Numerical methods of nonlinear optimal control based on mathematical programming (Q4374282)

In the first half of this paper, we study a primal-dual interior-point algorithm for convex quadratic programming problems in Hilbert space with the object of direct application to continuous-time LQ optimal control problems. Since our problems are infinite-dimensional, it is impossible in general to solve exactly the linear equations for finding a search direction at each iterate. We study in the latter half of the paper a dual sequential quadratic programming (SQP) framework for a general class of constrained optimization problems in Hilbert space. In this framework, quadratic programming subproblems, with the Hessian of the Lagrangean replaced by its approximation, are iteratively solved to generate a sequence of search directions. This paper includes two applications of the dual SQP method -- one is an application to optimal control problems with terminal-state equality constraints, and the other is an application to the so-called state-constrained optimal control problems. The method converts the former problem to a series of simultaneous systems of linear algebraic equations, and the latter problem to a series of nonnegatively constrained quadratic programming problems in a function space.