Can production systems be applied to hard real-time applications?

Abstract

Efficient match algorithms have influenced the acceptance of production systems in the industrial world. However, in embedded control systems they have not been applied intensively because of their non-deterministic run-time behavior. Until recently no method was known yielding rigorous upper bounds for run-time on match algorithms. Although the already published MLR (Micro Level Reasoner) method yields very good run-time estimations for basic actions in the RETE match algorithm it can only be applied to soft real-time problems. We present the UB (Upper Bound) and EUB (Extended Upper Bound) methods which in contrast to MLR guarantee accurate upper bounds for basic actions in RETE or TREAT match algorithms. Hence, both algorithms can be applied to solving hard real-time problem in a limited way.<>