H-MAS architecture and reinforcement learning method for autonomous robot path planning

This work provides a novel collaborative Q-Learning method using a Holonic Multi Agent System (H-MAS), with the use of two types of Q-table; Q-Master Table (QMT) and Q-Embedded Table (QET) instead of the classical use. The QMT is updated by a collaborative policy between head holons. Thus the QET is updated based on the individual decision of each agent, so that each and every robot in the environment can use and profit not only from its own experiences but also from the experiences of other agents. This method is supported by a H-MAS architecture based on holonic and internal agents, and a Fuzzy Logic System, in order to solve the problem of path planning for a mobile robot in a partially or completely unknown environment.