An improved H-infinity unscented FastSLAM with adaptive genetic resampling

Abstract

The FastSLAM is a typical tracking algorithm for SLAM, but it often suffers from the low tracking accuracy. To mitigate the problem, an improved H-Infinity unscented FastSLAM (IHUFastSLAM) with adaptive genetic resampling is proposed in this paper. Specifically, the H-Infinity unscented Kalman filter algorithm is improved using an adaptive factor and is employed as importance sampling in particle filter. Next, the process noise and the measurement noise are estimated by a time varying noise estimator. Moreover, an adaptive genetic algorithm is used to complete the resampling of particle filter. Finally, the improved H-Infinity UFastSLAM with adaptive genetic resampling is proposed to complete robot tracking. The proposed algorithm can track robot with good accuracy, and obtain reliable state estimation in SLAM. Simulation results reveal the validity of the proposed algorithm.