Match Algorithm for a Real Time Expert System

Rule based systems or production rule languages have become an established means of encoding expertise 2]. The evaluation of satisfaction of a set of production rules (match phase in the inference cycle) is computationally expensive 2]. RETE 1] and TREAT 3] are two algorithms that perform this evaluation eeciently. The complexity of these algorithms is O(W C), where W is the number of working memory elements and C is the number of conditions elements in the rule base. This exponential bound leads to unpredictable response times. This unpredictability in response times is not acceptable in real time applications. Alternate match algorithms with polynomial bounds and hence predictable response times are necessary for real time expert systems. We present a match algorithm with polynomial time bound. The polynomial time bound is obtained by exploiting the real time domain semantics. This algorithm is implemented in a real time expert system shell, REX 4, 5].