocation [4], spam filtering [5], feature extraction from media files [12], [16], automated web extraction [3], [31] etc. One can use different terminologies to define GP, but fundamentally: “At the most abstract level, GP is a systematic, domain independent method for getting computers to automatically solve problems starting from a high-level statement of what needs to be done” [18]. This paper focuses on a specific part of GP – the Genotype-Phenotype Mapping – thus other components of GP will not be covered here. Readers are encouraged to look at [31] for a discussion of the different GP components in relation to a real project, as well as [1], [8] and [9] for a large selection of GP papers and surveys of the history of evolutionary computing.

The Genotype-Phenotype Mapping relates to the way individuals in a population are represented, as this can have a significant effect on the performance of GP. A Genotype represents each individual in the search space, whereas its Phenotype represents the individual in the solution space [2]. Some research, particularly earlier GP research, do not make a distinction between Genotypes and Phenotypes [5], [17], [27]. Individuals in each genetic population remain the same throughout the evolution process. In these works the search space and the solution space are identical. In 1994, Banzhaf [2] suggested the separation of the two spaces and introduced his work on the GenotypePhenotype mapping. The separation involves the encoding of the individuals to a form known as the ‘Genotype’, which is later on decoded back to the corresponding program, referred to as the ‘Phenotype’. This separation simplifies and increases the efficiency of certain genetic operations such as: reproduction and mutation, because these would no longer be constrained by the parameters used in the program being evolved. In GenotypePhenotype based GP, genetic operators such as Crossover and Mutation would be performed on the Genotype, whereas other processes, such as the Fitness scoring, would be performed on the Phenotype. Sections 3.1 and 3.4 explain this concept further through examples. There are researchers who criticise the separation into Genotypes and Phenotypes [19]. The main concern expressed is that the conversion process of a mutated Genotype into the Phenotype may result in anomalies that could potentially lead to invalid solutions. A direct

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org mapping between the encoded program and the actual program is therefore vital to ensure the validity of the solutions [23], [28]. In our research, the Genotypes are presented as strings of integers. The direct mapping of these integers to the corresponding structures is achieved through an innovative approach involving XML rules as described in section 3.4. This research is part of a larger project in collaboration with an independent brokerage – Apricot Training Management (ATM) – which helps organisations to identify and analyse their training needs and recommend suitable courses for their employees. ATM currently uses time-consuming, labour-intensive, manual techniques to gather information related to training courses, however, this process often results in out-of-date courses being stored in the company’s database and many hours being wasted on Web browsing and data entry. Our overall project is to provide ATM with a system that will automate the extraction and storing of training course information into the company’s database, guaranteeing always up-to-date training data [29], [30]. Specifically, the research concentrates on the evolution of Regular Expressions (REs) for the extraction of pieces of information such as course names, prices, dates and locations from training web pages. The following section focuses on research and techniques related to Genotype-Phenotype based GP.

2. Related Research Many researchers have embraced the separation of the search space from the solution space through the use of Genotypes and Phenotypes [2], [15], [28]. This, however, adds an additional step to the genetic evolution process – the translation or mapping of the Genotypes to their corresponding Phenotypes. This step occurs after the genetic reproduction stage (i.e. the crossover and mutation) and before the Fitness test can take place.

2

concern about generating constant numbers. Koza [17] solved this problem by defining “random ephemeral constants” where constants are only generated once for a particular program and then reused wherever they are needed within that program. Keller [15] continued in the footsteps of Banzhaf, concentrating on providing experimental evidence for choosing the Genotype-Phenotype approach instead of the normal GP approach. Keller’s system however, could only evolve programs in languages defined by the LALR (Look Ahead Look Recursive) grammar, as this was the grammar chosen for the repairing stage of the Genotype-Phenotype mapping process. There was a certain amount of redundancy in the genetic coding in both Banzhaf’s and Keller’s works. They both admitted that, in their works, different binary strings could correspond to the same symbol, which could lead to inconsistencies e.g. 000 and 100 both mapping to ‘a’. A slightly different Genotype representation is seen in the work of Withall et al. [28]. In here Genotypes are represented as linear blocks of integers. Each block comprises exactly four integers, each representing a different gene. Although both research works used fixedlength genomes, in [2] the resulting Phenotypes could vary in length, whereas in [28] they remained fixed. The first integer in Withall’s Genotype determines the type of function that follows. Similarly, the Genotype in our research is represented as a string of integers. There are no fixed length genomes determined however; instead the Genotype can contain any number of genes. The unique feature of our research is the use of XML to define the necessary rules to achieve the GenotypePhenotype mapping. The first gene in the string determines the XML rule to be followed, which in itself guides the mapping of the rest of the genes into a valid Phenotype. This is explained in detail in section 3.4.

3. Genotype to Phenotype Mapping The following concentrates on the different methods that have been used to achieve the mapping process. Banzhaf [2] represented his Genotypes as linear binary strings. The mapping stage then processed these Genotypes from left to right in 5-bit sections, where each 5-bit code mapped to a pre-specified symbol. For example: 00000 mapped to PLUS, 00100 mapped to POW, 11000 mapped to variable X etc. The first bit indicated whether the code represented a function (PLUS, POW etc.) or a terminal (X, Y etc.). The research also discussed their

Before discussing our mapping approach, it is important to explain a few related topics in order for the reader to fully understand how the approach works. The following discusses the main representation chosen for the system, as well as the different domains on which this approach was tested. Specifically, the two domains relate to the Genotype-Phenotype mapping of: REs and Software Program Statements. The Software Program Statements domain was chosen because it is very different from that

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org

3

of REs and thus, it can truly illustrate the extent to which our system needs to be changed to apply to such a domain.

used as part of the tree-based GP [21], which is computationally expensive.

Section 3.4 explores the XML mapping process in detail.

3.2 Regular Expressions (REs)

3.1 GP Representation

REs [26], [10], [22], [25] are powerful tools used to detect patterns in data. They can range from basic to very complex, matching from just literal text1 to very specific instances of text based on certain criteria. For example: ^[A-Z][a-z]+ matches all instances that begin (^) with an uppercase letter ([A-Z]) followed by one or more (+) lowercase letters ([a-z]) such as “Regular” or “Expression” but not “RE” or “REs”.

There are two main representations in GP: the Tree-Based GP and the Linear GP [28]. The Tree-Based GP represents programs as syntax trees (Fig. 1), where program variables and constants make up the tree leaves (x, 1, 5, y, 2), whereas the program operators (*, +, –, /) are the internal nodes. The tree leaves are also known as the terminals, whereas the internal nodes are known as the functions. Fig. 1 shows the tree representation of program “(x+1)*(5(y/2))”. For complex programs, the main tree may contain many sub trees.

Fig. 1 Tree-based GP Representation

Linear GP represents programs as a linear sequence of instructions (Fig. 2). Based on biological evolution, the sets of instructions are known as Genomes and each individual instruction is called a gene. All the available Genomes for a particular program form the Genotype. Linear GP representations may have either fixed length Genomes i.e. the same number of genes for every instruction, or variable ones. This depends on the problem to be solved. 1st Instruction 2nd Instruction

...

nth Instruction

REs are very well known, particularly in the UNIX community and they feature largely in some programming languages such as Perl, PHP or AWK. However, the manual generation of REs can be a difficult, error-prone and time consuming undertaking, especially for complex patterns. This is due to the fact that although REs are built up from small building blocks, where each block is fairly simple; all the available blocks can be combined in an infinite number of ways [10], which may result in a highly complex RE. Tools have been developed to evaluate the validity of REs [11], [24], however, very little, if anything has been done towards the automatic generation of REs. Our research focuses on the automatic generation of REs through the use of genetic programming principles in order to automate web extraction. This paper concentrates on the mapping of Genotypes (strings of meaningless numbers) to Phenotypes (REs) through XML rules as explained in section 3.4.

3.3 Software Program Statements The research in this paper is based on the work of Withall et al. [28]. The examples are kept as close to the original as possible in order to ensure their integrity. The only difference is that the Software Program Statements in [28] were evolved to be valid in Perl, whereas in this paper their validity is ensured against VB.NET 2008. Fig. 3 shows an example of the syntax differences in both languages.

Fig. 2 Linear GP Representation

We have chosen the linear approach for our research, because it is more appropriate for the representation of non-standard models like REs. Furthermore, linear GP runs faster than tree-based GP [21]. This is because almost all computer architectures represent computer programs in a linear fashion – as a sequence of commands to execute. The execution of tree-shaped programs is not natural for computers, thus interpreters or compilers would need to be

1

/ all this is literal text and will be matched /


PERL:

if ($x != $y) { $z = $z + 1; }

VB.NET: IF x y THEN z=z+1 END IF Fig. 3 PERL vs. VB.NET

The Software Program Statements that are considered in this paper include simple structures that are commonly used in programming such as: FOR … NEXT; IF … THEN … ELSE, WHILE … END WHILE as well as useful statements such as Addition (x = y + z), Subtraction (x = y - z), Multiplication (x = y * z) and Division (x = y / z). The following section will explain in detail how Genotypes are converted to either REs or Software Program Statements using XML rules.

4

rigid when it comes to the number of components they contain. As mentioned previously, a unique attribute of this research is that it achieves the Genotype to Phenotype conversion through XML rules (Fig. 4). This technique has many advantages including: improved readability, compatibility with many programming languages, portability and extendibility (XML is not restricted to a limited set of keywords defined by the proprietary vendors, which aids the process of creating rules of different levels of complexity). The initial XML rules were created manually after an extensive analysis of a number of web pages. However, in the future, new rules will be able to be added to the XML file automatically (Hinde, Stone and Siau are currently working towards implementing this functionality). The rest of this section explains the way XML is used to guide the mapping of Genotypes to valid Phenotypes. The Genotype-Phenotype mapping process for the REs domain is shown below:

3.4 XML Mapping – REs It is very important for the Genotypes and Phenotypes to have a direct relationship between them, in order for essential characteristics to be preserved from the parents and inherited by the offspring [28]. Koza [17] did not encounter this problem, because the Genotypes were not separated from the Phenotypes in his work, however, representations such as grammatical evolution [20] could be in danger of creating offspring that do not inherit important qualities from their parents, due to a lack of direct mapping between the parent and offspring Phenotypes. Our Genotype to Phenotype mapping is similar to the work of Withall et al. [28], whereby the Genotypes are used for the genetic manipulation, whereas the Phenotypes are used for the Fitness Test. One of the differences, however, is the length of the Genomes. Withall et al. use fixed-length gene blocks or Genomes, where each block comprises four genes, however, they allow for variablelength Genomes through padding – in cases of shorter program structures or statements, left over genes are ignored. In this research, we only use variable-length Genomes. This is because different REs may contain a varying number of components (i.e. tags, RE structures, keywords etc.). One can create a regular expression that is a centimetre long or one that covers a whole A4 page. Withall et al., however, deal with the evolution of structures, as discussed in section 3.3, which are more

Pseudo-code: Genotype-Phenotype Mapping Process 1) Determine the XML rule to follow 2) Follow the chosen XML rule to the end 3) IF the Genotype has fewer genes than the rule requires a) Follow the rule for the number of genes available b) Repair outcome to create a valid partial solution. 4) IF the Genotype has enough genes for the XML rule a) Follow all the components within the rule b) Repair outcome (if necessary) to create a valid and complete solution. 5) IF the Genotype has more genes than the rule requires a) Follow the same steps as above (4a and 4b) b) Ignore the rest of the genes in the Genotype

Note that this is not a character by character evolution, because this would increase the search space and dramatically increase the execution time. Instead REs are divided into three collections: HTML tags (e.g. “title”, “tr” etc.), keywords (e.g. “course”, “title” etc.) and RE substructures (e.g. “.*?” or “[\s]?” etc.). Each evolved gene will be translated to an element of one of these collections. Each collection is further divided into a number of components e.g. the Start-Tag component determines an opening tag, the End-Tag determines a closing tag, the Start-REStructure determines an opening RE structure, an RE-Structure determines a normal structure that does not need closing etc.


5

knows that it needs to reuse this tag as an End-Tag when told so by the rules. We have also introduced some additional components that are not related to the specified collections e.g. the StartCapture component translates to the symbol ‘(‘ and indicates the beginning of a capturing group i.e. the part of the RE, which will capture the part of the results needed; the End-Capture component indicates the end of the capturing group etc. These additional components do not use any genes from the Genotype. They were introduced simply to help the mapping process to translate Genotypes into syntactically correct REs. Fig. 4 shows a partial list of the components needed for the Genotype-Phenotype Mapping process and the order in which they are used within the XML rules. Each XML rule (Fig. 4) determines the necessary components and the order in which they are to be chosen by the mapping stage in order to create a valid and efficient RE. The first gene in the Genotype is always associated with the RE rule choice. The remaining string of integers in the Genotype maps to the different components within that RE rule. The modulo function is used for this purpose, e.g. Table 1 shows the Genotype to be translated using the information in Fig. 4. The value of the first gene is 5. This represents the RE rule to be used. In this case, there are two different rules in the XML file, so 5 mod 2 = 1, which means that the second rule (id = 1) is chosen. This rule contains five different components. The first component number is 4 (Table 2). This corresponds to the Start-Tag component (Fig. 4), which means the next gene in the Genotype (gene 7) needs to be mapped to one of the tags in the ‘tags’ collection. In this case, the collection has three tags, thus 7 mod 3 = 1 gives the ‘td’ tag. The following component number is 6. This corresponds to the Start-Capture component, which maps to the symbol “(“ without using any genes from the Genotype. The remaining components are dealt with in the same manner (see Table 2). Note that in this example, only the first three genes of the Genotype were needed; the remaining two are simply ignored.

Table 1: Genotype Fig. 4 Sample of XML Rules for REs

5 These components are important for ensuring consistency and accuracy in the Phenotypes created. For example once a Start-Tag has been determined, the system automatically

7

15 22 43

Table 2: Genotype to RE Mapping

Component

Component

Gene

Modulo

Translation

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org No. 4 6 3 7 5

Start-Tag Start-Capture RE-Structure End-Capture End-Tag

Used 5 7 15 -

3.5 XML Mapping – Software Program Statements 1 1 0 -

Rule 2 td ( .*? ) td

Fig. 5 Phenotype

Once all the required genes have been decoded, the resulting RE is repaired to ensure its validity and efficiency. Fig. 5 shows the complete RE (Phenotype) for the above example. The additional symbols added by the repairing function are shown encircled. The repairing function is an independent function which scrutinises the Phenotype created in order to guarantee the validity of the RE2. This function is in charge of tasks like: closing opening tags, closing opening parentheses, adding/removing RE structures in cases when there are fewer genes in the Genotype than required by the XML rule, adding variable declarations at the beginning of the program, adding ‘footer’ information where necessary e.g. when returning the RE to a calling function etc. All this is achieved through the use of a STACK programming structure, which works in a LIFO (Last In First Out) manner. This is particularly helpful when closing nested tags and RE structures, for example: Tag1 + Tag2 + RE-Structure + C-RE + C-Tag2 + C-Tag1 Above, C stands for Closing. The Italic part shows the part of the RE added by the repairing function. The XML rules can also determine whether or not a closing tag is actually needed. This is useful in cases where the inclusion of a closing tag results in different results being extracted to the ones needed. For example, the RE: .*? contains two opening tags (‘tr’ and ‘td’), however, only the ‘td’ tag needs to be closed for this to work as intended. This is achieved by including no_end=”true” in the rule (Fig. 4).

2

6

Many HTML tags come in pairs. The system ensures that the same closing tag is chosen based on the opening tag evolved, and the nesting of the tags in the RE hierarchy is preserved.

This section discusses the changes that needed to be made to the system for it to work with Software Program Statements instead of REs. The areas changed were: the XML rules and the repairing function. The following explains the changes involved in order to show the minimum effort required to adapt our system to such an entirely different domain. Fig. 7 shows a sample of the XML rules and components needed to guide the Genotype-Phenotype stage of the genetic evolution of software program statements. When compared to the Genotype-Phenotype Mapping system for REs (Fig. 4), it is clear that the overall logic remains the same, with each XML rule using the available collections (in this case: ‘variables’ and ‘comparisons’) to guide the system through the different components needed for each rule. Identically, there are a number of components that have been introduced to simplify and ensure the consistency of the mapping process for Software Program Statements, but which do not use any genes from the Genotype since they do not need to be evolved. Such components include: the ‘Assign (=) operator, the ‘Addition’ (+) operator, etc. Table 3: Genotype

Table 4: Genotype to Software Statement Mapping

Component No. 1 2 1 1 11 1 7 1

Component Variable Comparison Variable Variable Assign Variable Addition Variable

Gene Used 10 27 7 13 19 9 63 4

Fig. 6 Phenotype

Modulo

Translation

0 0 3 1 4 0 0 1

If x < y Add x = x + y


7

For example, Table 3 shows the Genotype to be translated using the information in Fig. 7. The value of the first gene is 10. This represents the statement type to be used. In this case, there are five different rules (statements) in the XML file, so 10 mod 5 = 0 means that the statement is an ‘If’. This statement contains three different components each requiring the use of a gene to choose from the ‘variables’ and ‘comparisons’ collections. The ‘variables’ collection has three elements, whereas the ‘comparisons’ has four, therefore, 27 mod 3 = 0, 7 mod 4 = 3 and 13 mod 3 = 1 give elements ‘x’, ‘ union (m.DBMS object.Types) We consider the concepts of subtype/supertype (RM-ODP 2-9.9) and subclass/superclass (RM-ODP 2-9.10) as relations between types and classes respectively. Context m: model inv: m.types-> forall( t1: Type, t2: Type | t2.subtype -> includes(t1) implies t1.valid_for.satisfies_type=t2) m.types-> forall( t1: Type, t2: Type | t1.supertype >includes(t2) implies t1.valid_for.satisfies_type=t2) Context a: ActionTemplate inv: a.DBMS object.StartState a.DBMS object.EndState Context o: Object Template inv: iot (DBMS object template) is not parent of or child of itself not (iot.parents >includes (iot ) or iot.children->includes(iot)).

