Toward Distributing Autonomy over Robot Teams

This paper presents a way to command a system of systems robustly as a single entity. Instead of modeling each component system in isolation and then manually crafting interaction protocols, this approach starts with a model of the collective population as a single system. By compiling the model into separate elements for each component system and utilizing a teamwork model for coordination, it circumvents the complexities of manually crafting robust interaction protocols. The resulting system is both globally responsive by virtue of a team oriented interaction model and locally responsive by virtue of a distributed approach to planning as well as model-based fault detection, isolation, and recovery.