Thursday, April 20, 2023, 11:00 - 11:59
In this thesis, we propose an approximate reformulation for constrained Optimal Control Problems (OCP) by introducing penalty, barrier and squashing terms. We also provide a tailored algorithm to solve the reformulated OCP within the Sequential Quadratic Programming (SQP) framework. Additionally, we approximate the Hessian matrix of the SQP with General Gauss-Newton (GGN) and eXtended Gauss-Newton methods (XGN). We exploit the sparsity pattern of the Hessian matrix by using a Riccati recursion based method to compute the iterates. We implement three different
algorithms based on Multiple Shooting (MS), Single Shooting (SS) and Differential Dynamic Programming (DDP).