GLOBAL ENGINEERS & TECHNOLOGISTS REVIEW www.getview.org
SOLVING WIRELESS NETWORK SCHEDULING PROBLEM BY GENETIC ALGORITHM REHAN SHAMS1 and HANIF KHAN2 1
Department of Telecommunication 2 Department of Mathematics Sir Syed University of Engineering and Technology Karachi, PAKISTAN 1
[email protected] 2
[email protected] ABSTRACT This study presents a hybrid meta heuristic algorithm that finds promising solutions within an acceptable time frame and conditions. Based on hybrid approach, this study combined the exploratory power of genetic algorithm with an intensive exploitation of good visiting sequences as a case. This is an optimization problem which arises in the domain of wireless sensor networks. The optimization problem of the replacement of damaged sensor in an area of interest by limited cargo capacity robots also discuss as application option. This is one of the kind of selective salesman problem in which a certain amount of unique commodity is delivered under constraints. Keywords: Wireless sensor network, Genetic algorithm, Optimization problem, Meta heuristic approach, Robot trajectory.
1.0
INTRODUCTION
Advancement in technologies such as sensing, electronics and computing have attracted tremendous research interest in the field of wireless sensors network (WSNs). Apart from their enormous potential for both the commercial and military applications wireless sensors network is generally consist of large number of low cost, low power multifunctional energy constrained sensors nodes with limited computational and communication capabilities. WSNs are the collection of autonomous sensing devices that can be deployed in to physical environment like hospitals, forest, highways, etc to cooperatively monitor the region. Any wireless sensors robot network (WSRN) consist of an ensemble of sensors and robot nodes that communicate via wireless links to perform distributed and actuation task. While sensors are highly constrained devices (i.e, they posses limited computing power, battery, memory, transmission range etc). Robots are resource-rich usually mobile and meant to assist, maintain and optimize sensors network, for example, they may perform intelligent movement for the data collection or sensors placement. In this paper we are dealing with an optimization problem by using the meta heuristic approach in which the demand of delivery customer can be met by a relatively large number of pickup customers. In this situation it is not necessary to visit all the delivery spots, only demanded locations will be included in the tour so as to minimize its cost. This problem is similar to the prize distribution and pickup and delivery problem. As an application we are taking the coverage repair problem in wireless sensors robot network already deployed in some area of interest. Since the therefore in order to save the energy and prolonging the network life time and due to the sensors have limited battery lifetime therefore not all of the sensors are actually needed to provide reliable area coverage. Thus sensors may follow a scheduling algorithm in which some will go to sleep mode (passive sensors) and some will be responsible for monitoring the region (active sensors). In addition due to the failure of any active node the total network coverage can be collapse, thus a single mobile robot with a limited cargo capacity is required locate to the damage sensors for their replacement. This task is to be repeated so often, and the network coverage has to be repair by replacing all the damage sensors with a spare ones. By optimal tour we mean the shortest trajectory or the minimum distance among the nodes of wireless sensors network. To solve the problem, the meta heuristic genetic algorithm GA are introduced. Genetic algorithm is the branch of the evolutionary algorithm that is the change over a long period of time. This approach is very useful in solving most of the networking problem. As far as the significance of GA is concern it is most helpful tool for solving different kinds of optimizing problem regarding to wireless sensors network. For example GA is helpful to achieve the optimum in band width in wireless networking by investigating the channel allocation for radio transmission over wireless network (Pradeep et al., 2012). GA based multi-objective methodology can be implemented for self-organizing wireless sensor network for transmission signal strength (Amol et al., 2009). -
G.L.O.B.A.L E.N.G.I .N.E. E.R.S. .& . .T.E.C. H.N.O.L.O.G.I.S.T.S R.E.V.I.E. W
10
Global Engineers & Technologists Review, Vol.2 No.11
(2012)
