Abstract :
An approach to order reduction in Decentralized Systems is presented. In this approach, a measure of "channel effectiveness" is used to retain or drop certain states in a system model. Each control agent, thus ends up with a partial model of the system (possibly with over-lapping states) and uses it to calculate a feedback to minimize a quadratic cost. The composite feedback is suboptimal with respect to the global criterion, however, the suboptimality may be acceptable if the "channel effectiveness" approach is used in obtaining the local models.
