A rule-based inference engine which is optimal and VLSI implementable

An inference engine for rule-based expert systems which forms part of the EXPRES system is developed and presented. It is shown to be universal, correct, and optimal with respect to time. Its correctness is due to the following facts: the engine is invoked when attributes change, conditions are checked when their attributes take values, rules are examined whenever their conditions change, and actions are executed when their rules are fired. Optimality of execution time is primarily due to the fact that nothing is done which does not need doing. In particular, nothing happens until attributes change value, conditions are evaluated only if their attributes are defined, and rules are examined only when their conditions are true. Another reason for the efficiency of the inference engine is that it is a finite-state machine which operates on a graph rather than a tree. A VLSI implementation of the system that allows automatic design of universal as well as special-purpose expert systems on a chip is proposed