pith. sign in

arxiv: 2507.02098 · v2 · submitted 2025-07-02 · 📡 eess.SY · cs.LG· cs.SY· math.OC

A robust and adaptive MPC formulation for Gaussian process models

Mathieu Dubied , Amon Lahr , Melanie N. Zeilinger , Johannes K\"ohler This is my paper

Pith reviewed 2026-05-19 05:35 UTC · model grok-4.3

classification 📡 eess.SY cs.LGcs.SYmath.OC
keywords robust model predictive controlGaussian processescontraction metricsadaptive controluncertain nonlinear systemsquadrotoronline learning
0
0 comments X

The pith

A robust MPC framework uses contraction-metric bounds on Gaussian process predictions to guarantee recursive feasibility and high-probability constraint satisfaction for uncertain nonlinear systems.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops an adaptive model predictive controller that learns uncertain nonlinear dynamics with Gaussian processes from noisy data collected during operation. It derives robust future predictions for these learned models by applying contraction metrics to bound the effects of disturbances and model errors. These bounded predictions are embedded directly in the MPC optimization so that the resulting controller stays recursively feasible, satisfies constraints robustly, and drives the closed-loop system to a desired reference state, all with high probability. A reader would care because the method turns online learning into a safe, certifiable control law for real systems whose dynamics are only partially known in advance.

Core claim

The central claim is that robust predictions derived for Gaussian process models via contraction metrics can be incorporated into a model predictive control formulation for uncertain nonlinear systems. This yields a controller that guarantees recursive feasibility, robust constraint satisfaction, and convergence to a reference state with high probability while adapting online to new measurements.

What carries the argument

Robust predictions for GP models using contraction metrics, which generate explicit bounds on future states that are then inserted as constraints inside the MPC optimization problem.

If this is right

  • The controller can be applied to systems subject to both parametric uncertainty and unmodeled nonlinearities while still guaranteeing safety.
  • Online collection of noisy measurements during operation continuously improves the Gaussian process model and thereby reduces conservatism of the robust bounds.
  • The same contraction-metric technique can be used to certify stability and constraint satisfaction for other learning-based predictors beyond GPs.
  • Numerical results on the quadrotor show that the robust prediction method yields visibly tighter trajectories and faster convergence than non-robust GP-MPC.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The approach could be combined with event-triggered or periodic re-learning schedules to balance computational load against model accuracy in long-duration missions.
  • Because the bounds are derived from contraction metrics, the method may extend naturally to other contraction-based stability certificates already used in nonlinear control.
  • If the GP kernel or length-scale hyperparameters are also adapted online, the same robust-prediction pipeline would automatically tighten or relax the MPC constraints without manual retuning.

Load-bearing premise

The contraction-metric bounds computed from the Gaussian process model are assumed to remain valid and tight enough that they can be added to the MPC problem without destroying recursive feasibility or performance.

What would settle it

Run the planar quadrotor example with the proposed controller; if the actual trajectory violates the predicted robust bounds or the MPC problem becomes infeasible at any step despite the contraction-metric tightening, the guarantee does not hold.

Figures

Figures reproduced from arXiv: 2507.02098 by Amon Lahr, Johannes K\"ohler, Mathieu Dubied, Melanie N. Zeilinger.

Figure 1
Figure 1. Figure 1: Planar quadrotor. =             v1 cos(ϕ) − v2 sin(ϕ) v1 sin(ϕ) + v2 cos(ϕ) ϕ˙ v2ϕ˙ − g sin(ϕ) −v1ϕ˙ − g cos(ϕ) 0             +             0 0 0 0 0 0 0 0 1 m 1 m l J −l J             " u1 u2 # +             0 0 0 0 1 0             0.25 3.0 · height(p1, p2) + 1 | {z } g(x) +             0 0 0 cos(ϕ) − sin(ϕ) 0       … view at source ↗
Figure 4
Figure 4. Figure 4: Comparison between the quadrotor’s trajectories ob [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison between the size of the predicted [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Evolution of the GP models stored across the differ [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
read the original abstract

