Preconditioned conjugate gradient method for optimal control problems with control and state constraints