Finite‐difference gradients versus error‐quadrature gradients in the solution of parameterized optimal control problems

Abstract

SUMMARYT w o non-linear programming algorithms based on the Lagrangian function are compared with respect to their computational efficiency for solving parameterized optimal control problems. If the most efficient, constrained variable metric method together with forward-difference gradients is used, the formulation and implementation of the adjoint variables can be avoided. This is especially convenient in the design phase of large complex systems. KEY WORDS Parameterized optimal control problems Finite-difference gradientsError-quadrature gradients Recursive quadratic programming Multiplier methods Aircraft trajectory optimization