PARADISE: A parallel architecture for a distributed inference engine

A framework is described for the parallel implementation of a distributed inference engine (PARADISE). In the system architecture there are two kinds of nodes, the inference nodes and the communication nodes. The cooperation among the inference modules is achieved using a distributed blackboard technique. A real implementation of PARADISE can be based on a network of Inmos transputers connected according to a ring topology. A set of concurrent processes mapped onto each transputer implements the node operations. Differences and improvements in comparison with the traditional sequential solutions are discussed