Adaptive supervisory control based on a preference of agents for decentralized discrete event systems

A decentralized system consists of several local systems. Each local system exists independently and the total behavior of the decentralized system is made by the interaction of each local system. It is possible to regard the decentralized system as an event-driven system when we focus on the logical behavior of the system. In this paper, we consider a decentralized system as a Decentralized Discrete Event System (DDES) which consists of several local Discrete Event Systems (DESs). The DDES is supervised by the central controller, called the supervisor. In our previous work, we have proposed the optimal supervisory control method based on reinforcement learning for the DES. We extend it to the DDES, and propose a learning method of an optimal supervisor for the DDES. In addition, we consider the supervisor and local DESs are the agents which have their own preferences. Each preference is represented by a preference function. We introduce the new value function based on the preference functions. The supervisor learns how to assign the control pattern so as to maximize the value function for the DDES. We show the efficiency of the proposed method by computer simulation.