Collaborative Q-learning path planning for autonomous robots based on holonic multi-agent system

In this paper we present a novel collaborative Q-learning based path planning system using holonic multi agent system architecture, to use in autonomous mobile robot represented as a head-holon, for planing the optimal path between any starting point and a goal in a grid environment. The mobile robot has to explore the 2D grid randomly in order to update a local state action space Q-table relaying on a standalone decision. A global (Master) Q-table is then update based on collaborative policy between head holons, in which every holon has a preset confidence degree used as a decisive parameter in the Q-learning equation.