4.4. Semantics Domain The semantics of a UML model is given by constraining the relationship between a model and possible instances of that model (see Fig.3). It means constraining the relationship between expressions of the UML abstract syntax for models and expressions of the UML abstract syntax for instances. We define a model to specify the ODP Engineering viewpoint: a set of DBMS objects, their relationships and behaviors. This model defines DBMS Semantic Domain (Fig.3).

Fig.2 RM-ODP Foundation and DBMS concepts

In the following, we define context constraints for the defined syntax. Context m: Model inv: m.Specifier->includes All (m.DBMS object Templates. Dynamic Schema) m.Describer->includesAll (m. DBMS Template.StaticSchema) m.Constrainer->includesAll (m. DBMS object.Invariant Schema) m.ActionTemplates -> includesAll (m.DBMS object Templates.action) m.Types->includesAll(m.ActionTemplates.

Fig.3 DBMS Semantic Domains

In addition, a system can only be an instance of a single system model, because it is self contained and disjoint from other models. On the other side, objects are instances of one ore more object templates; they may be of one or several types. With no further constraints, it is possible for an object to change the templates of which it is an instance; thus this meta-model supports dynamic types. There is one well-formedness rule for instances, which

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org are given bellow: Context s: system inv: The source and target DBMS object of s'slinks is DBMS object in s: s.DBMS objects->includesAll(s.links.source -> union(s.links.target)) Links between two DBMS objectare unique per role s.links->forAll(l|s.links ->select (l'|l'.source=l.source&l'.target=l.target&l'.of=l.of)=l) Declaration of “Specification concepts” (RM-ODP 2.9) in Alloy [28], time dependence. Context Time inv: forall(o:DBMS object ,t:Time | t.instant ->notEmpty implies o.state ->notEmpty) Context Precondition inv: Forall (prec: Dynamicschema.Precondition, o: DBMS object | exists(s: State) | o.mappedTo = prec and o.state_start = s) Context Postcondition inv:forall (postc: dynamicschema.Postcondition , o :DBMS object | exists(s : State) | o.mappedTo = postc and a.state_end = s).

4.5. Meaning Function Other invariants are required to constraint the relationships between models and instances. These constitute the semantics which are the subject of this section. The semantics for the UML-based language are defined by the relationship between a system model and its possible instances (systems). The constraints are relatively simple, but they demonstrate the general principle. Firstly, there are two constraints related to DBMS object and links, respectively. The first shows how inheritance relationships can force a DBMS object to be of many DBMS object Template. Context o: object inv: The templates of o must be a single template and all the parents of that template o.of->exists (t | o.of=t->union (t.parents)) The second ensures that a link connects objects of templates as dictated by its role. Context l: link inv: DBMS object which are the source/target of link have templates which are the source/target of the corresponding roles. (l.of.source)->intersection (l.source.of) -> notEmpty and (l.of.target)->intersection(l.target.of)->notEmpty Secondly, there are four constraints which ensure that a model instance is a valid instance of the model, it is claimed to be an instance of: The first and second ensure that objects and links are associated with templates known in the model. Context s: system inv: The model, that s is an instance of, includes all object templates that s.objects are instances of.s.of. DBMS object

15

Templates->includes All (s.DBMS objects.of) The model, that s is an instance of, includes all DBMS object Class that s. DBMS Objects are instances of s.of.DBMS object Class ->includesAll(s.s. DBMS Objects.of). The third ensures that links are associated with roles known in the model. Context s: system inv: The model, that s is an instance of, includes all the role that s.links are instances of s.of.roles ->includesAll(s.roles.of) The fourth constraint ensures that the system cardinality constraints on roles are observed. Context s: system inv:The links of s respect cardinality constraints for their corresponding role s.links.of -> forAll( r | let links_in_s be r.instances >intersect ( s.links ) in ( r.upperBound -> notEmpty implies links_in_s ->size size >= r.upperbound) The fifth ensures that reverse links are in place for roles with inverses. If a link is of a role with an inverse, then there is a corresponding reverse link s.links->forAll (l | l.of.role.inverse ->notEmpty implies s.links ->select (l’ | l’.source=l.target & l’.target=l.source & l’.of = l..of.inverse) ->size=1.

5. DBMS Engineering Specifications. An engineering specification defines the infrastructure required to support functional distribution of an ODP system by: - Identifying the ODP functions necessary to manage physical distribution, communication, processing and storage. - Identifying the roles of different DBMS object supporting the ODP functions. In order to do this, we specify: 1. The activities that occur within those DBMS objects 2. The interaction of the DBMS objects (Fig.4). To achieve that, we respect the engineering language rules like interface reference rules, binding rules, cluster, capsule and node rules, etc

5.1. DBMS Object Activities The functions of a software entity are: - Transferring a software entity; - Creating a software entity; - Providing globally unique operator names and locations; - Supporting the concept of a domain; - Ensuring a secure environment for software entity operations. We specify these functions of a Software entity with the

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org ODP functions.

6.

5.2. DBMS Object Interactions We define three types of interactions related to interoperability: - Remote software entity creation; - Interaction needed for the software entity transfer; - Software entity method invocation.

Fig.4: organization of the DBMS objects

A client could be a non-software entity program or a software entity from a software entity having the same 7. system type as the destination agent or not. This client authenticates itself to the destination software entity system and interacts with the destination software entity to request the creation of a software entity. When a software entity transfers to another software entity, the software entity system creates a travel request providing information that identifies the destination place. In order to fulfill the travel request, the destination software entity transfers the software entity’s state, authority, security credential and the code. For example in database System server, a channel between system client manager Object and the system server Object can be defined as illustrated in Fig.5.

Fig.5: An example of a basic system client / system server channel.

A system client object invokes a method of another system client object or system server object if it has the authorization and a reference to the system client object.

16

Conclusion. In the present paper, we have seen that the Reference Model for Open Distributed Processing (RM-ODP) provides a framework which supports the integration of distribution, interworking and portability. In addition, the UML standard has adopted a meta-modeling approach to define the abstract syntax of UML. One approach to define the formal semantics of a language is denotational: essentially elaborating the value or instance denoted by an expression of the language in a particular context. However, the ODP viewpoint languages define what concepts should be supported and not how these concepts should be represented, So when we used the denotational meta-modeling approach in this paper, we have defined the UML/OCL based syntax and semantics of a language for a fragment of ODP object concepts described in the foundations part and in the Engineering viewpoint language. Indeed, these concepts are suitable to define and constrain ODP Engineering viewpoint specifications. In parallel, we have applied the same approach to define a language of DBMS concepts characterizing dynamic behavior of DBMS objects.

References [1] ISO/IEC, Basic Reference Model of Open Distributed Processing-Part1: Overview and Guide to Use, ISO/IEC CD 10746-1, July 1994. [2] ISO/IEC, RM-ODP-Part2: Descriptive Model, ISO/IEC DIS 10746-2, February 1994. [3] ISO/IEC, RM-ODP-Part3: Prescriptive Model, ISO/IEC DIS 10746-3, February 1994. [4] ISO/IEC, RM-ODP-Part4: Architectural Semantics, ISO/IEC DIS 10746-4, July 1994. [5] OMG, the Object Management Architecture, OMG, 1991. http://www.omg.org [6] Jalal Laassiri and al.“A Denotational Semantics of Concepts in ODP Information Language”, Proceedings of the International MultiConference of Engineers and Computer Scientists , London, UK, July 2009,pp 486-492 [7] ISO/IEC, ODP Type Repository Function, ISO/IEC JTC1/SC7 N2057, January 1999. [8] ISO/IEC, the ODP Trading Function, ISO/IEC JTC1/SC21, June 1995. [9] J.M. Spivey, The Z Reference manual, Prentice Hall, 1992. [10] IUT, SDL: Specification and Description Language, IUTT-Rec. Z.100, 1992. [11] ISO and IUT-T, LOTOS: A Formal Description Technique Based on the Temporal Ordering of Observational Behavior, ISO/IEC 8807, August 1998. [12] H. Bowman et al. FDTs for ODP, Computer Standards & Interfaces Journal, Elsevier Science Publishers, Vol.17, No.5-6, 1995, pp. 457-479. [13] J. Rumbaugh et al., The Unified Modeling Language, Addison Wesley, 1999.

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org [14] B. Rumpe, A Note on Semantics with an Emphasis on UML, Second ECOOP Workshop on Precise Behavioral Semantics, Technische Universitaty unchen publisher, 1998. [15] A. Evans et al., Making UML precise, OOPSLA'98, October 1998, [16] Evans et al. The UML as a formal notation, UML'98, France June 1998, LNCS 1618, Springer Berlin, 1999, pp. 336-348 [17] J. Warner and A. Kleppe, the Object Constraint Language: Precise Modeling with UML, Addison Wesley, 1998. [18] M. Bouhdadi et al, An UML-based Meta-language for the QoS-aware Enterprise Specification of Open Distributed Systems, IFIP TC5/WG5.5 Third Working Conference on Infrastructures for Virtual Enterprises (PRO-VE'02), May 13 Sesimbra Portugal, Kluwer Vol. 213 (IFIP Conference Proceeding series), 2002. [19] Jacques Saraydaryan and al, Comprehensive Security Framework for Global Threats Analysis, IJCSI Volume 2, August 2009. [20] S. Kent, S. Gaito, N. Ross. A meta-model semantics for structural constraints in UML,, In H. Kilov, B. Rumpe, and I. Simmonds, editors, Behavioral specifications for businesses and systems, chapter 9, pages 123-141. Kluwer Academic Publishers, Norwell, MA, September 1999. [21] D.A. Schmidt, Denotational semantics: A Methodology for Language Development, Allyn and Bacon, Massachusetts, 1986. [22] Myers, G. The art of Software Testing, John Wiley &Sons, New York, 1979 [23] Binder, R. Testing Object Oriented Systems. Models. Patterns, and Tools, Addison-Wesley, 1999 [24] Cockburn, A. Agile Software Development. AddisonWesley, 2002. [25] Bernhard Rumpe. Agile Modeling with UML. Habilitation Thesis, Germany, 2003. [26] Beck K. Column on Test-First Approach. IEEE Software, 18(5):87-89, 2001 [27] Briand L. and Labiche Y. A UML-based Approach to System testing. In M. Gogolla and C. Kobryn (eds): “UML” – The Unified Modeling Language, 4th Intl. Conference, LNCS 2185. Springer, 2001 pp. 194-208, [28] Bernhard Rumpe, Executable Modeling with UML. A vision or a Nightmare? In Issues & Trends of Information Technology Management in Contemporary Associations, Seattle. Idea group Publishing, Hershey, London, pp. 6977001. 2002 Author, Title of the Book, Publishing House, 200X. [29] A.Naumenko, A.Wegmann, “Proposal for a formal foundation of RM-ODP concepts » conference woodpecker 2001. [30] ISO/IEC, May 2006, Basic Reference Model of Open Distributed Processing-Use of UML for ODP system specifications, ISO/IEC CD 19793. [31] Naumenko, A., et al. A Viewpoint on Formal Foundation of RM-ODP Conceptual Framework, Technical report No. DSC/2001/040, July 2001, EPFL-DSC ICA. [32] Wegmann, A. and A. Naumenko. Conceptual Modeling of Complex Systems Using an RM-ODP Based Ontology. in 5th IEEE International Enterprise Distributed Object Computing Conference - EDOC 2001. 2001.

17

Received the B.S.and M.S. degree from the Jalal Laassiri: University of Mohamed V, Morocco, in 2002 and 2005 respectively. He is a Ph.D. student at the department of Computer Science of the Faculty of Sciences Rabat Morocco. His research interests include formal verification techniques and performance evaluation methods of concurrent and distributed systems with applications to computer and communication systems, In 2009, he received the Best Student Paper Awards of the WCE 2009.Is a member of the IAENG, Is a member of the ICIIC2010 International Program Committee.


18

Performance Comparison of SVM and ANN for Handwritten Devnagari Character Recognition Sandhya Arora1. Debotosh Bhattacharjee2, Mita Nasipuri2, L. Malik4 , M. Kundu2 and D. K. Basu3 1

Dept. of CSE & IT, Meghnad Saha Institute of Technology kolkata, 700150,India

2

Department of Computer Science and Engineering, Jadavpur University Kolkata, 700032,India

3

AICTE Emeritus Fellow, Department of Computer Science and Engineering, Jadavpur University Kolkata, 700032,India 4

Dept. of Computer science. G.H. Raisoni college of Engineering Nagpur, India

Abstract Classification methods based on learning from examples have been widely applied to character recognition from the 1990s and have brought forth significant improvements of recognition accuracies. This class of methods includes statistical methods, artificial neural networks, support vector machines (SVM), multiple classifier combination, etc. In this paper, we discuss the characteristics of the some classification methods that have been successfully applied to handwritten Devnagari character recognition and results of SVM and ANNs classification method, applied on Handwritten Devnagari characters. After preprocessing the character image, we extracted shadow features, chain code histogram features, view based features and longest run features. These features are then fed to Neural classifier and in support vector machine for classification. In neural classifier, we explored three ways of combining decisions of four MLP’s, designed for four different features. Keywords: Support Vector Machine, Neural Networks, Feature Extraction, Classification.

1. Introduction Over the years, computerization has taken over large number of manual operations, one such example is offline handwritten character recognition, which is the ability of a computer system to receive and interpret handwriting input present in the form of scanned images. In the early stage of OCR (optical character recognition) development, template matching based recognition techniques were used [16]. The templates or prototypes in these early methods were designed artificially, selected or averaged from few samples. As the number of samples increased, this simple design methodology, became insufficient to accommodate the shape variability of samples, and so, are not able to yield high recognition accuracies. To take full advantage of large volume of sample data, the character recognition

community has turned attention to classification methods based on learning from examples strategy, especially based on artificial neural networks (ANNs) from the late 1980s and the 1990s. New learning methods, using support vector machines (SVMs), are now actively studied and applied in pattern recognition problems. Learning methods have beneficiated character recognition methods tremendously. They relieve us from painful job of template selection and tuning, and the recognition accuracies get improved significantly because of learning from large sample data. Some excellent results have been reported [17, 18, 19]. Despite the improvements, the problem is far from being solved. The recognition accuracies of either machine-printed characters on degraded document image or freely handwritten characters are still insufficient. The recent spurt in the advancement in handwriting recognition has provided publications but do not involve the performance comparison of artificial neural networks and support vector machines on the same feature set for handwritten Devnagari characters. In this paper, we discuss the results of ANNs and SVM applied on handwritten Devnagari Characters. The strengths and weaknesses of these classification methods will also be discussed.

2. Challenges in Handwritten Devnagari Recognition The Devanagari script has descended from the Brahmi script sometime around the 11th century AD. It was originally developed to write Sanskrit but was later adapted to write many other languages like Bhojpuri, Bhili, Magahi, Maithili, Marwari, Newari, Pahari, Santhali, Tharu, Marathi, Mundari, Nepali and Hindi.


Hindi, is the official national language of India and also the third most popular language in the world. According to a recent survey, Hindi is being used by 551 million people in India. The basic set of symbols of Devnagari script consists of 36 consonants (or vyanjan) and 13 vowels (or swar) as shown in Figure 1. The characters may also have half forms. A half character in most of the cases is touched by the following character, resulting in a composite character , also known as compound character. The script has a set of modifier symbols which represent the modified shapes undertaken by the vowels, when they are combined with consonants, as own in Figure 2. These symbols are placed either on top, at the bottom, on the left, to the right or a combination of these. There are infinite variations of handwriting of individuals because of perceptual variability and generative variability. This variability effects in handwriting make the machine recognition of handwritten characters difficult. OCR of Devnagari script documents becomes further complicated due to the presence of compound characters and modifiers that make character separation and identification very difficult. All the individual characters are joined by a head line called “Shiro Rekha” in case of Devanagari Script. This makes it difficult to isolate individual characters from the words. There are various isolated dots, which are vowel modifiers, namely, “Anuswar”, “Visarga” and “Chandra Bindu”, which add up to the confusion. Ascenders and Descender recognition is also complex.

Figure 1: Vowels and Consonants of Devnagari Script

Figure 2: Vowels and corresponding modifiers of Devnagari Script

19

Figure 3: Combination of half consonant and consonant (compound characters)

3. State of the Art Here, we will confine the discussion strictly on, feature based classification methods for off-line Devnagari character recognition. These methods can be grouped into four broad categories namely, statistical methods, ANN based methods, Kernel based methods, and multiple classifier combination, which are discussed below. We discuss the work done on Devnagari characters, using some of these classification methods.

3.1 Statistical methods Statistical classifiers are rooted in the Bayes decision rule, and can be divided into parametric ones and nonparametric ones [20, 21]. Non-parametric methods, such as Parzen window, the nearest neighbor (1-NN) and kNN rules, the decision-tree, the subspace method, etc, are not much used, since all training samples are stored and compared. Sinha and Bansal [1,2] have explored various knowledge sources at all levels. Initial segmentation process use horizontal and vertical histogram for line and word separation. Horizontal zero crossings, moments, aspect ratio, pixel density in 9 zones, number and position of vertex points and structural description of character are taken as classifying features. Dictionary is used at post processing step. Sethi and Chatterjee [7] proposed a decision tree based approach for recognition of constrained hand printed Devnagari characters using primitive features. Parametric classifiers include the linear discriminant function (LDF), the quadratic discriminant function (QDF), the Gaussian mixture classifier, etc. An improvement to QDF, named regularized discriminant analysis (RDA), was shown to be effective to overcome inadequate sample size [50] and stabilizes the performance of QDF by smoothing the covariance matrices[25]. The modified quadratic discriminant function (MQDF) proposed by Kimura et al. was shown to improve the accuracy, memory, and computation efficiency of the QDF [22]. They used directional information obtained from arc tangent of the gradient. The directions are sampled down using Gaussian filter to get 392 dimensional feature-vector. This feature vector is


applied on MQDF classifier. For modeling multi-modal distributions, the mixture of Gaussians in high dimensional feature space does not necessarily give high classification accuracy, yet the mixture of linear subspaces has shown effects in handwritten character recognition [23, 24]. R. Kapoor, D. Bagai, T. Kamal, [6] proposed HMM based approach, using junction points of a character as the main feature. The character has been divided into three major zones. Three major features i.e. number of paths, direction of paths, and region of the node were extracted from the middle zone.