Several researchers have successfully implemented Gas in a sensor network design (Narasimhan and Sen, 1998; Ferentinos and Tsiligiridis, 2007) this led to the development of several other GA-based applications-specific approaches on WSN, mostly by the construction of fitness function. Due to the limited power recourses, the sensors network life time could be prolong by optimal dynamic base station therefore, GA can be serve as a dynamic optimum for base station replacement so that can save energy sensors and increase network life time (Mollanejad et al., 2010; Bani Hashim et al., 2011). GA also optimizes the operational modes for the sensors and transmission signal strength. GA algorithm is mainly probabilistic search algorithm based on the principles and concepts of natural selection and evaluation of individuals where each individual is an ordered sequence of nodes. According to the proposed problems these individuals is supposed to be the damage sensors nodes. The rest of the paper is organizes as follows section 2 will provide the description of genetic algorithm for the proposed problem, section 3 will give the brief summary of the proposed algorithm section 4 will show the experimental results section 5 will describe the conclusion and section 6 is for the references
2.0
STEPS OF THE GENETIC ALGORITHM 2.1 Initial Population The first step of GA is the creation of initial population to initiate the algorithm. In this study we are generating the initial population with a concept of random number generation algorithm. At the beginning of iteration the initial population can be start with different location from which they will start adding nodes to their tour. The initial population is created from the matrix representation that showing the distances among the active nodes. First indicate the profitable nodes and create initial random population among them, since our main task is to minimize the total tour therefore the selection is depending upon the nearest neighborhood criteria, i.e in case of delivery the robot must be drop to that location which nearer the visited node. Whenever there is a tie between the nodes then select the damage sensors node but it might affect the final tour because the unvisited node might be closer to the visited node that is why in the genetic algorithm there is a fitness criteria as a next step 2.2 Fitness Creteria The fitness criteria is the cost function associated with each individual generated by random procedure. Only those sub tour will undergo for the process of crossover procedure which is supposed to be the minimum. The effectiveness of the tour depends upon the numbers of sensors deployed in the network. In order to explain the fitness function here we are using these notations. F value = [(nact)(dact) - (nsen)(dp1) - (npas)] / N
(1)
Where; nact = number of active nodes in the network. dact = shortest distance between the active nodes. nsen = number of nodes where sensor replacement is required. dp1 = distance of the parent individual generated in the initial step. npas = number of passive nodes in the network. N = total number of nodes in the network. nlon = longest tour generated in the initial population procedure. By the help of above F values the fitness function is defined as, Fitness function = F value / nlon
(2)
If the above fitness measure is close to 1 then this individual is selected as a parent chromosomes for the crossover procedure. 2.3 Crossover According to the GA, crossover indicates the combination of two parent chromosomes to produce an offspring which is batter then the parent. For this algorithm, we are considering the two point crossover discussed in (Chakraborty et al., 2011). P1= {3, 2, 1, 4, 6, 5} P2= {5, 2, 1, 4, 3, 6} © 2012 GETview Limited. All rights reserved
11
Global Engineers & Technologists Review, Vol.2 No.11
(2012)
By randomly selecting the a slot i.e {2, 1, 4} from P1 and insert in the same position of P2 and repeated nodes must remove after insertion. {3, 2, 1, 4, 6, 5} 2.4 Mutation Mutation adds variation in the obtained offspring from the crossover procedure. For the mutation we select two random points on the chromosomes and flip the nodes of the selected bits. If there is any further improvement in the current tour then stop otherwise after applying the fitness criteria it will again send for the crossover procedure. The mutation operation can help the optimization problem to jump out of the local optimization by sharing the global information about the population. GA optimization highly depends upon the crossover and mutation methodologies. The crossover methodologies are available in GA toolbox of MATLAB such as heuristic, two point, scattered and single points. However the point methodologies give us optimum performance in terms of cost and speed.
3.0
STEPS OF THE PROPOSED ALGORITHM (i) (ii) (iii) (iv) (v)
Generate the initial population by the nearest neighbourhood criteria. Find the fitness value of each individual by using the above mentioned criteria. According to the fitness value select the chromosomes for the two point crossover as mentioned. Apply the mutation procedure. The resultant tour is supposed to be optimal after mutation.
