Robust Position Tracking for Mobile Robots with Adaptive Evolutionary Particle Filter

Abstract

Robust position tracking is a challengeable issue for mobile robot in presence of faults. In the paper, an adaptive evolutionary particle filter is designed to achieve robust position tracking for wheeled mobile robot when the robot is subjected to faults such as sensor faults and wheel slippage. Firstly, the kinematics models of wheeled mobile robots and the measurement models of laser range finder are derived, five kinds of residual features are extracted and faults are detected according residual features. Secondly, an adaptive evolutionary particle filter is designed for robust localization, which includes two key steps: (1) adapting the proposal distribution according to residual features, (2) evolutionary operators, which are tuned with unnormalized weights of particles, are designed to recover the diversity of particle sets. Lastly, the presented method is testified in a real mobile robot.