A GA-fuzzy logic based extended Kalman filter for mobile robot localization

Abstract

The basic requirement of mobile robot localization is to know the information about its position and direction. The extended Kalman filter is an excellent tool to estimate the robot's posture in its work environment. Traditional extended Kalman filter uses fixed error covariance matrices Q and R, which does not conform the real situation. In this paper, GA-fuzzy logic controller is developed to adjust the error covariance matrices on-line. To improve the accuracy of fuzzy logic controller, a genetic algorithm is developed to tune the membership functions. The simulation results show that the proposed approach has good performance.