Architectural Analysis for Supervisory Control of Discrete-Event System

Using probabilistic automata (PA) within the Ramadge and Wonham (RW) framework, the model of discrete-event systems ( DES) is extended to include probabilistic information. The extended model can provide useful numerical results for performance analysis of DES by exploiting Markov Chain properties, as well as allowing the use of logical synthesis techniques. Using the extended model, performance evaluation and pros and cons analysis of centralized and decentralized architectures are provided for two typical DES, Small Factory and Transfer Line. Given estimated cost and profit data, the centralized and decentralized architectures are evaluated in these two examples, and various summary rules as to which architecture to implement are obtained. For the Transfer Line example, 'material feedback' in the loop and only partial knowledge of the global system state in the decentralized architecture can result in blocking. Blocking removal strategies are introduced and compared.