Abstract

In this thesis, decentralized supervisory control problems which cannot be solved without some communication among the controllers are studied. Recent work has focused on finding minimal communication sets (events or state information) required to satisfy the specifications. A quantitative analysis for the decentralized supervisory control and communication problem is pursued through which an optimal communication strategy is obtained. Finding an optimal strategy for a controller in the decentralized control setting is challenging because the best strategy depends on the choices of other controllers, all of whom are also trying to optimize their own strategies. A locally-optimal strategy is one that minimizes the cost of the communication protocol for each controller. Two important solution concepts in game theory, namely Nash equilibrium and Pareto optimality, are used to analyze optimal interactions in multi-agent systems. These concepts are adapted for the decentralized supervisory control and communication problem.

A communication protocol may help to realize the exact control solution in decentralized supervisory control problem; however, the cost may be high. In certain circumstances, it can be advantageous, from a cost perspective, to reduce communication, but incur a penalty for synthesizing an approximate control solution. An exploration of the trade-off between the cost and accuracy of a decentralized discrete-event control solution with synchronously communicating controllers in a multi-objective optimization problem is presented. A widely-used evolutionary algorithm (NSGA-II) is adapted to examine the set of Pareto-optimal solutions that arise for this family of decentralized discrete-event systems (DES).

The decentralized control problem is synthesized first by considering synchronous communication among the controllers. In practice, there are non-negligible delays in communication channels which lead to undesirable effects on controller decisions. Recent work on modeling communication delay between controllers only considers the case when all observations are communicated. When this condition is relaxed, it may still be possible to formulate communicating decentralized controllers that can solve the control problem with reduced communications. Instead of synthesizing reduced communication protocols under bounded delay, a procedure is developed for testing protocols designed for synchronous communications (where not all observations are communicated) for their robustness under conditions when only an upper bound for channel delay is known.

Finally a decentralized discrete-event control problem is defined in timed DES (TDES) with known upper-bound for communication delay. It is shown that the TDES control problem with bounded delay communication can be converted to an equivalent problem with no delay in communication. The latter problem can be solved using the algorithms proposed for untimed DES with synchronous communication.