The Challenge: Exception-capable Agent Systems

A critical challenge to creating effective agent-based systems is allowing them to operate effectively when the operating environment is complex, dynamic, and error-prone. In this paper we will review the limitations of current “agent-local” approaches to exception handling in agent systems, and propose an alternative approach based on a shared exception handling service that is “plugged”, with little or no customization, into existing agent systems. This service can be viewed as a kind of “coordination doctor”; it knows about the different ways multi-agent systems can get “sick”, actively looks system-wide for symptoms of such “illnesses”, and prescribes specific interventions instantiated for this particular context from a body of general treatment procedures. Agents need only implement their normative behavior plus a minimal set of interfaces. We claim that this approach offers simplified agent development as well as more effective and easier to modify exception handling behavior. The Challenge: Exception-Capable Agent Systems A critical challenge to creating effective agent-based systems is allowing them to operate effectively when, as is typical for many domains ranging from manufacturing to office work to military information gathering, the operating environment is complex, dynamic, and error-prone (Suchman 1983; Auramaki and Leppanen 1989; Karbe and Ramsberger 1990; Strong 1992; Mi and Scacchi 1993). In such domains, we can expect to utilize a highly diverse set of agents; some have fairly sophisticated coordination capabilities, but many will be simple encapsulations of legacy applications. New tasks, agents and other resources can be expected to appear and disappear in unpredictable ways. Communication channels can fail or be compromised, agents can “die” (break down) or make mistakes, inadequate responses to the appearance of new tasks or resources can lead to missed opportunities or inappropriate resource allocations, unanticipated agent interdependencies can lead to systemic problems like multi-agent conflicts, “circular wait” deadlocks, and so on. All of these departures from “ideal” collaborative behavior can be called exceptions. The result of inadequate exception handling is the potential for systemic problems such as clogged networks, wasted resources, poor performance, system shutdowns, and security vulnerabilities. In this paper we will review the limitations of current “agent-local” approaches to exception handling in agent systems, and propose an alternative “shared service” approach that offers simplified agent development as well as more effective and easier to modify exception handling behavior. Initial versions of this service have been developed and tested in the multi-agent collaborative design conflict management domain; we will describe our preliminary results as well as our future plans. 1 The authors gratefully acknowledge the support of the DARPA CoABS Program (contract F30602-98-2-0099) while preparing this paper.