Toward Opportunistic Collaboration in Target Pursuit Problems

Abstract

This paper proposes an opportunistic framework to modeling and control of multi-robots moving-target pursuit in dynamic environment with partial observability. The partial observability is achieved via the introduction of third party agent (referred to as mediator) that transforms the target’s as well as group members’ positioning information into the robotic agents’ common knowledge, thereby eliminating the necessity of direct inter-robots communication. The robotic agents involved are modeled as fully autonomous entities, capable of determining their corresponding action profiles, using a strategy inference engine. The robot’s inference engine consists of two sub-rating components viz. fixed or predefined sub-rating, and the variable or the opportunistic sub-rating. The action profiles at individual level are further analyzed by the mediator that finalizes the agent’s action assignment at every execution cycle. It has been proven that addition of the third party mediator guarantees the optimality of group level performance.