A decomposition-based control strategy for large, sparse dynamic systems

Abstract

We propose a new output control design for large, sparse dynamic systems. A graph-theoretic decomposition is used to cluster the states, inputs, and outputs, and to identify an appropriate bordered block diagonal structure for the gain matrix. The resulting control law can be easily implemented in a multiprocessor environment with a minimum of communication. A large-scale problem is considered to demonstrate the validity of the proposed approach.