Closed-loop multilevel control of large nonlinear systems via invariant imbedding techniques

Abstract

An efficient multilevel control technique is developed whereby the closed-loop control policy is derived for a wide class of large non-linear dynamic systems using the invariant imbedding algorithms. The basic idea behind the developed technique is the decomposition-coordination procedure of hierarchical multilevel systems theory. Modification of the local subproblems at the first level results in the requirement of solving essentially the dynamic state equations as well as the Ricatti-like ordinary matrix equations. At the second level the coordinator updates the intervention parameters using a conjugate gradient algorithm. A principal feature of the developed technique is reflected in the possibility of being implemented in a real-time. Computer simulation of practical system applications greatly supports the potentiality of the developed two-level optimization technique.