摘要
传统Rao-Blackwellized粒子滤波器(RBPF)在移动机器人同步定位与地图构建(SLAM)研究中,存在算法复杂度过高、占用内存空间过多导致实时性不理想的问题,因此提出一种改进算法。在某一特定状态的一组粒子集中,粒子的统计特性是一致的,改进算法从中选取一个代表粒子,进行卡尔曼更新步骤,并在同一粒子集中重复使用。同时结合Gmapping算法的建议分布和自适应重采样技术。实际Pioneer III移动机器人在机器人操作系统(ROS)平台上进行的实验表明,该方法在保证栅格地图精度的同时能提高系统的实时性,降低复杂度,提高运算速度。
As in the research of simultaneous localization and mapping（ SLAM） of mobile robot applying traditional Rao-Blackwellized particle filter,the computational complexity is too high and memory space usage is too large,which causes poor real-time performance,an improved approach is proposed. Among a group of particles gathering in a particular state,the statistical properties of particles are identical. By applying the Kalman updating step to one representative particle in the group of particles,and using it repeatedly in the same group,the complexity is reduced and arithmetic speed is improved. Combining the proposed distribution and adaptive resampling methods from the Gmapping algorithm,the results of actual experiment carried out with Pioneer III robot and ROS platform illustrate that the real-time performance of the proposal could be enhanced while ensuring the quality of grid map.
出处
《智能系统学报》
CSCD
北大核心
2015年第3期460-464,共5页
CAAI Transactions on Intelligent Systems
基金
国家自然科学基金资助项目(51075420)
重庆市教委科学技术研究项目(KJ120519)