In this paper, we present a robust and adaptive model predictive control (MPC) framework for uncertain nonlinear systems affected by bounded disturbances and unmodeled nonlinearities. We use Gaussian Processes (GPs) to learn the uncertain dynamics based on noisy measurements, including those collected during system operation. As a key contribution, we derive robust predictions for GP models using contraction metrics, which are incorporated in the MPC formulation. The proposed design guarantees recursive feasibility, robust constraint satisfaction and convergence to a reference state, with high probability. We provide a numerical example of a planar quadrotor subject to difficult-to-model ground effects, which highlights significant improvements achieved through the proposed robust prediction method and through online learning.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes a robust and adaptive MPC framework for uncertain nonlinear systems with bounded disturbances and unmodeled dynamics. Gaussian processes are used to learn the system from noisy measurements, including online data. Robust predictions are derived via contraction metrics and incorporated into the MPC optimization. The design is claimed to guarantee recursive feasibility, robust constraint satisfaction, and convergence to a reference state with high probability. A numerical example on a planar quadrotor with ground effects is provided to illustrate performance gains from the robust predictions and online learning.

Significance. If the high-probability guarantees on recursive feasibility and constraint satisfaction hold under online GP updates, the work would represent a meaningful advance in safe learning-based control by combining contraction-metric tubes with adaptive GP models. The approach addresses a practical gap in handling time-varying uncertainty bounds within MPC while preserving feasibility, and the quadrotor example demonstrates tangible improvements for systems with difficult-to-model effects.

major comments (2)
  1. [§4] §4 (Robust Predictions via Contraction Metrics): The derivation of time-varying robust tubes around the GP posterior mean must explicitly show that the chosen contraction metric remains contracting after each online GP update; the current argument appears to fix the metric on the initial posterior, which does not automatically extend to the updated variance ball at subsequent steps.
  2. [§5.2] §5.2 (Recursive Feasibility Theorem): The terminal set and cost construction for recursive feasibility relies on the robust prediction sets remaining invariant under the true dynamics; because both GP mean and variance evolve with new measurements, it is not shown that a single terminal ingredient computed at the initial step continues to certify feasibility for the updated robust sets at future time steps.
minor comments (2)
  1. [§5] The high-probability statements in the main theorems should include explicit dependence on the GP confidence parameter δ and the number of online samples to make the overall probability bound transparent.
  2. [Numerical Example] Figure 3 (quadrotor trajectories): the legend and axis labels are too small for readability; consider enlarging or splitting into subplots.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major comment below, indicating where revisions will be made to strengthen the manuscript.

read point-by-point responses

  1. Referee: [§4] §4 (Robust Predictions via Contraction Metrics): The derivation of time-varying robust tubes around the GP posterior mean must explicitly show that the chosen contraction metric remains contracting after each online GP update; the current argument appears to fix the metric on the initial posterior, which does not automatically extend to the updated variance ball at subsequent steps.

    Authors: We thank the referee for highlighting this point. The contraction metric is selected to satisfy the differential contraction condition uniformly over a compact set that bounds all possible GP posteriors, including those arising from online updates. Because additional measurements can only reduce (or leave unchanged) the posterior variance under our fixed kernel and prior, the initial metric remains valid for all subsequent posteriors. Nevertheless, we agree that this invariance should be stated explicitly. In the revised manuscript we will insert a supporting lemma in §4 proving that any metric contracting for the initial posterior mean and variance ball remains contracting for all updated posteriors whose variance is component-wise smaller. The proof relies on the monotonicity of the GP variance and the Lipschitz continuity of the dynamics. revision: yes

  2. Referee: [§5.2] §5.2 (Recursive Feasibility Theorem): The terminal set and cost construction for recursive feasibility relies on the robust prediction sets remaining invariant under the true dynamics; because both GP mean and variance evolve with new measurements, it is not shown that a single terminal ingredient computed at the initial step continues to certify feasibility for the updated robust sets at future time steps.

    Authors: This observation is correct and we will strengthen the argument. The terminal set is computed from the initial robust prediction tube, which is constructed as an over-approximation that contains every possible future updated tube (owing to the fact that posterior variance is monotonically non-increasing). Consequently, any trajectory that enters the initial terminal set under the true dynamics also satisfies the updated, smaller tubes. We will revise the proof of the recursive feasibility theorem in §5.2 to include an explicit invariance argument together with a short lemma establishing that the updated robust sets are nested subsets of the initial set. These additions will be placed immediately before the statement of the theorem. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation builds on external contraction metrics and GP bounds

full rationale

The paper's central contribution derives robust prediction tubes for GP models via contraction metrics and embeds them in an adaptive MPC scheme. These steps rely on standard properties of contraction metrics (external to the paper) and high-probability GP posterior bounds rather than re-using fitted parameters as predictions or closing a self-citation loop. Recursive feasibility and constraint satisfaction follow from the tube construction and terminal ingredients, which are not shown to reduce to the MPC cost or online data by construction. No load-bearing self-citation or ansatz smuggling is evident in the provided derivation chain.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The framework rests on standard assumptions from robust MPC and Gaussian process regression plus a few design choices for the contraction metric and disturbance bounds.

