Study on Online εNN Q-Learning for Multi-Robot System Based on Community Perception Network

摘要

Considering the curse of dimensionality and inadequate learning of the historical information provided by the other robots, an online ε-radius Nearest Neighbors (eNN) classification method is employed in the community perception network environment. In order to increase learning efficiency, a new multi-robot reinforcement learning strategy is proposed by employing classification of historical states and information sharing mechanism. Finally, the analysis of convergence of Q values matrix is done, and simulations are shown the effectiveness and efficiency of the proposed scheme.