自主式水下机器人(AUV)采用的传统导航方式，都存在需要外部声学基阵或定期上浮修正累积误差的问题。同步定位与建图方法（SLAM）在AUV水下潜航过程中，可根据位置估计和多波束声纳获得的海底地形信息，预估自身位姿并完成数据关联，在形成海底地形图的同时，利用地图进行自主定位和导航。. 针对海底地形相邻帧数据无法准确关联所带来的一系列难点问题，开展多波束测深数据滤波、空间归位与畸变修正方法探索，建立强弱关联相结合的海底地形SLAM位姿图模型；研究以全局路径修正为主、局部路径修正为辅的海底地形图一致性保持与定位误差约束方法；构建地形匹配结果置信度评估模型，探索伪定位点判别方法，进行考虑地图一致性的鲁棒滤波方法研究。. 通过半物理仿真、船载多波束声纳海中试验、AUV集成多波束声纳海中试验，验证海底地形SLAM方法的可行性和可靠性，为实现AUV“长期、独立、全天候、高精度”水下导航提供理论依据。

Conventional forms of navigation on autonomous underwater vehicle (AUV), relies on external acoustic arrays or regular surface of the vessel to correct accumulated navigation error. Simultaneous localization and mapping (SLAM) can acquire bathymetric data according to location estimation and data received from multi-beam sonar to predict pose along with location and complete data association, and this means the map could be created for localization and navigation while underwater bathymetric charts are continued to form.. To conquer the series of issues with the absence of accurate data associations between adjacent pings, researches and done in methods for filtering, spatial homing and distortion correction, and established models to combined strong and weak data association into pose-graph; researches are also done in methods keeping bathymetric charts consistent and constraining localization errors via global path planning supported with local path planning; constructing model for evaluation of terrain matching belief results, exploring methods for identifications of fake matching, and researches on robust filtering methods with the consideration of map consistency. . By conducting experiments on semi-physical platform, sea trails with shipborne multi-beam sonar, and sea trails of AUV-embedded multi-beam sonar, the reliability and stability of bathymetric SLAM method is validated, and this research provides theoretic and technique foundation for achieving AUV’s goal of ‘Long term, Independent, All-weathered, High precision’ navigation.