Distributed artificial intelligence: a necessary paradigm for supervising production management activities

摘要

Each unit-controller is a knowledge-based system, built on top of a blackboard mechanism. Indeed, each unit-controller must react to about the state of its internal activities as well asq to arrivals of messages sent by other unit-controllers. The blackboard mechanism allows the unit-controller to take such events into account in an opportunistic way and in real-time.

A set of knowledge sources describes the possible behaviours of the unit-controller toward the possible situations. The action of the knowledge sources can be: to start specific internal activities, to send messages to other unit-controllers, or to create new events on the blackboard (i.e. modification of the internal situation).

Each unit-controller uses two kinds of knowledge : specific knowledge about the island it is in charge of, and more generic knowledge about supervision tasks in general (internal and external). This generic knowledge about supervision tasks (internal and external) is used to guide the basic functionning of the blackboard mechanism while specific behavioural knowledge-bases help it to react correctly to events relative to its internal activities or to messages sent by other unit-controllers.

Because, a large part of the knowledge which is used by a unit-controller is generic, we have build a unit-controller shell. Each unit-controller is thus obtained by instantiation of this unit-controller shell and by addition of specific informations about a specific island.

Finally, in the system outcoming from the ESPRIT 932 Project, the task of supervising the Production Management activities is performed by a set of knowledge-based systems using blackboard mechanisms for cooperating with each other and this actually makes it a typical DAI system.