Robert Moldenhauer

Monday, May 04, 2026, 10:00 - 11:30

Room 02-012, Georges-Köhler Allee 102, Freiburg 79110, Germany

Abstract: Optimal control methods are typically subject to plant-model mismatch. Such discrepancies may arise from external disturbances, parametric uncertainty, numerical discretization, the use of data-driven surrogate models, or the need to rely on simplified models for computational tractability. This motivates to investigate the robustness of properties ensured by an optimal controller designed using a surrogate model when it is applied to the actual plant.

We employ a unified framework based on quadratic stage costs to analyse both finite- and infinite-horizon problems, encompassing discounted and undiscounted scenarios alike. Discounting, which is commonly used in reinforcement learning, prioritizes costs in the near future over the far future and has the benefit of reducing the accumulation of prediction errors. Under continuity of the model, cost-controllability and sufficiently small plant-model mismatch, exponential stability of the closed loop can be guaranteed. Furthermore, we give a suboptimality bound for the closed-loop cost recovering the optimal cost of the surrogate. The results reveal a trade-off between horizon length, discounting and plant-model mismatch. The robustness guarantees are uniform over the horizon length, meaning that larger horizons do not require successively more accurate models.

Bio: Robert Moldenhauer received his B.Sc. and M.Sc in Technical Cybernetics and Systems Theory from TU Ilmenau (Germany) in 2021 and 2023, respectively, and his B.Sc. in Mathematics from TU Ilmenau in 2023. He is currently undertaking a joint PhD at the University of Melbourne (Australia) and Université de Lorraine, Nancy (France). His research interests include nonlinear systems, optimal control, dynamic programming and wind turbine control. He received the Young Author Award at the 13th IFAC Symposium on Nonlinear Control Systems (NOLCOS), 2025.