free parameters (2)
  • Contraction metric parameters
    Chosen to ensure the robust prediction bounds hold for the learned GP model.
  • GP kernel hyperparameters
    Fitted from noisy measurements collected during operation.
axioms (2)
  • domain assumption System dynamics are affected by bounded disturbances and unmodeled nonlinearities that can be captured by a GP.
    Stated directly in the abstract as the setting for the uncertain nonlinear systems.
  • domain assumption Contraction metrics can be used to derive valid robust predictions for GP models.
    Central technical step claimed in the abstract.

pith-pipeline@v0.9.0 · 5658 in / 1340 out tokens · 39951 ms · 2026-05-19T05:35:27.536258+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

62 extracted references · 62 canonical work pages

  1. [1]

    Online learning for linearly parametrized control problems

    Yasin Abbasi-Yadkori. Online learning for linearly parametrized control problems. PhD thesis, University of Alberta, 2013

    work page 2013

  2. [2]

    Robust adaptive mpc for constrained uncertain nonlinear systems

    Veronica Adetola and Martin Guay. Robust adaptive mpc for constrained uncertain nonlinear systems. International Journal of Adaptive Control and Signal Processing , 25(2):155–167, 2011

    work page 2011

  3. [3]

    Themosekinterior point optimizer for linear programming: an implementation of the homogeneous algorithm

    ErlingDAndersenandKnudDAndersen. Themosekinterior point optimizer for linear programming: an implementation of the homogeneous algorithm. In High performance optimization, pages 197–232. Springer, 2000

    work page 2000

  4. [4]

    Stochastic model predictive control for sub-gaussian noise

    Yunke Ao, Johannes Köhler, Manish Prajapat, Yarden As, Melanie Zeilinger, Philipp Fürnstahl, and Andreas Krause. Stochastic model predictive control for sub-gaussian noise. arXiv preprint arXiv:2503.08795, 2025

    work page arXiv 2025

  5. [5]

    Actsafe: Active exploration with safety constraints for reinforcement learning

    Yarden As, Bhavya Sukhija, Lenart Treven, Carmelo Sferrazza, Stelian Coros, and Andreas Krause. Actsafe: Active exploration with safety constraints for reinforcement learning. arXiv preprint arXiv:2410.09486, 2024

    work page arXiv 2024

  6. [6]

    Discrete-time incremental ISS: A framework for robust nmpc

    Florian Bayer, Mathias Bürger, and Frank Allgöwer. Discrete-time incremental ISS: A framework for robust nmpc. In Proc. European Control Conference (ECC), pages 2068–

    work page 2068

  7. [7]

    Safe learning-based predictive control from efficient reachability

    Michael E Cao and Samuel Coogan. Safe learning-based predictive control from efficient reachability. In Proc. American Control Conference (ACC) , pages 1832–1837. IEEE, 2023

    work page 2023

  8. [8]

    An optimistic approach to cost-aware predictive control

    Michael Enqi Cao, Matthieu Bloch, and Samuel Coogan. An optimistic approach to cost-aware predictive control. Automatica, 176:112263, 2025

    work page 2025

  9. [9]

    Data-driven modelpredictivecontrolfortrajectorytrackingwitharobotic arm

    Andrea Carron, Elena Arcari, Martin Wermelinger, Lukas Hewing, Marco Hutter, and Melanie N Zeilinger. Data-driven modelpredictivecontrolfortrajectorytrackingwitharobotic arm. IEEE Robotics and Automation Letters, 4(4):3758– 3765, 2019

    work page 2019

  10. [10]

    On kernelized multi-armed bandits

    Sayak Ray Chowdhury and Aditya Gopalan. On kernelized multi-armed bandits. In International Conference on Machine Learning, pages 844–853. PMLR, 2017

    work page 2017

  11. [11]

