Pith. sign in

REVIEW 3 major objections 4 minor 39 references

Context-Aware Deep Lagrangian Networks for Model Predictive Control

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A context-conditioned Deep Lagrangian Network, trained entirely in simulation, performs online system identification for model predictive control and reduces real-robot end-effector tracking error by 39%, compared with 21% for an…

desk verdict Interesting architecture, but the SysID encoder is trained with the current torque and deployed without it, so the headline 39% improvement is not established. read the letter →

arxiv 2506.15249 v3 pith:CZVLBBMX submitted 2025-06-18 cs.RO cs.LG

Lucas Schulze , Jan Peters , Oleg Arenz This is my paper · ORCID
classification cs.ROcs.LG
keywords modelpredictivecontrolDeepLagrangianNetworksonlinesystemidentificationresidualdynamicslatentcontextphysics-consistentlearningrobotmanipulatorzero-shotsim-to-real
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that online system identification can be done with a physics-consistent neural model that learns a latent context for the current environment, and that putting this context-conditioned model inside model predictive control improves trajectory tracking under unknown payloads. It extends Deep Lagrangian Networks so the learned inertia and potential energy depend on a latent variable inferred by a recurrent encoder from recent joint states and residual torques, avoiding retraining when the load changes. On a 7-DOF robot arm, a model trained only in simulation reduces real-robot end-effector tracking RMSE by 39% in a pick-and-place task with changing loads, while an EKF-based adaptive MPC achieves 21%. This matters because it suggests physics-consistent learned dynamics can adapt to changing environments without real-world data or online weight updates.

What carries the argument

The central object is the context-aware residual DeLaN model: an inertia matrix written through a Cholesky decomposition, $\hat{H}(q,z)=\hat{L}(q,z)\hat{L}(q,z)^T$, and a potential energy $\hat{P}(q,z)$, both conditioned on a latent environment embedding $z$. The residual torque takes the same Lagrangian form as the nominal dynamics, $\tilde\tau=\tilde f^{-1}(q,\dot q,\ddot q,\theta_H,\theta_P,z)$, so the sum of the nominal model and the learned residual remains a physically consistent Euler-Lagrange system for each fixed $z$. An LSTM encoder produces $z$ from the recent sequence of joint positions, velocities, and measured residual torques, allowing online identification between MPC iterations while keeping the model fixed inside each horizon.

What would settle it

Run CaDeLaC on the real robot with a payload mass outside the 0-4 kg training range, such as 5 kg, or with a center-of-mass offset beyond 0.3 m, and compare end-effector tracking RMSE against the nominal MPC; the central claim would be falsified if the learned model shows no improvement, or if the latent context inferred by the LSTM fails to shift when the payload is swapped mid-task.

Watch

Extended reading notes

Core claim

The central claim is that residual dynamics, learned in a physically consistent Lagrangian form and conditioned on a latent context embedding, can be identified online fast enough for receding-horizon control and transfers zero-shot from simulation to hardware. The paper combines a nominal rigid-body model with a residual Deep Lagrangian Network, so only the unmodeled torque is learned; an LSTM encoder infers the latent environment variable from the recent history of joint positions, velocities, and residual torques before each MPC iteration. Because the environment is assumed fixed over the short prediction horizon, the same physically plausible Euler-Lagrange structure holds within each iteration while still allowing swift adaptation when the load changes. In hardware experiments, this method reduces end-effector tracking RMSE by 39% in a pick-and-place task and by 28-62% in high-speed tracking with different loads, consistently outperforming an EKF that only estimates an external end-effector force.

Load-bearing premise

The load-bearing premise is that residual dynamics learned purely in simulation, on payloads drawn from the same range as the test loads, transfer zero-shot to the real robot; if the sim-to-real gap from friction, cables, and imperfect load attachment is large, the claimed tracking improvement will not reproduce.

