Second order methods for the optimal control problems (Q1792547)

The present paper discusses second order methods for optimal control problems. The area of application is in bioeconomics. The functional \(I\) associated with the control problem is augmented by introducing an additional functional \(J\) to the problem formulation. The resulting auxiliary functional is a convex combination of \(I\) and \(J\). The contribution of these functionals is controlled by a parameter \(\alpha \in [0,1]\). Two methods for the solution of the control problem are considered: The ``method of weak improvement'' and the ``method of strong improvement.'' They differ in terms of the functional \(J\). To be able to control the influence of different components of the control vector, two additional diagonal matrices \(\beta\) and \(\gamma\) are introduced into the problem formulation. Detailed steps on how to implement an algorithm to iteratively solve for the new states and control variables for both formulations is provided. These algorithms include a greedy search for the parameters \(\alpha\), \(\beta\), and \(\gamma\). To improve parameter selection (and by that convergence of the algorithm), it is suggested to search for optimal parameters \(\alpha\), \(\beta\), and \(\gamma\) using gradient information (i.e., derivatives of the considered functionals with respect to these parameters). Numerical results for the improved scheme are provided.