Catalogue of Artificial Intelligence Techniques
Keywords: Knowledge Sources, scheduler
Categories: Computer Architecture , Problem Solving
Author(s): Martin Bennet, John Lumley
An architectural technique developed principally for continuous reasoning `real-time' systems. Based on the concept of `experts' sitting round a blackboard, it attempts to co-ordinate the activities of a number of different Knowledge Sources (KS) by providing a global database between them, to which partial solutions to the problem under examination are posted. The blackboard is divided into a number of abstraction levels, each one containing hypotheses for partial solutions linked to other levels by logical relationships; a monitor controls access to these hypotheses and inspects any changes in order to notify KSs of those of interest. Each KS is independent of all others and interfaces externally in a uniform way that is identical across KSs and in which no KS knows which or how many others exist. In general, a KS monitors a level of the blackboard for conditions for which its knowledge is applicable, it then proposes to process those conditions by placing an item on the Agenda. The agenda is a list of possible processing events from which the scheduler must choose the one most likely to lead to a complete problem solution. This decision making process can be controlled by a static set of rules about problem solving and a dynamic goal structure, which change as the solution progresses, to focus attention in a data-directed fashion. The chosen event is passed for execution to its instantiating KS. The blackboard idea has also been extended to a hierarchical model, with differing concepts on different blackboards and knowledge sources `piping' information between blackboards in the form of expectations, supports, refutations, etc.
- Erman, L.D. and Lesser, V.R., A multi-level organisation for problem solving using many, diverse, cooperating sources of knowledge
, Proceedings of IJCAI-75, pp.483--490.