基于模糊神经网络Sarsa学习的多机器人任务分配

摘要

Aiming at that issue of mutli-robot task allocation in dynamic environment, a utility function model based on fuzzy-neural Sarsa learning networks is proposed, which is to combine the fuzzy inference system, the neural networks model and the Sarsa learning algorithm together. The networks structure, the learning algorithm and the confirming steps of final utility value are designed and finalised. In simulation experiment, to use this model can fast converge and realise the task allocation, and can incessantly optimise the object and path as well.%针对动态环境下多机器人任务分配的问题,提出一种基于模糊神经Sarsa学习网络的效用函数模型,将模糊推理系统,神经网络模型与Sarsa学习算法相结合.设计确定了网络的结构、学习算法以及最终效用值的确定步骤.在仿真实验中,利用该模型能快速收敛实现任务分配,并且能不断优化目标和路径.