Computation of optimal feedforward and feedback control by a modified reduced gradient method

摘要

A computational scheme for solving problems of optimal open-loop control policy and optimal state feedback control was introduced. The problems were first converted into a nonlinear programming problem. The conjugate reduced gradient method was then applied to find the minimum solution. The sensitivity matrices of the state trajectories with respect to the coefficients of control variables could be recursively calculated so that the reduced gradients could be easily evaluated in this computational algorithm. The H-orthogonal property for linear optimal control problems was demonstrated such that the global minimum solution of such problems was thus achieved in finite iterations using this algorithm. The computational strategy was easily and efficiently extended to determine the optimal state feedback gain matrices.