Fuzzy Logic and Neural Network-aided Extended Kalman Filter for Mobile Robot Localization

Abstract

In this thesis, an algorithm that improves the performance of the extended Kalman filter (EKF) on the mobile robot localization issue is proposed, which is aided by the cooperation of neural network and fuzzy logic. An EKF is used to fuse the information acquired from both the robot optical encoders and the external sensors in order to estimate the current robot position and orientation. Then the error covariance of the EKF is tracked by the covariance matching technique. When the output of the matching technique does not meet the desired condition, a fuzzy logic is employed to adjust the error covariance matrix to modify it back to the desired value range. Since the fuzzy logic is lack of the capability of learning, a neural network is presented in the algorithm to train the EKF. The simulation results demonstrate that, with the comparison to the odometry and the standard EKF method under the same error divergence condition, the proposed extended Kalman filter effectively improves the accuracy of the localization of the mobile robot system and effectively prevents the filter divergence.