    Robust adaptive model predictive control of quadrotors

    Alexandre Didier, Anilkumar Parsi, Jeremy Coulson, and Roy S Smith. Robust adaptive model predictive control of quadrotors. In Proc. European Control Conference (ECC), pages 657–662. IEEE, 2021

    work page 2021

  12. [12]

    A comparison on robust mpc methods for nonlinear systems

    Alexander Erdin. A comparison on robust mpc methods for nonlinear systems. Master’s thesis, ETH Zurich, 2024. https://github.com/Gedlex/nonlinear-robust-MPC

    work page 2024

  13. [13]

    Practical and rigorous uncertainty bounds for gaussian process regression

    Christian Fiedler, Carsten W Scherer, and Sebastian Trimpe. Practical and rigorous uncertainty bounds for gaussian process regression. In Proceedings of the AAAI conference on artificial intelligence, volume 35, pages 7439–7447, 2021. corrected version on arxiv

    work page 2021

  14. [14]

    A general framework to design stabilizing nonlinear model predictive controllers.Systems & Control Letters, 42(2):127–143, 2001

    Fernando ACC Fontes. A general framework to design stabilizing nonlinear model predictive controllers.Systems & Control Letters, 42(2):127–143, 2001

    work page 2001

  15. [15]

    Gpytorch: Blackbox matrix-matrix gaussian process inference with gpu acceleration

    Jacob R Gardner, Geoff Pleiss, David Bindel, Kilian Q Weinberger, and Andrew Gordon Wilson. Gpytorch: Blackbox matrix-matrix gaussian process inference with gpu acceleration. In Advances in Neural Information Processing Systems, 2018

    work page 2018

  16. [16]

    Gaussian process priors with 11 uncertain inputs application to multiple-step ahead time series forecasting.Advances in neural information processing systems, 15, 2002

    Agathe Girard, Carl Rasmussen, Joaquin Q Candela, and Roderick Murray-Smith. Gaussian process priors with 11 uncertain inputs application to multiple-step ahead time series forecasting.Advances in neural information processing systems, 15, 2002

    work page 2002

  17. [17]

    Cautious model predictive control using gaussian process regression

    Lukas Hewing, Juraj Kabzan, and Melanie N Zeilinger. Cautious model predictive control using gaussian process regression. IEEE Transactions on Control Systems Technology, 28(6):2736–2743, 2019

    work page 2019

  18. [18]

    Learning-based model predictive control: Toward safe learning in control.Annual Review of Control, Robotics, and Autonomous Systems, 3(1):269–296, 2020

    Lukas Hewing, Kim P Wabersich, Marcel Menner, and Melanie N Zeilinger. Learning-based model predictive control: Toward safe learning in control.Annual Review of Control, Robotics, and Autonomous Systems, 3(1):269–296, 2020

    work page 2020

  19. [19]

    Robust optimization for MPC

    Boris Houska and Mario E Villanueva. Robust optimization for MPC. Handbook of model predictive control, pages 413– 443, 2019

    work page 2019

  20. [20]

    Robust and kernelized data-enabled predictive control for nonlinear systems

    Linbin Huang, John Lygeros, and Florian Dörfler. Robust and kernelized data-enabled predictive control for nonlinear systems. IEEE Transactions on Control Systems Technology, 32(2):611–624, 2023

    work page 2023

  21. [21]

    Learning-based model predictive control for autonomous racing

    Juraj Kabzan, Lukas Hewing, Alexander Liniger, and Melanie N Zeilinger. Learning-based model predictive control for autonomous racing. IEEE Robotics and Automation Letters, 4(4):3363–3370, 2019

    work page 2019

  22. [22]

    Linear robust adaptivemodelpredictivecontrol:Computationalcomplexity and conservatism

    Johannes Köhler, Elisa Andina, Raffaele Soloperto, Matthias A Müller, and Frank Allgöwer. Linear robust adaptivemodelpredictivecontrol:Computationalcomplexity and conservatism. In Proc. IEEE 58th Conference on Decision and Control (CDC), pages 1383–1388. IEEE, 2019

    work page 2019

  23. [23]

    A robust adaptive model predictivecontrolframeworkfornonlinearuncertainsystems

    Johannes Köhler, Peter Kötting, Raffaele Soloperto, Frank Allgöwer, and Matthias A Müller. A robust adaptive model predictivecontrolframeworkfornonlinearuncertainsystems. International Journal of Robust and Nonlinear Control , 31(18):8725–8749, 2021

    work page 2021

  24. [24]

