Handling events during multiagent cooperative problem solving is a very difficult task. Events can arise due to a single agent or a group of agents; an event can be detected by the agent that gave rise to it or by others. Deciding who should handle the event, as well as the process of diagnosing it, assessing its importance, and selecting a response may all need to be done cooperatively.
In this paper, we have sketched an approach to handling events in multiagent cooperative systems. The process relies on knowledge about the world, oneself and one's intentions, and others' beliefs and intentions, as well as on additional a priori contextual knowledge, possibly built from past experiences in similar situations. The process also depends on knowing when and what to communicate with others, based on what an agent knows about their current beliefs and intentions, to inform them of pertinent information, to ask for information or assistance, and to negotiate about handling the event.
The work reported here is preliminary. Though we will be developing and evaluating the approach in the AUV domain as part of the ORCA project, we believe it is applicable for handling unanticipated events in most cooperative problem-solving situations in the real world.