Editorial extensions

If this is right

  • When a nominal model exists, only the residual dynamics need to be learned, keeping the network small enough for real-time MPC with model evaluation around 8.5 ms.
  • A single context-conditioned network covers many environments, so changing payloads do not require retraining; the LSTM infers a new context online.
  • Unlike an EKF that estimates only an end-effector force, full residual dynamics identification also improves joints whose axes are parallel to gravity, where inertial effects dominate.
  • Zero-shot transfer from simulation to hardware is sufficient to reduce tracking error under unknown loads, without real-world data or domain randomization.
  • The added model complexity raises MPC computation time by roughly a factor of four, but the total time stays below the 20 ms control period.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the latent context reliably encodes load mass and center-of-mass offset, the same architecture could be extended to output these physical parameters explicitly, making the learned model more interpretable for safety monitoring.
  • The method assumes the environment is constant over the prediction horizon, which suits sparse changes like pick-and-place; a natural extension the paper does not explore is using the LSTM's latent dynamics to detect and flag context switches online.
  • Because training payloads are limited to masses up to 4 kg and offsets up to 0.3 m, a cheap test of the method's limits is to evaluate outside that range; the paper leaves this untested.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes CaDeLaC, a model predictive controller that uses a Context-Aware Deep Lagrangian Network (DeLaN) residual model conditioned on a latent context inferred from recent states and measured residual torques by an LSTM encoder. The contextual DeLaN is trained in MuJoCo across random payloads and then applied zero-shot to a Franka Emika Panda for joint and end-effector trajectory tracking under varying loads. Against a nominal MPC and an EKF-based adaptive MPC, the authors report improved tracking in simulation (1800 trajectories) and on hardware (high-speed trajectories and a pick-and-place task), with a headline 39% reduction in end-effector position RMSE versus 21% for the EKF baseline.

Significance. If the causal encoder works as claimed, the paper offers a practical integration of physics-consistent residual learning with real-time MPC: small network sizes, a nominal-model inductive bias, zero-shot transfer from simulation, and use of standard frameworks (acados, HPIPM, Pinocchio, L4CasADi). The breadth of the simulation evaluation and the inclusion of hardware experiments are strengths, as is the explicit computational-time analysis. However, the central online-identification claim rests on a train/deploy mismatch in the encoder objective that must be resolved before the reported results can be interpreted unambiguously.

major comments (3)
  1. [Section II-C, Algorithm 1 vs Algorithm 2, Eq. (5)] The online system identification is trained under a non-causal conditioning that does not match deployment. In Algorithm 1 (lines 4-5), z is computed as z_t = LSTM(h_0,...,h_t), and h_t includes the current measured residual torque tau_tilde_t (Eq. 5), so the loss can be minimized by copying the target tau_tilde_t into z_t rather than by learning a state-dependent residual model. At deployment, Algorithm 2 (line 2) infers z_k from h_{k-n_h},...,h_{k-1}, i.e., without the current torque, and Section III-B additionally low-pass filters the LSTM output at 2 Hz. The encoder is therefore evaluated under a causal conditioning never seen during training. No one-step-ahead prediction metric (residual torque prediction from past data only) is reported, so the claimed online identification capability and the resulting 39% hardware improvement are not established. Please report causal one-step-ahead residual torque RMSE, and either train with the same causal masking or demonstrate that the current torque does not affect z.
  2. [Section III-D, III-E, Tables II, IV, VII, Figure 5] The headline improvements are not supported by statistical evidence. In simulation, only mean RMSE values over the 600 trajectories per controller are reported, without variances, confidence intervals, or per-environment breakdowns. On hardware, each controller is run once per condition (nine high-speed trajectories and one pick-and-place trajectory), so the 39% versus 21% comparison in Figure 5 is a single-run observation. This matters because the tables show mixed per-joint results: CaDeLaC has larger velocity RMSE than the nominal controller on joints 1-3 in Table II, and larger position RMSE on q2 and q6 in the high-speed hardware experiment (Table IV). Reporting distributions or repeated trials is necessary to support the central tracking-improvement claim.
  3. [Section III-E-1, Table III] The residual-torque results on hardware are not uniformly in favor of the learned model, and this weakens the interpretation of the tracking gains. For the 1 kg condition, the CaDeLaC residual model has higher RMSE than the nominal model on joints 1, 2, 3, 5, 6, and 7 (Table III); for 2 kg and 3 kg it is worse than nominal on the distal joints 5-7. The paper attributes differences to noise, friction, and imperfect attachment, but if the residual model is inaccurate in several joints, the improved task-space tracking may be due to the MPC cost tradeoff rather than to accurate online identification. Please analyze the relationship between residual-torque accuracy and closed-loop tracking, and report residual-torque metrics for the pick-and-place task in a causal setting.
minor comments (4)
  1. [Section III-E-1] The sentence 'The EKF MPC still presented an overall performance than the nominal one' appears to be missing the word 'better'.
  2. [Section III-E-1] The statement 'As the last joint was the only one with a constant reference, its error analysis is irrelevant as they are very small' is confusing; please clarify whether joint 7 is excluded from the analysis and why.
  3. [Tables II, IV, VII] The bar notation in the table headers is not defined in the captions; please state explicitly that the entries are means over trajectories or runs.
  4. [Section II-C and Algorithms 1-2] Please specify how often the latent z_k is updated relative to the 50 Hz MPC loop and whether the 2 Hz low-pass filter mentioned in Section III-B is applied to the latent vector or to the predicted residual torque.