    A computationally efficient robust model predictivecontrolframeworkforuncertainnonlinearsystems

    Johannes Köhler, Raffaele Soloperto, Matthias A Müller, and Frank Allgöwer. A computationally efficient robust model predictivecontrolframeworkforuncertainnonlinearsystems. IEEE Transactions on Automatic Control, 66(2):794–801, 2020

    work page 2020

  25. [25]

    State space models vs

    Johannes Köhler, Kim P Wabersich, Julian Berberich, and Melanie N Zeilinger. State space models vs. multi-step predictors in predictive control: Are state space models complicating safe data-driven designs? InProc. IEEE 61st Conference on Decision and Control (CDC), pages 491–498. IEEE, 2022

    work page 2022

  26. [26]

    Learning-based model predictive control for safe exploration

    Torsten Koller, Felix Berkenkamp, Matteo Turchetta, and Andreas Krause. Learning-based model predictive control for safe exploration. InProc. IEEE conference on decision and control (CDC), pages 6059–6066. IEEE, 2018

    work page 2018

  27. [27]

    Optimal kernel regression bounds under energy-bounded noise

    Amon Lahr, Johannes Köhler*, Anna Scampicchio*, and Melanie N Zeilinger. Optimal kernel regression bounds under energy-bounded noise. arXiv preprint arXiv:2505.22235, 2025

    work page arXiv 2025

  28. [28]

    L4acados: Learning-based models for acados, applied to gaussian process-based predictive control.arXiv preprint arXiv:2411.19258, 2024

    Amon Lahr, Joshua Näf, Kim P Wabersich, Jonathan Frey, Pascal Siehl, Andrea Carron, Moritz Diehl, and Melanie N Zeilinger. L4acados: Learning-based models for acados, applied to gaussian process-based predictive control.arXiv preprint arXiv:2411.19258, 2024

    work page arXiv 2024

  29. [29]

    Zero-order optimization for gaussian process-based model predictive control

    Amon Lahr, Andrea Zanelli, Andrea Carron, and Melanie N Zeilinger. Zero-order optimization for gaussian process-based model predictive control. European Journal of Control, 74:100862, 2023

    work page 2023

  30. [30]

    Yalmip: A toolbox for modeling and optimization in matlab

    Johan Lofberg. Yalmip: A toolbox for modeling and optimization in matlab. In 2004 IEEE international conference on robotics and automation (IEEE Cat. No. 04CH37508), pages 284–289. IEEE, 2004

    work page 2004

  31. [31]

    Adaptive robust model predictive control for nonlinear systems

    Brett Thomas Lopez. Adaptive robust model predictive control for nonlinear systems. PhD thesis, Massachusetts Institute of Technology, 2019

    work page 2019

  32. [32]

    Robust MPC with recursive model update

    Matthias Lorenzen, Mark Cannon, and Frank Allgöwer. Robust MPC with recursive model update. Automatica, 103:461–471, 2019

    work page 2019

  33. [33]

    Robust adaptive model predictive control: Performance and parameter estimation

    Xiaonan Lu, Mark Cannon, and Denis Koksal-Rivet. Robust adaptive model predictive control: Performance and parameter estimation. International Journal of Robust and Nonlinear Control, 31(18):8703–8724, 2021

    work page 2021

  34. [34]

    Kpc: Learning-based model predictive control with deterministic guarantees

    Emilio T Maddalena, Paul Scharnhorst, Yuning Jiang, and Colin N Jones. Kpc: Learning-based model predictive control with deterministic guarantees. InLearning for Dynamics and Control, pages 1015–1026. PMLR, 2021

    work page 2021

  35. [35]

    Deterministic error bounds for kernel-based learning techniques under bounded noise

    Emilio Tanowe Maddalena, Paul Scharnhorst, and Colin N Jones. Deterministic error bounds for kernel-based learning techniques under bounded noise. Automatica, 134:109896, 2021

    work page 2021

  36. [36]

    Online learning-basedmodelpredictivecontrolwithgaussianprocess models and stability guarantees

    Michael Maiworm, Daniel Limon, and Rolf Findeisen. Online learning-basedmodelpredictivecontrolwithgaussianprocess models and stability guarantees. International Journal of Robust and Nonlinear Control, 31(18):8785–8812, 2021

    work page 2021

  37. [37]

