Local Preferential Bayesian Optimization

Bayesian optimization (BO) is a popular and effective approach for tuning expensive, noisy experiments, but requires the formulation of an explicit objective function. Preferential BO (PBO) removes this requirement by learning from pairwise human feedback, yet existing methods struggle to efficiently optimize beyond low- and medium-dimensional problems due to their global search approaches. We address this limitation by developing a family of local PBO methods that transfer key ideas from high-dimensional BO to the preferential setting. In particular, we introduce local PBO methods which adapt trust-region and derivative-informed local search to pairwise preference feedback, where the latter exploits first- and second-order derivatives of the Laplace-approximated GP posterior. Our benchmark on GP sample paths, standard optimization benchmark functions, and policy-search tasks shows that local PBO methods are especially effective in high-dimensional and complex landscapes with steep optima. Compared with global preference-based baselines, they can substantially reduce cumulative regret, making them particularly useful for real-world preference-based optimization tasks such as policy search.

All Models are Wrong, Knowing Where is Useful: On Model Uncertainty in Reinforcement Learning

Model-based reinforcement learning (MBRL) infers information about the environment from a learned dynamics model and bears the potential to address open problems such as data efficient and safe learning in robotics. However, inaccuracies of the learned dynamics model are typically exploited by the agent, substantially hampering the capabilities of MBRL methods. We present a framework for dealing with inaccuracies of probabilistic models through targeted handling of uncertainty that effectively mitigates model exploitation. We present recent successes in learning directly on hardware and safe exploration, and discuss future directions for uncertainty-aware MBRL.

Biased Dreams: Limitations to Epistemic Uncertainty Quantification in Latent Space Models

Model-Based Reinforcement Learning distinguishes between physical dynamics models operating on proprioceptive inputs and latent dynamics models operating on high-dimensional image observations. A prominent latent approach is the Recurrent State Space Model used in the Dreamer family. While epistemic uncertainty quantification to inform exploration and mitigate model exploitation is well established for physical dynamics models, its transfer to latent dynamics models has received limited scrutiny. We empirically demonstrate that latent transitions are biased toward well-represented regions of latent space, exhibiting an attractor behavior that can deviate from true environment dynamics. As a result, discrepancies in environment dynamics may not manifest in latent space, undermining the reliability of epistemic uncertainty estimates. Because these attractors often lie in high-reward regions, latent rollouts systematically overestimate predicted rewards. Our findings highlight key limitations of epistemic uncertainty estimation in latent dynamics models and motivate more critical evaluation of this method.

Dyna-Style Safety Augmented Reinforcement Learning: Staying Safe in the Face of Uncertainty

Safety remains an open problem in reinforcement learning (RL), especially during training. While safety filters are promising to address safe exploration, they are generally poorly suited for high-dimensional systems with unknown dynamics. We propose Dyna-style Safety Augmented Reinforcement Learning (Dyna-SAuR), a novel algorithm that learns both a scalable safety filter and a control policy using a learned uncertainty-aware dynamics model, while requiring minimal domain knowledge. The filter avoids failures and high uncertainty regions. Thus, better models expand the set of safe and certain states, reducing filter conservatism. We present the effectiveness of Dyna-SAuR on goal-reaching CartPole as well as MuJoCo Walker, reducing failures compared to state-of-the-art methods by 2 orders of magnitude.

Preferential Bayesian Optimization with Crash Feedback

Bayesian optimization is a popular black-box optimization method for parameter learning in control and robotics. It typically requires an objective function that reflects the user's optimization goal. However, in practical applications, this objective function is often inaccessible due to complex or unmeasurable performance metrics. Preferential Bayesian optimization (PBO) overcomes this limitation by leveraging human feedback through pairwise comparisons, eliminating the need for explicit performance quantification. When applying PBO to hardware systems, such as in quadcopter control, crashes can cause time-consuming experimental resets, wear and tear, or otherwise undesired outcomes. Standard PBO methods cannot incorporate feedback from such crashed experiments, resulting in the exploration of parameters that frequently lead to experimental crashes. We thus introduce CrashPBO, a user-friendly mechanism that enables users to both express preferences and report crashes during the optimization process. Benchmarking on synthetic functions shows that this mechanism reduces crashes by 63% and increases data efficiency. Through experiments on three robotics platforms, we demonstrate the wide applicability and transferability of CrashPBO, highlighting that it provides a flexible, user-friendly framework for parameter learning with human feedback on preferences and crashes.

