Inductive inference of optimal controllers for uncertain logical discrete event systems

Abstract

This paper summarizes preliminary results on the implementation and convergence of an inductive learning algorithm used to identify optimal logical DES controllers. The algorithm is a modification of Angluin's L* procedure. It uses queries to an equivalence oracle and incomplete membership oracle to learn a minimal deterministic finite automaton consistent with the supremal controllable sublanguage of a known specification language and unknown plant. This paper shows how the proposed algorithm is implemented, provides results on the procedure's convergence properties, and presents an example illustrating its use.