一种多假设联合相容SLAM数据的关联方法

An Approach Combined Multiple Hypothesis with Joint Compatibility Data Association for SLAM

针对复杂环境下移动机器人同时定位与地图创建(SLAM)中的数据关联问题,提出了一种多假设联合相容分枝定界算法(MHJCBB)。该算法结合了多假设跟踪(MHT)算法与联合相容分支定界(JCBB)算法的优点,利用在关联时保留的多个联合相容关联假设形成多个机器人航迹假设分支。定义了航迹假设分支的评价函数,根据评价函数计算结果进行剪枝,将机器人航迹限制在一定范围内,以减小计算量,并输出得分最高的航迹假设分支。不同测量误差条件下的数据关联试验结果表明,与经典的最邻近(NN)算法、JCBB算法相比,MHJCBB算法能够获得更准确的关联结果。

Considering the data association problem of SLAM under complex environment,this paper proposes a multiple hypothesis joint compatibility branch and bound(MHJCBB)algorithm.The MHJCBB combines the advantages of multiple hypothesis tracking(MHT)and joint compatibility branch and bound(JCBB).It keeps several joint compatible association hypotheses to generate track hypotheses.According to the scores of the track cost function,the number of track hypotheses can be limited by pruning the track hypotheses which have low scores.Thus,the computational need is reduced.Finally,the track hypothesis with the highest score is selected.Simulation experiments under different measurement errors are carried out.The results indicate that MHJCBB can get more accurate results than nearest neighbor(NN)and JCBB.Therefore,the validity of MHJCBB is obtained.