3.2 Artificial neural networks Feedforward neural networks, including multilayer perceptron (MLP) [51], radial basis function (RBF) network [52], the probabilistic neural network (PNN) [53], higher-order neural network (HONN) [54], etc., have been widely applied to pattern recognition. The connecting weights are usually adjusted to minimize the squared error on training samples in supervised learning. Using a modular network for each class was shown to improve the classification accuracy [26]. A network using local connection and shared weights, called convolutional neural network, has reported great success in character recognition [27]. Kumar and Singh [12] proposed a Zernike moment feature based approach for Devnagari handwritten character recognition. They used an artificial neural network for classification. Bhattacharya et al [13,15] proposed a Multi-Layer Perceptron (MLP) neural network based classification approach for the recognition of Devnagari handwritten numerals. S.Arora [55] proposed a MLP designed on on some statistical features for handwritten devnagari characters recognition. The RBF network can yield competitive accuracy with the MLP when training all parameters by error minimization [28]. The HONN is also called as functional-link network, polynomial network or polynomial classifier (PC). Its complexity can be reduced by dimensionality reduction before polynomial expansion [29] or polynomial term selection [30]. Vector quantization (VQ) networks and autoassociation networks, with the sub-net of each class trained independently in unsupervised learning, are also useful for classification. The learning vector quantization (LVQ) of Kohonen [31] is a supervised learning method and can give higher classification accuracy than VQ. Some improvements of LVQ learn prototypes by error minimization instead of heuristic adjustment [32].

3.3 Kernel based methods Kernel based methods, including support vector machines (SVMs) [33, 34] primarily and kernel principal component analysis (KPCA), kernel Fisher discriminant analysis (KFDA), etc., are receiving increasing attention and have shown superior performance in pattern

20

recognition. An SVM is a binary classifier with discriminant function being the weighted combination of kernel functions over all training samples. After learning by quadratic programming (QP), the samples of non-zero weights are called support vectors (SVs). For multi-class classification, binary SVMs are combined in either oneagainst-others or one-against-one (pair wise) scheme [35]. Due to the high complexity of training and execution, SVM classifiers have been mostly applied to small category set problems. A strategy to alleviate the computation cost is to use a statistical or neural classifier for selecting two candidate classes, which are then discriminated by SVM [30]. Dong et al. used a oneagainst-others scheme for large set Chinese character recognition with fast training [37]. They used a coarse classifier for acceleration but the large storage of SVs was not avoided.

3.4 Multiple classifier combination Combining multiple classifiers has been long pursued for improving the accuracy of single classifiers [38]. Parallel (horizontal) combination is more often adopted for high accuracy, while sequential (cascaded, vertical) combination is mainly used for accelerating large category set classification. The decision fusion methods are categorized into abstract-level, rank-level, and measurement-level combination [39, 40]. Many fusion methods have been proposed [41, 42]. The complementariness (also called as independence or diversity) of classifiers is important to yield high combination performance. For character recognition, combining classifiers based on different techniques of pre-processing, feature extraction, and classifier models is effective. Bajaj et al [11] employed three features namely, density features, moment features and descriptive component features for classification of Devnagari Numerals. They proposed multi-classifier connectionist architecture for increasing the recognition reliability for handwritten Devnagari numerals. S. Arora et al [49] has worked on Chain code histogram, view based, intersection, shadow, momentum based and some curve fitting based features and classifiers combination method for handwritten Devnagari characters. Another effective method, called perturbation, uses a single classifier to classify multiple deformations of the input pattern and combine the decisions on multiple deformations [43, 44]. The deformations of training samples can also be used to train the classifier for higher generalization performance [44, 27].

4. Performance comparison The experiments of character recognition reported in the literature vary in many factors such as the sample data, pre-processing technique, feature representation,


classifier structure and learning algorithm. Only a few works have compared different classification/learning methods based on the same feature data. In the following, we first discuss the extracted feature data used for all classifiers, and then summarize some classification results on this feature data. As handwritten Devnagari characters have wide applicability in India, we tested the performance on it. As till date no standard dataset is available for Handwritten Devnagari Characters we collected some samples from ISI, Kolkata and some samples we created within our organization. Total of 7154 data samples are used out of which 4900 samples are provided by ISI, Kolkata. The database contains 4900 training samples and 2254 test samples. Each sample was normalized to binary image of 100 X 100 pixels and the features discussed below are extracted. These extracted features are used in MLP and SVM classifiers, discussed in section 5.

4.1 Shadow Features of character Shadow is basically the length of the projection on the sides as shown in Figure 4. For computing shadow features [45] on scaled binary image, the rectangular boundary enclosing the character image is divided into eight octants. For each octant shadows or projections of character segment on three sides of the octant dividing triangles are computed so, a total of 24 shadow features are obtained. Each of these features is divided by the length of the corresponding side of the triangle to get a normalized value.

21

Given a scaled binary image, we first find the contour points of the character image. We consider a 3 × 3 window surrounded by the object points of the image. If any of the 4-connected neighbor points is a background point then the object point (P), as shown in Figure 5 is considered as contour point.

X

X

P

X

X

Figure 5. Contour point detection

The contour following procedure uses a contour representation called “chain coding” that is used for contour following proposed by Freeman[14], shown in Figure 6a. Each pixel of the contour is assigned a different code that indicates the direction of the next pixel that belongs to the contour in some given direction. Chain code provides the points in relative position to one another, independent of the coordinate system. In this methodology of using a chain coding of connecting neighboring contour pixels, the points and the outline coding are captured. Contour following procedure may proceed in clockwise or in counter clockwise direction. Here, we have chosen to proceed in a clockwise direction.

2

3

4

1

0

5

6 (a)

Figure 4. Shadow features

4.2 Chain Code Histogram of Character Contour

(c)

7

(b)

Figure 6. Chain Coding: (a) direction of connectivity, (b) 4connectivity, (c) 8-connectivity. Generate the chain code by detecting the direction of the next-in-line pixel


The chain code for the character contour will yield a smooth, unbroken curve as it grows along the perimeter of the character and completely encompasses the character. When there is multiple connectivity in the character, then there can be multiple chain codes to represent the contour of the character. We chose to move with minimum chain code number first. We divide the contour image in 5 × 5 blocks. In each of these blocks, the frequency of the direction code is computed and a histogram of chain code is prepared for each block. Thus for 5 × 5 blocks we get 5 × 5 × 8 = 200 features for recognition.

4.3 View based features This method is based on the fact, that for correct character-recognition a human usually needs only partial information about it – its shape and contour. This feature extraction method, which works on scaled, thinned binarized image, examines four “views” of each character extracting from them a characteristic vector, which describes the given character. The view is a set of points that plot one of four projections of the object (top, bottom, left and right) – it consists of pixels belonging to the contour of the character and having extreme values of one of its coordinates. For example, the top view of a letter is a set of points having maximal y coordinate for a given x coordinate. Next, characteristic points are marked out on the surface of each view to describe the shape of that view (Figure.7) The method of selecting these points and their number may vary and can be decided on experiment bases. In the considered examples, eleven uniformly distributed characteristic points are taken for each view.

22

4.4 Longest-run Features For computing longest-run features from a character image, the minimum square enclosing the image is divided into 25 rectangular regions. In each region, 4 longest-run features are computed row wise, column wise and along of its major diagonals. The row wise longest-run feature is computed by considering the sum of the lengths of the longest run bars that fit consecutive black pixels along each of all the rows of a rectangular region, as illustrated in Figure 8. The three other longestrun features are computed in the same way but along all column wise and two major diagonal wise directions within the rectangular separately. Thus in all, 25x4=100 longest-run features are computed from each character image.

Length of the Longest Bar

1 1 1 1 0 0

0 0 0 0 1 0

1 0 0 0 0 1

1 1 1 0 0 1

1 1 1 1 1 0

1 0 0 0 0 0

4 2 2 1 1 2

Sum=12 Figure 8: Longest run bar

5. Evaluated classifiers 5.1 Neural classifier

Figure 7. Selecting characteristic points for four views

The next step is calculating the y coordinates for the points on the top and down views, and x coordinates for the points on left and right views. These quantities are normalized so that their values are in the range . Now, from 44 obtained values the feature vector is created to describe the given character, and which is the base for further analysis and classification.

Combination of multiple classifiers is used to improve the accuracy in many pattern recognition tasks. We explored three ways of combining decision from multiple classifiers as a viable way of delivering a robust performance. All three ways are variations of majority voting scheme. We designed the same MLP with 3 layers including one hidden layer for four different feature sets consisting of 100 longest run features, 24 shadow features, 44 view based features and 200 chain code histogram features. The classifier is trained with standard back propagation. It minimizes the sum of squared errors for the training samples by conducting a gradient descent search in the weight space. As activation function we used sigmoid


function. Learning rate and momentum term are set to 0.8 and 0.7 respectively. As activation function we used the sigmoid function. Numbers of neurons in input layer of MLPs are 100, 24, 44 or 200, for longest run features, shadow features, view based features and chain code histogram features respectively. Number of neurons in Hidden layer is not fixed, we experimented on the values between 20-70 to get optimal result. The output layer contained one node for each class, so the number of neurons in output layer is 49. And classification was accomplished by a simple maximum response strategy.

5.1.1 Majority Voting Majority Voting systems for decision combination, choices between selecting either the “consensus decision” or the “decision delivered by the most competent expert” strategy. This section presents some of the principal techniques based on Majority Voting System. Max Voting: If there are n independent experts having the same probability of being correct and each of these experts produce a unique decision regarding the identity of the unknown sample, then the sample is assigned to the class for which all n experts agrees. Assuming that each expert makes a decision on an individual basis, without being influenced by any other expert in the decision making process. Min Voting: If there are n independent experts having the same probability of being correct and each of these experts produce a unique decision regarding the identity of the unknown sample, then the sample is assigned to the class for which any one of n experts agrees. Assuming that each expert makes a decision on an individual basis, without being influenced by any other expert in the decision making process

5.1.2 Weighted Majority Voting

23

therefore: dcom = max i=1,2,...,m dcom. We worked on four features namely: Chain code histogram, shadow, view based and longest run features so we have ω1, ω2, ω3 and ω4 as 0.316, 0.303, 0.241 and 0.140 where ωk is dk ωk = ------- -----------4

∑ dk k=1 where m = 49 and dk is the result of classifiers trained with Chain code histogram, shadow based, view based features and longest run features.

5.2 Support Vector Machines The objective of any machine capable of learning is to achieve good generalization performance, given a finite amount of training data, by striking a balance between the goodness of fit attained on a given training dataset and the ability of the machine to achieve error-free recognition on other datasets. With this concept as the basis, support vector machines have proved to achieve good generalization performance with no prior knowledge of the data. The principle of an SVM is to map the input data onto a higher dimensional feature space nonlinearly related to the input space and determine a separating hyperplane with maximum margin between the two classes in the feature space[47]. A support vector machine is a maximal margin hyperplane in feature space built by using a kernel function in gene space. This results in a nonlinear boundary in the input space. The optimal separating hyperplane can be determined without any computations in the higher dimensional feature space by using kernel functions in the input space. Commonly used kernels include:1. Linear Kernel :

A simple enhancement to the simple majority systems can be made if the decisions of each classifier are multiplied by a weight to reflect the individual confidences of these decisions. In this case, Weighting Factors, ωk, expressing the comparative competence of the cooperating experts, are expressed as a list of fractions, with 1 ≤ k ≤ n,



n k 1

ωk = 1, n being the

number of participating experts. The higher the competence, the higher is the value of ω. So if the decision by the kth expert to assign the unknown to the ith class is denoted by dik with 1 ≤ i ≤ m, m being the number of classes, then the final combined decision dicom supporting assignment to the ith class takes the form of: dicom =



k 1,2,...n

ωk * dik. The final decision dcom is

K(x, y) = x.y 2. Radial Basis Function (Gaussian) Kernel : K(x,y) = exp(-||x – y||2/2σ2) 3. Polynomial Kernel : K(x, y) = (x.y + 1)d An SVM in its elementary form can be used for binary classification. It may, however, be extended to multiclass problems using the one-against-the-rest approach or by using the one-against-one approach. We begin our experiment with SVM’s that use the Linear Kernel because they are simple and can be computed quickly. There are no kernel parameter choices needed to create a linear SVM, but it is necessary to choose a value for the soft margin (C) in advance. Then,given training data with feature vectors xi assigned to class yi € {-1,1} for


24

i = 1,……l, the support vector machines solve

Subject to yi(K(ω,xi)+b) ≥ 1 – ξi ξi ≥ 0 where ξ is an l-dimensional vector, and ω is a vector in the same feature space as the xi . The values ω and b determine a hyper plane in the original feature space, giving a linear classifier. A priori, one does not know which value of soft margin will yield the classifier with the best generalization ability. We optimize this choice for best performance on the selection portion of our data.

6. Performance Evaluation We tested the performance on Handwritten Devnagari characters. As till date no standard dataset is available for Handwritten Devnagari Characters we collected some samples from ISI, Kolkata and some samples we created within our organization. Total of 7154 data samples are used out of which 4900 samples are provided by ISI, Kolkata. We tested Neural Networks and Support vector machines on ISI data samples and also on our own created data samples. For ISI dataset we considered 3430 data samples for training and 1470 for testing and for our own dataset of 2254 samples, we considered 1470 data samples for training and 784 for testing. Some samples of Devnagari characters are given in fig.9 Table 1. Results of SVM and ANN on ISI dataset

Classifier SVM Multiple Neural Network Classifier Combination (ANNs)

Min Max Weighted majority

Test set 80.67% 60.92% 74.65% 70.38% (top1) 90.74% (top5)

Training set 94.77% 76.75% 84.12% 82.15%(top1) 93.31%(top5)

Table 2. Results of SVM and ANN on Our dataset

Classifier SVM Multiple Neural Network Classifier Combination (ANNs)

Min Max Weighte d majority

Test set 92.38% 78.49% 93.93% 90.44% (top1) 99.08% (top5)

Training set 99.62% 91.08% 98.24% 97.94%(top1) 99.51%(top5)

Figure 9: Some Devnagari Characters Sample set

7. Neural networks Vs. SVM Neural classifiers and SVMs show different properties in the following respects. Complexity of training. The parameters of neural classifiers are generally adjusted by gradient descent. By feeding the training samples a fixed number of sweeps, the training time is linear with the number of samples. SVMs are trained by quadratic programming (QP), and the training time is generally proportional to the square of number of samples. Some fast SVM training algorithms with nearly linear complexity are available, however. Flexibility of training. The parameters of neural classifiers can be adjusted in string-level or layout-level training by gradient descent with the aim of optimizing the global performance [48]. In this case, the neural classifier is embedded in the string or layout recognizer for character recognition. On the other hand, SVMs can only be trained at the level of holistic patterns. Model selection. The generalization performance of neural classifiers is sensitive to the size of structure, and the selection of an appropriate structure relies on crossvalidation. The convergence of neural network training suffers from local minima of error surface. On the other hand, the QP learning of SVMs guarantees finding the global optimum. The performance of SVMs depends on the selection of kernel type and kernel parameters, but this dependence is less influential. Classification accuracy. SVMs have been demonstrated superior classification accuracies to neural classifiers in many experiments. Storage and execution complexity. SVM learning by QP often results in a large number of SVs, which should be stored and computed in classification. Neural classifiers have much less parameters, and the number of


parameters is easy to control. In a word, neural classifiers consume less storage and computation than SVMs.

8. Conclusion The result obtained for recognition of Devnagari characters show that reliable classification is possible using SVMs. We applied SVMs and ANNs classifiers on same feature data namely Shadow based, Chain code Histogram, Longest Run, and View based features. The SVM-based method described here for offline Devnagari can be easily extended to other Indian scripts and Handwritten Devnagari numerals also. Acknowledgments Authors are thankful to the “Centre for Microprocessor Application for Training Education and Research” and “Project on Storage Retrieval and Understanding of Video for Multimedia”, at the Department of Computer Science and Engineering, Jadavpur University, Kolkata700032 for providing the necessary facilities for carrying out this work. Authors are also thankful to the CVPR Unit, ISI Kolkata for providing the dataset of Handwritten Devnagari Characters. First author gratefully acknowledge the support of the Meghnad Saha Institute of Technology for carrying out this research work.

References [1] Bansal V, Sinha R. M. K., ”Integrating Knowledge Resources in Devanagari. Text Recognition System,” IEEE Transaction on System, Man & Cybernetics Part A: Systems & Humans, Vol 30 , 4 July 2000,pp 500-505. [2] R.M.K. Sinha, Veena Bansal, Thesis on “Integrating Knowledge Sources in Devnagari Text Recognition”, Ph.D. thesis, Indian Institute of Technology, Kanpur, India, March 1999. [3] Kompalli, S. Nayak, S. Setlur, V. Govindaraju, “Challenges in OCR of Devnagari Documents”, ICDAR 05 [4] B. Shaw, S. Pauri, M. Shridhar, “Off Line Handwritten Word Recognition: A Segmentation Based Approach” ,LN CS Springer, ISSN 0302-9743, Vol 4815/2007, pp 528-535. [5] N. Sharma, U. Pal, F. Kimura, S. pal, “Recognition of Off Line Handwritten Devnagari Characters using Quadratic Classifier”, ICCGIP 2006, LNCS 4338, pp 805-816, 2006. [6] R. Kapoor, D. Bagai and T.S. Kamal, “Representation and Extraction of Nodal Features of DevNagri Letters”, Proceedings of the 3rd Indian Conference on Computer Vision, Graphics and Image Processing [7] I.K. Sethi and B. Chatterjee, “Machine Recognition of Constrained Hand Printed Devnagari”, Pattern Recognition, Vol. 9, pp. 69-75, 1977. [8] U. Pal and B.B. Chaudhuri, “Indian script character recognition: ASurvey”, Pattern Recognition, Vol. 37, pp. 1887-1899, 2004. [9] B. B. Chaudhuri and U. Pal, “A complete printed Bangla OCR system”,Pattern Recognition, vol. 31, pp. 531-549, 1998.

