This research, funded as an Engineering Research Initiation Award, develops methods of logic inference for parallel computers. The proposed scheme is based on maintaining logical data in related clusters. Reasoning (resolutions) on these clusters can be performed concurrently on multiple processors. This approach will utilize existing multicomputers; no special-purpose architecture is assumed. Optimal methods of of partitioning clause sets and managing clause migrations will be studied analytically and experimentally. Knowledge based systems cannot effectively use logical inference unless reasoning can be carried out at high speed. The importance of this research is that parallel processing offers one promising route to the necessary speedup; logical clustering is a method to "divide and conquer" a logical inference problem, spreading the computation effectively among many separate processors.
