深海未知环境下的AUV仿生导航信息融合机理与算法

Long endurance navigation is the first problem to be solved in the application of the deep-sea autonomous vehicle. Long endurance survey in unknown deep-sea area by autonomous vehicle is taken as the research background of this project. Aiming at the problem of serious dependence on a prior knowledge of the geophysical field during deep-sea autonomous vehicle long endurance navigation, a novel method of bio-inspired long endurance navigation is proposed. Major research topics are as follows: analysis of marine navigation information elements and their utilization mechanism; bio-inspired navigation model by multi-geophysical field information fusion under unknown deep-sea environment; bio-inspired navigation estimation algorithm under deep-sea autonomous vehicle performance constraints. By imitating utilization and switching mechanisms of the marine navigation information, the proposed bio-inspired navigation model does not require a priori database of the geophysical field and could take full advantage of multi geophysical field information fusion to improve the navigation accuracy. Meanwhile, the presented navigation parameter estimation algorithm based on data representation and association optimization will overcome the problem of the inconsistency of spatio-temporal attributes of multi geophysical field data measured in real time and reduce the calculation burden of the autonomous vehicle. Accordingly, the ability of an autonomous vehicle to carry out long endurance survey in unknown deep-sea area has been greatly improved. The expected findings of this project will provide a theoretical basis and information fusion algorithm reference for the research and development of navigation systems for deep-sea applications and promote the development of applications such as deep-sea autonomous vehicles and autonomous mobile robots.

长航时导航是深海自主航行器应用中需要解决的首要问题。本项目以自主航行器在深海未知区域开展长时间调查为研究背景，针对目前深海自主航行器长航时导航对地球物理场的先验知识依赖严重的问题，提出长航时仿生导航的新思路，研究内容主要包括：海洋生物导航的信息要素及其利用机理分析；深海未知环境下多物理场融合仿生导航模型；深海自主航行器载体性能约束下的仿生导航估计算法。通过模仿海洋生物导航信息利用及切换机制，提出的仿生导航模型不需要地球物理场先验数据库，能够充分利用多物理场信息融合来提高导航精度，基于数据表示和关联优化的导航参数估计算法能够克服多物理场数据时空属性不一致的问题和降低自主航行器计算负担，从而大大提高自主航行器在深海未知区域开展长时间调查的能力。预期研究成果将为深海领域导航系统的研究和开发提供理论支撑和信息融合算法参考，促进深海自主航行器以及自主移动机器人等应用领域的发展。

针对深海未知区域AUV长航时导航中的难点问题，通过本项目研究，开展了海洋生物导航全过程中多物理场信息的利用机理分析，通过文献调研，将海洋生物的导航信息要素归纳为地磁场、与嗅觉相关的化学线索、太阳罗盘、以及海流与波浪线索四大类，多源导航线索的使用能够让海龟归巢效率大大提高，在海洋生物导航信息利用机理方面，分析得出获取导航信息的同时“磁图”也在与环境发生信息交换，学习新的“导航知识/信息”与导航策略。在此基础上，结合多物理场融合数据库，设计了基于分布式异质传感器的仿生航姿模型与基于LSTM的多地球物理场SLAM信息分航段匹配方法，实现了受生物导航启发的航向与姿态估计方法和多源信息分阶段利用机制，其中基于“自由基对”模型的仿生罗盘与现有的磁罗盘算法相比具有较强的鲁棒性，进行了未知环境下地磁场敏感距离、基于可观测度的AUV接近初始基准时航迹优化分析，进而完成了适用于未知区域建立初始基准和返回初始基准进行导航校正过程的AUV仿生导航模型研究，深入开展了多物理场数据的表示及关联优化分析，从提高计算效率、降低算法流程与计算复杂度等角度出发，提出了不依赖先验信息的深海AUV长航时仿生导航方法，所提出的地磁反演定位模型为深海长航时惯性导航误差抑制提供了新的约束手段，悬臂梁后端轨迹校正模型的使用使得AUV全航程导航误差得到了有效约束，形成了深海AUV载体约束下的仿生导航优化估计算法，对项目提出的各模型与算法进行了实验验证，项目成果能够为深海未知环境下AUV长航时导航提供理论和算法支撑。