Multirobot communication and task coordination

Abstract

This thesis is concerned with multi-robot systems. Robots are increasingly being used in applications such as patrolling or exploration whose natures necessitate the consideration of multiple cooperating robots. The cooperation may vary from sole communication to collaboration. In this perspective, the cooperation problem is examined from three critical and related aspects. First, the problem of achieving effective communication topologies is considered. Two alternative approaches - namely decentralized or centralized network topologies - are proposed based on pairwise games. Depending on the robots’ objectives, the pairwise stability of the resulting networks and the convergence of the associated games are analyzed. Next, the problem of determining the assisting robots in tasks requiring the cooperation of robot pairs is considered. For this, the concept of assistance networks is introduced. Assistance networks designate potential helpmates in an implicit manner. The network topology is determined by a coordinator in a manner similar to that of centralized network topologies. Finally, the more general case of tasks requiring a multitude of resources is considered. A task coordinator is held responsible for finding robot coalitions endowed with these resources with minimal cost. This process is modeled as a series of coalition formation games. Extensive simulation and experimental results demonstrate the practical applicability of the proposed approaches in real-time multi-robot applications.