Discrete Event Modeling and Control Using Finite State Machines with Parameters

Dr. Yi-Liang Chen

Rockwell Science Center

 

Control theories of discrete event systems (DES) modeled as finite state machines have been well developed over the years in addressing various fundamental control issues. However, most of the approaches have suffered several deficiencies that render them unsuitable for many practical, especially complex, applications. First, the demon of state explosion has haunted constantly with unthinkable state space to manage and analyze. Secondly, venerable system configuration/decomposition schemes (i.e., single-level centralized/decentralized, two-level hierarchical) are often inadequate in supporting efficient (and flexible) model and control generation. Finally, the legacy of "supervisory" viewpoint cannot satisfy the need to actively "drive" the systems toward specific objectives/states.

Motivated by the air operation (i.e. command and control) systems, we propose a new hierarchical modeling and control approach in response to the challenges for practical discrete event system applications. We employ both finite state machines and sets of parameters in our modeling primitives, called Finite State Machine with Parameters (FSMwP). Based on this modeling primitive, we describe a multiple (2+) level hierarchical modeling framework with automatic model abstraction capabilities. By introducing event enforcement, we then present a hierarchical control mechanism that combines online safety control with optimal (effectiveness) control of discrete event systems. We describe the current progress on this approach and discuss the specific issues arisen from this approach.