arXiv subjects

Katrin Baumgärtner

Publications and source records attributed to Katrin Baumgärtner.

18 recordsLinked to original sources

arXiv2026

Bayesian Optimization with Lower Confidence Bounds for Minimization Problems with Known Outer Structure

This paper considers Bayesian optimization (BO) for problems with known outer problem structure. In contrast to the classic BO setting, where the objective function itself is unknown and needs to be iteratively estimated from noisy observations, we analyze the case where the objective has a known outer structure - given in terms of a loss function - while the inner structure - an unknown input-output model - is again iteratively estimated from noisy observations of the model outputs. We introduce a novel lower confidence bound algorithm for this particular problem class which exploits the known outer problem structure. The proposed method is analyzed in terms of regret for the special case of convex loss functions and probabilistic parametric models which are linear in the uncertain parameters. Numerical examples illustrate the superior performance of structure-exploiting methods compared to structure-agnostic approaches.

math.OC↗

arXiv2026

Gray-Box Optimization using Optimism in the Face of Uncertainty

This paper considers sequential gray-box optimization where the objective function is given as the composition of a loss function and a parametric model. Crucially, the parameters of the model are unknown and need to be iteratively estimated from noisy observations of the model outputs. This problem setup generalizes the parametric black-box optimization problem known as (contextual) stochastic linear bandit. To address the sequential gray-box optimization problem, we propose a structure-exploiting method that leverages known problem structure given in terms of the loss function and an a priori set of admissible parameters. The method is based on the principle of optimism in the face of uncertainty and trades off exploration and exploitation by minimizing a lower confidence bound on the true objective function. We provide a detailed regret analysis of the novel method, improving on state-of-the-art results for the special case of linear stochastic bandits due to the use of a recently published bound for the parameter confidence sets arising in multi-output linear least-squares estimation. Numerical examples illustrate the superior performance of structure-exploiting methods compared to structure-agnostic approaches.

math.OC↗

arXiv2026

Anderson Acceleration for Linearly Converging SQP-Type Methods

Although Anderson acceleration (AA) is known to speed up fixed-point iterations, it is rarely applied in constrained optimization, in particular sequential quadratic programming (SQP). We show that the local convergence behavior of a general family of (inexact) SQP-type methods can benefit from AA and introduce a simple heuristic to alleviate slower convergence farther from the solution. The method is implemented in the software framework acados. Numerical examples from optimal control illustrate consistent improvements in convergence of different SQP-type methods.

math.OC↗

arXiv2026

Rollout Then Optimize: A One-Step Newton Refinement of Learned Policies for Nonlinear Model Predictive Control

We propose a computationally efficient rollout-then-optimize method to improve a learned control policy at deployment time. A learned policy provides a nominal trajectory, which is refined online by a single Newton step implemented via a Riccati recursion within a model predictive control (MPC) scheme. This refinement combines model knowledge with the learned policy at minimal additional computational cost. We establish bounds on the approximation error of the learned policy relative to the MPC policy and show that one Newton step reduces the suboptimality of the learned rollout quadratically in the policy approximation error. The proposed controller is validated in simulation on a constrained trajectory-tracking task for a quadcopter with nonlinear dynamics. Results highlight that the Newton step significantly improves the learned policy, achieving performance close to a fully converged MPC solution while requiring roughly half of the computational time. The code is available at https://github.com/aghezz1/rl-riccati.

math.OC↗

arXiv2026

Finite-sample bounds for multi-output system identification

This paper presents uniform-in-time finite-sample bounds for regularized linear regression with vector-valued outputs and conditionally zero-mean subgaussian noise. By revisiting classical self-normalized martingale arguments, we obtain bounds that apply directly to multi-output regression, unlike most of the prior work. Compared to the state of the art, the new results are more general and yield tighter bounds, even for scalar-valued outputs. The mild assumptions we use allow for unknown dependencies between regressors and past noise terms, typically induced by system dynamics or feedback mechanisms. Therefore, these novel finite-sample bounds can be applied to many affine-in-parameter system identification problems, including the identification of a linear time-invariant system from full-state measurements. These new results may lead to significant improvements in stochastic learning-based controllers for safety-critical applications.