    Control contraction metrics: Convex and intrinsic criteria for nonlinear feedback design.IEEE Transactions on Automatic Control, 62(6):3046–3053, 2017

    Ian R Manchester and Jean-Jacques E Slotine. Control contraction metrics: Convex and intrinsic criteria for nonlinear feedback design.IEEE Transactions on Automatic Control, 62(6):3046–3053, 2017

    work page 2017

  38. [38]

    Robust control contraction metrics: A convex approach to nonlinear state-feedback H ∞ control

    Ian R Manchester and Jean-Jacques E Slotine. Robust control contraction metrics: A convex approach to nonlinear state-feedback H ∞ control. IEEE Control Systems Letters, 2(3):333–338, 2018

    work page 2018

  39. [39]

    Towards safe bayesian optimization with wiener kernel regression

    OleksiiMolodchyk,JohannesTeutsch,andTimmFaulwasser. Towards safe bayesian optimization with wiener kernel regression. arXiv preprint arXiv:2411.02253, 2025

    work page arXiv 2025

  40. [40]

    Robust constrained learning-based nmpc enabling reliable mobile robot path tracking

    ChrisJOstafew,AngelaPSchoellig,andTimothyDBarfoot. Robust constrained learning-based nmpc enabling reliable mobile robot path tracking. The International Journal of Robotics Research, 35(13):1547–1563, 2016

    work page 2016

  41. [41]

    Exact multiple-step predictions in gaussian process-based model predictive control: observations, possibilities, and challenges

    Maik Pfefferkorn, Michael Maiworm, and Rolf Findeisen. Exact multiple-step predictions in gaussian process-based model predictive control: observations, possibilities, and challenges. In Proc. American Control Conference (ACC), pages 2829–2836. IEEE, 2022

    work page 2022

  42. [42]

    Efficient implementation of gaussian process–based predictive control by quadratic programming

    Péter Polcz, Tamás Péni, and Roland Tóth. Efficient implementation of gaussian process–based predictive control by quadratic programming. IET Control Theory & Applications, 17(8):968–984, 2023

    work page 2023

  43. [43]

    Robust contractive model predictive control based on gaussian processes

    Marco Polver, Daniel Limon, Mirko Mazzoleni, Fabio Previdi, and Antonio Ferramosca. Robust contractive model predictive control based on gaussian processes. 2025

    work page 2025

  44. [44]

    Finite-sample-based reachability for safe control with gaussian process dynamics.arXiv preprint arXiv:2505.07594, 2025.(Cited on pages 2, 5, 15, 16, and 17)

    Manish Prajapat, Johannes Köhler, Amon Lahr, Andreas Krause, and Melanie N Zeilinger. Finite-sample-based reachability for safe control with gaussian process dynamics. arXiv preprint arXiv:2505.07594, 2025

    work page arXiv 2025

  45. [45]

    Safe guaranteed exploration for non-linear systems

    Manish Prajapat, Johannes Köhler, Matteo Turchetta, Andreas Krause, and Melanie N Zeilinger. Safe guaranteed exploration for non-linear systems. IEEE Transactions on Automatic Control, 2025

    work page 2025

  46. [46]

    Towards safe and tractable gaussian process-based MPC: Efficient sampling within a sequential quadratic programming framework

    Manish Prajapat*, Amon Lahr*, Johannes Köhler, Andreas Krause, and Melanie N Zeilinger. Towards safe and tractable gaussian process-based MPC: Efficient sampling within a sequential quadratic programming framework. InProc.IEEE 63rd Conference on Decision and Control (CDC), 2024. 12

    work page 2024

  47. [47]

    Homothetic tube model predictive control for nonlinear systems.IEEE Transactions on Automatic Control, 68(8):4554–4569, 2022

    Saša V Raković, Li Dai, and Yuanqing Xia. Homothetic tube model predictive control for nonlinear systems.IEEE Transactions on Automatic Control, 68(8):4554–4569, 2022

    work page 2022

  48. [48]

    Gaussian processes for machine learning, volume 1

    Carl Edward Rasmussen, Christopher KI Williams, et al. Gaussian processes for machine learning, volume 1. Springer, 2006

    work page 2006

  49. [49]

    Model predictive control: theory, computation, and design, volume 2

    James Blake Rawlings, David Q Mayne, and Moritz Diehl. Model predictive control: theory, computation, and design, volume 2. Nob Hill Publishing Madison, WI, 2017

    work page 2017

  50. [50]