25

[10] M. Hanmandlu and O.V. Ramana Murthy, “Fuzzy Model Based Recognition of Handwritten Hindi Numerals”, Intl.Conf. on Cognition and Recognition, pp. 490-496, 2005. [11] Reena Bajaj, Lipika Dey, and S. Chaudhury, “Devnagari numeral recognition by combining decision of multiple connectionist classifiers”,Sadhana, Vol.27, part. 1, pp.-59-72, 2002 [12] Satish Kumar and Chandan Singh, “A Study of Zernike Moments and its use in Devnagari Handwritten Character Recognition”, Intl.Conf. on Cognition and Recognition, pp. 514- 520, 2005. [13] U. Bhattacharya, B. B. Chaudhuri, R. Ghosh and M. Ghosh, “On Recognition of Handwritten Devnagari Numerals”, In Proc. of the Workshop on Learning Algorithms for Pattern Recognition (in conjunction with the 18th Australian Joint Conference on Artificial Intelligence), Sydney, pp.1-7, 2005. [14] Freeman, H., On the Encoding of Arbitrary Geometric Configurations, IRE Trans. on Electr. Comp. or TC(10), No. 2, June, 1961, pp. 260-268. [15] J. Hertz, A. Krogh, R.G. Palmer, ”An Introduction to neural Computation”, Addison-Wesley (1991) [16] S. Mori, C.Y. Suen, K.Yamamoto, Historical review of OCR research and development, Proc. IEEE, 80(7):1029-1058, 1992. [17] Y. LeCun, L. Bottou, Y. Bengio, P. Haffner, Gradient-based learning applied to document recognition, Proc.IEEE, 86(11): 2278-2324, 1998. [18] C.Y. Suen, K. Kiu, N.W. Strathy, Sorting and recognizing cheques and financial documents, Document Analysis Systems: Theory and Practice, S.-W. Lee andY. Nakano (eds.), LNCS 1655, Springer, 1999, pp. 173-187. [19] C.-L. Liu, K. Nakashima, H. Sako, H. Fujisawa, Handwritten digit recognition: benchmarking of state-of-the-art techniques, Pattern Recognition, 36(10): 2271-2285, 2003. [20] K. Fukunaga, Introduction to Statistical Pattern Recognition, 2nd edition, Academic Press, 1990. [21] R.O. Duda, P.E. Hart, D.G. Stork, Pattern Classification, second edition, Wiley Interscience, 2001. [22] F. Kimura, K. Takashina, S. Tsuruoka, Y. Miyake,Modified quadratic discriminant functions and the application to Chinese character recognition, IEEETrans. Pattern Anal. Mach. Intell., 9(1): 149-153,1987. [23] G.E. Hinton, P. Dayan, M. Revow, Modeling the manifolds of images of handwritten digits, IEEE Trans.Neural Networks, 8(1): 65-74, 1997. [24]H.-C. Kim, D. Kim, S.Y. Bang, A numeral character recognition using the PCA mixture model, PatternRecognition Letters, 23: 103-111, 2002. [25] J.H. Friedman, Regularized discriminant analysis, J.Am. Statist. Ass., 84(405): 165-175, 1989. [26] I.-S. Oh, C.Y. Suen, A class-modular feedforward neural network for handwriting recognition, PatternRecognition, 35(1): 229-244, 2002. [27] P.Y. Simard, D. Steinkraus, J.C. Platt, Best practices for convolutional neural networks applied to visual document analysis, Proc. 7th ICDAR, Edinburgh,UK, 2003, Vol.2, pp.958962. [28] C.M. Bishop, Neural Networks for Pattern Recognition, Claderon Press, Oxford, 1995. [29] U. Kre¼el, J. SchÄurmann, Pattern classification techniques based on function approximation, Handbook of Character Recognition and Document Image Analysis,H. Bunke and P.S.P. Wang (eds.), World Scientific,1997, pp.49-78. [30] J. Franke, Isolated handprinted digit recognition,Handbook of Character Recognition and Document Image Analysis, H. Bunke and P.S.P. Wang (eds.), World Scientific, 1997, pp.103-121. [31] T. Kohonen, The self-organizing map, Proc. IEEE, 78(9): 14641480, 1990. [32] C.-L. Liu, M. Nakagawa, Evaluation of prototype learning algorithms for nearest neighbor classifier in application to handwritten character recognition, Pattern Recognition, 34(3): 601-615, 2001.

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org [33] V. Vapnik, The Nature of Statistical Learning Theory, SpringerVerlag, New Work, 1995. [34] C.J.C. Burges. A tutorial on support vector machines for pattern recognition, Knowledge Discovery and Data Mining, 2(2): 1-43, 1998. [35] U. Kressel, Pairwise classification and support vector machines, Advances in Kernel Methods: Support Vector Learning, B. SchÄolkopf, C.J.C. Burges, A.J. Smola (eds.), MIT Press, 1999, pp.255-268. [36] A. Bellili, M. Gilloux, P. Gallinari, An MLP-SVM combination architecture for o²ine handwritten digit recognition: reduction of recognition errors by support vector machines rejection mechanisms, Int. J. Document Analysis and Recognition, 5(4): 244-252, 2003. [37] J.X. Dong, A. Krzyzak, C.Y. Suen, High accuracy handwritten Chinese character recognition using support vector machine, Proc. Int. Workshop on ArtificialNeural Networks for Pattern Recognition, Florence, Italy, 2003. [38] A.F.R. Rahman, M.C. Fairhurst, Multiple classifier decision combination strategies for character recognition: a review, Int. J. Document Analysis and Recognition, 5(4): 166-194, 2003. [39] L. Xu, A. Krzyzak, C.Y. Suen, Methods of combining multiple classifiers and their applications to handwriting recognition, IEEE Trans. System Man Cybernet.,22(3): 418-435, 1992. [40] C.Y. Suen, L. Lam, Multiple classifier combination methodologies for different output levels, Multiple Classifier Systems, J. Kittler and F. Roli (eds.), LNCS 1857, Springer, 2000, pp.5266. [41] J. Kittler, M. Hatef, R.P.W. Duin, J. Matas, On combining classifiers, IEEE Trans. Pattern Anal. Mach.Intell., 20(3): 226239, 1998. [42] R.P.W. Duin, The combining classifiers: to train or not to train, Proc. 16th ICPR, Quebec, Canada, 2002, Vol.2, pp.765-770. [43] T. Ha, H. Bunke, Off-line handwritten numeral recognition by perturbation method, IEEE Trans. Pattern Anal. Mach. Intell., 19(5): 535-539, 1997. [44] J. Dahmen, D. Keysers, H. Ney, Combined classification of handwritten digits using the virtual test sample method, Multiple Classifier Systems, J. Kittler and F.Roli (eds.), LNCS 2096, Springer, 2001, pp.99-108. [45] S. Basu, N.Das, R. Sarkar, M. Kundu, M. Nasipuri, D.K. Basu, “Handwritten Bangla alphabet recognition using MLP based classifier”, NCCPB, Bangladesh, 2005 [46] M. Hanmandlu, O.V. Ramana Murthy, Vamsi Krishna Madasu, “Fuzzy Model based recognition of Handwritten Hindi characters”, IEEE Computer society, Digital Image Computing Techniques and Applications , 2007 [47] C. J. C. Burges, .A tutorial on support vector machines for pattern recognition., Data Mining and Knowledge Discovery, 1998, pp 121-167. [48] C.-L. Liu, K. Marukawa, Handwritten numeral string recognition: character-level training vs. string-level training, Proc.7th ICPR, Cambridge, UK, 2004,Vol.1, pp.405-408. [49] S.Arora, D. Bhattacharjee, Mita Nasipuri, M. Kundu, D.K. Basu, “Combining Multiple Feature extraction techniques for Handwritten Devnagari Character Recognition”, ICIIS, IEEE International Conference, IITKGP, 978-1-42442806-9/08/$25.00© 2008 IEEE [50] H. Friedman: Regularized discriminant analysis. J. Am. Stat. Assoc. 84(405):166–175 (1989) [51] D.E. Rumelhart, G.E. Hinton, R.J. Williams: Learning representations by back-propagation errors. Nature 323(9):533– 536 1986) [52]C.M. Bishop: Neural networks for pattern recognition.Clarendon, Oxford (1995) [53] D.F. Specht: Probabilistic neural networks. Neural Networks 3:109–118 (1990) [54] U. Kreßel, J. Schurmann: Pattern classification techniques based on function approximation. In: H. Bunke,P.S.P. Wang (eds.),

26

Handbook of Character Recognition and Document Image Analysis, World Scientific, Singapore, 1997, pp. 49–78 [55] S. Arora, D. Bhattacharjee, M. Nasipuri, M.Kundu, D.K. Basu, “Application of Statistical Features in Handwritten Devnagari Character Recognition”, International Journal of Recent Trends in Engineering[ISSN 1797-9617], IJRTE Nov 2009

SANDHYA ARORA has completed M.Tech. (Computer Science & Engineering) from Banasthali Vidyapith, Rajasthan, India and B.E.(Computer Engineering) from University of Rajasthan, India .She is currently working as Assistant Professor in Department of Computer Science & Engineering at Meghnad Saha Institute of Technology, Kolkata, WB, India. She has teaching experience of 13 years. Mrs.Sandhya Arora presented 5 papers in national and 4 papers in international conferences She is life member of CSI. DEBOTOSH BHATTACHARJEE received the MCSE and Ph.D. (Eng.) degrees from Jadavpur University, India, in 1997 and 2004 respectively. He was associated with different institutes in various capacities until March 2007. After that he joined his Alma Mater, Jadavpur University. His research interests pertain to the applications of computational intelligence techniques like Fuzzy logic, Artificial Neural Network, Genetic Algorithm, Rough Set Theory, Cellular Automata etc. in Face Recognition, OCR, and Information Security. He is a life member of Indian Society for Technical Education (ISTE, New Delhi), Indian Unit for Pattern Recognition and Artificial Intelligence (IUPRAI), and member of IEEE (USA). MITA NASIPURI received his B.E.Tel.E., M.E.Tel.E. and Ph.D. (Eng.) degrees from Jadavpur University, in 1979,1981 and 1990 respectively. Prof. Masipuri has been a faculty member of J.U. since 1987. Her current research interest include pattern recognition, image processing and multimedia systems. She is a senior member of the IEEE, U.S.A., Fellow of I.E. (India) and W.B.A.S.T., kolkata , India. MAHANTAPAS KUNDU received his B.E.E., M.E.Tel.E. and Ph.D.(Eng.) degrees from Jadavpur University, in 1983,1985 and 1995respectively. Prof. Kundu has been a faculty member of J.U. since 1988. His areas of current research interest include pattern recognition, image processing , multimedia database, and artificial intelligence. DIPAK KUMAR BASU received his B.E.Tel.E., M.E.Tel., and Ph.D. (Eng.) degrees from Jadavpur University, in 1964,1966 and 1969 respectively. Prof. Basu has been a faculty member of J.U.since 1968. His current fields of research interest include pattern recognition, image processing, and multimedia systems. He is a senior member of the IEEE, U.S.A., Fellow of I.E. (India) and W.B.A.S.T., kolkata , India and a former Fellow, Alexander von Humboldt Foundation, Germany. LATESH MALIK became a Member (M) of IEEE in 2006. She has completed M.Tech. (Computer Science & Engineering) from Banasthali Vidyapith, Rajasthan, India and B.E. (Computer Engineering) from University of Rajasthan,India . She is gold medallist in B.E. and M.Tech. She is currently working as Assistant Professor & Head of Department in Department of Computer Science & Engineering at G.H. Raisoni College of Engineering, Nagpur University, Nagpur, MS, India. She has teaching experience of 13 years.Mrs. Latesh Malik is life member of ISTE and presented 5 papers in international journal, 8 papers in national and 20 papers in international conferences.


27

Improving Network Performance using ACO Based Redundant Link Avoidance Algorithm Chandra Mohan B1, and Baskaran R2 1

2

Principal, John Bosco Engineering College Chennai, Tamil Nadu, India

Assistant Professor, Department of CSE, Anna University Chennai, Tamil Nadu, India

Abstract In the wide spread internet, response time and pocket loss are inappropriate due to network traffic, as a result the network efficiency becomes worst and the system provides poor Quality of Service (QoS). An optimal routing protocol, especially multipath may avoid such traffic in the network. But existing routing protocols, both single path and multi path, concentrates only on finding the routes based on any one or some set of metrics, that not always suitable for dynamic, cloud natured network environment. Ant Colony Optimization (ACO) based multipath routing protocol was suggested as an alternate to this problem by many researchers. The multipath ACO also provides same set of link(s) for the source to destination, so that traffic merging again becomes a critical problem. This paper proposes an optimal solution to avoid the problem of traffic merging in the network by removing redundant link in the route. The Proposed algorithm, called ‘Redundant Link Avoidance (RLA) algorithm’, is an ACO based multi path routing methodology, avoiding copious link in the suggested routes of ACO multipath protocol.

Keywords: Computer Networks, Routing, QoS, Swarm

[2]. It is unambiguous from these experimental results that the Average packet loss is reaches up to 17%. When considering response time, less than 150ms is an appreciate value for data communication, but all continents except North America have more than 150ms. Especially in Australia, the packet loss is 0% but response time is 169ms. Table 1 Average Response Time and Average Packet Loss of each continent Avg. Response Time (in ms)

Avg. Packet Loss (%)

Asia

386

17%

Australia

169

0%

Europe

205

5%

North America

142

8%

South America

138

0%

Continent

Intelligence, Ant Colony Optimization.

1. Introduction In October 1986, network had a series of congestion collapses, the data throughput from Lawrence Berkley Lab to University of California, sites separated by only two IMP hops, dropped from 32 kbps to 40 bps, i.e., 99.88% packet drop [1]. Due to this incident, the traffic management in computer network becomes critical problem in data communication. Several research efforts have proposed different approaches for traffic management problems, each one has its own strengths and limitations but optimality not yet achieved. The following Table 1 shows the average response time and the average packet loss of each continent recorded on 12th February of 2010 at 12:40 Indian Standard Time (IST)

The loss of data packets and delay in data transmission in the computer network becomes obvious problem due to these reasons: {a} Bottleneck problem due to the growth of networks and increased link speeds; {b} Packet arrival rate exceeds the outgoing link capacity; {c} insufficient memory or buffer to store arriving packets, i.e., buffer of receiving node is not sufficient to store the incoming data packets; {d} network Burst; {e} Demand temporarily exceeds available resources at some point in the network; {f} Same route is used for more transmission, overloading; and {g} traffic merging. Congestion in the computer network may be controlled [3] either by traffic management algorithms and / or by

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org optimal routing. Traffic management provide optimal solution to solve the facts {a – c} but optimal routing may solve all the listed facts. All the existing traffic management and routing protocol concentrates to avoid the traffic by the way of solving one or few facts in the {a – e} except {f} and {g}. In which, the facts {d} and {e} may be controlled, only by multipath routing algorithm but when applying multipath there are high probability that the network becomes congested by the facts {f} and {g}. The combination of {f} and {g}, i.e., when two or more sources would simultaneously try to transmit packets to one destination via a single link, a high probability that the number of packets would exceed the packets handling capacity of the network and lead to heavy congestion. In this paper, the proposed algorithm, swarm intelligence based multi path routing protocol with Redundant Link Avoidance (RLA) algorithm, focused to solve this important facts {f}, {g} and also the other facts {a-e}. This routing methodology ensures traffic free routing by avoiding copious link. Perspective of this routing methodology is sharing load on multiple routes to obtain optimal response time and at the same time avoiding packet drop due to overloading in the network. The remaining of this paper is organized in the following manner: Survey of various related work, its classification, and existing algorithm is explained in the Section Two. The proposed work and the algorithm used in this paper are described in Section Three. Conclusion and further studies are discussed in Section Five followed by the performance analysis of the proposed work is shown in Section Four.