Circularity Check

1 steps flagged · score 6.0 of 10

Training objective leaks the target torque into the LSTM input, so the claimed online system identification is a copy rather than a causal prediction; the hardware result is not independently supported by the training signal.

  1. self definitional [Section II-B Eq. (5); Algorithm 1 lines 4-5; Algorithm 2 line 2]
    "each entry i in the nh length sequence is: hi = [q_i^T qdot_i^T tilde_tau_i^T]^T. ... z <- LSTM(h_0,...,h_t); theta <- theta - alpha grad_theta ||tilde_tau_t - f^{-1}(q_t, qdot_t, qddot_t; theta, z)||^2. ... z_k <- LSTM(h_{k-n_h}, ..., h_{k-1})."

    During training, the LSTM input h_t contains the target tilde_tau_t, and z_t is computed from h_0,...,h_t. The loss can be driven to zero by encoding tilde_tau_t into z_t and having the contextual DeLaN return that copied value, without learning any state-dependent or context-dependent dynamics. This is target leakage: the 'prediction' is available as an input by construction. At deployment, Algorithm 2 infers z_k only from past h entries, a conditioning distribution not seen in training, and the output is low-pass filtered at 2 Hz. Thus the claimed online system identification and the residual-torque predictions in Tables I, III, and VI are not supported by the training objective; the encoder was never trained to predict the current torque from past torque alone.

full rationale

The paper's central advertised contribution is a recurrent encoder that online-identifies a latent context from recently observed states and actions, combined with a residual DeLaN for MPC. The training loop in Algorithm 1 defines z_t as the LSTM output over h_0,...,h_t, and h_t includes the residual torque tilde_tau_t that is also the regression target of the loss. This makes the training-time residual-torque prediction reducible to copying the input, so the encoder is not forced to learn a causal latent representation. The control phase (Algorithm 2) then uses only past data (and a low-pass filter), a distribution mismatch that is never evaluated with a one-step-ahead metric. This is a concrete, equation-level reduction rather than a vague concern. However, the headline hardware tracking improvement (39% versus 21% EKF) is an emergent closed-loop measurement, not itself a fitted output; it could in principle arise from other properties of the architecture. The DeLaN and residual-dynamics components, while taken from prior work by co-authors, are used as external building blocks with published validation and are not circular. The paper is self-contained against external benchmarks for the control evaluation, but the training signal for the online SysID component is circular by construction. Score 6 reflects one load-bearing prediction that reduces by construction, while the final hardware claim retains some independent empirical content.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central claim rests on a learned model; the ledger lists the key hand-chosen hyperparameters, the physical and architectural assumptions about the residual dynamics and latent context, and the sim-to-real assumption. No new physical entities are introduced; the latent z is a learned internal state.

