认知框架下远程AUV多波束测深主动定位算法研究

Navigation system based on inertial sensors is normally adopted as the main navigation system of long range AUV, whereas such kind of navigation system is not suitable for long range, large scale accurate underwater positioning, just because of its accumulated positioning error. Considering the concealment performance, long endurance and constrained by the energy and sensors of long range AUV, the positioning error of inertial navigation system can be effectively reduced by Terrain-Referenced Navigation system (TRN) based on multibeam sonar. However, there are still some challenges faced by TRN when it is used in remote underwater navigation. First, the amount of navigation cells in a priori sea chart is normally small, which restricts the correction frequcncy and positioning accuracy of TRN. Second, if the distance between navigation cells is large, meanwhile the TRN cannot restrain the divergence of position error by repeatedly getting highly navigable terrain information, that all can reasonably make the TRN have poor robustness. Considering the comparatively high similarity of the unstructured undersea terrain which will easily induce aliasing to the terrain information perceived by multibeam sonar, this project will research on the TRN algorithm in a cognitive framework, including constructing the active positioning algorithm framework constrainted by navigability, research on the representation of undersea terrain information and the real time assessment of its navigability as well as the acquisition strategy and solidification strategy of the navigation cells, which makes the TRN have the ability of active navigation and active perception so as to improve the accuracy, robustness and intelligence level of the navigation system of long range AUV.

远程AUV核心导航系统多为惯性基导航系统，此类导航系统的定位误差随时间累积,不适于远距离、大范围的水下准确定位。考虑到远程AUV的隐蔽性、远程性并受其能源和传感器约束，基于多波束测深信息的地形参考导航系统是抑制惯性基导航系统定位误差的可行方法。但目前该系统的水下远程应用还存在以下主要问题：一、先验海图中的可导航区通常较少,制约了地形参考导航系统的校正频率和定位精度；二、若导航区之间相距较远，地形参考导航系统又不能通过反复获取可导航性高的地形信息来抑制定位结果的发散,故鲁棒性较差。考虑到非结构化海底地形的高度相似性，极易导致多波束声纳对海底地形产生感知歧义，本项目拟在认知框架下构建可导航性约束的主动定位算法，研究海底地形信息的表达及其可导航性的实时评估算法以及导航区的获取策略与固化策略，使地形参考导航系统具有主动导航和主动感知的能力，提升远程AUV导航系统的精度、鲁棒性和智能水平。

远程AUV核心导航系统多为惯性基导航系统，此类导航系统的定位误差随时间累积,不适于远距离、大范围的水下准确定位。考虑到远程AUV的隐蔽性、远程性并受其能源和传感器约束，基于多波束测深信息的地形参考导航系统是抑制惯性基导航系统定位误差的可行方法。但目前该系统的水下远程应用还存在以下主要问题：一、先验海图中的可导航区通常较少,制约了地形参考导航系统的校正频率和定位精度；二、若导航区之间相距较远，地形参考导航系统又不能通过反复获取可导航性高的地形信息来抑制定位结果的发散,故鲁棒性较差。考虑到非结构化海底地形的高度相似性，极易导致多波束声纳对海底地形产生感知歧义。针对上述问题，本项目基于混合式多波束线数据实时空间分解算法，建立三维海底数字高程模型，实现了导航地图动态在线构建机制；利用分形扫描算法，基于Hausdorff分形维数和条带地形起伏量分布概率，实时评估条带状地形的可导航性，实现了海底地形表达与可导行性评估相融合的海底地形特征认知机制，为导航区的获取策略与固化提供决策依据。在此基础上采用朴素贝叶斯分类器，模拟视觉关注机制，将无人水下航行器引导到地形信息更加丰富的地形区域， AUV利用混合式多波束线数据实时空间分解算法对地形信息丰富的区域进行在线制图，完成可导航区域的获取，并借助先验地形信息将在线的制图结果固化下来，构成分辨率更高的海底地形，在此基础上，AUV在全局规划路线的基础上，主动选取全局规划路径途经的地形粗糙区域，基于强跟踪平方根UKF算法和已知海底地图抑制水下导航系统定位误差发散。使地形参考导航系统具有主动定位的能力，提升远程AUV导航系统的精度、鲁棒性和智能水平。