Adaptive Supervisory Control of Multi-agent Systems

Multi-Agent Systems (MAS) provide for the modeling of practical systems in the elds of communications, exible manufacturing, and air-traac management 1]. Herein, we treat the general MAS as an Interconnected Discrete Event System (IDES). By design or due to failure, the actual system is often subject to change and so is its IDES model. For example, at the highest level, an existing subsystem may fail or a new subsystem may be connected. At a lower level, the structure of a subsystem may change as well. In spite of these changes, we want the MAS to preserve important properties which describe correct overall behavior of the system. For example, we want agent actions to be non-connicting. Formal veriication followed by self-repair is a solution, albeit a complex one. The complexity is due to the credit assignment, i.e., if veriication fails we must determine which agents' actions are responsible for the failure of the MAS to satisfy the properties, and what is the best way to resolve the connicts. This could entail highly complex reasoning to resolve. A simpler solution may be found by recasting the multi-agent paradigm in the context of supervisory control theory { a method well-known in the realm of discrete event systems that is applicable to the class of IDES. Unfortunately, structural changes in an IDES hamper the application of current supervisory control theory. At present, the literature ooers but a few works on adaptive or robust supervisory control to tackle the problem of uncertainty in the IDES model 2]. Here we present a novel approach that tackles the issue of adaptive supervisory control in a multi-agent context. Two architectures are presented. The rst of our two MAS architectures is one where a single supervisor guarantees the system-wide properties of the MAS. The second is a distributed architecture in which supervisors are themselves agents. For this second architecture , each supervisor agent supervises a MAS subsystem. In this latter case, there exist interactions between agents via subsystem interconnections. Our objective is to develop a method for adaptive control of the aforemen-tioned MAS architectures. The proposed approach combines supervisory control,