POD-based feedback control of the Burgers equation by solving the evolutionary HJB equation (Q2387351)

A numerical method is proposed for solving finite-time horizon suboptimal feedback control problems of distributed parameter systems. The method is based on model reduction by proper orthogonal decomposition (POD), and a local Lax-Friedrichs scheme is used to solve the resulting evolutionary Hamilton-Jacobi-Bellman (HJB) equation. The latter scheme for solving the HJB equation is first tested on three 1-D examples. The whole procedure is then applied to a stabilization problem of the Burgers equation, where the POD technique is utilized to obtain a 4-D reduced-order optimal control problem and it is solved with a third-order local Lax-Friedrichs scheme. A parallel algorithm is used for computational speedup since the local Lax-Friedrichs flux is independently constructed at each point of the grid. The robustness of the proposed feedback design is shown in comparison with the open-loop design.