free parameters (4)
  • Latent context dimension = 10
    Output dimension of LSTM encoder and conditioning input; chosen without reported ablation.
  • LSTM history length (n_h) = 15
    Number of past timesteps fed to the LSTM; chosen without reported ablation.
  • Network sizes = MLP: 30/20; LSTM: 5 layers of 10 units
    Hand-selected architecture; no ablation study reported.
  • Training noise variances = var(q̇)=1e-3, var(q̈)=[0.05..0.65], var(τ)=[0.5..0.01], var(τ̃)=[1.0..0.02] (Table IX)
    Noise injected into inputs during training; values chosen by hand to improve robustness.
assumptions (5)
  • domain assumption The residual torque τ̃ = τ - τ̂ follows the same Lagrangian structure as the full dynamics (Eq. 4).
    This is the core architectural assumption enabling the residual DeLaN; it assumes the model mismatch can be expressed as an inertia/potential difference of the same form.
  • domain assumption The latent context z captures all environment-dependent dynamics, and z is constant over the MPC prediction horizon.
    Used in Algorithm 2: z is inferred once per MPC iteration and the model is held fixed during the horizon.
  • domain assumption The nominal model (Pinocchio-based) exactly captures the robot's own dynamics; only the payload-induced residual needs learning.
    The residual formulation relies on this split.
  • domain assumption Simulation-trained model transfers to the real robot.
    Zero-shot hardware experiments assume sufficient fidelity of MuJoCo for the residual dynamics.
  • standard math Standard rigid-body Lagrangian mechanics with holonomic constraints (Euler-Lagrange equation).
    Foundation of DeLaN and the residual model; standard mathematical background.
invented entities (1)
  • Latent context vector z
    purpose: A 10-dimensional embedding produced by an LSTM that summarizes recent observations and conditions the DeLaN residual model on the inferred environment.
    z is a learned latent variable; the paper provides no direct external measurement or falsifiable prediction for z itself, only for the downstream torque prediction and tracking error.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Context-Aware Deep Lagrangian Networks for Model Predictive Control." pith.science (2026). https://pith.science/paper/CZVLBBMX

@misc{pith2026250615249,
  author       = {Pith},
  title        = {Pith review of: Context-Aware Deep Lagrangian Networks for Model Predictive Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CZVLBBMX}},
  note         = {Machine review of arXiv:2506.15249}
}
read the original abstract

Controlling a robot based on physics-consistent dynamic models, such as Deep Lagrangian Networks (DeLaN), can improve the generalizability and interpretability of the resulting behavior. However, in complex environments, the number of objects to potentially interact with is vast, and their physical properties are often uncertain. This complexity makes it infeasible to employ a single global model. Therefore, we need to resort to online system identification of context-aware models that capture only the currently relevant aspects of the environment. While physical principles such as the conservation of energy may not hold across varying contexts, ensuring physical plausibility for any individual context-aware model can still be highly desirable, particularly when using it for receding horizon control methods such as model predictive control (MPC). Hence, in this work, we extend DeLaN to make it context-aware, combine it with a recurrent network for online system identification, and integrate it with an MPC for adaptive, physics-consistent control. We also combine DeLaN with a residual dynamics model to leverage the fact that a nominal model of the robot is typically available. We evaluate our method on a 7-DOF robot arm for trajectory tracking under varying loads. Our method reduces the end-effector tracking error by 39%, compared to a 21% improvement achieved by a baseline that uses an extended Kalman filter.

Figures

Figures reproduced from arXiv: 2506.15249 by the authors.

Figure 2
Figure 2. CaDeLaC: Context-Aware Deep Lagrangian Model [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Hardware experiments: (a) 3 kg gym weight attached [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. End-effector trajectory in high-speed experiments on the real robot. (a) 1 kg: CaDeLaC reduced the RMSE by 28.24% [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: End-effector trajectory in the pick-and-place task for [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 26 canonical work pages

  1. [1]

    On the adaptive control of robot manipulators,

    J.-J. E. Slotine and W. Li, “On the adaptive control of robot manipulators,” The International Journal of Robotics Research, vol. 6, no. 3, pp. 49–59, 1987. [Online]. Available: https://doi.org/10.1177/027836498700600303

    doi:10.1177/027836498700600303 1987

  2. [2]

