COORDINATION OF MULTIAGENT SYSTEMS WITH FUZZY CLOCKS

An Object Petri Net (OPN) model for coordination of multiagent systems based on fuzzy clocks is introduced. The OPN model takes a refined de sign-to-time approach in real-time scheduling of Concurrent Engineerin g (CE) projects. Agent tasks are subdivided into required and optional subtasks to take advantage of the imprecise nature of the computation performed by an agent. Concurrent with each agent is a fuzzy clock wh ich measures the timeliness of an agent-task relative to temporal inte rvals and prescribed deadlines. In the context of a CE project, agents cooperate with a coordinator in managing their time, and in periodica lly supplying tasking status reports for a blackboard. The blackboard is handled by a subagent of the coordinator, Performance is measured i n terms of elapsed times, deadlines, Quality-oi-Product statistics, ca pability, consumption of resources, amount of work done, and difficult y of a task. The coordinator also maintains a collection of concurrent , hybrid Fuzzy Watchdog Clocks (hFWC) which provide useful statistics (current fuzzy time, elapsed time, and percent of allowed time for eac h agent) posted on the blackboard. The contribution of this article is the introduction of fuzzy clocks in constructing timetables for proje ct tasks and in making appropriate adjustments in the overall schedule whenever agents fail, disappear, or are in conflict.