Team Cooperation for Plan Recovery in Multi-agent Systems

The paper addresses the problem of recovering the execution of a multi-agent plan (MAP) when the occurrence of unexpected events (e.g., faults) may cause the failure of some actions. We deal with MAPs where actions are executed concurrently by a group of agents organized in teams; each team has to achieve a specific portion of the global goal of the given MAP. The MAP is decomposed into sub-plans and distributed among the agents in the system; each agent performs a local control loop on the progress of the sub-plan it is responsible for. At this local level of control, each agent has to on-line monitor the execution of the actions and, in case the occurrence of a fault causes an action failure, the agent has to repair (if possible) its own sub-plan by means of a replanning step. When such a local recovery mechanism is unable to recover from the action failure, a more powerful recovery strategy at team level is invoked. Such a strategy is based on the cooperation of agents within the same team: the agent in trouble asks another teammate, properly selected, to cooperate for recovering from a particular action failure. The cooperation is aimed at achieving the goal assigned to the agents’ team despite the action failure and to this end the agents exchange each other their own sub-plans and sub-goals. A new plan for each cooperating agent is obtained (if possible) by means of a replanning step.