    Estimation of inertial parameters of rigid body links of manipulators,

    C. H. An, C. G. Atkeson, and J. M. Hollerbach, “Estimation of inertial parameters of rigid body links of manipulators,” in 1985 24th IEEE Conference on Decision and Control , 1985, pp. 990–995

    work page 1985

  3. [3]

    Adaptive control based friction estimation for tracking control of robot manipulators,

    J. Huang, D. Tateo, P. Liu, and J. Peters, “Adaptive control based friction estimation for tracking control of robot manipulators,” IEEE Robotics and Automation Letters , vol. 10, no. 3, pp. 2454–2461, 2025

    work page 2025

  4. [4]

    Physically consistent online inertial adaptation for humanoid loco- manipulation,

    J. Foster, S. McCrory, C. DeBuys, S. Bertrand, and R. Griffin, “Physically consistent online inertial adaptation for humanoid loco- manipulation,” in 2024 IEEE/RSJ International Conference on Intel- ligent Robots and Systems (IROS) , 2024, pp. 11 278–11 285

    work page 2024

  5. [5]

    Model learning for robot control: A survey,

    D. Nguyen-Tuong and J. Peters, “Model learning for robot control: A survey,” Cognitive processing, vol. 12, pp. 319–40, 04 2011

    work page 2011

  6. [6]

    Model learning with local gaussian process regression,

    M. S. Duy Nguyen-Tuong and J. Peters, “Model learning with local gaussian process regression,” Advanced Robotics , vol. 23, no. 15, pp. 2015–2034, 2009

    work page 2015

  7. [7]

    Cautious model predictive control using gaussian process regression,

    L. Hewing, J. Kabzan, and M. N. Zeilinger, “Cautious model predictive control using gaussian process regression,” IEEE Transactions on Control Systems Technology , vol. 28, no. 6, pp. 2736–2743, 2020

    work page 2020

  8. [8]

    Safe machine-learning- supported model predictive force and motion control in robotics,

    J. Matschek, J. Bethge, and R. Findeisen, “Safe machine-learning- supported model predictive force and motion control in robotics,” IEEE Transactions on Control Systems Technology , vol. 31, no. 6, pp. 2380–2392, 2023

    work page 2023

Show all 39 references
  1. [9]

    Load estimation and control using learned dynamics models,

    G. Petkos and S. Vijayakumar, “Load estimation and control using learned dynamics models,” in 2007 IEEE/RSJ International Confer- ence on Intelligent Robots and Systems , 2007, pp. 1527–1532

    2007

  2. [10]

    Adaptive nmpc-rbf with application to manipulator robots,

    L. F. Recalde, J. Varela, B. S. Guevara, V . Andaluz, and D. Gandolfo, “Adaptive nmpc-rbf with application to manipulator robots,” in 2023 9th International Conference on Control, Decision and Information Technologies (CoDIT), 2023, pp. 2475–2482

    2023

  3. [11]

    Machine learning with physics knowledge for prediction: A survey,

    J. Watson et al. , “Machine learning with physics knowledge for prediction: A survey,” 2024. [Online]. Available: https://arxiv.org/abs/2408.09840

    2024 arXiv

  4. [12]

    Deep lagrangian networks: Using physics as model prior for deep learning,

    M. Lutter, C. Ritter, and J. Peters, “Deep lagrangian networks: Using physics as model prior for deep learning,” in International Conference on Learning Representations , 2019. [Online]. Available: https://openreview.net/forum?id=BklHpjCqKm

    2019

  5. [14]

    Hamiltonian neural networks,

    S. Greydanus, M. Dzamba, and J. Yosinski, “Hamiltonian neural networks,” in Advances in Neural Information Processing Systems , vol. 32. Curran Associates, Inc., 2019

    2019

  6. [15]

    Deep lagrangian networks for end-to-end learning of energy-based control for under-actuated systems,

    M. Lutter, K. D. Listmann, and J. Peters, “Deep lagrangian networks for end-to-end learning of energy-based control for under-actuated systems,” 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 7718–7725, 2019. [Online]. Available: https://a...