math.ST↗

arXiv2026

Gauss-Newton accelerated MPPI Control

Model Predictive Path Integral (MPPI) control is a sampling-based optimization method that has recently attracted attention, particularly in the robotics and reinforcement learning communities. MPPI has been widely applied as a GPU-accelerated random search method to deterministic direct single-shooting optimal control problems arising in model predictive control (MPC) formulations. MPPI offers several key advantages, including flexibility, robustness, ease of implementation, and inherent parallelizability. However, its performance can deteriorate in high-dimensional settings since the optimal control problem is solved via Monte Carlo sampling. To address this limitation, this paper proposes an enhanced MPPI method that incorporates a Jacobian reconstruction technique and the second-order Generalized Gauss-Newton method. This novel approach is called \textit{Gauss-Newton accelerated MPPI}. The numerical results show that the Gauss-Newton accelerated MPPI approach substantially improves MPPI scalability and computational efficiency while preserving the key benefits of the classical MPPI framework, making it a promising approach even for high-dimensional problems.

eess.SY↗

arXiv2025

Differentiable Nonlinear Model Predictive Control

The efficient computation of parametric solution sensitivities is a key challenge in the integration of learning-enhanced methods with nonlinear model predictive control (MPC), as their availability is crucial for many learning algorithms. This paper discusses the computation of solution sensitivities of general nonlinear programs (NLPs) using the implicit function theorem (IFT) and smoothed optimality conditions treated in interior-point methods (IPM). We detail sensitivity computation within a sequential quadratic programming (SQP) method which employs an IPM for the quadratic subproblems. Previous works presented in the machine learning community are limited to convex or unconstrained formulations, or lack an implementation for efficient sensitivity evaluation. The publication is accompanied by an efficient open-source implementation within the acados framework, providing both forward and adjoint sensitivities for general optimal control problems, achieving speedups exceeding 3x over the state-of-the-art solvers mpc.pytorch and cvxpygen.

math.OC↗

arXiv2025

Incorporating a Deep Neural Network into Moving Horizon Estimation for Embedded Thermal Torque Derating of an Electric Machine

This study presents a novel state estimation approach integrating Deep Neural Networks (DNNs) into Moving Horizon Estimation (MHE). This is a shift from using traditional physics-based models within MHE towards data-driven techniques. Specifically, a Long Short-Term Memory (LSTM)-based DNN is trained using synthetic data derived from a high-fidelity thermal model of a Permanent Magnet Synchronous Machine (PMSM), applied within a thermal derating torque control strategy for battery electric vehicles. The trained DNN is directly embedded within an MHE formulation, forming a discrete-time nonlinear optimal control problem (OCP) solved via the acados optimization framework. Model-in-the-Loop simulations demonstrate accurate temperature estimation even under noisy sensor conditions and simulated sensor failures. Real-time implementation on embedded hardware confirms practical feasibility, achieving computational performance exceeding real-time requirements threefold. By integrating the learned LSTM-based dynamics directly into MHE, this work achieves state estimation accuracy, robustness, and adaptability while reducing modeling efforts and complexity. Overall, the results highlight the effectiveness of combining model-based and data-driven methods in safety-critical automotive control systems.

eess.SY↗

arXiv2025

Synthesis of Model Predictive Control and Reinforcement Learning: Survey and Classification

The fields of MPC and RL consider two successful control techniques for Markov decision processes. Both approaches are derived from similar fundamental principles, and both are widely used in practical applications, including robotics, process control, energy systems, and autonomous driving. Despite their similarities, MPC and RL follow distinct paradigms that emerged from diverse communities and different requirements. Various technical discrepancies, particularly the role of an environment model as part of the algorithm, lead to methodologies with nearly complementary advantages. Due to their orthogonal benefits, research interest in combination methods has recently increased significantly, leading to a large and growing set of complex ideas leveraging MPC and RL. This work illuminates the differences, similarities, and fundamentals that allow for different combination algorithms and categorizes existing work accordingly. Particularly, we focus on the versatile actor-critic RL approach as a basis for our categorization and examine how the online optimization approach of MPC can be used to improve the overall closed-loop performance of a policy.