2. Related Work The basic problem of routing is, to find the path of lowest cost between any two nodes. Routing algorithms can be grouped into three major classes [4]: non-adaptive, adaptive and distributed. Non-adaptive algorithm, also static routing is computed in advance, offline using local information, their routing decision not based on measurements or estimates of the current traffic and topology. Adaptive algorithms, centralized routing will collect information from the entire subnet using global information in an attempt to make optimal solution. Distributed algorithm uses a mixture of local and global information. Based on the discovery of no of path, the routing algorithms further divided as single-path and multi-path. Single path routing algorithm chooses only one ‘best’ shortest route between a given pair of source to destination for all of its traffic. Multi-path, the bifurcated routing

28

algorithm finds several shortest paths that are almost equally good, between given source to destination. As the traffic is split over several paths, better performance can be obtained in multi-path routing. There are several research papers [5 - 13] implemented variety of multi-path algorithms for various applications, in which Ant Colony Optimization (ACO) based multipath [13] routing provide better result than others due to its real time computation. In which, [5], [6] are few application of multi path that provides better performance in multimedia and data communication. In order to offer high video quality, a certain minimum end-to-end bandwidth has to be provided. In bandwidth-limited networks, it is difficult to guarantee such a bandwidth with a single path. To meet the bandwidth requirement, a multipath approach is proposed by Jiancong Chen et al [5]. Ka-Cheong Leung et al [6], explains implementation of multipath for avoiding two different traffic types, connection-oriented traffic and mixed traffic. Connection-oriented traffic consists of traffic from TCP connections only, while connectionless traffic comes from non-TCP connections. Mixed traffic is composed of both connectionless and connection-oriented traffic. For both multimedia and data communication, multipath avoids traffic induced by the facts {a – e} shown in section 1, but there are some more situations {f – g} that the network becomes congested when implementing multi path. This paper, further continuing with literature survey of the various multipath routing algorithms developed recently. The selection of the routing paths is major design consideration that has a drastic effect on the resulting performance. Therefore, although many flow-control algorithms are optimal for a given set of routing paths, their performance can significantly differ for different sets of paths. Previous studies and proposals on multi-path routing in the previous context have focused on heuristic methods. In [7], a multi-path routing scheme, termed equal cost multi-path (ECMP), has been proposed for balancing the load along multiple shortest paths using a simple round-robin distribution. By limiting itself to shortest paths, ECMP considerably reduces the load balancing capabilities of multi-path routing; moreover, the equal partition of flows along the shortest paths, resulting from the round robin distribution, further limits the ability to decrease congestion through load balancing. OSPF-OMP [8] allows splitting traffic among paths unevenly; however, the traffic distribution mechanism is based on a heuristic scheme that often results in an inefficient flow distribution. A. E. I. Widjaja [9] considered multi-path routing as an optimization problem with an objective function that minimizes the congestion

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org of the most utilized link in the network; however, they focused on heuristics and did not consider the quality of the selected paths. Israel Cidon et al [10] concluded two phenomena in the implementation of multi path routing protocol: 1) as the link capacity increases, the relative performance of multipath routing decreases, 2) as the number of possible routes increases, the relative performance of multi-path routing also increases till k number of paths and when it exceeds the limit, the performance will be degraded. So to choose only k path is an important consideration for implementing multi path routing and the optimal value of k, may change in practice. Weijia Jia et al [11] proposed an approach that named as integrated routing, to select a subset of routers to carry out multi path routing (MPR), the rest of the routers execute Single path routing (SPR). With proper selection of MPR routers, shortcomings in the MPR will be rectified. This approach may provide better result but to implement this protocol in the network requires two tasks: router assignment and routing interface selection. The performance of this approach lies only in these two tasks and implementing these two tasks in the internet like environment is highly impractical. Dijkstra-old-touch-first (Dijkstra-OTF) with multipath extension [12] is an extended version of conventional Dijikstra’s shortest path algorithm that computing all lexicographic- lightest paths from a source to every other node in the network, but it requires additional computational efforts. DiCaro et al [13] developed Swarm Intelligence (SI) based routing algorithm, called Ant-Net Routing using ant colony optimization technique with multi path capability. The ACO era starts with Schoonderwoerd et al [14], they proposed Ant based control (ABC) for telephone network, it continues with ant based algorithm for packet switched network proposed by Subramanian et al [15] and a co-operative Asymmetric Forward (CAF) model for routing with asymmetric costs are ant colony optimization based single path routing algorithm developed by Heusse et al [16]. ACO is a stochastic approach that has been proposed to solve different hard combinatorial optimization problems in almost all engineering fields such as traveling salesman problems, graph coloring problems, quadratic assignment problems, resource-constrained project scheduling problem [17], knowledge representation tools[18], Computer-assisted testing systems[19], Grid Workflow Scheduling Problem[20], Self-Structured P2P Information System[21], Intelligent transportation systems[22], Robot Path Integration[23], and Clustering and Data mining[24].

29

Initially, three different versions of AS were proposed, namely ant density, ant quantity and ant cycle [25]. In the ant density and ant quantity versions, the ants updated the pheromone directly after a move from one city to an adjacent city, in the ant cycle version the pheromone update was only done after all the ants had constructed the tours and the amount of pheromone deposited by each ant was sent to be a function of the tour quality. Now a day, when referring to ACO, it means to ant cycle since the two other variants were abandoned because of their inferior performance. Solution construction and Pheromone Update are the two phases of ACO implementation. In ACO based routing algorithm, solution construction handles formation of routing table, Pheromone update used to choose the single optimal path based on density of pheromone, in case of single path routing; and the pheromone update desires the probability of load sharing along multiple path, in case of multi path routing. The ACO routing may provide feasible solution for traffic management problem, even though the overloading {f} and traffic merging {g} still is an issue in computer network. This paper concentrates to solve overloading and traffic merging by implementing RLA algorithm as creamy layer over existing ACO. The following section describes our proposed work in detail.

3. Proposed Work The proposed routing protocol has two components, namely ACO Based Multipath Routing algorithm and redundant link avoidance algorithm. The main idea of ACO is to model the problem as the search for a minimum cost path in a graph. Artificial ants walk through this graph, looking for good paths. Each ant has a rather simple behavior so that it will typically only find rather poor-quality paths on its own. Better paths are found as the emergent result of the global cooperation among ants in the colony. The behavior of artificial ants is inspired from real ants, they lay pheromone trails on the graph edges and choose their path with respect to probabilities that depend on pheromone trails and these pheromone trails progressively decrease by evaporation. In addition, artificial ants have some extra features that do not find their counterpart in real ants. In particular, they live in a discrete world and their moves consist of transitions from nodes to nodes. Also, they are usually associated with data structures that contain the memory of their previous actions. In most cases, pheromone trails are updated only after having constructed a complete path and not during the walk, and the amount of pheromone deposited is usually a function of the quality of the path. Finally, the probability for an

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org artificial ant to choose an edge often depends not only on pheromones, but also on some problem-specific local heuristics. The ACO based multipath algorithm implemented in this paper, using forward and backward ants. Ants in each set possess the same structure, but they are differently situated in the environment, i.e., they can sense different inputs and they can produce different, independent outputs. Ants communicate in an indirect way, according to the dynamic paradigm, through the information they concurrently read and write on the network nodes they visit. While travelling toward their destination nodes, the forward ants keep memory of their paths and of the traffic conditions found. In single path routing protocol, an optimal path recorded by the ant was chosen as path between given source to destination. In multi path routing protocol, all paths are chosen for communicating from source to destination. As discussed in the previous section, some of the available path in the multi path routing protocol, heavily congested due to traffic merging problem. The redundant link avoidance algorithm removes such links called replicating links from destination towards source using two data structures named available route array and selected route array. All available routes found by multipath routing protocol from source to destination is stored in the available routes, a two dimensional array, followed by initializing a selected route (one more two dimensional array), and two queues namely selected_ relationship and available_ relationship. The first optimal route is copied from available route to selected route, then all links of this route is divided from destination towards source and stored into selected_ relationship queue and the same was removed available route. The same way, next route from available route is selected and divided. If front of available_relationship and front of selected_relationship are equal then this route is removed from available route, and deleted all association from available_relationship and the previous process is repeated until reaches the tail. If front of available_relationship and front of selected_relationship are not equal then next link is processed, and the above

30

step is carried out until front reaches rear of the queue. The detailed algorithm is shown in the following section 3.1 and various test cases for the proposed work is shown in section 3.2 to 3.4.

3.1 Redundant Link Avoidance (RLA) Algorithm: Procedure RLA(t, tend, t) input t // current time input tend // time length of simulation input t // time interval between ants generation // concurrent activity over the network for each i  C do M  InitLocalTrafficModel T A  InitNodeAvailRoutingTable TS  InitNodeSelectedRoutingTable // concurrent activity on each node while t  tend do in_parallel if (t mod t) = 0 then destinationSelectDestination (traffic_distribution_at_source) Launch_Forward_Ant (source, destination) end-if for each (Active_Forward_Ant [source, current, destination] ) do while (current  destination) do next_hop  SelectLink (current, destination, link_queues, T A) PutAntOnLinkQueue (current, next_hop) WaitOnDataLinkQueue(current, next_hop) CrossLink (current, next_hop) Memorize (next_hop, elapsed_time) current  next_hop end-while

1.3 1.1

1.2

Fig 1. Data Flow in ACO based Routing Where 1.1 Forward Ant Generation 1.2 Pheromone Matrix Formation 1.3 Removing Cyclic Path 1.4 Backward Ant Generation 1.5 Formatting Routing Table

1.4

1.5

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org LaunchBackwardAnt memory_data)

(destination,

31

source,

end for each

end-procedure

3.2 Test case 1:

for each (ActiveBackwardAnt [source, current, destination]) do while (current  destination) do next_hop  PopMemory WaitOnHighPriorityLinkQueue (current, next_hop) CrossLink (current, next_hop) from  next_hop current  next_hop UpdateTrafficModel (M,current,from,source, memory_data) rGetNewPheromone (M, current, from, source, memory data) UpdateLocalRoutingTable (T A, current, source, r) end-while end-foreach end-in_parallel T S [0] = T A [0] T SL = each link from Destination to source on T S for each (T A [source, intermediate, destination]) do T AL[] = each link from Destination to source on current T A[] Current1  First Link in T AL Current2  First Link in T SL for each (Current1  Null, Current2  Null) if (Current1  Current2) do Current1  next Link in T AL Current2  next Link in T SL else-if Flag = 0 end-if end-foreach if (Flag = 1) T S = T A [current] end-if end-for each end-while end-for each

Fig 2 Data flow from source 13 to destination 5 (Test Case 1) using three selected path

Let consider Fig 2, the following table shows the available routes, directed paths of existing ACO and selected routes, directed routes of ACO with RLA from source 13 to destination 5: Available Route (ACO):  13 – 7 – 8 – 1 – 6 – 5  13 – 7 – 8 – 1 – 3 – 5  13 – 7 – 8 – 1 – 2 – 5  13 – 7 – 8 – 1 – 3 – 2 – 5  13 – 7 – 8 – 1 – 3 – 6 – 5  13 – 7 – 8 – 1 – 6 – 3 – 5  13 – 7 – 8 – 1 – 2 – 3 – 5  13 – 7 – 8 – 1 – 6 – 3 – 2 – 5  13 – 7 – 8 – 1 – 2 – 3 – 6 – 5  13 – 7 – 12 – 10 – 8 – 1 – 6 – 5  13 – 7 – 12 – 10 – 8 – 1 – 3 – 5  13 – 7 – 12 – 10 – 8 – 1 – 2 – 5  13 – 7 – 12 – 10 – 8 – 1 – 3 – 2 – 5  13 – 7 – 12 – 10 – 8 – 1 – 3 – 6 – 5  13 – 7 – 12 – 10 – 8 – 1 – 6 – 3 – 5  13 – 7 – 12 – 10 – 8 – 1 – 2 – 3 – 5  13 – 7 – 12 – 10 – 8 – 1 – 6 – 3 – 2 – 5  13 – 7 – 12 – 10 – 8 – 1 – 2 – 3 – 6 – 5 Selected Route (ACO with RLA):  13 – 7 – 8 – 1 – 6 – 5  13 – 7 – 8 – 1 – 3 – 5  13 – 7 – 8 – 1 – 2 – 5 If packets are transmitted in all above 18 routes of available routes, then the link 3 – 5, 2 – 3 and 6 – 5 become heavy loaded and it may lead to packet drop. In order to avoid traffic on these links, our redundant link avoidance algorithm removes such link from routing table.


32

3.4 Test case 3: 3.3 Test case 2: The following table 3, shows the available routes, directed routes of existing ACO and selected routes, directed routes of ACO with RLA In multiple route routing the following are available routes from source 11 to destination 4: Available Route (ACO):  11 – 9– 14– 0 – 2– 5– 4  11– 9– 14– 0 – 2 – 3– 5 – 4  11 – 9– 14– 0 – 2– 1– 6– 5– 4  11 – 9 – 14 – 0 – 2 – 1 – 3 – 5 – 4  11 – 9 – 14 – 0 – 2 – 3 – 6 – 5 – 4  11 – 9 – 14 – 0 – 2– 1 – 6 – 3 – 5 – 4  11 – 9 – 14 – 0 – 2 – 3 – 1 – 6 – 5 – 4  11 – 9 – 14 – 0 – 2 – 1 – 3 – 6 – 5 – 4

The following table 4, shows the available routes, directed routes of existing ACO and selected routes, directed routes of ACO with RLA In multiple route routing the following are available routes from source 0 to destination 1: Available Route (ACO):  0–2– 1  0 – 2 – 3 –1  0–2–5–6–1  0–2– 3–6– 1  0–2– 5–3– 1 Selected Route (ACO with RLA):  0–2– 1  0 – 2 – 3 –1  0–2–5–6–1

Selected Route (ACO with RLA):

4. Result and Analysis Response Time in Ant-Net Vs RLA on different Load condition

Response Time in ms

160 140 120 100 in Ant-Net

80

in RLA

60 40 20 0 100

150

200

500

1000

Load in Packets

(a) Ant Net Vs RLA

We implemented the proposed work in NS2 simulation, a widely used open source discrete event Network Simulator tool. There are many network simulation tools available, for example, GloMoSim, Omnet, Opnet, Qualnet. In which NS2 is preferred by almost all researchers and academicians, because it covers a very large set of applications, protocols, network types, network elements and traffic models.

Response Time in OSPF Vs RLA on different Load condition

Response Time in ms

250 200 150

in OSPF in RLA

100 50 0 100

150

200

500

1000

Load in Packets

(b) OSPF Vs RLA Response Time in BF Vs RLA on different Load condition

Response Time in ms

300 250 200 in RIP (BF)

150

in RLA

100 50

The following events are scheduled in the Implementation, 1) Data communication from Source1 to destination for 8 sec, 2) the proposed work implemented for 8 Sec. By using AWK Programming, the packet loss and average response time of before and after implementing the proposed algorithm is measured from the trace file. These values are listed here:  100 packets loss in 8 sec simulation implementing existing ACO based multi-path algorithm  0 packets loss in 8 sec simulation implementing proposed ACO with RLA  RTT is 0.1581 Sec in existing ACO multi-path algorithm  RTT is 0.1586 Sec in proposed ACO with RLA

0 100

150

200

500

1000

Load in Packets

(c) BF (RIP) Vs RLA

Fig. 3: Response Time between various routing protocols Vs RLA on different Load condition 

11 – 9 – 14 – 0 – 2 – 5 – 4

The detailed performance analysis is shown in the following section 4.1

4.1 Simulation Result of Redundant Link Avoidance Algorithm Based Routing

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org In this section, the result of RLA algorithm from source 13 to destination 5, described in Test case 1 is shown Fig 2. The data communication from source node 13 to destination node 5 through three selected route explained in case 1, also shown in the Fig 2. Comparison of response time in existing ACO (Ant Net) Vs proposed ACO with RLA algorithm is shown in the Fig. 3(a). Similarly, Comparison of response time in OSPF and SPI Vs proposed ACO with RLA algorithm is shown in the Fig.3(b) and Fig. 3(c) respectively.

5. Conclusion Network traffic is a common problem in all type of network, of topology, of protocol. It requires more attention, as it will diminish all major metric of network. Especially in internet like huge, dynamic and cloud like network, traffic is an unavoidable condition. This paper analyses the various factors that causes for congestion and provides optimal solution called redundant link avoidance algorithm over the existing ant colony optimization based routing protocol. When comparing the existing routing protocol with the proposed work, the response time get delayed and this delay in transmission time is a negligible, only 5 ms but all packet loss is avoided. It will perform even better, when implementing the same work for some complex network with different topologies and high packet loss. As shown in the test case 2 of section 3.3, there are few situations, the multi route routing protocol may act as single route routing protocol. It may lead to poor response time. But in the huge resource centric world wide web, it is promised that it never will ensue.

Acknowledgement: The First Author, Chandra Mohan B would like to thanks Dr. Sridharan R, Assistant Professor, Department of Electronics and Communication Engineering, College of Engineering, Anna University, Chennai for his constant support and help. Both Authors would like to thanks Dr. Srivatsa, Professor (Retired), Department of Electronics and Communication Engineering, Madras Institute of Technology, Anna University, Chennai and Dr. Kannan. A, Professor, Department of Computer Science and Engineering, College of Engineering, Anna University, Chennai

33

References [1] Jacobson. V, “Congestion Avoidance and Control”, ACM SIGCOMM Computer Communication Review, Vol 25, p157 – 187, 1995 [2] Real time traffic monitoring available at http://www.internettrafficreport.com/details.htm [3] Andrew S. Tanenbaum, “Computer Networks”, Prentice Hall of India, 3rd Edition, 1998 [4] Larry L. Peterson & Bruce S. Davie, “Computer Networks – A Systems approach”, Morgan Kaufmann Series, 2000 [5] Jiancong Chen, S.-H. Gary Chan, and Victor O. K. Li, “Multipath Routing for Video Delivery Over BandwidthLimited Networks”, IEEE Journal on Selected Areas in Communications, Vol. 22, No. 10, December 2004 [6] Ka-Cheong Leung, and Victor O.K. Li, “Generalized Load Sharing for Packet-Switching Networks I: Theory and Packet-Based Algorithm”, IEEE Transactions on Parallel and Distributed Systems, Vol. 17, No. 7, July 2006 [7] J. Moy, “OSPF Version 2,” IETF RFC 2328, Apr. 1998 [8] C. Villamizar, “OSPF optimized multi-path (OSPF-OMP),” Internet Draft, 1998. [9] A. E. I. Widjaja, “Mate: MPLS adaptive traffic engineering,” Internet Draft, Aug. 1998. [10] Israel Cidon, Raphael Rom, and Yuval Shavitt, “Analysis of Multi-Path Routing”, IEEE/ACM Transactions On Networking, Vol. 7, No. 6, December 1999 [11] Weijia Jia, Dong Xuan and Wei Zhao, “Integrated Routing Algorithms for any cast Messages”, IEEE Communications Magazine, January 2000 [12] Joao Luis Sobrinho, “Algebra and Algorithms for QoS Path Computation and Hop-by-Hop Routing in the Internet”, IEEE / ACM TRANSACTIONS ON NETWORKING, VOL. 10, NO. 4, AUGUST 2002 [13] Di caro G., and Dorigo M. (1998), ‘Extending AntNet for best effort quality of service routing’, First international workshop on ant colony optimization [14] Schoonderwoerd R., Holland O., Bruten J., (1997) ‘Ant like agents for load balancing in telecommunication networks’, In proceedings of the first int. conf. on autonomous agents, pp209-216, New York, ACM Press [15] Subramanian D., Druschel P., and Chen J., (1997) ‘Ants and reinforcement learning: A case study in routing in dynamic networks’, In proceedings of the 15th int. joint conf. on artificial intelligence, pp823-838, San Francisco, Morgan Kaufmann [16] Heusse M., Snyers D., Guerin S., and Kuntz P. (1998), ‘Adaptive agent driven routing and load balancing in communication networks’, Advances in complex systems, 1(2), pp 237-254. [17] Wei-Neng Chen, Jun Zhang, Henry Shu-Hung Chung, RuiZhang Huang, and Ou Liu, “Optimizing Discounted Cash Flows in Project Scheduling—An Ant Colony Optimization Approach”, IEEE TRANSACTIONS ON SYSTEMS,

