Available online at www.sciencedirect.com
ScienceDirect Procedia Computer Science 76 (2015) 449 – 454
2015 IEEE International Symposium on Robotics and Intelligent Sensors (IRIS 2015)
A Modified Hybrid Fuzzy Controller for Real-Time Mobile Robot Navigation J. Hossena*, S. Sayeedb, A. K. M. Parvez Iqbalc a Faculty of Engineering and Technology (FET), Multimedia University (MMU), Malaysia, Faculty of Information Science and Technology (FIST), Multimedia University (MMU), Malaysia c Department of Mechanical Engineering, College of Engineering, University Tenaga Nasional (UNITEN), Malaysia b
Abstract The Fuzzy hybridization technique for intelligent systems have become of research interests in a variety of research areas over past the decades. There are limitations faced by all popular fuzzy systems architectures when they are applied to applications with a large number of inputs (Multi-sensors) systems. In this paper, proposes a modified hybrid fuzzy controller for multi-sensors (Large number of inputs) real-time mobile robot navigation. A modified hybrid fuzzy controller is constructed by the automatic generation of membership functions (MFs) and formed a minimal numbers of rules using hybrid fuzzy clustering algorithm (Combination of Fuzzy C-means and Subtractive clustering algorithm) and the modified apriori algorithm, respectively. The generated hybrid fuzzy controller is then adjusted by the least square method and the gradient descent algorithm towards better performance with a minimal set of rules. The performance is observed in an application for real-time mobile robot navigation has been found to be very impressive and competitive. ©©2015 Publishedby byElsevier ElsevierB.V. B.V. This is an open access article under the CC BY-NC-ND license 2015 The The Authors. Authors. Published (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of organizing committee of the 2015 IEEE International Symposium on Robotics and Intelligent Sensors (IRIS Peer-review under responsibility of organizing committee of the 2015 IEEE International Symposium on Robotics and Intelligent 2015). Sensors (IRIS 2015) Keywords: Apriori algorithm; Fuzzy C-means; Subtractive clustering, and ANFIS __________________________________________________________________________________________________________________
1. Introduction In the past decades, fuzzy systems have been combined with neural networks mainly for performing mobile robot navigation1,2,3,4, pattern recognitions5,6, modeling and control systems7,8. __________ *
Corresponding author. Email address:
[email protected]
1877-0509 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of organizing committee of the 2015 IEEE International Symposium on Robotics and Intelligent Sensors (IRIS 2015) doi:10.1016/j.procs.2015.12.307
450
J. Hossen et al. / Procedia Computer Science 76 (2015) 449 – 454
Many approaches have been proposed to address the issue of automatic generation of fuzzy membership functions and a fuzzy rule base from an input-output data set and also subsequent adjustment of them towards more satisfactory performance9,10,11,12. Most of these schemes that incorporate the learning property of neural networks within a fuzzy system framework provide encouraging results. However, most of these techniques also have difficulties associated with the number of resulting fuzzy rules, which increase exponentially when high numbers of input attributes are employed. The computational load required to search for a corresponding rule becomes very heavy as the number of fuzzy rules in a complicated situation is increased. Fuzzy clustering (Fuzzy C-Means and Subtractive Clustering) is a process of assigning membership levels, and then using them to assign data elements to one or more clusters13,14,15. The purpose of clustering is to identify natural groupings of data from a large data set to produce a concise representation of a system's behavior. Clustering can also be thought of as a form of data compression where a large number of samples are converted into a small number of representative prototypes or clusters. apriori algorithm16 (a shortened form of a priori algorithm) from data mining field has been used with fuzzy inference system to obtain more compact information from a data set or clusters to produce a minimum number of rules. This is a popular algorithm used in data mining using associative rules16. Real-time experiments have been conducted to test the performance of the developed hybrid fuzzy controller for mobile robot navigation task. The proposed hybrid fuzzy controller was evaluated subjectively and objectively with other fuzzy controllers and also the processing and training times was taken in consideration. The performance has been compared with other existing approaches such as ANFIS17 (in terms of RMSE, rules, processing time etc.) in an application of real-time mobile robot navigation and shown to be very competitive and impressive results. This paper is organized as follows: Section 2 provides a framework of modified hybrid fuzzy controller. Section 3 presents the results and discussion along with other subsections. In subsection 3.1, the prototype mobile robot has been shown. The experiment of real-time mobile robot navigation tasks are shown in graphically in subsection 3.2. Finally conclusions are drawn in Section 4. 