基于群体智能的多自主水下航行器三维路径规划可靠优化研究

Autonomous Underwater Vehicles (AUV) have important application values in military investigation, environmental monitoring and resource exploration, etc. The three-dimensional (3D) path planning of AUV has the effects of stealth and danger avoidance, and it is a complex optimization problem. Because there are many barriers and high dynamics in the underwater environment, and the new challenges of multi-AUV and multiple monitoring targets in the optimization of path planning, this project proposes the research on reliable optimization of 3D path planning for multi-AUV based on swarm intelligence techniques: 1) To establish a unified 3D path evaluation mechanism to reflect the complex underwater characteristics, and to meet the need of multi-AUV collaboration in practical applications. 2) To study the single AUV path planning optimization by the swarm intelligence method, and to propose an adaptive grid encoding strategy for the feasible path construction and constraint optimization in the multi-barrier environment. 3) To study the swarm intelligence method in solving the multiple monitoring targets and multi-AUV cooperative path planning problem, so as to break the bottle neck of multi-AUV path co-optimization by the cooperative coevolution strategies. 4) To study the reliable path planning method in the dynamic environment. By using the Monte Carlo simplified method with the hypothetical test and the two-objective swarm intelligence optimization algorithm, the effectiveness and reliability of the algorithm in the underwater dynamic environment will be further improved. This project is not only a frontier exploration for the swarm intelligence algorithms in multi-constrained, cooperative and dynamic environments, but also an innovative application of swarm intelligence algorithms in AUV path planning.

自主水下航行器（AUV）在军事侦查、环境监测和资源勘探等方面有重要价值。AUV三维路径规划起到隐身避险等作用，是复杂的优化问题。针对水下环境障碍多、动态性强等特点，面向多AUV协同工作、同时监测多个目标对路径规划提出新挑战，本项目开展基于群体智能的多AUV三维路径规划可靠优化研究：1）建立三维路径评价机制，反映水下复杂环境特性和多AUV协同的实际应用需求；2）研究单AUV路径优化的群体智能方法，提出自适应网格编码策略解决多障碍环境中构建可行路径和约束优化的难题；3）研究多监测目标、多AUV协同的路径优化群体智能方法，通过协同进化策略突破多AUV路径协同优化的瓶颈；4）研究动态环境下可靠路径规划方法，利用假设检验简化蒙特卡罗方法和两目标群体智能优化进一步提高算法在水下动态环境中的有效性和可靠性。本项目既是多约束、协同和动态环境下群体智能算法的前沿探索，也是算法在AUV路径规划领域的创新应用。

围绕自主水下航行器（Autonomous Underwater Vehicles，AUV）的三维路径规划难题，提出了基于群体智能的可靠路径规划方法，突破路径规划优化算法在多约束、多AUV协同和动态环境下的适用性瓶颈。在路径规划的性能评价方面，提出了AUV的三维路径规划统一评价机制，既反映水下环境中障碍稠密、AUV移动方式受限等复杂特性，又体现多AUV协同监测多个水下目标的实际需求；针对单个AUV的三维路径规划问题，提出了基于群体智能的路径优化方法，通过对环境自适应网格化，设计有效的路径编码策略，规避障碍物，实现算法的高效路径构建和约束处理；针对多个AUV协同监测多个目标的路径规划难题，提出了基于协同进化框架的群体智能路径优化方法，实现多AUV的高效协同；针对水下动态环境中的路径可靠性问题，综合考虑路径的最优化与动态可靠性这两个目标，提出了基于多峰优化多目标群体智能算法的路径优化方法，实现对多AUV路径的可靠、高效优化。在算法应用上，仿真测试了提出的算法在新西兰三维海底地形下的多AUV协同路径优化。在项目执行过程中，课题组围绕研究目标和内容开展了工作，共发表标注了本项目基金号的学术论文18篇，其中IEEE Transactions系列等国际核心期刊论文4篇，国际会议论文10篇，国内核心期刊论文4篇，申请中国发明专利9项，顺利达到了项目的预期目标。