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS, VOL. 40, NO. 1, p64 – 77, JANUARY 2010 [18] Pedro C. Pinto, Andreas Nägele, Mathäus Dejori, Thomas A. Runkler, and João M. C. Sousa, “Using a Local Discovery Ant Algorithm for Bayesian Network Structure Learning”, IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION, VOL. 13, NO. 4, p767, AUGUST 2009 [19] Xiao-Min Hu, Jun Zhang, Henry Shu-Hung Chung, Ou Liu, and Jing Xiao, “An Intelligent Testing System Embedded With an Ant-Colony-Optimization-Based Test Composition Method”, IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS, VOL. 39, NO. 6, NOVEMBER 2009 [20] Wei-Neng Chen, and Jun Zhang, “An Ant Colony Optimization Approach to a Grid Workflow Scheduling Problem With Various QoS Requirements”, IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS, VOL. 39, NO. 1, JANUARY 2009 [21] Agostino Forestiero, and Carlo Mastroianni, “A Swarm Algorithm for a Self-Structured P2P Information System”, IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION, VOL. 13, NO. 4, AUGUST 2009 [22] Minglu Li, Hongzi Zhu, Member, Yanmin Zhu, and Lionel M. Ni, “ANTS: Efficient Vehicle Locating Based on Ant Search in Shanghai Grid”, IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 58, NO. 8, OCTOBER 2009 [23] Girma S. Tewolde, and Weihua Sheng, “Robot Path Integration in Manufacturing Processes: Genetic Algorithm Versus Ant Colony Optimization”, IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 38, NO. 2, MARCH 2008 [24] Parag M. Kanade and Lawrence O. Hall, “Fuzzy Ants and Clustering”, IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 37, NO. 5, SEPTEMBER 2007 [25] Dorigo M and Stutzle T, “Ant Colony Optimization”, MIT Press, First Edition

Authors [*] Mr. B. Chandra Mohan received his Diploma in Electrical and Electronics Engineering, B.Tech Degree in Information Technology and Master degree in Computer Science and Engineering from Directorate of Technical Education, Madras University and Anna University in the years 1993, 2003 and 2006 respectively. Presently he is pursuing Research in Anna University Chennai. He worked both industry and academic in India and Abroad

34

about 15 years. Presently he is serving as Principal in John Bosco Engineering College, Tiruvallur, Tamilnadu, India. His experience includes as Faculty in Anna University Chennai, as Tech In-charge in Kuwait National Petroleum Company, as Principal in PTR College of Engg and Tech. His present research included Networks, Data Mining and Swarm intelligence. He has published dozen of papers in National, International Conferences and Book series. His paper is awarded as best paper in the IEEE International Conference (SNPD 2008). He is a reviewer in IEEE JSAC, Elsevier and few international journals. He delivered more than 50 Special Lectures in National, International Seminars, Workshops and Faculty Development Programs. He has worked in Kuwait. He is a life member of Institution of Electronics and Telecommunication Engineers (IETE), Indian Society for Technical Education (ISTE), Computer Society of India (CSI) and International Association for Engineers (IAEng). [**]Dr. R. Baskaran received his B.Tech in Electrical and Electronics Engineering, Master Degree in Computer Science and Engineering, Madras University and Doctorate from Anna University in the years 2000, 2001, and 2007 respectively. He is now working as an Assistant Professor in Department of Computer Science and Engineering, College of Engineering Guindy, Anna University Chennai. His present research includes Networks, Database, Data mining and warehousing and Image Processing. He presented more than 50 Special Lectures in National, International Seminars, Workshops and Development Programs. He is an expert in Data mining. He has published 35+ papers in International, National Journals and Conferences. He is a reviewer in IEEE/ ACM JSAC, Elsevier and many international journals. He is a life member of Institution of Electronics and Telecommunication Engineers (IETE), Indian Society for Technical Education (ISTE), Computer Society of

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org India (CSI) and International Association for Engineers (IAEng).

35

35


An Investigation about Performance Comparison of Multi-Hop Wireless Ad-Hoc Network Routing Protocols in MANET Dr.S.Karthik1

S.Kannan2

Dr.V.P.Arunachalam3

Dr.T.Ravichandran4

Dr.M.L.Valarmathi5

1 Professor & Head, 2 Final Year M.E-CSE, Department of Computer Science and Engineering, SNS College of Technology, Sathy Main Road, Coimbatore-641035, Tamil Nadu, India.

3

4

Principal, SNS College of Technology, Coimbatore, Tamil Nadu, India.

Principal, Hindustan Institute of Technology, Coimbatore, Tamilnadu, India. 5

Assistant Professor, in the Department of computer Science & Engineering, Government College of Technology, Coimbatore, Tamilnadu, India

Abstract Mobile Ad-Hoc Network (MANET) is a collection of wireless mobile hosts forming a temporary network without the aid of any stand-alone infrastructure or centralized administration. Mobile Ad-hoc networks are self-organizing and self-configuring multihop wireless networks where, the structure of the network changes dynamically. This is mainly due to the mobility of nodes. The Nodes in the network not only acts as hosts but also as routers that route data to or from other nodes in network. In mobile ad-hoc networks a routing procedure is always needed to find a path so as to forward the packets appropriately between the source and the destination. The main aim of any ad-hoc network routing protocol is to meet the challenges of the dynamically changing topology and establish a correct and an efficient communication path between any two nodes with minimum routing overhead and bandwidth consumption. The design problem of such a routing protocol is not simple since an ad hoc environment introduces new challenges that are not present in fixed networks. A number of routing protocols have been proposed for this purpose like Ad Hoc On Demand Distance Vector (AODV), Dynamic Source Routing (DSR), DestinationSequenced Distance Vector (DSDV). In this paper, we study and compare the performance of the following three routing protocols AODV, DSR and DSDV. Key words: On-demand, Table driven, DSR, AODV, DSDV, Adhoc, MANET

1. Introduction The Internet Engineering Task Force (IETF) created a Mobile Ad hoc Network (MANET) working group to standardize IP routing protocol functionality suitable for wireless routing application within both static and dynamic topologies with increased dynamics due to node motion and other factors. The vision of ad hoc networks is wireless internet, where users can move

anywhere anytime and still remaining connected with the rest of the world. The Mobile Ad-Hoc Network is characterized by energy constrained nodes, bandwidth constrained links and dynamic topology. In real-time applications, such as audio, video, and real-time data, the ad hoc networks need for Quality of Service (QoS) in terms of delay, bandwidth, and packet loss is becoming important. Providing QoS in ad-hoc networks is a challenging task because of dynamic nature of network topology and imprecise state information. Hence it is important to have a dynamic routing protocol with fast re-routing capability, which also provides stable route during the life-time of the flows. Generally there are two distinct approaches for enabling wireless mobile units to communicate with each other: Infrastructure-based - Wireless mobile networks have traditionally been based on the cellular concept and relied on good infrastructure support. Here mobile devices communicate with access points like base stations connected to the fixed network infrastructure. Infrastructure-less – In Figure.1 infrastructure-less approach, the mobile wireless network is commonly known as a mobile ad hoc network (MANET). A MANET is a collection of wireless nodes that can dynamically form a network to exchange information without using any preexisting fixed network infrastructure. The infrastructure-less approach is increasingly becoming a very important part of communication technology, because in many contexts information exchange between mobile units cannot rely on any fixed network infrastructure, but on rapid configuration of a wireless connections on-the-fly.


36

natural/environmental disaster that demolished previous communication infrastructure.

the

Law enforcement activities: Rapid installation of a communication infrastructure during special operations. Commercial projects: Simple installation of a communication infrastructure for commercial gatherings such as conferences, exhibitions, workshops and meetings. Educational classrooms: Simple installation of a communication infrastructure to create an interactive classroom on demand. Fig 1: MANET Approaches

1. 1 MANET Characteristics The fundamental difference between fixed networks and MANET is that the computers in a MANET are mobile. Due to the mobility of these nodes, there are some characteristics that are only applicable to MANET. Some of the key characteristics are described below [9] Dynamic Network Topologies: Nodes are free to move arbitrarily, meaning that the network topology, which is typically multi-hop, may change randomly and rapidly at unpredictable times. Bandwidth constrained links: Wireless links have significantly lower capacity than their hardwired counterparts. They are also less reliable due to the nature of signal propagation. Energy constrained operation: Devices in a mobile network may rely on batteries or other exhaustible means as their power source. For these nodes, the conservation and efficient use of energy may be the most important system design criteria. The MANET characteristics described above imply different assumptions for routing algorithms as the routing protocol must be able to adapt to rapid changes in the network topology.

1.2 Applications of MANET There are numerous scenarios that do not have an available network infrastructure and could benefit from the creation of an ad hoc network. [15] Rescue/Emergency operations: Rapid installation of a communication infrastructure during a

Military battlefield: Ad hoc networking would allow the military to take advantage of commonplace network technology to maintain an information network between the soldiers, vehicles, and military information head quarters. Commercial sector: Emergency rescue operations (like fire, flood, earthquake, etc.,) must take place where nonexisting or damaged communications infrastructure and rapid deployment of a communication network is needed. Local level: Ad hoc networks can autonomously link an instant and temporary multimedia network using notebook computers or palmtop computers to spread and share information among participants at a e.g. conference or classroom.

2. Related work Several researchers have done the qualitative and quantitative analysis of Ad Hoc Routing Protocols by means of different performance metrics. They have used different simulators for this purpose. ♦ J. Broch et al. [4], in their paper have compared the DSDV, TORA, DSR and AODV Protocols using ns-2 simulator. The simulation was done with 50 nodes with varying pause times. The results were obtained for the metrics: Packet delivery ratio, Routing overhead, Number of hops taken by the packet to reach the destination. ♦ Samir R. Das , Charles E. Perkins et al. [14] evaluated the DSR and AODV on-demand routing protocols with three performance metrics : Packet delivery fraction, Average End-End Delay and Normalized routing load with varying pause times. They have used ns-2 simulator. Based on the observations, recommendations were made as to how the performance of either protocol can be improved.


♦ Jyoti Raju and Garcia-Luna-Aceves [5], in their paper have compared WRP-Lite a revised version of Wireless Routing Protocol with DSR. The performance parameters used are end-end delay, control overhead, percentage of packets delivered and hop distribution. The evaluation of the performance metrics was done with respect to varying pause time. It was observed that WRP-lite has much better delay and hop performance while having comparable overhead to DSR.

3. FEATURES OF MANET Some of the salient features that describe the MANET clearly are [7] Dynamic network topology: Since the nodes are mobile, the network topology may change rapidly and unpredictably and the connectivity among the terminals may vary with time

37

relaying packets. Further, wireless link characteristics introduce also reliability problems, because of the limited wireless transmission range, the broadcast nature of the wireless medium (e.g. hidden terminal problem), mobilityinduced packet losses, and data transmission errors. Routing: Since the topology of the network is constantly changing, the issue of routing packets between any pair of nodes becomes a challenging task. Most protocols should be based on reactive routing instead of proactive. Quality of Service (QoS): Providing different quality of service levels in a constantly changing environment will be a challenge. Power Consumption:. For most of the lightweight mobile terminals, the communication-related functions should be optimized for less power consumption.

3.2 Performance Metrics of MANET

Autonomous terminal: In MANET, each mobile terminal is an autonomous node, which may function as both a host and a router (to perform switching functions).

The following metrics are considered for simulating and analyzing the performance of routing protocols and characteristics of MANET [12,13]

Multi hop routing: When delivering data packets from a source to its destination (i.e., only when the nodes are not directly linked), the packets should be forwarded via one or more intermediate nodes.

Jitter: Jitter describes standard deviation of packet delay between all nodes.

Distributed operation: Since there is no background network, the control and management of the network is distributed among the terminals. Light-weight terminals: In most cases, the MANET nodes are mobile devices with less CPU processing capability, small memory size, and low power storage. Such devices need optimized algorithms and mechanisms that implement the computing and communicating functions.

3.1 Challenges faced in MANET Regardless of the attractive applications, the features of MANET introduce several challenges that must be studied carefully before a wide commercial deployment can be expected. These include [6]: Internetworking: The coexistence of routing protocols, for the sake of internetworking a MANET with a fixed network, in a mobile device is a challenge for the mobility management. Security and Reliability: An ad hoc network has its particular security problems due to e.g. nasty neighbor

Throughput: The throughput metric measures how well the network can constantly provide data to the sink. Throughput is the number of packet arriving at the sink per milliseconds. Power consumption: The total consumed energy divided by the number of delivered packet. Packet Delivery Ratio (PDR) : PDR is the ratio of the number of packets successfully received by all destinations to the total number of packets injected into the network by all sources. The PDR is a number between 0 and 1. Average Packet Delay: It is sum of the times taken by the successful data packets to travel from their sources to destination divided by the total number of successful packet. The average packet delay is measured in seconds. Average Hop Count: It is the sum of the number of hops taken by the successful data packets to travel from their sources to destination divided by the total number of successful packets. The average hop count is measured in number of hops. Node Expiration time (NET): It is the time for which a node has been alive before it must halt transmission due to battery depletion. The node expiration is plotted as number

38


of nodes alive at a given time, for different point in time during the simulation.

called as the route discovery process. Examples include DSR and AODV.

End-to-End Delay: The average time interval between the generation of a packet in a source node and the successfully delivery of the packet at the destination node. It counts all possible delays that can occur in the source and all intermediate nodes, including queuing time, packet transmission and propagation, and retransmissions at the MAC layer. The queuing time can be caused by network congestion or unavailability of valid routes.

4.1. Destination-Sequenced Distance Vector (DSDV) routing protocol

4. Routing Protocols in MANET There are different criteria for designing and classifying routing protocols for wireless ad hoc networks. For example, what routing information is exchanged; when and how the routing information is exchanged, when and how routes are computed etc. Proactive vs. Reactive Routing: Proactive Schemes determine the routes to various nodes in the network in advance, so that the route is already present whenever needed. Route Discovery overheads are large in such schemes as one has to discover all the routes. Examples of such schemes are the conventional routing schemes, Destination Sequenced Distance Vector (DSDV).[8] Reactive Schemes determine the route when needed. Therefore they have smaller Route Discovery overheads. Examples of such Single path vs. Multi path: There are several criteria for comparing single-path routing and multi-path routing in ad hoc networks. First, the overhead of route discovery in multi-path routing is much more than that of single-path routing.[10] On the other hand, the frequency of route discovery is much less in a network which uses multi-path routing, since the system can still operate even if one or a few of the multiple paths between a source and a destination fail. Second, it is commonly believed that using multi-path routing results in a higher throughput. Table driven vs. Source Initiated: In Table Driven Routing protocols, up-to-date routing information from each node to every other node in the network is maintained on each node of the network. The changes in network topology are then propagated in the entire network by means of updates. Destination Sequenced Distance Vector Routing (DSDV) and Wireless Routing Protocol (WRP) are two schemes classified under the table driven routing protocols head. The routing protocols classified under Source Initiated On-Demand Routing, create routes only when desired by the source node.[11] When a node requires a route to a certain destination, it initiates what is

DSDV is a table-driven routing scheme for ad hoc mobile networks based on the Bellman-Ford algorithm. It was developed by C. Perkins and P.Bhagwat in 1994. The main contribution of the algorithm was to solve the Routing Loop problem. Each entry in the routing table contains a sequence number, the sequence numbers are generally even if a link is present; else, an odd number is used. The number is generated by the destination, and the emitter needs to send out the next update with this number. Routing information is distributed between nodes by sending full dumps infrequently and smaller incremental updates more frequently [1]. A comparison of the characteristics of the above three ad hoc routing protocols DSDV, DSR, AODV is given in following Table. PROTOCOL PROPERTY Loop Free Multicast Routes Distributed Unidirectional Link Support Multicast Periodic Broadcast QoS Support Routes Maintained in Route Cache /Table Timer Reactive

DSDV

DSR

AODV

Yes

Yes

Yes

No

Yes

No

Yes

Yes

Yes

No

Yes

No

No

No

Yes

Yes

No

Yes

No Route Table

No Route Cache

No Route Table

Yes

No

Yes

No

Yes

Yes

Table1: property comparison of DSDV,DSR & AODV

4.2. Dynamic Source Routing (DSR) Protocol DSR is a routing protocol for wireless mesh networks. It is similar to AODV in that it forms a route on-demand when a transmitting computer requests one. However, it uses source routing instead of relying on the routing table at each intermediate device. Determining source routes requires accumulating the address of each device between the source and destination during route discovery. The accumulated path information is cached by nodes processing the route discovery packets. The learned paths are used to route packets [3].

39

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814 Fig 2. Dropped Packets for AODV, DSR, DSDV

This protocol is truly based on source routing whereby all the routing information is maintained (continually updated) at mobile nodes. It has only 2 major phases which are Route Discovery and Route Maintenance. Route Reply would only be generated if the message has reached the intended destination node.

4.3. Ad Hoc on Demand Distance Vector (AODV) Routing Protocol AODV is capable of both unicast and multicast routing. It is a reactive routing protocol, meaning that it establishes a route to a destination only on demand. In contrast, the most common routing protocols of the Internet are proactive, meaning they find routing paths independently of the usage of the paths. AODV is, as the name indicates, a distance-vector routing protocol. AODV avoids the counting-to-infinity problem of other distance-vector protocols by using sequence numbers on route updates, a technique pioneered by DSDV [2].

5. Performance Results of AODV, DSR, DSDV The graphs given here are the performance analysis of the routing protocol with respect to different metric considered above. The X- Axis shows the number of nodes and the y axis shows the Metric considered.

