一种移动机器人SLAM中的多假设数据关联方法

针对移动机器人同时定位与建图(SLAM)中的局部数据关联问题,提出一种基于粒子滤波的多假设数据关联方法.该方法将数据关联问题转换成离散优化问题,利用多个粒子来维持多种数据关联假设,通过计算关联代价来获得粒子权重,用基本剪枝技术在粒子重采样过程中滤除错误的数据关联假设.研究结果表明:该方法弥补了经典的数据关联方法中关联假设一旦确定就不能修改的不足；与ICNN和JCBB数据关联方法相比,该方法能获得更正确的数据关联结果和更高的定位精度.

A multiple hypotheses data association method in mobile robot SLAM

According to the local data association problem in mobile robot SLAM process,a new multiple hypotheses data association method based on the particle filter was presented.In the method,the data association problem was transformed as the discrete optimization,and multiple particles were used to maintain the multiple data association hypotheses and every particle’s weight was calculated by association cost.During the resample,the wrong hypotheses were discarded through basic branch and bound approach.The results show that the method resolves the problem where the classic method cannot modify the previous association hypothesis.By experimental results analysis and comparison,the new method can reach more correct data association results and higher location precision than the classic ICNN and JCBB method.