高斯混合概率假设密度SLAM算法

Gaussian mixture probability hypothesis density SLAM algorithm

研究了同步定位与地图创建(SLAM)中的数据关联问题。针对环境特征数未知时,数据关联的误关联率增加,导致SLAM的定位精度偏低的问题,提出了高斯混合概率假设密度SLAM算法。首先采用UFastSLAM解决SLAM中的粒子退化和耗尽问题,其次针对地图特征数未知的情况,将UFastSLAM算法中的数据关联问题转换成有限集统计理论跟踪算法的高斯混合问题,利用高斯混合概率假设密度(Gaussian Mixture Probability Hypothesis Density,GMPHD)算法解决UFastSLAM中数据关联问题。仿真实验结果表明本文提出的GMPHD-UFastSLAM算法在地图特征个数未知的情况下,数据关联准确率和定位精度都得到了提高。

This paper focus on the data association problem in the Simultaneous Localization and Mapping （SLAM）.Incorrect rate of data association would be increased under the enviromment with unknown：umber of the features,whereby leading to a decrease in the positioning accuracy of SLAM.To solve this problem,a SLAM algorithm based on Gaussian Mixture Probability Hypothesis Density is proposed.Firstly,UFastSLAM algorithm is used to deal with the problem of the particle degradation and exhaustion.Secondly,for the unknown number of the map features,the data association problem in the UFastSLAM is converted into the gaussian problem of finite set statistics theory tracking algorithm,then GMPHD algorithm is adopted to solve the data association problem in the UFastSLAM algorithm.Simulation results show that the proposed GMPHD-UFastSLAM algorithm can improve the correct rate of data association and position accuracy of the robot under the enviroment with unknown number of the map features.