2. Framework of Modified Hybrid Fuzzy Controller A Framework of a Modified Hybrid Fuzzy Controller is shown in Fig. 1. As a brief overview, a pre-processing has been applied to the given input and output data to generate two major components of the proposed system; first, adaptive fuzzy clusters for the inputs using the hybrid fuzzy clustering algorithm, and second, minimal number of decision rules of the given information using the rules generation algorithm using data mining algorithm. The obtained fuzzy clusters and minimal rules have been used to model membership functions and fuzzy rules in the knowledge-base. Once the modified hybrid controller is constructed with input/output data, a system evaluation process has been carried out as a post-processing. Suppose the fuzzy controller’s performance is not satisfactory, an advanced parameters adjustment mechanisms has been applied to the knowledge-based towards better accuracy. The architecture has been designed with the help of hybrid learning algorithms and these learning algorithms are applied to the developed fuzzy model using adaptive network. These algorithms are implemented and constructed a modified hybrid controller for real time mobile robot navigation18.
J. Hossen et al. / Procedia Computer Science 76 (2015) 449 – 454
__
Wi __
Wi Yi
Fig. 1. A Framework of a Modified Hybrid Fuzzy Controller
3. Result And Discussion 3.1. Prototype The overall behavior of the system is presented as well as the data obtained. Once the robot was completed mechanically, electronically and programmed, it was ready for testing. In Figure 2 shows a three dimensional view of the prototype developed and navigated in our lab at Multimedia University.
Fig. 2. Mobile Robot Prototype The prototype allows real investigation and analysis for the results. These results and findings are obtained by conducting real world tests on the system and then monitoring robot reaction in real time. 3.2. Real-Time Mobile Robot Navigation Performance The slope is used as factor in the program to conclude the distance by knowing the output voltage which comes from the sensor. When the obstacle is detected by the center sensor within 18 cm to 20 cm, if the left sensor detect the obstacle the mobile robot will move right forward with normal speed, else if right sensor detect the obstacle the mobile robot will move left forward with normal speed. When the obstacle within 6 cm to 18 cm, the react as same as the operation in 18 cm to 20 cm, but the speed is slow. However, when the center, left and right sensors at 6 cm,
451
452
J. Hossen et al. / Procedia Computer Science 76 (2015) 449 – 454
the mobile will stop and turn to left, if the right sensor detects, or turn to right, if the left sensor detects.
Fig. 3. Description of Complex Scenario As seen in Figure 3, the environment to mimic a warehouse with rooms or zones, where there are some rooms allows mobile robot to navigate in and some other restricted areas that the robot should not enter. Scenario description as follows: (A) The mobile robot (B) Wall or obstacles (C) A zone that robot is allowed to enter and exit from it (D) An area that robot is allowed to enter but not allowed to exit from it (E) Restricted zone and (F) The Goal. First the robot is scanning the environment using ultrasonic sensor, then move toward the farther obstacle or the longest path. As shown in Figure 4.
Fig. 4. Navigating of Mobile Robot prototype In Figure 5, the robot is navigating in zone C and will exit soon since it is small.
Fig. 5. Mobile Robot At Zone C In Figure 6, the robot is navigating and avoiding the obstacles.
J. Hossen et al. / Procedia Computer Science 76 (2015) 449 – 454
Fig. 6. Mobile Robot Navigating In Figure 7, still the robot is navigating and avoiding the obstacles.
Fig. 7. Mobile Robot Navigating to Goal In Figure 8, the robot detect red color which is the goal.
Fig. 8. Mobile Robot Reached in the Goal In Figure 9, after reaching the goal the robot will go to parking and stop there.