    2019

  7. [16]

    Port- hamiltonian neural ode networks on lie groups for robot dynamics learning and control,

    T. Duong, A. Altawaitan, J. Stanley, and N. Atanasov, “Port- hamiltonian neural ode networks on lie groups for robot dynamics learning and control,” IEEE Transactions on Robotics , vol. 40, pp. 3695–3715, 2024

    2024

  8. [17]

    Using model knowledge for learn- ing inverse dynamics,

    D. Nguyen-Tuong and J. Peters, “Using model knowledge for learn- ing inverse dynamics,” in 2010 IEEE International Conference on Robotics and Automation , 2010, pp. 2677–2682

    2010

  9. [18]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 770–778, 2015. [Online]. Available: https://api.semanticscholar.org/CorpusID:206594692

    2016

  10. [19]

    Meta- Learning in Neural Networks: A Survey,

    T. Hospedales, A. Antoniou, P. Micaelli, and A. Storkey, “Meta- Learning in Neural Networks: A Survey,” IEEE Transactions on Pattern Analysis & Machine Intelligence , vol. 44, no. 09, pp. 5149–5169, Sept. 2022. [Online]. Available: https://doi.ieeecomputersociety.org/10.1109/T...

    2022

  11. [20]

    Nonparametric adaptive control using gaussian processes with online hyperparameter estimation,

    R. C. Grande, G. V . Chowdhary, and J. P. How, “Nonparametric adaptive control using gaussian processes with online hyperparameter estimation,” 52nd IEEE Conference on Decision and Control , pp. 861–867, 2013. [Online]. Available: https://api.semanticscholar.org/CorpusID:7528974

    2013

  12. [21]

    Meta learning mpc using finite-dimensional gaussian process approximations,

    E. Arcari, A. Carron, and M. N. Zeilinger, “Meta learning mpc using finite-dimensional gaussian process approximations,” ArXiv, vol. abs/2008.05984, 2020. [Online]. Available: https://api.semanticscholar.org/CorpusID:221112284

    2008 arXiv

  13. [22]

    Data-driven model predictive control for trajectory tracking with a robotic arm,

    A. Carron, E. Arcari, M. Wermelinger, L. Hewing, M. Hutter, and M. N. Zeilinger, “Data-driven model predictive control for trajectory tracking with a robotic arm,” IEEE Robotics and Automation Letters , vol. 4, no. 4, pp. 3758–3765, 2019

    2019

  14. [23]

    Adaptive-control-oriented meta-learning for nonlinear systems,

    S. M. Richards, N. Azizan, J.-J. E. Slotine, and M. Pavone, “Adaptive-control-oriented meta-learning for nonlinear systems,” ArXiv, vol. abs/2103.04490, 2021. [Online]. Available: https://api.semanticscholar.org/CorpusID:232147745

    2021 arXiv

  15. [24]

    Bayesian multi-task learning mpc for robotic mobile manipulation,

    E. Arcari, M. V . Minniti, A. Scampicchio, A. Carron, F. Farshidian, M. Hutter, and M. N. Zeilinger, “Bayesian multi-task learning mpc for robotic mobile manipulation,” IEEE Robotics and Automation Letters , vol. 8, no. 6, pp. 3222–3229, 2023

    2023

  16. [25]

    Meta-learning augmented mpc for disturbance- aware motion planning and control of quadrotors,

    D. Lapandi ´c, F. Xie, C. K. Verginis, S.-J. Chung, D. V . Dimarogonas, and B. Wahlberg, “Meta-learning augmented mpc for disturbance- aware motion planning and control of quadrotors,” IEEE Control Systems Letters , vol. 8, pp. 3045–3050, 2024. [Online]. Available: https://api...

    2024

  17. [26]

    Preparing for the unknown: Learning a universal policy with online system identification,

    W. Yu, J. Tan, C. Karen Liu, and G. Turk, “Preparing for the unknown: Learning a universal policy with online system identification,” in Robotics: Science and Systems XIII , ser. RSS2017. Robotics: Science and Systems Foundation, July 2017. [Online]. Available: http://dx.doi.o...

