An Improved Algorithm of UKF-SLAM Based on RBF Neural Network Adaptive Robot

Abstract

Abstract For Extend Kalman Filtering (EKF) algorithm in the robot simultaneous localization and mapping (SLAM), the linearization of the nonlinear system leads to the error of the system state equation, the large amount of computation caused by the Jacob matrix calculation, and the noise model uncertainty. Filtering instability and other issues. A modified UKF-SLAM algorithm based on RBF network adaptive is proposed. In the absence of prior noise information, the algorithm identifies the process noise and observation noise through RBF network adaptive identification unit, and iteratively corrects the noise covariance and the mean filter new covariance, so that the real-time positioning accuracy of the robot is improved. The experimental results show that this method has higher positioning precision and adaptive ability compared with the EKF algorithm and UKF algorithm.