自主水下潜器全局规划及其仿真方法研究

Research on a global planning method and its simulation for autonomous underwater vehicle

应用遗传算法设计出二维/三维全局规划器,用于解决自主水下潜器在大范围海洋环境中自主全局路径规划的问题.规划空间采用栅格法离散建模,染色体采用十进制可变长编码方式,基于领域知识设计了初始种群生成算法和五种遗传算子;基于三维建模软件与视景技术建立了半实物虚拟仿真系统,通过设计不同的使命案例对全局规划算法进行了仿真验证.结果表明:该规划算法具有路径描述简单清晰、算法收敛速度快、求解效率高的特点,满足了使命规划的安全性和可行性要求.

A global planner was designed based on a genetic algorithm （GA） and used to solve the planning problem of 2-degree and 3-degree for an autonomous underwater vehicle （AUV） in the large-range ocean environment. The planning space was modeled using a grid-based approach, so that the digital elevation property was stored in each discrete node of the grid. An encoding method adopted a decimal and variable-length code mechanism for the GA chromosome. To avoid generating any non-feasible chromosomes, domain knowledge was taken into account to generate initial population and design five kinds of generic operators. Using the software MultiGen Creator Pro and Vega, the semi-physical virtual simulation system of an AUV was constructed and a demonstration was done to simulate the process of AUV steering as the result of GA path planning. The simulation result shows that the global planner provides a simpler planned path, has the character of high speed global convergence, and can more efficiently and effectively solve the problem of path planning for an AUV. It meets the safety and feasibility needs of minion planning demand for an AUV.