An Improved Adaptive Unscented FastSLAM with Genetic Resampling

The simultaneous localization and mapping (SLAM) is a significant topic in intelligent robot. In this paper, an improved adaptive unscented FastSLAM with genetic resampling is proposed. Specifically, the adaptive unscented Kalman filter (IAUKF) algorithm is improved as importance sampling of particle filter, where the adaptive factor is used to improve the tracking ability of system and the Huber cost function is constructed to decrease the measurement covariance. Next, the process noise and the measurement noise are assessed by a time varying estimator. Moreover, the resampling in particle filter is carried out by an improved genetic algorithm (GA). Finally, the improved adaptive unscented FastSLAM (IAUFastSLAM) is proposed to complete robot tracking. The proposed algorithm has good tracking performance and obtains reliable state estimation in SLAM. Simulation results reveal the validity of the proposed algorithm.

Automated summary

This paper proposes an improved adaptive unscented FastSLAM algorithm, which enhances system tracking ability by using an adaptive factor and constructing a Huber cost function, resampling in particle filter with an improved genetic algorithm to complete robot tracking, achieving good tracking performance and reliable state estimation.