Some tools for the direct solution of optimal control problems

Abstract

This paper describes two ANSI C subroutines for the direct solution of finite time optimal control problems. Here the control variable is parameterized using a piecewise linear approximation. The resultant nonlinear programming problem (NLP) is solved using the sequential unconstrained minimization technique (SUMT) and the sequential quadratic programming (SQP) method. These methods are embedded in the computer codes dyn_sumt and dyn_sqp, respectively. Both subroutines are used to solve more than 26 optimal problems that have appeared in the literature. The paper compares the computational efficiency of both codes as well as techniques for computing the gradient of the cost function and constraints of the optimal control problem.