CoMutaR: A framework for multi-robot coordination and task allocation

In multi-robot systems, task allocation and coordination are two fundamental problems that share high synergy. Although multi-robot architectures typically separate them into distinct layers, relevant improvement may be expected from solutions that are able to concurrently handle them at the same ¿level¿. This paper proposes a novel framework, called CoMutaR (coalition formation based on multitasking robots), which is used for both tackle task distribution among teams of mobile robots, and to guarantee the coordination within the formed teams. Robot capabilities are modelled as actions, independent modules whose inputs do not depend on the robot that generated it. Solutions to tasks are devised as coalitions of actions, that can be spread amongst the available robots. We also define the concept of share-restricted resources, which are periodically checked and updated by the actions in the system. In contrast to prior approaches, this mechanism enables to quickly determine if two actions can be executed simultaneously, allowing a single robot to execute multiple tasks concurrently. A single-round auction protocol is used to automatically discover and form coalitions. Once a coalition is formed, coordination among robots is modelled as constraints imposed over the share-restricted resources. Finally, we have successfully implemented and applied CoMutaR in typical scenarios like object transportation, area surveillance, and multi-robot box pushing. Experimental results demonstrate that the system is able to provide good solutions even in the case of severe failures in participating robots.