A Distributed Supervisor Synthesis Approach Based on Weak Bisimulation

Abstract

It is well known that space and time complexity is one of the main bottlenecks of supervisor synthesis. The model of a large system is usually too big to be explicitly expressed, owing to composition of local components' models, making supervisor synthesis difficult, if still possible. In this paper, we propose a distributed supervisor synthesis approach, where instead of computing a model of the entire system, a local abstraction of the system with respect to each local component is computed based on weak bisimulation. Then a local supervisor for each component is computed based on the component model and the abstract system model, by using well developed centralized supervisor synthesis approaches. The collection of all resulting local supervisors forms a distributed supervisor