eess.SY↗

arXiv2024

Multi-Phase Optimal Control Problems for Efficient Nonlinear Model Predictive Control with acados

Computationally efficient nonlinear model predictive control relies on elaborate discrete-time optimal control problem (OCP) formulations trading off accuracy with respect to the continuous-time problem and associated computational burden. Such formulations, however, are in general not easy to implement within specialized software frameworks tailored to numerical optimal control. This paper introduces a new multi-phase OCP interface for the open-source software acados allowing to conveniently formulate such problems and generate fast solvers that can be used for nonlinear model predictive control (NMPC). While multi-phase OCP (MOCP) formulations occur naturally in many applications, this work focuses on MOCP formulations that can be used to efficiently approximate standard continuous-time OCPs in the context of NMPC. To this end, the paper discusses advanced control parametrizations, such as closed-loop costing and piecewise polynomials with varying degree, as well as partial tightening and formulations that leverage models of different fidelity. An introductory example is presented to showcase the usability of the new interface. Finally, three numerical experiments demonstrate that NMPC controllers based on multi-phase formulations can efficiently trade-off computation time and control performance.

math.OC↗

arXiv2024

Fast Generation of Feasible Trajectories in Direct Optimal Control

This paper examines the question of finding feasible points to discrete-time optimal control problems. The optimization problem of finding a feasible trajectory is transcribed to an unconstrained optimal control problem. An efficient algorithm, called FP-DDP, is proposed that solves the resulting problem using Differential Dynamic Programming preserving feasibility with respect to the system dynamics in every iteration. Notably, FP-DDP admits global and rapid local convergence properties induced by a combination of a Levenberg-Marquardt method and an Armijo-type line search. The efficiency of FP-DDP is demonstrated against established methods such as Direct Multiple Shooting, Direct Single Shooting, and state-of-the-art solvers.

math.OC↗

arXiv2024

AC4MPC: Actor-Critic Reinforcement Learning for Nonlinear Model Predictive Control

\Ac{MPC} and \ac{RL} are two powerful control strategies with, arguably, complementary advantages. In this work, we show how actor-critic \ac{RL} techniques can be leveraged to improve the performance of \ac{MPC}. The \ac{RL} critic is used as an approximation of the optimal value function, and an actor roll-out provides an initial guess for primal variables of the \ac{MPC}. A parallel control architecture is proposed where each \ac{MPC} instance is solved twice for different initial guesses. Besides the actor roll-out initialization, a shifted initialization from the previous solution is used. Thereafter, the actor and the critic are again used to approximately evaluate the infinite horizon cost of these trajectories. The control actions from the lowest-cost trajectory are applied to the system at each time step. We establish that the proposed algorithm is guaranteed to outperform the original \ac{RL} policy plus an error term that depends on the accuracy of the critic and decays with the horizon length of the \ac{MPC} formulation. Moreover, we do not require globally optimal solutions for these guarantees to hold. The approach is demonstrated on an illustrative toy example and an \ac{AD} overtaking scenario.

eess.SY↗

arXiv2024

Fourth-order suboptimality of nominal model predictive control in the presence of uncertainty

We investigate the suboptimality resulting from the application of nominal model predictive control (MPC) to a nonlinear discrete time stochastic system. The suboptimality is defined with respect to the corresponding stochastic optimal control problem (OCP) that minimizes the expected cost of the closed loop system. In this context, nominal MPC corresponds to a form of certainty-equivalent control (CEC). We prove that, in a smooth and unconstrained setting, the suboptimality growth is of fourth order with respect to the level of uncertainty, a parameter which we can think of as a standard deviation. This implies that the suboptimality does not grow very quickly as the level of uncertainty is increased, providing further insight into the practical success of nominal MPC. Similarly, the difference between the optimal and suboptimal control inputs is of second order. We illustrate the result on a simple numerical example, which we also use to show how the proven relationship may cease to hold in the presence of state constraints.

