A Hybrid System Geno-Fuzzified Neural Network for Mobile Robot Control

Abstract

Aims: The goal of mobile robot is build system able to achieve tasks without human intervention in cluttered unknown environments. A main issue of an autonomous mobile robot is the design of an intelligent controller which enables the robot to navigate in a real world environment and avoiding obstacles especially in crowded and changing environment. Study Design: The controller uses genetic, fuzzy and neural to control of mobile robot. Place and Duration of Study: College Science, computer department, between September 2011 and December 2012. Methodology: In this search, fuzzy logic, genetic algorithm, and neural network (soft computing) are used to design an intelligent controller. This is due to the fact that fuzzy if-then rules are well suited for capturing the imprecise nature of human knowledge and reasoning processes. On the other hand, the neural networks are equipped for learning. Genetic algorithm has active role in the generating of fuzzy rules, it is designed to minimize the number of rules to minimum number. It is also helped to improve membership functions. Neural network is trained by using back propagation to increase efficiency of the work in time of arrive and get the shortest path to goal, it is obtained the steer angle of robot to the appropriate direction (avoid obstacles or get target). Results: The efficiency and robust of this work is appeared by using many different unknown environments that have different numbers, sizes and shapes of obstacles. The controller enables robot to avoid obstacles and reach goal with shortest distance (757 pixels) compared with other techniques(fuzzy controller and neuro-fuzzy controller),which owns the largest distance from same start position to the same end position and also less time(14 seconds). Conclusion: Geno – fuzzified – neural controller is efficient with different numbers, shapes, sizes of obstacles in unknown environments.