A GENETIC-BASED OPTIMIZED FUZZY-TABU CONTROLLER FOR MOBILE ROBOT RANDOMIZED NAVIGATION IN UNKNOWN ENVIRONMENT

Although sampling-based path planning algorithms have been shown robust performances in many cases, they still suffer from many drawbacks. Inefficiency in online/sensor-based planning purposes, generating far-from-optimum paths, high running time, failure in difficult environments and generating unstable variant results in different runs are some of the existing problems with sampling-based methods. We propose an online sampling-based path planning algorithm which employs a fuzzy-tabu controller to evaluate the generated samples and select better ones in order to improve the performance of the planner. Furthermore, a genetic algorithm-based optimization framework is developed to optimize the parameters of the controller. The genetic optimizer takes place before the planner starts any planning task and attempts to optimize the fuzzy-tabu controller parameters and also the parameters of the planner. We simulate the proposed algorithm to analyze its performance and to compare it with existing sampling-based algorithms. The proposed algorithm generates relatively short paths in less than few seconds while results variations between different runs are considerably low.