Towards an Online approach for Knowledge Communication Planning

Abstract

Explicit coordination among robot usually requires a method to exchange information regarding their intentions and their knowledge of the world. Most of the research found in the multi-robot coordination literature uses ad-hoc approaches for communication. This work is an initial attempt to tackle the communication problem in a generic way. It tries to answer the question of when to communicate, what to communicate and with whom to communicate. It proposes the use of dynamic action sets combined with a Determinized Sparse Partially Observable Tree structure that can address the communication decision problem in an online fashion. Compared with a state-of-the-art approach used for sharing knowledge among robotic teams it shows a constantly better performance.