4.0 EXPERIMENTAL RESULTS We have performed several experiments to check the feasibility of our algorithm. Simulation were run in MATLAB and we compare our benchmarks from (Falcon et al., 2012) in which they have used the ant colony approach different techniques. The time factor is also very important in the algorithm since we might not afford such delay in the calculation of robot trajectory, especially when immediate coverage repair is demanded. Table 1: Comparison between the proposed algorithm and (Falcon et al., 2010) s.no 1 2 3 4 5 6 7 8 9 10
N 20 30 40 50 60 100 200 300 400 500
cost of the pro. Algo 235.27 315.71 4123.21 5031.91 6201.56 999.87 21289.1 35428.72 52463.85 65234.1
Cost of the old algo 257.59 3572.71 4343.67 5211.06 6432.87 10629.87 23125.5 37865.17 54204.18 65559.9
Time of pro. Algo 13.17 23.17 18.75 22.83 22.25 78.67 140.21 197.13 238.71 325.26
Time taken by old algo 20.6 25.07 26.77 33.68 34.43 98.02 160.89 222.16 331.51 397.35
5.0 CONCLUSIONS This study has given the approach to find the optimal tour in the wireless sensors networking system by genetic algorithm in which the mobile robots and sensors are communicated via wireless links to perform distribution and actuation task. In case of any failure a single mobile robot has to replace these damage sensors. Therefore the main task is to produce a minimum cost trajectory for the robot by genetic algorithm.
REFERENCES [1] Amol, P.B., Renu, V., Madan L.S., Ghanshyam, C. and Pawan, K. (2009): Genetic Algorithm Based Node Replacement Methodology For Wireless Sensors Network, Proceedings of international conference of Engineers and Computer Sciences 2009 Vol.1, IMECS, March 18-20, 2009, Hong Kong. [2] Chakraborty, A., Mitra, S.K. and Naskar, M.K. (2011): A Genetic Algorithm inspired Routing Protocol for Wireless Sensors Network, International Journal of Computional Intelligence Theory and Practice, Vol.1, No. 6, pp.1-10. [3] Falcon, R., Li, X., Nayak, A. and Stojmenovic, I. (2010): The One-Commodity Travelling Salesman Problem With Selective Pickup And Delivery: Any colony Approach, WCCI IEEE world Congress on computational intelligence July, 18-23, 2010-CCIB, Spain. © 2012 GETview Limited. All rights reserved
12
Global Engineers & Technologists Review, Vol.2 No.11
(2012)
[4] Ferentinos, K.P. and Tsiligiridis, T.A. (2007): Adaptive Design Optimization of Wireless Sensors Network Using Genetic Algorithm. Computer networks, Vol.51, pp.1031-1051. [5] Mollanejad, A., Khanli, L.M. and Zeynali, M. (2010): DBSR: Dynamic Base Station Repositioning Using Genetic Algorithm In Wireless Sensors Network. International Journal of Computer Science, Vol.7, Iss.2, pp.24-29. [6] Narasimhan, K. and Sen, S. (1998): Sensors Network Design of Linear Processes Using Genetic Algorithm. Comput. Chem. Eng., Vol.22, No.3, pp.385-390. [7] Pradeep, K.T., Sanjay, S. and Satyanarayan, M. (2012): Bandwidth Optimization using Genetic Algorithm for Video Wireless Network. International Journal of Management and business Studies, Vol.2, Iss.1, pp.47-51. [8] Bani Hashim, A.Y., Abu Osman, N.A. and Wan Abas, A.A.B. (2011): Determining Foot Ankle Mechanism Design by Mapping the Relationships Among Bones, Joints and Ground Reaction Force. Global Engineers and Technologists Review, Vol.1, No.2, pp.39-47.
© 2012 GETview Limited. All rights reserved
13