The Pontryagin maximum principle for solving Fokker-Planck optimal control problems (Q2181601): Difference between revisions

In this paper, a class of non-smooth optimal control problems governed by Fokker-Planck (FP) equations are investigated using the framework of Pontryagin's maximum principle (PMP). The FP equations model the dynamics of a Ito random process in a deterministic framework through the evolution of its probability density function. In this context, it is natural to consider control problems, where the job of the controller is to drive the process to a desired state or along a desired trajectory. Such problems arise in the context of pedestrian motion or traffic flows. The objective functionals to be minimized, usually, are discontinuous with respect to the controls. Thus, PMP frameworks are appropriate for the characterization of the optimality system as they do not involve derivatives with respect to the control functions. However, a fast and accurate numerical realization of the PMP optimality system is a major challenge. The authors in this paper develop and analyze a sequential quadratic Hamiltonian (SQH) scheme for solving non-smooth FP optimal control problems in the PMP framework. Such a scheme was earlier devised by the same authors in context of simple linear and distributed optimal control problems. For the FP control problem, the bilinear control structure introduces additional challenges in the theoretical analysis of SQH schemes. The authors provide a comprehensive theoretical analysis of the PMP framework and convergence analysis of their proposed SQH scheme applied to open-loop and closed loop control problems. Further, they demonstrate the robustness of their PMP optimal control framework and the SQH scheme through several numerical experiments. The results of this paper show that the PMP framework is more convenient in solving optimal control problems compared to the Lagrangian and non-smooth frameworks and, further, that the SQH scheme is easy to implement and provide fast and accurate solutions of the optimality system.