Figure 3. Average End-to-End delay for AODV, DSR, DSDV

In terms of packet delivery ratio (Figure1), DSR performs well when the number of nodes is less as the load will be less. However its performance declines with increased number of nodes due to more traffic in the network. The performance of DSDV is better with more number of nodes than in comparison with the other two protocols. The performance of AODV is consistently uniform. In terms of dropped packets (Figure2), DSDV’s performance is the worst. The performance degrades with the increase in the number of nodes. AODV and DSR performs consistently well with increase in the number of nodes. For average end-to-end delay (Figure3), the performance of DSR and AODV are almost uniform. However, the performance of DSDV is degrading due to increase in the number of nodes the load of exchange of routing tables becomes high and the frequency of exchange also increases due to the mobility of nodes.

6. Conclusion

Fig 1. Packet delivery ratio for AODV, DSR, DSDV

It is difficult for the quantitative comparison of the most of the ad hoc routing protocols due to the fact that simulations have been done independent of one another using different metrics and using different simulators. In this paper, we have presented comparison studies about On-Demand (DSR and AODV) and Table-Driven (DSDV) routing protocols. Our comparison indicate that the performance of the two on demand protocols namely DSR and AODV is superior to the DSDV in conformance with the work done by other researchers as mentioned in section 2. It is also observed that DSR outperforms AODV in less stressful situations, i.e smaller number of nodes.


AODV outperforms DSR in more stressful situations. The routing overhead is consistently low for DSR and AODV than in comparison with DSDV especially for large number of nodes. This is due to the fact that in DSDV the routing table exchanges would increase with larger number of nodes. Our comparison also indicate that as the number of nodes in the network increases DSDV would be better with regard to the packet delivery ratio, but it may have considerable routing overhead.

[8].

[9].

[10].

As far as packet delay and dropped packets ratio are concerned, DSR/AODV performs better than DSDV with large number of nodes. Hence for real time traffic AODV is preferred over DSR and DSDV. For less number of nodes and less mobility, DSDV’s performance is superior. A general observation is that protocol performance is linked closely to the type of MAC protocol used. In conclusion, the design of the routing protocol must take into consideration the features of the lower layer protocols.

Acknowledgement The authors would like to thank the Director cum Secretary, Correspondent, Principal, SNS College of Technology, Coimbatore for their motivation and constant encouragement. The authors would like to thank the Faculty Members of Department of Computer Science and Engineering for critical review of this manuscript and for his valuable input and fruitful discussions. Also, he takes privilege in extending gratitude to his family members and friends who rendered their support throughout this research work.

References [1]. [2]. [3]. [4].

[5].

[6]. [7].

Charles E. Perkins, Pravin Bhagwat, “Highly Dynamic Destination-Sequenced Distance-Vector Routing (DSDV) for Mobile Computers”.2004. Charles E. Perkins, rElizabeth M.Royer ,”Ad_hoc On_Demand Distance Vector Routing”.2005. David B. Johnson, David A. Maltz, “Dynamic Source Routing in Ad Hoc Wireless Networks”.2003. J. Broch, D. A. Maltz, D. B. Johnson, Y-C. Hu, and J Jetcheva,“A performance comparison of multi-hop wireless ad hoc network routing protocols”, Mobicom’ 98, October 1998, pages 85-97,. J. Raju and J. Garcia-Luna-Aceves, "A comparison of on-demand and table driven routing for ad-hoc wireless networks", Proc. of IEEE ICC, June -2000, pages 1702- 1706. Jochen Schiller.Mobile Communications. AddisonWesley, 2000. Karthik, S., V.P. Arunachalam and T. Ravichandran, 2008. A comparative study of various ip trace back strategies and simulation of ip trace back. Asian J. Inform. Technol., ISSN: 1682-3915, 7(10):454-458.

[11].

[12].

[13]. [14].

[15].

40

Karthik. S., V.P. Arunachalam and T. Ravichandran, 2008. Multi directional geographical trace back within directions generalization. J. Compute. Sci., ISSN: 1549-3636, 4(8):646-651. Karthik. S., V.P. Arunachalam and R.M.Bhavdharini, 2008. Analyzing interaction between denial of service (dos) attacks and threats. Proceeding of the IEEE Int. Conf. on Computing, Communication and Networking (ICCCN 2008) 978-1-4244-3595-1/08/2008. DOI:10.1109/ICCCNET.2008.4787663 Karthik. S., V.P. Arunachalam and T. Ravichandran, 2006. Implementation and application of location discovery in enterprise-based wireless networks. Natl. J. Eng., ISSN: 0974-8377, 8(12), 22-26. Karthik. S., V.P. Arunachalam and T. Ravichandran, 2009. Analyzing interaction between denial of service (dos) attacks and threats. Int. J. Soft Computing, ISSN: 1816-9503, 68-75. Karthik. S., V.P. Arunachalam and T. Ravichandran, 2009, A Novel Direction Ratio Sampling Algorithm (DRSA) Approach for Multi Directional Geographical Traceback. Int. J. Computer Science and Security, ISSN: 1985-1553,3(3), 272-279 S. Corson and J. Macker, Mobile Ad hoc Networking: Routing Protocol Performance issues and Evaluation Considerations, RFC 2501, January 1999. S.R. Das, C.E. Perkins, and E.E. Royer, “Performance Comparison of Two On-Demand Routing Protocols for Ad Hoc Networks,” Proc. INFOCOM, 2000, pp. 3- 12. Evaluation for High Density Ad Hoc Stefano Basagni (Editor), Marco Conti (Editor), ilvia Giordano (Editor), Ivan Stojmenović (Editor) Mobile Ad Hoc Networking ISBN: 978-0-471-37313-1, Wiley-IEEE Press, August 2004

Dr.S.Karthik is presently Professor & Head, Department of Computer Science & Engineering, SNS College of Technology, affiliated to Anna University- Coimbatore, Tamilnadu, India. He received the B.Sc degree from Bharathiar University in 1998 and M.Sc degree from the Periyar University in 2000, Anna University Chennai and Ph.D degree from International University, Washington. His research interests include network security, web services and wireless systems. In particular, he is currently working in a research group developing new Internet security architectures and active defense systems against DDoS attacks. Dr.S.Karthik published papers in international journal and conference papers and has been involved many international conferences as Technical Chair and tutorial presenter. He is a active member of IEEE, ISTE and Indian Computer Society.

Kannan.S is presently a Final Year M.E candidate at Department of Computer Science & Engineering, SNS College of Technology, Coimbatore, Tamilnadu, India. He received his B.E degree from Anna University in 2008. Before he came to PG Study in the year of 2008, he was a Lecturer in SNS College of Technology, Coimbatore, Tamilnadu, India


Professor Dr.V.P.Arunachalam received the B.E and M.E degrees from Madras University, Madras, India in 1971 and 1972, respectively, and PhD degree from the Bharathiar University, Coimbatore, India, in 1988. He is currently the Principal of SNS College of Technology, Coimbatore, Tamilnadu, India. Before joining SNS College of Technology, Professor Arunachalam has been a Professor and Vice Principal in Government College of Technology, Coimbatore, Tamilnadu, India. His research interests include theory and practical issues of building distributed systems, Internet computing and security, mobile computing, performance evaluation, and fault tolerant computing, CAD,CAM,CAE. Professor Arunachalam is a member of the IE and ISTE. Professor Arunachalam has published more than 50 papers in refereed international journals and refereed international conferences proceedings. Since 1990 Professor Arunachalam has been involved in research and produces more than 15 PhD Candidates.

Professor Dr.T.Ravichandran received the B.E degrees from Bharathiar University, Tamilnadu, India and M.E degrees from Madurai Kamaraj University, Tamilnadu, India in 1994 and 1997, respectively, and PhD degree from the Periyar University, Salem, India, in 1988. He is currently the Principal of Hindustan Institute of Technology, Coimbatore,Tamilnadu, India. Before joining Hindustan Institute of Technology, Professor Ravichandran has been a Professor and Vice Principal in Vellalar College of Engineering & Technology, Erode, Tamilnadu, India. His research interests include theory and practical issues of building distributed systems, Internet computing and security, mobile computing, performance evaluation, and fault tolerant computing. Professor Ravichandran is a member of the IEEE, CSI and ISTE. Professor Ravichandran has published more than 30 papers in refereed international journals and refereed international conferences proceedings. Professor Dr.M.L.Valarmathi received the B.E degree from Madurai Kamraj University and M.E degree from Bharathiar University, Tamilnadu, India in 1983 and 1990, respectively, and PhD degree from the Bharathiar University, Coimbatore, India, in 2007. She is currently the Assistant Professor, in the Department of computer Science & Engineering, Government College of Technology, Coimbatore, Tamilnadu, India. Before joining Government College of Technology, Professor Valarmathi has been a Lecturer in Allagappa Chettiyar College of Technology, Karaikudi, Tamilnadu, India. Her research interests include theory and practical issues of building distributed systems, Internet computing and security, mobile computing, performance evaluation, and fault tolerant computing, Image Processing, Optimization Techniques. She is a member of the ISTE. Professor Valarmathi has published more than 20 papers in refereed international journals and refereed international conferences proceedings.

41


42

Latency Optimized Data Aggregation Timing Model for Wireless Sensor Networks A.Sivagami1, K. Pavai2 and D. Sridharan3 1

Research Scholar, Department of Electronics and Communication Engineering, Anna University Chennai, 600 025, Tamil Nadu, India

2

Research Scholar, Department of Electronics and Communication Engineering, Anna University Chennai, 600 025, Tamil Nadu, India

3

Assistant Professor, Department of Electronics and Communication Engineering, Anna University Chennai, 600 025, Tamil Nadu, India

Abstract In Wireless Sensor Networks (WSN), the data aggregation schemes are extensively used to avoid the redundant transmission of correlated data from densely deployed sensor nodes. Even though, it enhances the network lifetime, it seriously suffers from the increased data delivery time. The high end-to-end delay experienced by the packets is not acceptable in the delay-constrained applications like seismic activity monitoring, military field monitoring, etc. In this work, we propose a novel data aggregation-timing model for node’s aggregation time out to reduce the data delivery time.

Keywords: Wireless Sensor Networks, Data Aggregation, latency minimization, timing model, TOSSIM.

1. Introduction The recent development in the field of Micro-ElectroMechanical Systems (MEMS), brings the dream of developing a low cost, tiny and autonomous device called wireless sensor node, capable of sensing, processing and communicating the field data, into reality. The low processing and limited communication capabilities of the nodes demand the dense deployment of sensor nodes in the monitoring terrain to cover the entire area and provide fault tolerance to node failure. The densely deployed nodes will sense the similar data and there is a high correlation among these data. It is not worthy to transmit the similar information by many nodes, since the communication cost is the dominant energy consumer in WSN. Many efforts are taken to reduce the number of unwanted transmissions in sensor networks. The data aggregation techniques gained more attention in achieving energy saving in WSN. The data aggregation is a technique to combine the data from various sensor nodes to eliminate the redundant information and provide the rich and multi- dimensional view of the monitoring environment [1]. Many data aggregation protocols were

proposed in literature [2-4] to reduce the energy consumption. But the data aggregation algorithms suffer from the increased data delivery time due to the fact that the aggregator nodes have to wait for the data from its children nodes. If the aggregator waits for more time, it collects more data from its children and hence aggregation gain increases. This will increase the gain with the increased delay and vice versa. Hence, there is a trade off between the energy and the delay [5]. The Wireless Sensor Networks are mainly deployed for transferring real world data wirelessly and hence it is not worthy to transmit the information to the base station belatedly. There are many applications, which need the time stringent data delivery, pose the challenge in developing data aggregation method that guarantee the delay requirement. If the aggregator node waits long time for the data from all its children, the data delivery time to the sink increases and the data of the present round interfere with that of next round. The timing model defines how long an aggregator node should wait for the data from its children. The aggregation time out should be such a way that the waiting time should be optimum to optimize the data aggregation and deliver the data to the sink within the stipulated time bound. Some of the works to reduce the delay while aggregating data are considered here. The rest of the paper is organized as follows. In Section 2 the related work in this domain is outlined, in the Section 3 the proposed algorithm is explained, the protocol implementation is given in the section 4, in the Section 5, the performance of the algorithm is presented and the Section 6 gives the final conclusion of the work.


2. Related Work The aggregation time out may be periodic simple, that all the nodes can wait for the fixed predetermined time period or periodic per hop, the aggregator waits to hear from all of its children or cascading timeout, the time out depends on its position in the data aggregation tree (DAT). In cascading timeout [6], nodes schedule their time out based on its position in the DAT. A node’s time out happens after its children’s time out and thus enables a node to collect information from all its children. But, all the nodes in the particular level are given with same time out irrespective of its number of children. The advantages of this algorithm are it does not require any time synchronization or any centralized control. It doesn’t consider the number of children in each tree and thus the nodes with more children will loose some of its children data and also the nodes in the same level will send the data at almost same time, leading to traffic congestion. In [7], the nodes time out is determined dynamically based on the aggregation tree structure and the number of children it has. The node’s aggregation time out may be increased when a node detects the dead line miss. The update process is more complex and the agent nodes, which are one hop neighbors to the sink, will be congested more since the data arrive at the agent node almost simultaneously. The Adaptive Time Control (ATC) [8] determines the aggregation time out for a node based on the level of the sensor node in the data aggregation tree and the number of children it has. The nodes with more children are given with more time and thus maximizing the opportunity for data aggregation from its children. Thus the nodes in the same level will get different aggregation time out. The authors claimed that this algorithm provides higher data delivery rate and lower energy costs compared to cascading time out. In their simulation, modified IEEE 802.11 protocol is used as medium access protocol and the node’s sleeping schedule is not considered. This will have the impact on the energy consumption and the delay calculation. In [9], the time efficient data aggregation on clustered WSN is considered. The time out is calculated for each sub tree in the cluster, which is based on the packet transmission delay and the cascading delay. The performance is compared for various modulation techniques, which is the key factor for the packet transmission delay. The above protocols are simulated either using network simulator NS2 or discrete simulator based on C++. Also, they have used IEEE802.11 as channel access mechanism with or with out some modifications. The performance of these protocols may be tested for the real time deployment of wireless sensor nodes. The wireless sensor nodes of

43

our consideration are IRIS motes from Crossbow Technologies [10]. It is good practice to test the performance of any protocol using a simulator before it is implemented in the real world. The simulators like ns2 and other similar simulators do not reflect the real world scenario properly. Therefore we have chosen the WSN simulator TOSSIM to test the behavior of our proposed timing model. The simulator TOSSIM can simulate the program written in NesC, which is the native language for WSN mote IRIS, and the same the code can be fused into the mote with minor modifications. [11]

3. Timing Model The objective of this work is to develop a protocol, which delivers the data to the sink with in the deadline while adapting the data aggregation methods for reducing the energy consumption. This protocol estimates the time out of each node in the tree in a distributed manner, so that the data generated by each node should be delivered to the sink before the deadline. This work aims at the target platform as the wireless sensor mote called IRIS. These motes use TinyOS operating system, which is one of the most widely adapted operating system for the resourceconstrained mote network. The motes send the data to the sink using the collection tree, which is formed and maintained by the Collection Tree Protocol (CTP) [12]. The CTP protocol uses the wireless link quality between the motes as the metric to construct the tree and it is a dynamic tree. If the link quality changes, the tree structure will also change. Hence, the aggregation time out should be dynamic and updated as the tree structure changes. In cascading timeout, when an aggregator node receives the request from the sink, it calculates its time out based on the level in the tree. The staggered time out takes place between the levels and time disarticulation between them is just a single hop delay. The nodes in the same level are having same time out and hence they try to gain access to the channel simultaneously. Therefore, the transmission of the packets by the nodes in the same level will be deferred by the MAC and the parent time out will takes place before the child. Hence, the aggregator node will miss some of the packets from its children nodes and the aggregation gain decreases. To improve the performance of the DAT, the aggregator’s node time out should be assigned in such a way that it could collect more information efficiently. In our proposed algorithm, each node calculates the initial time out for data aggregation, which is based on the hop distance from the sink and the number of children it has in its sub tree rooted from it. This initial time out will be different for each node and it increases from the leaf node to sink node. The children nodes time out first, followed by its parent and hence the data generated by the children nodes are collected, processed and forwarded by its

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org parent node in staggered manner. After the initial staggered timeout, the nodes follow the fixed time out of duration T, which is data generation period. This ensures that the data generated by the nodes reach the sink before the next round begins. The Fig. 1 shows the timeout model of the proposed system called Delay Efficient Aggregation Timing Model (DEATM).

Where n is the set of aggregator nodes in the path from Lj to Ai. The maximum path cost from any leaf node to an aggregator node is Pi is the maximum path cost from leaf node to the aggregator. Pi = max {Path_cost (Lj, Ai)}  j

LEAF

Sub Tree 2

(2)

The maximum path cost from any leaf node to sink is Psink is the maximum path cost from leaf node to the aggregator. Psink = max {Path_cost (Lj ,sink )}  i, j

Sub Tree 1

44

(3)

The initial staggered time out is calculated as follows. Let Ti is the stagger time out for the aggregator node i, which is equal to Ti = Tci + Tai

(4)

Hop Distance

Where Tci is the cascading time out which depends on the level in which the aggregator is in. This gives the initial timeout as in cascade time out for each node and it is same for all the nodes in the same level. The Tai is the aggregation time out of the node, which depends on the number of children it has. Tci = 2 *[T – (TTD * h)] Tai = (Pi /Psink)*(T-TTD*D)

SINK

(5)

Time in Seconds Fig. 1 Timing Model of DEATM

The Fig. 1 shows the initial time out of the nodes at different level from the sink in two different sub trees. Each color in the figure shows the aggregation timer for the nodes in the same sub tree and the figure shows two such a sub trees. The nodes in the same level are represented with different color and their initial time out will depend on the total number of children in the sub tree in which it resides. The leaf nodes get the least waiting time and the nodes near to the sink gets more staggering time. All the leaf nodes get the same staggering time but the nodes in the same higher levels will get different stagger time. The leaf node’s time out takes place early and the time out of the nodes in the different levels increases as it approaches the sink. The nodes near to the sink get more time out and it is less than the deadline T. The nodes in the same level will have different time out and hence the collision in the same level be avoided and it ensures that the data wave will reaches the sink through that sub tree with in the dead line.

