【摘要】 针对水面无人艇(unmanned surface vessel, USV)在复杂环境下的局部路径规划问题,对USV路径规划问题进行了数学建模,提出了基于增强拓扑神经演化(neuroevolution of augmenting topologies, NEAT)算法的局部路径规划方法;设计了神经网络初始结构和演化参数,对初始神经网络结构进行演化实现避障及到达指定目标的路径规划任务;通过设计适应度函数,实现路径点数目的优化。仿真结果表明:利用NEAT算法演化神经网络的方法能够使USV在复杂的环境中准确避开障碍物并到达目标点,且在路径点数目和鲁棒性方面优于传统的模糊逻辑算法与人工势场算法。更多还原

【Abstract】 Aiming at the local path planning problem of the unmanned surface vessel(USV) in complex environment, a model of USV path planning problem was built, and a local path planning method was proposed based on the neuroevolution of augmenting topologies(NEAT) algorithm. The initial structure and evolution parameters of the neural network were designed. Through the evolution of the initial neural network structure, the path-planning task of avoiding obstacles and reaching the designated target was realized. By designing the fitness function, the number of path points was optimized. Simulation shows that the method of using NEAT algorithm to evolve neural network could make USV to bypass obstacles and reach the target point accurately in complex environment. In addition, the NEAT algorithm was shown superior to the traditional fuzzy logic algorithm and artificial potential field algorithm in the number of path points and robustness.更多还原