基于图像声纳SLAM算法在AUV组合导航中的应用

Application of SLAM Algorithm Based on Image Sonar to AUV Integrated Navigation

为了在有限的体积、功耗及成本范围内提高自主水下航行器(AUV)远程、深海导航定位精度,提出了将基于图像声纳的同时定位与地图创建(SLAM)方法应用于AUV水下组合导航系统,利用图像声纳获取AUV与地形特征点之间的距离和3D方位数据,结合捷联惯性导航系统(SINS)得到的导航数据,通过扩展卡尔曼滤波(EFK)方法对载体状态和地图状态进行连续并行估计和量测,将得到的误差估计反馈回SINS进行修正,可抑制其随航行时间和距离增加的姿态、速度和位置误差。此外,在地形特征点向量中加入声学尺寸元素,可提高特征识别的准确性。仿真结果表明,在持续观测到有效的地形特征点条件下,惯导误差得到了较好抑制,特别是在AUV返程或往复巡航过程中,重复观测到同一地标时,可大幅提高水下组合导航的定位精度。

To improve the long-distance or deep-sea navigation and localization precision for an autonomous underwater vehicles（AUV） within limited volume, power and cost, an algorithm of simultaneous localization and mapping（SLAM） based on image sonar is presented for AUV underwater integrated navigation system. According to the data of real-time distance between each feature on map and an AUV, the data of three-dimensional bearing from image sonar and the data of navigation from strapdown inertial navigation system（SINS）, the cartier state and map state can be estimated and measured simultaneously and continuously by extended Kalman filter（EKF）. The errors are fed back to SINS for correction, hence the accumulated errors of attitude, velocity and position increasing with nmning distance and time can be effectively inhibited. Furthermore, the accuracy of features identification can be improved by putting acoustics size element into map features vector. Simulation result shows that this method can effectively inhibit errors of inertial navigation by continuously measuring valid features on map. Particularly, the localization accuracy of an underwater integrated navigation is significantly enhanced when same landmarks are observed repeatedly in cruise or return voyage of an AUV.