Fig. 9. Mobile Robot Parked in the Goal
453
454
J. Hossen et al. / Procedia Computer Science 76 (2015) 449 – 454
4. Conclusion A modified hybrid fuzzy controller has been proposed for real-time mobile robot navigation task which automatically generates fuzzy membership functions and minimal number of fuzzy rules via the hybrid fuzzy clustering and the modified apriori algorithm based on input sensors data. The performance evaluation is done to achieve better performance through the adjustment of antecedent membership functions using gradient descent method. It is significant that the number of rules generated by the proposed method were reduced effectively using the modified apriori technique towards better performance. The performance of modified hybrid fuzzy controller was found to be encouraging and satisfactory for mobile robot navigation. Research is continuing on the refinement process of the fuzzy rules to achieve better accuracy of a real time mobile robot navigation task.
References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.
S. K. Pradhan, “Navigation tecniques for Control of Multiple Mobile Robots”, PhD Thesis, National Institute of Technology (NIT), 2007. K. K. Tahboub, and M. S. N. Al-Din, “ A Neuro-Fuzzy Reasoning System for Mobile Robot Navigation”, Jordan Journal of Mechanical and Industrial Engineering, 2009; 3{1)1: 77 – 88. Munaf S. N. Al-Din, “Decomposed Fuzzy Controller for Reactive Mobile Robot Navigation”, International Journal of Soft Computing and Engineering (IJSCE), 2012; 2(4). Chi-Wen Lo, Kun-Lin Wu, Yue-Chen Lin, Jing-Sin Liu “An Intelligent Control System for Mobile Robot Navigation Tasks in Surveillance”, Advances in Intelligent Systems and Computing, 2014;274:449-462. B. Chang, H. Tsui, and P. Yu, “ Handwritten Characters Recognition using a Neuro Fuzzy System”, International Journal of Innovative Computing, Information and Control, Sept. 2008; 4(9):2345 – 2362. J. C. Obi, and A. A. Imianvan, “ Breast Cancer Recognition using Fuzzy Classifier”, International Journal of Academic Research,2011; 3:3. S. Chopra, R. Mitra and V. Kumar, “A Neurofuzzy Learning and its Application to Control system”, World Academy of Science, Engineering and Technology, 2007;34: 475 – 481. R-E. Precup, and H. Hellendoorn, “A survey on industrial application of fuzzy control”, Computer in industry, 2011;62: 213 – 226. Sandeep C., and Rene V. M., “RANFIS: Rough Adaptive Neuro-Fuzzy Inference System”, International Journal of Computational Intelligence, 2006;3(4):10. ChangSu L., Anthony. Z., Tomas B., “An Adaptive T-S type Rough-Fuzzy Inference System (ARFIS) for Pattern Classification”, Fuzzy Information Society, IEEE Explorer, 2007;117-122. Pradhana, S.K., D.R. Parhib and A.K. Pandac, “Fuzzy logic techniques for navigation of several mobile robots”. J. Soft Comput., 2009;9: 290-304. Parasuraman, S., B. Shirinzadeh and V. Ganapathy, “Mobile robot navigation using alpha level fuzzy logic system: Experimental investigations”. Proceedings of the IEEE International Conference on Systems, Man and Cybernetics, Oct. 12-15, IEEE Xplore Press, Bandar Sunway, 2008;1878-1884. J. C. Bezdek, Pattern recognition with fuzzy objective function algorithms, Plenum Press, New York, 1981. Mohanad A., Mohammad M., Abdullah R., “ Optimizing of Fuzzy C-Means Clustering Algorithm Using GA”, World Academy of Science, Engineering and Technology, 2008;39. S. L. Chiu, “Fuzzy model identification based on cluster estimation,” Journal of Intelligent Fuzzy Systems, 1994;2:267 – 278. R. Agrawal, R. Srikant, “Fast Algorithms for Mining Association Rules”, Proceedings of the 20th VLDB Conference, Santiago, Chile, 1994 Prases K. Mohanty and Dayal R. Parh, “A New Intelligent Motion Planning for Mobile Robot Navigation using Multiple Adaptive NeuroFuzzy Inference System”, Applied Mathematics & Information Sciences, 2014;8(5): 2527-2535. J. Hossen, A. Rahman, K. Samsuddin, F. Rokhani, S. Sayeed, R. Hasan, ”A Novel Modified Adaptive Fuzzy Inference Engine and Its Application to Pattern Classification”, World Academy of Science, Engineering and Technology, 2011;75:1201-1207.
