一种基于NEAT的智能体路径规划方法

A NEAT-based algorithm for Agent Path Planning

针对非结构化、未知环境下的智能体路径规划问题,提出了一种基于增强拓扑神经进化(Neuro-Evolution ofAugmenting Topologies)的方法。通过对智能体进行仿真建模,为智能体提供感知模型和记忆模型,将感知模型和记忆模型的数值作为神经网络的输入来指导智能体在仿真环境中的行为。设定合理的适应性函数,对智能体在仿真环境中执行路径规划任务的表现进行评价。通过NEAT算法对指导智能体行为的神经网络进行结构和权值优化,并生成一个最佳神经网络。仿真实验显示,基于NEAT算法进化出的神经网络,可以指导智能体快速地寻找到一条有效路径。

Based on Neuro-Evolution of Augmenting Topologies(NEAT),this paper proposes a path planning algorithm for unstructured and unknown environment.Through the simulation modeling of the agent,the agent is provided with the perception model and the memory model,and the values of the perception model and the memory model are used as the input of the neural network to guide the behavior of the agent in the simulation environment.A reasonable adaptive function is set to evaluate the performance of agents in path planning tasks in simulation environment.NEAT algorithm is used to optimize the structure and weight of the neural network that guides agent behavior and generate an optimal neural network.The simulation results show that the neural network based on the NEAT algorithm can guide the agent to find a valid path quickly.