Rao-Blackwellised粒子滤波SLAM的一致性研究

Rao-Blackwellised粒子滤波SLAM(RBPF SLAM)算法的复杂度与特征个数呈线性关系,对于大规模SLAM有明显的计算优势,但是该算法不能长时间满足一致性要求,必须进行改进。采用归一化估计方差NEES对算法的一致性进行了分析,得出粒子耗尽是造成算法不一致的原因,并分别采用辅助粒子滤波及正则粒子滤波对算法进行改进,以得到一致的RBPF SLAM。最后,通过大量的Monte-Carlo仿真实验,验证了方法的有效性。

Consistency of Simultaneous Localization and Map Building (SLAM)with Rao-Blackwellised Particle Filter

Rao-Blackwellised particle filtering (RBPF) SLAM is a linear time algorithm proportional to number of landmarks,and the algorithm has obvious computational superiority for dense map or large-scale SLAM,but is inconsistent for long-term.The inconsistency problem of the algorithm was analyzed using the normalized estimation error square (NEES).The result shows that it is a sample impoverishment of particle filter which causes the inconsistency.So it is necessary to reduce the impact of resampling.Auxiliary particle filtering and regularized particle filtering were used to improve the RBPF SLAM in order to obtain consistent RBPF SLAM.Finally,plentiful Monte-Carlo simulations were carried out to evaluate the algorithm's performance and the simulation results indicate that the methods are valid.