math.OC↗

arXiv2024

Approximate propagation of normal distributions for stochastic optimal control of nonsmooth systems

We present a method for the approximate propagation of mean and covariance of a probability distribution through ordinary differential equations (ODE) with discontinous right-hand side. For piecewise affine systems, a normalization of the propagated probability distribution at every time step allows us to analytically compute the expectation integrals of the mean and covariance dynamics while explicitly taking into account the discontinuity. This leads to a natural smoothing of the discontinuity such that for relevant levels of uncertainty the resulting ODE can be integrated directly with standard schemes and it is neither necessary to prespecify the switching sequence nor to use a switch detection method. We then show how this result can be employed in the more general case of piecewise smooth functions based on a structure preserving linearization scheme. The resulting dynamics can be straightforwardly used within standard formulations of stochastic optimal control problems with chance constraints.

math.OC↗

arXiv2024

Progressive Smoothing for Motion Planning in Real-Time NMPC

Nonlinear model predictive control (NMPC) is a popular strategy for solving motion planning problems, including obstacle avoidance constraints, in autonomous driving applications. Non-smooth obstacle shapes, such as rectangles, introduce additional local minima in the underlying optimization problem. Smooth over-approximations, e.g., ellipsoidal shapes, limit the performance due to their conservativeness. We propose to vary the smoothness and the related over-approximation by a homotopy. Instead of varying the smoothness in consecutive sequential quadratic programming iterations, we use formulations that decrease the smooth over-approximation from the end towards the beginning of the prediction horizon. Thus, the real-time iterations algorithm is applicable to the proposed NMPC formulation. Different formulations are compared in simulation experiments and shown to successfully improve performance indicators without increasing the computation time.

eess.SY↗

arXiv2023

Gauss-Newton Runge-Kutta Integration for Efficient Discretization of Optimal Control Problems with Long Horizons and Least-Squares Costs

This work proposes an efficient treatment of continuous-time optimal control problem (OCP) with long horizons and nonlinear least-squares costs. The Gauss-Newton Runge-Kutta (GNRK) integrator is presented which provides a high-order cost integration. Crucially, the Hessian of the cost terms required within an SQP-type algorithm is approximated with a Gauss-Newton Hessian. Moreover, L2 penalty formulations for constraints are shown to be particularly effective for optimization with GNRK. An efficient implementation of GNRK is provided in the open-source software framework acados. We demonstrate the effectiveness of the proposed approach and its implementation on an illustrative example showing a reduction of relative suboptimality by a factor greater than 10 while increasing the runtime by only 10 %.

math.OC↗

arXiv2023

Local Convergence Behaviour of Generalized Gauss-Newton Multiple Shooting, Single Shooting and Differential Dynamic Programming

We revisit three classical numerical methods for solving unconstrained optimal control problems - multiple shooting, single shooting, and differential dynamic programming - and examine their local convergence behaviour. In particular, we show that all three methods converge with the same linear rate if a Gauss-Newton (GN), or more general a Generalized Gauss-Newton (GGN), Hessian approximation is used, which is the case in widely used implementations such as iLQR.

math.OC↗

arXiv2023

A dual-control effect preserving formulation for nonlinear output-feedback stochastic model predictive control with constraints

We propose a formulation for approximate constrained nonlinear output-feedback stochastic model predictive control. Starting from the ideal but intractable stochastic optimal control problem (OCP), which involves the optimization over output-dependent policies, we use linearization with respect to the uncertainty to derive a tractable approximation which includes knowledge of the output model. This allows us to compute the expected value for the outer functions of the OCP exactly. Crucially, the dual control effect is preserved by this approximation. In consequence, the resulting controller is aware of how the choice of inputs affects the information available in the future which in turn influences subsequent controls.

math.OC↗