论文部分内容阅读
针对连续环境下的多机器人追捕-逃跑问题,提出一种基于人工神经网络的方法,利用增强拓扑神经演化(NEAT)方法自动生成并优化控制单个机器人行为的神经网络,并通过联合单个机器人之间的行为实现多机器人协调控制.该方法避免了传统的人工设计神经网络所存在的缺陷.仿真实验显示:迭代50代后即可实现有效的围捕行为,其结果证明了该方法可成功实现多机器人合作完成对入侵者的围捕.
Aiming at the multi-robot hunt-and-run problem in continuous environment, a method based on artificial neural network (ANN) is proposed to automatically generate and optimize the neural network that controls the behavior of a single robot by using the enhanced topological neural evolution (NEAT) method. This method avoids the defects of the traditional artificial neural network.The simulation results show that after 50 iterations, the effective capture behavior can be realized, and the results show that this method can successfully implement multi-robot Cooperate to complete the round of invaders.