    2017 doi

  18. [27]

    Physics- informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,

    M. Raissi, P. Perdikaris, and G. Karniadakis, “Physics- informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,” Journal of Computational Physics, vol. 378, pp. 686–707, 2019. [Online]. Av...

    2019

  19. [28]

    Metalearning generalizable dynamics from trajectories,

    Q. Li, T. Wang, V . Roychowdhury, and M. K. Jawed, “Metalearning generalizable dynamics from trajectories,” Physical Review Letters , vol. 131, no. 6, p. 067301, 2023

    2023

  20. [29]

    Knode-mpc: A knowledge- based data-driven predictive control framework for aerial robots,

    K. Y . Chee, T. Z. Jiahao, and M. A. Hsieh, “Knode-mpc: A knowledge- based data-driven predictive control framework for aerial robots,” IEEE Robotics and Automation Letters , vol. 7, no. 2, pp. 2819–2826, 2022

    2022

  21. [30]

    Real-time neural-mpc: Deep learning model predictive control for quadrotors and agile robotic platforms,

    T. Salzmann, E. Kaufmann, J. Arrizabalaga, M. Pavone, D. Scara- muzza, and M. Ryll, “Real-time neural-mpc: Deep learning model predictive control for quadrotors and agile robotic platforms,” IEEE Robotics and Automation Letters , 2023

    2023

  22. [31]

    Adaptive control of se(3) hamiltonian dynamics with learned disturbance features,

    T. P. Duong and N. A. Atanasov, “Adaptive control of se(3) hamiltonian dynamics with learned disturbance features,” IEEE Control Systems Letters , vol. 6, pp. 2773–2778, 2021. [Online]. Available: https://api.semanticscholar.org/CorpusID:247597120

    2021

  23. [32]

    acados – a modular open-source framework for fast embedded optimal control,

    R. Verschueren et al., “acados – a modular open-source framework for fast embedded optimal control,” Mathematical Programming Compu- tation, 2021

    2021

  24. [33]

    HPIPM: a high-performance quadratic programming framework for model predictive control,

    G. Frison and M. Diehl, “HPIPM: a high-performance quadratic programming framework for model predictive control,” IF AC- PapersOnLine, vol. 53, no. 2, pp. 6563–6569, 2020, 21st IFAC World Congress

    2020

  25. [34]

    The pinocchio c++ library – a fast and flexible implementation of rigid body dynamics algorithms and their analytical derivatives,

    J. Carpentier et al. , “The pinocchio c++ library – a fast and flexible implementation of rigid body dynamics algorithms and their analytical derivatives,” in IEEE International Symposium on System Integrations (SII), 2019

    2019

  26. [35]

    PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transformation and Graph Compilation,

    J. Ansel et al. , “PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transformation and Graph Compilation,” in 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, V olume 2 (ASPLOS ’24) . ACM, Apr. 2024. ...

    2024

  27. [36]

    Long short-term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural computation, vol. 9, no. 8, pp. 1735–1780, 1997

    1997

  28. [37]

    E. F. Camacho and C. Bordons Alba, Model predictive control . Springer-Verlag London, 2007

    2007

  29. [38]

    Mujoco: A physics engine for model-based control,

    E. Todorov, T. Erez, and Y . Tassa, “Mujoco: A physics engine for model-based control,” in 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems . IEEE, 2012, pp. 5026–5033

    2012

  30. [39]

    External joint torques es- timation for a position-controlled manipulator employing an extended kalman filter,

    L. Roveda, D. Riva, G. Bucca, and D. Piga, “External joint torques es- timation for a position-controlled manipulator employing an extended kalman filter,” in 2021 18th International Conference on Ubiquitous Robots (UR) , 2021, pp. 101–107

    2021

  31. [2021]

    Available: https://arxiv.org/abs/2110.01894

    [Online]. Available: https://arxiv.org/abs/2110.01894

    arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.