    Robust adaptive MPC using control contraction metrics

    András Sasfi, Melanie N Zeilinger, and Johannes Köhler. Robust adaptive MPC using control contraction metrics. Automatica, 155:111169, 2023

    work page 2023

  51. [51]

    Gaussian processes for dynamics learning in model predictive control

    Anna Scampicchio, Elena Arcari, Amon Lahr, and Melanie N Zeilinger. Gaussian processes for dynamics learning in model predictive control. arXiv preprint arXiv:2502.02310, 2025

    work page arXiv 2025

  52. [52]

    Robust feedback motion planning via contraction theory.The International Journal of Robotics Research, 42(9):655–688, 2023

    Sumeet Singh, Benoit Landry, Anirudha Majumdar, Jean- Jacques Slotine, and Marco Pavone. Robust feedback motion planning via contraction theory.The International Journal of Robotics Research, 42(9):655–688, 2023

    work page 2023

  53. [53]

    Robust online motion planning via contraction theory and convex optimization

    Sumeet Singh, Anirudha Majumdar, Jean-Jacques Slotine, and Marco Pavone. Robust online motion planning via contraction theory and convex optimization. In Proc. IEEE International Conference on Robotics and Automation (ICRA), pages 5883–5890, 2017

    work page 2017

  54. [54]

    Prentice hall Englewood Cliffs, NJ, 1991

    Jean-Jacques E Slotine and Weiping Li.Applied nonlinear control. Prentice hall Englewood Cliffs, NJ, 1991

    work page 1991

  55. [55]

    Information-theoretic regret bounds for gaussian process optimization in the bandit setting

    Niranjan Srinivas, Andreas Krause, Sham M Kakade, and Matthias W Seeger. Information-theoretic regret bounds for gaussian process optimization in the bandit setting. IEEE transactions on information theory, 58(5):3250–3265, 2012

    work page 2012

  56. [56]

    Support vector machines

    Ingo Steinwart and Andreas Christmann. Support vector machines. Springer Science & Business Media, 2008

    work page 2008

  57. [57]

    Safe exploration for optimization with gaussian processes

    YananSui,AlkisGotovos,JoelBurdick,andAndreasKrause. Safe exploration for optimization with gaussian processes. In International conference on machine learning, pages 997–

    work page

  58. [58]

    Automatic nonlinear MPC approximation with closed-loop guarantees

    Abdullah Tokmak, Christian Fiedler, Melanie N Zeilinger, Sebastian Trimpe, and Johannes Köhler. Automatic nonlinear MPC approximation with closed-loop guarantees. IEEE Transactions on Automatic Control, 2025

    work page 2025

  59. [59]

    Data-driven MPC for quadrotors

    Guillem Torrente, Elia Kaufmann, Philipp Föhn, and Davide Scaramuzza. Data-driven MPC for quadrotors. IEEE Robotics and Automation Letters, 6(2):3769–3776, 2021

    work page 2021

  60. [60]

    acados – a modular open-source framework for fast embedded optimal control.Mathematical Programming Computation, 2021

    Robin Verschueren, Gianluca Frison, Dimitris Kouzoupis, JonathanFrey,NielsvanDuijkeren,AndreaZanelli,Branimir Novoselnik, Thivaharan Albin, Rien Quirynen, and Moritz Diehl. acados – a modular open-source framework for fast embedded optimal control.Mathematical Programming Computation, 2021

    work page 2021

  61. [61]

    High-dimensional probability: An introduction with applications in data science, volume 47

    Roman Vershynin. High-dimensional probability: An introduction with applications in data science, volume 47. Cambridge university press, 2018

    work page 2018

  62. [62]

    Tube- certifiedtrajectorytrackingfornonlinearsystemswithrobust control contraction metrics.IEEE Robotics and Automation Letters, 7(2):5528–5535, 2022

    Pan Zhao, Arun Lakshmanan, Kasey Ackerman, Aditya Gahlawat, Marco Pavone, and Naira Hovakimyan. Tube- certifiedtrajectorytrackingfornonlinearsystemswithrobust control contraction metrics.IEEE Robotics and Automation Letters, 7(2):5528–5535, 2022. A Details of the Proofs First, we present the proofs of the theoretical results of Section 4. We start with th...

    work page 2022