Multi-robot planning and execution for surveillance missions

This paper presents an architecture for planning and executing a multi-robot mission in presence of disturbances, including intermittent communication. The mission we consider is a surveillance mission in which the team of robots must explore an area and observe a set of zones of interest. The planning algorithm we have developed, HiPOP, takes into account the mission constraints, including task durations, communication constraints, visibility of zones of interest, capabilities of robots. . . From the mission description and additional user knowledge represented by abstract methods, it builds a temporal plan using partialorder planning techniques. The execution of this temporal plan is then made on a distributed way: each robot is responsible of executing its own tasks. If some delays occur, the robots will try to maintain the feasibility of the plan by exchanging macroscopic temporal information between their communication constraints. When the plan is not feasible anymore, or if a higher level disturbance occurs (like a robot being out of order), HiPOP is asked to repair the plan in order to reallocate the remaining observation tasks. This architecture has been deployed on a team of heterogeneous robots, and an application to a surveillance mission by two aerial vehicles and two ground vehicles is shown at the end of the paper.