基于粒子滤波器的室内移动机器人自定位

Indoor mobile robot self-location based on particle filter

针对里程计和超声波传感器构建地图时由于累积误差易造成的地图扭曲失真,引入红外定位传感器作为绝对路标信息,生成全局拓扑地图,并在此基础上利用贝叶斯理论进行局部栅格地图的构建,混合地图减小了里程计的累积误差,提高了地图的稳定性.在此栅格地图中,采用改进的粒子滤波器进行定位,基于大权值粒子及周围空间描述机器人位姿置的概率更大的思想,提出了大权值自适应算法,较好地解决了传统粒子滤波器迭代过程中的退化问题.实验结果表明,在250 cm×500 cm的区域内,绝对路标栅格定位方法能够准确生成地图,改进的粒子滤波器的定位误差小于2 cm.

Focusing on issues of map-distortion caused by accumulated error in mapping with only odometer and sonar sensors,new infrared location lags as absolute landmarks are introduced and build a topological map.Then a grid map is also built in the topological map and landmark information updates related incremental map on Bayesian theory.As result,the hybrid map reduces the accumulated error of sensors and improves the stability of the built map.Moreover,improved particle filter is used for robot localization,whi...