3.1 Mathematical Model Let Ai is the aggregator node i and Lj is the leaf node j. Let Ni is the number of children nodes for the aggregator node Ai. i.e Ni = degree(Ai). The path cost from any leaf node Lj to an aggregator node Ai is given by Path_cost (Lj, Ai) =



Akn

deg ree( Ak )

(1)

Here, h denotes the hop distance of the node Ai, D is the depth of the tree, T is the data generation period or the dead line and TTD is the one hop delay between the levels. It depends on the queuing delay, MAC delay, processing delay for aggregation function and the transmission delay. It is assumed to be 0.1 seconds as used in [8]. After introducing this initial delay in the aggregation timer, the aggregation timer is fired for every T seconds thus enabling the collection of packets generated in that round from all the nodes in the collection tree.

3.2 Update Phase The aggregation timer is updated whenever there is a change in the topology or the aggregation gain is reduced due to time synchronization between the nodes. The beacon messages are exchanged at regular intervals between the nodes. If there is a change in the topology, the routing engine will send the message to its neighbouring nodes and hence all the nodes in the network will be updated their information. The aggregation time out Tai is recomputed every time it receives the beacon. If the new value and the previous timer value are differing more than by the given threshold , then the aggregation timer will be re adjusted. Also, the optimum number of responses for each aggregator node per round is Ni, which is equal to number of children it has. If the aggregator receives less than Ni, the timer value is increased by TTD and if it receives more than Ni, it reduces the time by TTD.


4. Implementation The default CTP routing protocol sets up the data collection tree by exchanging the beacon messages, which contains the information about its parent, and the cumulative link quality to reach the sink. Each node sends the beacon messages periodically. The CTP’s beacon message is modified to carry the additional information about the hop distance and Pi. The neighbour table is also modified to keep the record of neighbour’s hop distance and Pi. Each aggregator node finds its own children from the parent field of beacon message sent by its children. When a node receives the beacon message from its neighbor node with its node id as parent, it increments the number of children field in its neighbor table. Aggregator node also finds its hop distance using the beacon information from its parent. Each aggregator node calculates Path cost for each sub tree by adding the Pi of its children node and its number of children. The Pi of the node is calculated by finding the maximum value among all Pi’s. This information is passed to aggregator function, which calculates the value of Ti. The initial time displacement is achieved and the data wave will reach the sink with in the stipulated time. Each node uses two timers, one timer for the data generation and the other for the aggregation. After the completion of setting up the collection tree phase, the node starts both the timer. The data timer is fired every T seconds. The initial time out for the aggregation timer is calculated from the beacon message and the one-shot aggregation timer is started with the calculated time. When the data timer fires, the node read the default sensor and keep it in the buffer. When the aggregation timer fires first time, it restarts the aggregation timer with T seconds, which is a periodic timer. When the periodic aggregation timer fires, the node does the aggregation of the packets from its children with its own value and send the aggregated packet up in the tree. The simple aggregation operator average is used.

5. Simulation and Result The proposed algorithm is simulated using TOSSIM simulator under Linux platform, which is a simulator for TinyOS2.x developed by University of California at Berekely, USA that can run the actual TinyOS code without any real motes. The 100 nodes are deployed uniformly in the area 200  200 m2. The TOSSIM uses SNR based simulation, the simulator parameters are such that, it simulates the indoor environment. The nodes are placed in uniform topology, where the sensor filed is divided into grid of equal size and the node is placed randomly in each grid. The default MAC IEEE 802.15.4 is used for Channel access and CTP with four-bit link estimation as data collection tree. The nodes generate traffic for every 20 seconds.

45

The performance parameters of our consideration are the aggregation gain, accuracy and miss ratio. The aggregation gain is defined as the measure of reduction in the communication traffic due to aggregation, which is related to the energy of the node [13]. It is the ratio of traffic reduction due to aggregation to the total traffic without aggregation.

AG  1 

ta t

Where t is the number of transmissions by all nodes without aggregation. Without any aggregation, an aggregator node should forward all the packets coming from its children. Hence the value of t is calculated by adding the total number of transmissions and reception by all the nodes. ta is the total number of transmissions with aggregation, which is equal to number of packets transmitted by all the nodes. The aggregation gain directly related to the energy consumption of the nodes and hence it has been chosen as one of the metric for analysing the protocol. An aggregator node receives and processes Nr packets from its children, and sent the information in one packet. Without aggregation, the aggregator node has to forward all the packets irrespective of its content. Hence the total number of transmission by a node is reduced by the factor of Nr/ (Nr +1). The aggregation gain is the ratio of reduction in number of transmission due to aggregation to the number of transmission without aggregation. If a node is waiting for an optimum time period, it could collect the information from all of its children, then the aggregation gain will be maximum. If an aggregator’s time out is less than the optimum time out, then the aggregator node could not collect information from some of its children. Thus, these packets, which are arriving after the deadline, are not considered for finding the aggregation result and hence discarded. This will reduce the accuracy of the aggregation process. The data accuracy is a measure of how much data is used to extract the information. It is defined as the number of readings received to the total number of packets generated in the network. The aggregated packet contains how many children’s data are used to find the aggregated value. Hence the sink could find the accuracy of the aggregation for a given simulation time. The accuracy of the aggregation process depends on the number of packets received from its children and in turn depends on the deadline requirement of the application and the node density of the network. The miss ratio is another parameter to judge the efficiency of the protocol. It is defined as the ratio of number of packets missing the deadline to the number of packets received. In each data collection round, the packets of the present round arriving after the aggregation time out will be dropped. Each node will calculate the total number of packets dropped and the total number of packets it received from its children for the given simulation period. The total number of dropped and received packets by all

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org the nodes are taken to measure the miss ratio. These performance parameters depend on the deadline, the number of hops, the number of nodes in the sub tree and the node density in the surveillance field.

46

1 0.8 Fixed Simple Cascade Timeout

0.6

DEATM

The proposed algorithm is compared with simple per hop, and cascade time out. In simple per hop, the nodes wait for fixed pre determined time and do the aggregation and in cascade timeout the staggered one hop delay time out is followed. The figure 2 to figure 7 show the results obtained from our simulation tests.

0.4 0.2 0 8

12

14

16

18

20

22

24

26

D e a d Line in m s

Fig. 3 Impact of Deadline on Miss Ratio

90 80 Aggregation Gain

10

The Fig. 3 shows that the miss ratio of the proposed algorithm is very less compared to other two schemes because, each node waits appropriate time to collect the data from all its children and hence the data reaches the sink with in the deadline. As deadline increases, the miss ratio decreases.

70 60

Fixed Simple

50

Cascade Timeout DEATM

40 30

90

20

80

10

12

14

16

18

20

22

24

26

Dealine in ms

Fig. 2 Impact of Deadline on Aggregation Gain

The Fig. 2 shows the aggregation gain for the different timing models as a function of deadline. The nodes sample the sensor and generate the data for every 20 seconds. This data is aggregated with the data received from its children during the time out period and the aggregated packet is transmitted out when the aggregation timer fires. The data generation period or deadline is varied from 10 seconds to 25 seconds and the aggregation gain is measured. The aggregation gain increases as the deadline increases. If T is small, more packets will miss the deadline due to small waiting period and increased data traffic. As the deadline increase the data traffic is reduced as well as the nodes will be given more opportunities to do the aggregation and hence the gain increases. Our proposed algorithm gives better aggregation gain compared to cascade time out since in our method, the aggregator node’s timeout includes the number of child nodes and hence it could collect more data from its children. In simple and cascade time out, the nodes, which are very close to the sink could send the data with in the deadline and hence the gain is less compared that of our proposed scheme.

70 Accuracy in %

8

60 50 40 30 Fixed Simple

20

Cascade Timeout

10

DEATM

0 8

13

18 Dead line in ms

23

28

Fig. 4 Impact of Deadline on Accuracy

The fig. 4 shows the data accuracy of the proposed algorithm as a function of deadline. The accuracy of the proposed system is generally high because it could collect more data from its children within the deadline. Accuracy increases with increase in deadline. In order to find the impact of node density on the performance parameters, the simulation is performed for the duration of 100 seconds and the number of nodes varies from 25 to 100. The area of the network is same for all the cases and hence the density of the nodes in the sensor field varies.

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org 90

95

80 Accuracy in %

85 Aggregation Gain

47

75 65 55

Fixed Simple Cascade Timeout

45

70 60 50 40 30

Fixed Simple Cascade Timeout

20

DEATM

35

DEATM

10 0

25 20

40

60

80

20

100

40

60 80 Number of Nodes

100

Number of Nodes

Fig. 7 Effect of Node density on Data Accuracy Fig. 5 Effect of Node density on Aggregation Gain

The Fig. 5 shows dependency of aggregation gain on the network size. As the network size increases, the aggregator nodes could collect more packets from its children and reduce the traffic by aggregation. Hence the gain increases as the number of nodes increases. This is not linear due to the fact that the increased number of nodes causes collision and the children nodes have to wait more time to get the channel, which leads to the decrease in gain.

1.2

Fixed Simple

Miss Ratio

DEATM

0.8 0.6 0.4 0.2 0 20

40

5. Conclusion and Future work The proposed protocol gives more aggregation gain, which leads to less energy consumption, less miss ratio, which delivers the data with in the stipulated time bound and the good data accuracy. Thus our protocol can deliver more accurate and fresh information to the sink in an energy efficient manner. Also, our proposed algorithm is independent of time synchronization and it doesn’t need any centralized control. It also adjusts the time dynamically according to the change in topology or change in synchronization. This proposed algorithm gives more aggregation gain compared to that of cascading time out scheme and the data generated in a round is delivered to the sink in the same round. Thus the data freshness is maintained.

Cascade Timeout

1

From these figures, our proposed algorithm can do better than other two algorithms even in high-density networks.

60

80

100

Number of Nodes

Fig. 6 Effect of Node density on Miss Ratio

The Fig. 6 shows the impact of node density on the miss ratio. The miss ratio of the proposed scheme is very less compared to the other two schemes. Also, for the small node densities, the miss ratio does not change much for all the schemes but the miss ratio increases as the node density increases. This is due to the fact that if the node density increases, the number of competitors for a node to access the channel increases. This will increase the MAC deferring time and hence the packets will miss the dead line and dropped by the aggregators. The fig. 7 shows the data accuracy of the proposed algorithm as a function of network size. The accuracy of the proposed system is generally high and increases with decrease in network size.

The gain, miss ratio and the accuracy of the protocol depend on the deadline and the size of the network. From our observations, for a given network size, a minimum deadline has to be fixed so that the data can be delivered to the sink with in the deadline. The proposed work will be tested with the real test bed consists of IRIS motes.

References [1]

[2]

[3]

R Rajagopalan and P. K. Varshney, “Data Aggregation techniques in sensor networks: A Survey”, Journal on IEEE Communications, Surveys and Tutorials, Vol. 8, Issue 4, 2006, pp.48-63. Mohamed A. Sharaf, Jonathan Beaver, Alexandros Labrinidis and Panos K. Chrysanthis, “TiNA: A Scheme for Temporal Coherency Aware in Network Aggregation”, in 3rd ACM International Workshop on Data Engineering for Wireless and Mobile Access, 2003, pp. 69-76. Tian He, Brian M. Blum, John A. Stankovic and Tarek Abdelzaher, “AIDA: Adaptive Application-independent Data Aggregation in wireless Sensor Networks”, ACM Transactions on Embedded Computing Systems, Vol. 3, No. 2, May 2004, pp. 426–457.

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 6, May 2010 www.IJCSI.org [4]

[5]

[6]

[7]

[8]

[9]

Yujie Zhu, Ramanuja Vedantham, Seung-Jong Partk, Raghupathy Sivakumar, “A Scalable correlation aware aggregation strategy for wireless sensor networks”, Journal of Information Fusion, Elsevier publishers, Vol. 9, Issue 3, 2008, pp. 354-369. Yang Yu, Bhaskar Krishnamachari, and Viktor K. Prasanna, “Energy-latency tradeoffs for data gathering in wireless sensor networks”. in IEEE INFOCOMM, 2004. Ignacio Solis and Katia Obraczka, “The impact of timing in data aggregation for sensor networks", in IEEE International Conference on Communication (ICC), 2004. Jae Young Choi, Jong Wook Lee, Kamrok Lee, Sunghyun Choi, Wook Hyun Kwon and Hong Seong Park, “Aggregation Time Control Algorithm for Time constrained Data Delivery in Wireless Sensor Networks”, in IEEE Vehicular Technology Conference, 2006, Vol. 2, pp. 563 – 567. Hong Li, HongYi Yu, BaiWei Yang and Ana Liu, “Timing control for delay-constrained data aggregation in wireless sensor networks”, International Journal of communication systems (Wiley), Vol. 20, issue 7, 2007, pp. 875- 887. Shan Guo Quan and Young Yong Kim, “Fast Data Aggregation Algorithm for Minimum Delay in Clustered Ubiquitous Sensor Networks”, in International Conference on Convergence and Hybrid Information Technology (ICHIT), 2008, pp. 327-333.

[10] Crossbow Technologies http://www.xbow.com [11] Philip Levis, Nelson Lee, Matt Welsh, and David Culler, “TOSSIM: Accurate and Scalable Simulation of Entire TinyOS Applications”, in the First ACM International Conference on Embedded Networked Sensor Systems (SenSys), 2003, pp. 126-137. [12] O. Gnawali, R. Fonseca, K. Jamieson, D. Moss, and P. Levis, “Collection Tree Protocol”, in the 7th ACM International Conference on Embedded Networked Sensor Systems (SenSys), 2009, pp. 1-14. [13] U. Roedig, A. Barroso, and C. Screenan, “Determination of Aggregation Points in Wireless Sensor Networks”, in the 30th EUROMICRO conference, 2004, pp. 503 – 510.

A. Sivagami received her B. E. degree in Electronics and Communication Engineering from University of Madras in the year 1990 and obtained her Master’s degree in Communication Systems from Bharathiyar University in the year 2000. Currently she is doing her research in the field of Wireless Sensor Networks in Anna University, Chennai, India. Her fields of interests are Wireless Networks, Wireless Sensor Networks, Digital Communication, etc. She has published 4 papers in International conferences and 3 in International Journals K. Pavai received her B.E. degree in Electronics and Instrumentation Engineering in the year 2004 from university of Madras and M.E. in Applied Electronics in the year 2006 from Anna University, Chennai. Currently she is doing her research in the field of Wireless Sensor Networks in Anna University, Chennai. She has 4 publications in the National conferences, 3 in International conferences and three in International Journals. Her current fields of interests are Wireless Sensor Networks, VLSI, Mobile Communication,

48

Instrumentation, etc. Dr. D. Sridharan received his B.Tech. Degree in Electronics Engineering and M.E. degree in Electronics Engineering from Madras Institute of Technology, Anna University in the years 1991 and 1993 respectively. He got his Ph.D degree in the Faculty of Information and Communication Engineering, Anna University in 2005. He is currently working as Assistant Professor in the Department of Electronics and Communication Engineering, CEG Campus, Anna University, Chennai, India. His present research interests include Internet Technology, Network Security, Distributed Computing and VLSI for wireless Communications. He has published more than 25 papers in National/International Conferences and Journals.

IJCSI CALL FOR PAPERS NOVEMBER 2010 ISSUE Volume 7, Issue 6 The topics suggested by this issue can be discussed in term of concepts, surveys, state of the art, research, standards, implementations, running experiments, applications, and industrial case studies. Authors are invited to submit complete unpublished papers, which are not under review in any other conference or journal in the following, but not limited to, topic areas. See authors guide for manuscript preparation and submission guidelines. Accepted papers will be published online and authors will be provided with printed copies and indexed by Google Scholar, Cornell’s University Library, ScientificCommons, CiteSeerX, Bielefeld Academic Search Engine (BASE), SCIRUS and more. Deadline: 30th September 2010 Notification: 31st October 2010 Revision: 10th November 2010 Online Publication: 30th November 2010                

Evolutionary computation Industrial systems Evolutionary computation Autonomic and autonomous systems Bio-technologies Knowledge data systems Mobile and distance education Intelligent techniques, logics, and systems Knowledge processing Information technologies Internet and web technologies Digital information processing Cognitive science and knowledge agent-based systems Mobility and multimedia systems Systems performance Networking and telecommunications

              

Software development and deployment Knowledge virtualization Systems and networks on the chip Context-aware systems Networking technologies Security in network, systems, and applications Knowledge for global defense Information Systems [IS] IPv6 Today - Technology and deployment Modeling Optimization Complexity Natural Language Processing Speech Synthesis Data Mining

For more topics, please see http://www.ijcsi.org/call-for-papers.php All submitted papers will be judged based on their quality by the technical committee and reviewers. Papers that describe research and experimentation are encouraged. All paper submissions will be handled electronically and detailed instructions on submission procedure are available on IJCSI website (www.IJCSI.org). For more information, please visit the journal website (www.IJCSI.org)

© IJCSI PUBLICATION 2010 www.IJCSI.org

IJCSI The International Journal of Computer Science Issues (IJCSI) is a refereed journal for scientific papers dealing with any area of computer science research. The purpose of establishing the scientific journal is the assistance in development of science, fast operative publication and storage of materials and results of scientific researches and representation of the scientific conception of the society. It also provides a venue for researchers, students and professionals to submit ongoing research and developments in these areas. Authors are encouraged to contribute to the journal by submitting articles that illustrate new research results, projects, surveying works and industrial experiences that describe significant advances in field of computer science.

Indexing of IJCSI: 1. Google Scholar 2. Bielefeld Academic Search Engine (BASE) 3. CiteSeerX 4. SCIRUS 5. Docstoc 6. Scribd 7. Cornell’s University Library 8. SciRate 9. ScientificCommons

© IJCSI PUBLICATION www.IJCSI.org