融合测深/测距信息的水下同步定位与建图技术

Underwater Simultaneous Localization and Mapping Fusing Bathymetric and Range Information

针对测深信息同步定位与建图技术(SLAM)应用于AUV中由于使用的多波束测深声呐相邻测量条带无信息冗余导致的帧间运动估计缺失和回访历史位置前无法准确定位问题,提出了一种融合测深/测距信息的水下SLAM技术。此方法在测深数据关联基础上,通过AUV和信标之间的声学通信增加测距信息约束,实现对AUV位置的实时估计。针对距离量测中的异常值易导致SLAM算法失效这一困难,提出了一种概率图形分割算法,同时,提出一种应用信息论节点约简算法,解决了测深/测距SLAM中位姿图中节点过多导致的算法实时性差难题。结合海上实测数据开展回放式仿真试验,结果表明:概率图形分割算法能够准确地识别出距离量测异常值,应用信息论节点约简算法能够剔除位姿图中的多余节点,缩短约60%程序运行时间;所提出的融合测深/测距信息的水下SLAM技术的绝对轨迹误差平均值为9.55 m,定位效果相较于测深信息SLAM(11.58 m)和仅测距SLAM(21.40 m)有明显提升。

Bathymetric simultaneous localization and mapping(BSLAM)method has been widely used on localization for AUV,but the method can’t estimate the inter-frame motion of the vehicle and locate a vehicle accurately before revisiting since the lack of overlap between adjacent bathymetric swath collected using a multi-beam echosounder(MBES).In order to solve the problem,an underwater SLAM method fusing bathymetric and range information is proposed.The method adds range measuremnt data assoications through acoustic signal exchanges between the vehicle and beacons,based on bathymetric data associations,to achieve real-time estimation of AUV position.In view of the difficulty that outlier in range measurement easily leads to SLAM algorithm failure,a probabilistic graph partitioning algorithm is proposed.Meanwhile,an Informationtheoretic node reduction method is proposed to solve the problem of poor real-time performance of the algorithm caused by too many nodes in the pose graph.Simulation experiments combined with the field data were conducted,and the results showed that probabilistic graph partitioning algorithm can identify outlier in range measurements accurately;information-theoretic node reduction method can eliminate redundant nodes in the pose graph,reduce Program runtime by approximately 60%;the average absolute trajectory error of our proposed method is 9.55 m,the localization result is significantly improved compared with bathymetric SLAM(11.58 m)and RO-SLAM(21.40 m).