Belief Management for HRI Planning

Abstract

This paper presents an extension of a hierarchical planning approach designed to handle multi-agent problems and, more especially, Human-Robot Interaction problems in which a robot and a human have to collaborate in order to achieve a joint goal. Our method allows to reason and plan for agents that have different or incomplete beliefs in order to produce feasible and comprehensible plans. It is based on a new description of the agent's beliefs and a mechanism that produces and inserts some communication actions into the current plan.