Linear Time-Varying Dynamic Systems Optimization via Higher-Order Method: A Sub-Domain Approach

Abstract

A new procedure for optimization of linear time-varying dynamic systems has been proposed that uses transformations to embed the dynamic equations explicitly into the cost functional. This leads to elimination of Lagrange multipliers and characterization of the optimality equations by high-order differential equations in the same number of variables as number of control inputs. This procedure requires that the transformation matrix be nonsingular at all time within the domain. This paper extends this procedure to problems where a single nonsingular transformation matrix does not exist over the entire domain. In this paper, the time domain is partitioned into intervals such that a nonsingular transformation exists over each interval. The transformations are used to embed the dynamic equations into the cost functional. Variational analysis of the unconstrained cost functionals results in the optimality equations, which are solved efficiently by weighted residual methods. [S0739-3717(00)00601-2]