BayeSQP: Bayesian Optimization through Sequential Quadratic Programming

We introduce BayeSQP, a novel algorithm for general black-box optimization that merges the structure of sequential quadratic programming with concepts from Bayesian optimization. BayeSQP employs second-order Gaussian process surrogates for both the objective and constraints to jointly model the function values, gradients, and Hessian from only zero-order information. At each iteration, a local subproblem is constructed using the GP posterior estimates and solved to obtain a search direction. Crucially, the formulation of the subproblem explicitly incorporates uncertainty in both the function and derivative estimates, resulting in a tractable second-order cone program for high probability improvements under model uncertainty. A subsequent one-dimensional line search via constrained Thompson sampling selects the next evaluation point. Empirical results show thatBayeSQP outperforms state-of-the-art methods in specific high-dimensional settings. Our algorithm offers a principled and flexible framework that bridges classical optimization techniques with modern approaches to black-box optimization.

The Mini Wheelbot Dataset: High-Fidelity Data for Robot Learning

The development of robust learning-based control algorithms for unstable systems requires high-quality, real-world data, yet access to specialized robotic hardware remains a significant barrier for many researchers. This paper introduces a comprehensive dynamics dataset for the Mini Wheelbot, an open-source, quasi-symmetric balancing reaction wheel unicycle. The dataset provides 1 kHz synchronized data encompassing all onboard sensor readings, state estimates, ground-truth poses from a motion capture system, and third-person video logs. To ensure data diversity, we include experiments across multiple hardware instances and surfaces using various control paradigms, including pseudo-random binary excitation, nonlinear model predictive control, and reinforcement learning agents. We include several example applications in dynamics model learning, state estimation, and time-series classification to illustrate common robotics algorithms that can be benchmarked on our dataset.

Fine-Tuning of Neural Network Approximate MPC without Retraining via Bayesian Optimization

Approximate model-predictive control (AMPC) aims to imitate an MPC's behavior with a neural network, removing the need to solve an expensive optimization problem at runtime. However, during deployment, the parameters of the underlying MPC must usually be fine-tuned. This often renders AMPC impractical as it requires repeatedly generating a new dataset and retraining the neural network. Recent work addresses this problem by adapting AMPC without retraining using approximated sensitivities of the MPC's optimization problem. Currently, this adaption must be done by hand, which is labor-intensive and can be unintuitive for high-dimensional systems. To solve this issue, we propose using Bayesian optimization to tune the parameters of AMPC policies based on experimental data. By combining model-based control with direct and local learning, our approach achieves superior performance to nominal AMPC on hardware, with minimal experimentation. This allows automatic and data-efficient adaptation of AMPC to new system instances and fine-tuning to cost functions that are difficult to directly implement in MPC. We demonstrate the proposed method in hardware experiments for the swing-up maneuver on an inverted cartpole and yaw control of an under-actuated balancing unicycle robot, a challenging control problem.

A Systems-Theoretic View on the Convergence of Algorithms under Disturbances

Algorithms increasingly operate within complex physical, social, and engineering systems where they are exposed to disturbances, noise, and interconnections with other dynamical systems. This article extends known convergence guarantees of an algorithm operating in isolation (i.e., without disturbances) and systematically derives stability bounds and convergence rates in the presence of such disturbances. By leveraging converse Lyapunov theorems, we derive key inequalities that quantify the impact of disturbances. We further demonstrate how our result can be utilized to assess the effects of disturbances on algorithmic performance in a wide variety of applications, including communication constraints in distributed learning, sensitivity in machine learning generalization, and intentional noise injection for privacy. This underpins the role of our result as a unifying tool for algorithm analysis in the presence of noise, disturbances, and interconnections with other dynamical systems.

Sailing Towards Zero-Shot State Estimation using Foundation Models Combined with a UKF

State estimation in control and systems engineering traditionally requires extensive manual system identification or data-collection effort. However, transformer-based foundation models in other domains have reduced data requirements by leveraging pre-trained generalist models. Ultimately, developing zero-shot foundation models of system dynamics could drastically reduce manual deployment effort. While recent work shows that transformer-based end-to-end approaches can achieve zero-shot performance on unseen systems, they are limited to sensor models seen during training. We introduce the foundation model unscented Kalman filter (FM-UKF), which combines a transformer-based model of system dynamics with analytically known sensor models via an UKF, enabling generalization across varying dynamics without retraining for new sensor configurations. We evaluate FM-UKF on a new benchmark of container ship models with complex dynamics, demonstrating a competitive accuracy, effort, and robustness trade-off compared to classical methods with approximate system knowledge and to an end-to-end approach. The benchmark and dataset are open sourced to further support future research in zero-shot state estimation via foundation models.