基于重采样技术的激光SLAM系统优化设计

Optimal design of laser SLAM system based on resampling technology

基于Rao-Blackwellized粒子滤波(Rao-Blackwellized partical filter,RBPF)算法的移动机器人在同时定位与地图构建(simultaneous localization and mapping,SLAM)过程中存在计算量大、粒子耗尽问题。基于重采样技术对激光SLAM系统开展优化设计:在采样过程中加入最近一帧的激光观测模型,减少构建地图所需要的粒子数;同时提出一种自适应优化组合重采样方法,以缓解粒子耗尽现象,保持粒子的多样性。利用Turtlebot 2和Rplidar A2搭建的平台进行实验验证,结果显示,改进的RBPF-SLAM系统优化方法能够以更少的粒子数生成精度更高的全局一致性的地图。

In the process of simultaneous localization and mapping(SLAM),the mobile robot based on the Rao Blackwellized particle filter(RBPF)algorithm has the problems of large amount of computation and particle exhaustion.In this work,based on resampling technology,the optimization design of laser slam system was carried out.In the process of sampling,the laser observation model of the latest frame was added to reduce the number of particles needed to build a map.At the same time,an adaptive optimization combined resampling method was proposed to alleviate the phenomenon of particle depletion and maintain the diversity of particles.Experimental verification was conducted by the Turnlebot 2 and Rplidar A2 platform.The results show that the improved RBPF-SLAM system optimization method can generate more accurate global consistent map with less particles.