Coordination through Joint Intentions in Industrial Multiagent Systems

My Ph.D. dissertation (Jennings 1992a)1 develops and implements a new model of multiagent coordination, called JOINT RESPONSIBILITY (Jennings 1992b), based on the notion of joint intentions. The responsibility framework was devised specifically for coordinating behavior in complex, unpredictable, and dynamic environments, such as industrial control. The need for such a principled model became apparent during the development and the application of a general-purpose cooperation framework (GRATE) to two real-world industrial applications. These experiments were successful in that it was possible to instantiate useful cooperation schemes; however, when anything unexpected happened (for example, new information invalidated existing goals, synchronization between actions was disrupted, or agents had misinterpreted the situation), the multiagent community acted incoherently. For example, agents would continue to work on a goal even though one community member knew their processing was obsolete; agents would stop processing requests if a more important task arose, without informing the originator; and agents would wait for the results of a task that had been abandoned. This incoherence occurred because the GRATE agents did not embody sufficient knowledge about the process of team problem solving. Therefore, it was decided to provide GRATE simplifies this process by providing a shell that contains built-in generic knowledge related to cooperation and control. The application designer can then build on this preexisting base of knowledge rather than construct the system completely from scratch (as is currently the case). To substantiate the claim that the knowledge is generic, GRATE was used to build two industrial applications: detection and location of faults in an electricity transportation network (Jennings et al. 1992) and cooperative diagnosis of a particle accelerator beam controller (Jennings et al. 1993). In both cases, the designer was able to construct a working multiagent system in a relatively short time span and did not need to augment the built-in knowledge.