A sequential quadratic programming algorithm for discrete optimal control problems with control inequality constraints

The authors exploit the structure of optimal control problem and present a version of a conventional sequential quadratic programming algorithm which merely requires, at each iteration, the solution of N quadratic subproblems of dimension m (equal to that of the control) rather than the solution of one quadratic subproblem of dimension Nm as required by the mathematical programming formulation. This is achieved by extending a stagewise implementation of Newton´s method for unconstrained discrete-time optimal control problems to discrete-time optimal control problems with control constraints