Computation of optimal feedforward and feedback control by a modified reduced gradient method (Q1294234)

A new modified reduced gradient method is proposed to compute the solution of optimal open-loop control and optimal state feedback control. The differential equations are converted into algebraic equality constraints by using generalized block pulse functions. The initial conditions are included in these constraints. The sensitivity matrices of the state trajectories with respect to the coefficients of the control variables are recursively computed in such a manner that the reduced gradients are easily evaluated.