426 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 29, NO. 3, JUNE 1999 angle every 1.5 s such that this approach can deal in reasonable response time with common obstacles that might cause collisions in indoor environments. VI. CONCLUSIONS In this study, a vision-based obstacle avoidance approach for ALV navigation has been proposed. The vehicle can detect obstacles, including walls and objects in the way, in an unknown indoor environment and safe collision-free paths can be generated from quadratic classifier design in real time. According to the collision- free path, the vehicle can modify the turning angle of the wheels to achieve the purpose of collision avoidance. Besides, a systematic method has been proposed for generating input patterns for classifier design to compute safe quadratic paths. The use of quadratic paths instead of linear ones produces smoother paths and prevents dead-reckoning navigation to increase the flexi- bility of ALV applications in unknown complex environments with obstacles. Additionally, quadratic paths also match the ALV trajectory better than linear ones. A method for computing the optimal turning angle to avoid collisions in real time has also been proposed. The proposed approach has been implemented on a real ALV and a lot of successful navigations confirm the feasibility of the approach. REFERENCES [1] J. C. Hyland and S. R. Fox, “A comparison of two obstacle avoidance path plannings for autonomous underwater vehicles,” in Proc. Symp. Autonomous Underwater Vehicle Technology, Washington, DC, June 1990, pp. 216–222. [2] J. Cesarone and K. F. Eman, “Mobile robot routing with dynamic programming,” J. Manufact. Syst., vol. 8, no. 4, pp. 257–266, 1989. [3] C. Acosta and R. G. Moras, “Path planning simulator for a mobile robot,” Comput. Indust. Eng., vol. 19, no. 1–4, pp. 346–350, 1990. [4] J. O. Kim and K. Khosla, “Real-time obstacle avoidance using hormontic potential functions,” IEEE Trans. Robot. Automat., vol. 8, pp. 338–349, June 1992. [5] K. Onoguchi, M. Watanabe, Y. Okamoto, Y. Kuno, and H. Asada, “A visual navigation system using a multi-infomation local map,” in Proc. 1990 IEEE Int. Conf. Robotics Automation, Cincinnati, OH, vol. 2, pp. 767–774, May 1990. [6] D. C. H. Yang, “Collision-free path planning by using nonperiodic B-spline curves,” J. Mech. Design, vol. 115, pp. 679–684, Sept. 1993. [7] L. L. Wang and W. H. Tsai, “Collision avoidance by a modified least- mean-square-error classification scheme for indoor autonomous land vehicle navigation,” J. Robot. Syst., vol. 8, no. 5, pp. 677–798, Oct. 1991. [8] J. Borenstein and Y. Koren, “Real-time obstacle avoidance for fast mo- bile robots,” IEEE Trans. Syst., Man, Cybern., vol. 19, pp. 1179–1187, Sep./Oct. 1989. [9] T. Skewis and V. Lumelsky, “Experiments with a mobile robot operating in a cluttered unknown environment,” in Proc. 1992 IEEE Int. Conf. Robotics Automation, Nice, France, vol. 2, pp. 1482–1487, May 12–14, 1992. [10] R. Bauer, W. Feitern, and G. Lawitzky, “Steer angle fields: An approach to robust manoeuvring in cluttered, unknown environments,” Robot. Auton. Systems, vol. 12, pp. 209–212, 1994. [11] K. Fukunaga, Introduction to Statistical Pattern Recognition, 2nd ed. San Diego, CA: Academic, 1990. [12] C. C. Lai, “Outdoor autonomous land vehicle guidance by road infor- mation using computer vision and fuzzy wheel adjustment techniques,” M.S. thesis, Inst. Comput. Inf. Sci., National Chiao Tung Univ., Hsinchu, Taiwan, R.O.C., June 1993. [13] S. D. Cheng and W. H. Tsai, “Model-based guidance of autonomous land vehicles in indoor environments by structured light using vertical line information,” J. Elect. Eng., vol. 34, pp. 441–452, Dec. 1991. [14] A. Ohya, A. Kosaka, and A. Kak, “Vision-based navigation of mobile robot with obstacle avoidance by single camera vision and ultrasonic sensing,” in Proc. 1997 IEEE/RSJ Int. Conf. Intelligent Robot Systems, Grenoble, France, Sept. 1997, vol. 2, pp. 704–711. [15] L. M. Lorigo, R. A. Brooks, and W. E. L. Grimsou, “Visually- guided obstacle avoidance in unstructured environments,” in Proc. 1997 IEEE/RSJ Int. Conf. Intelligent Robot .Systems, Grenoble, France, vol. 1, pp. 373–379, Sept. 1997. [16] Y. G. Yang and G. K. Lee, “Path planning using an adaptive-network- based fuzzy classifier algorithm,” 13th Int. Conf. Computers Applica- tions, Honolulu, HI, Mar. 1998, pp. 326–329. [17] R. Biewald, “Real-time navigation and obstacle avoidance for nonholo- nomic mobile robots using a human-like conception and neural parallel computing,” in Int. Workshop Parallel Processing Cellular Automata and Array, Berlin, Germany, Sept. 1996, pp. 232–240. Dynamic Fuzzy Control of Genetic Algorithm Parameter Coding Robert J. Streifel, Robert J. Marks, II, Russell Reed, Jai J. Choi, and Michael Healy Abstract— An algorithm for adaptively controlling genetic algorithm parameter (GAP) coding using fuzzy rules is presented. The fuzzy GAP coding algorithm is compared to the dynamic parameter encoding scheme proposed by Schraudolph and Belew. The performance of the algorithm on a hydraulic brake emulator parameter identification problem is investigated. Fuzzy GAP coding control is shown to dramatically increase the rate of convergence and accuracy of genetic algorithms. I. INTRODUCTION Genetic algorithms are powerful search techniques which have been applied to many practical problems. However, the accuracy of the final solution found by binary coded genetic algorithms is limited by the number of bits used to code search parameters into strings. The low resolution of binary coding does not seriously affect the solution for many problems (e.g., integer and combinatorial searches). Accuracy becomes a more important consideration when 1) the search space consists of floating point parameters; 2) the parameters have a large dynamic range; 3) a relatively small number of bits are used to code the param- eters. The standard genetic algorithm uses no problem specific informa- tion except the relative fitness of the coded binary strings. Lack of gradient information can cause slow progress in search regions where the objective function has nearly zero gradient. The combination of low slope areas and low resolution binary coding can cause slow convergence on many practical problems. The fuzzy genetic algorithm parameter (GAP) coding methodology presented in this paper is specifically designed to improve the search performance on a parameter identification problem. Conventional genetic algorithm parameter coding is static, the coding is constant for the entire search. This results in slow convergence. Greater accuracy Manuscript received April 15, 1996; revised July 5, 1997. This paper was recommended by Associate Editor L. O. Hall. R. J. Streifel, R. J. Marks, II, and R. Reed are with the University of Washington, Seattle, WA 98195 USA (e-mail: robert.j.streifel@boeing.com). J. J. Choi and M. Healy are with the Boeing Information and Support Services, Seattle, WA 98195 USA. Publisher Item Identifier S 1083-4419(99)00903-6. 1083–4419/99$10.00  1999 IEEE