Variable abstraction and approximations in supervisory control synthesis

This paper proposes a method to simplify Extended Finite-state Automata (EFA) in such a way the least restrictive controllable supervisor is preserved. The method is based on variable abstraction, which involves the identification and removal of irrelevant variables from a model. Variable abstraction preserves controllability, and the paper shows how approximations can be used to ascertain least restrictiveness of the synthesis result. The approach has the modelling benefits of Extended Finite-state Automata, leads to optimal control solutions, and reduces the synthesis cost. An example of a manufacturing system illustrates the contributions.