Pith. sign in

REVIEW 3 major objections 5 minor 30 references

VertiAKD: Adaptive Off-Road Kinodynamics on Vertically Challenging Terrain

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read With about one minute of new trajectory data, VertiAKD transfers terrain-aware off-road kinodynamic knowledge from a training fleet to a novel vehicle and adapts it online, cutting long-horizon prediction error by up to 34.52%.

desk verdict A coherent integration of the group's own prior pieces into a terrain-aware cross-vehicle transfer system; the physical trailer experiment is new and valuable, but the 'one minute' claim is not tied to a disjoint evaluation split. read the letter →

arxiv 2608.00945 v1 pith:RUFEIKRC submitted 2026-08-02 cs.RO

Tong Xu , Chenhui Pan , Francesco Cancelliere , Xuesu Xiao This is my paper
classification cs.RO
keywords off-roadkinodynamicscross-vehicletransferterrain-awaremodelingfunctionencoderneuralODErecursiveleastsquaresMPPInavigationverticallychallengingterrain
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

VertiAKD aims to solve a fleet-scale version of off-road motion modeling: instead of collecting hours of platform-specific data for each new robot, a new vehicle inherits terrain-aware kinodynamic knowledge from an existing heterogeneous fleet. The paper builds a shared latent space of vehicle-terrain mobility descriptors that encode how configurations such as mass, tire friction, and suspension stiffness move over different elevation and semantic terrain. For a new vehicle, roughly one minute of trajectory data is enough to identify which descriptors are relevant, form a weighted coefficient prior for a function encoder built from neural ordinary differential equation basis functions, and then refine the coefficients online with recursive least squares, without gradient retraining. The paper reports prediction error reductions of up to 34.52% over direct descriptor transfer and 94.43% over strong few-shot baselines, and closed-loop MPPI navigation on unseen terrain where a fixed prior fails. If true, this would make per-platform retraining unnecessary for off-road robot fleets.

What carries the argument

The load-bearing object is the shared vehicle-terrain mobility latent space together with the terrain-conditioned function encoder. The mobility encoder is a Transformer with cross-attention between trajectory transition tokens and terrain tokens, modulated by vehicle configuration through adaptive layer normalization, and trained with a triplet objective that organizes embeddings by joint vehicle-terrain kinodynamic similarity. Trained fleet embeddings are clustered into mobility descriptors; for a new vehicle, sliced Wasserstein distance assigns relevance weights $w_k$. These weights combine per-descriptor coefficient vectors $\alpha^*_k$, obtained by ridge-regularized least squares, into

What would settle it

Run VertiAKD on a new vehicle whose configuration is deliberately far outside the training fleet (for example, a trailer-towing or tracked platform) on terrain whose elevation and semantic features are far from every training descriptor, and collect a long dataset. If after many RLS updates the 64-step prediction error remains above the best single descriptor's error and above a from-scratch model trained on the same long data, then the frozen-basis coefficient representation is the bottleneck rather than the descriptor prior.

Watch

Extended reading notes

Core claim

The central claim is that kinodynamic modeling for a new off-road vehicle on complex terrain can be bootstrapped from about one minute of data by recombining knowledge from related vehicles in a latent mobility space. VertiAKD encodes trajectory transitions, the vehicle's physical configuration, and local elevation and semantic terrain features into a shared representation; clusters that representation into mobility descriptors; and uses a distribution-aware distance to weight the descriptors most relevant to the new vehicle. Those weights produce a coefficient prior for a terrain-conditioned function encoder, whose forward model is a linear combination of neural-ODE basis functions. The coe

Load-bearing premise

The novel vehicle's actual motion on unseen terrain must be approximately reproducible as a weighted blend of motion patterns seen in the training fleet; if the new platform moves in a way no trained descriptor captures, the transferred model starts biased and the online update cannot fully correct it because the learned basis functions are frozen.

Editorial extensions

If this is right

  • A fleet operator can add a novel vehicle with about one minute of driving data instead of platform-specific dataset collection and retraining, because the function-encoder basis functions are shared and only low-dimensional coefficients adapt.
  • Online adaptation remains cheap throughout deployment: RLS updates run on a fixed-size buffer with no gradient backpropagation, so per-update cost does not grow with history.
  • Terrain conditioning through elevation and semantic features lets the transferred model respond to local geometry and surface type, extending cross-vehicle transfer beyond flat-terrain settings.
  • In closed-loop MPPI evaluations, the RLS-refined prior succeeds in unseen environments where the fixed prior fails, so the online step is load-bearing for navigation, not just for prediction metrics.
  • Under a matched three-trajectory data budget, the method reports lower long-horizon error than gradient-based meta-learning and a universal dynamics baseline, with more data giving those baselines only small gains.

Reading between the lines

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

  • The descriptor relevance weights double as an out-of-distribution indicator: if a new vehicle's embeddings are far from every training descriptor, operators could flag it as outside fleet coverage and avoid trusting its model before more data is collected.
  • The same latent space could support active data collection: an operator could choose which training vehicle to drive on which terrain to most reduce uncertainty about an unknown platform, rather than collecting undirected exploration data.
  • Because adaptation is confined to coefficients over frozen basis functions, the clearest path to handling genuinely novel dynamics would be to unfreeze a small part of the basis during deployment; this is a natural next experiment, not something the paper claims.
  • If the one-minute result holds across wider configuration ranges, fleet maintenance becomes a data-sharing problem: the value of any new vehicle's data is mostly in selecting and weighting existing descriptors, not in learning new basis functions.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes VertiAKD, a framework for cross-vehicle transfer and online adaptation of terrain-aware kinodynamic models. It learns a shared mobility latent space from trajectory transitions, vehicle configurations, and elevation/semantic terrain patches using a Transformer with cross-attention and adaptive layer normalization, trained with a triplet objective. For a novel vehicle, VertiAKD computes sliced Wasserstein distances between the new data and clustered training mobility descriptors, builds a weighted coefficient prior for a terrain-conditioned neural ODE function encoder, and refines the coefficients online with recursive least squares. Experiments in the Verti-Bench simulator and on physical Verti-4-Wheeler configurations report lower 64-step prediction MSE than direct descriptor transfer and several baselines, plus improved closed-loop MPPI trajectory tracking.

Significance. If the central claims hold, VertiAKD addresses a real and growing need: transferring kinodynamic models across heterogeneous robot fleets without per-platform retraining, while accounting for terrain geometry and semantics. The paper has concrete strengths: a complete system design, validation in both simulation and physical hardware, use of an open benchmark (Verti-Bench), and an online adaptation mechanism with bounded memory and no gradient-based retraining. The reported improvements are substantial. However, the headline data-efficiency claim---'one minute of new trajectory data'---is not yet supported by a well-defined and disjoint evaluation protocol, and the small number of trials with large variances makes the quantitative reductions difficult to interpret. These issues are fixable and do not invalidate the overall approach, but they are load-bearing for the paper's main contribution.

major comments (3)
  1. [Abstract and Sec. V-A] The central 'one minute' claim is not tied to a concrete, disjoint data budget. The paper says the prior is built from 'only three trajectories' (Sec. V-A) but never defines trajectory length, number of transitions, or how three trajectories relate to 'one minute' in the abstract. At the stated 10 Hz recording (Sec. IV-A), the 64-step horizon is 6.4 s, so three length-64 trajectories would be about 19.2 s, not one minute. More importantly, the paper does not state whether the three prior-construction trajectories are disjoint from the trajectories used to compute the MSE in Tables I, II, and IV, or from the navigation trials. If they overlap, the reported 34.52% and 94.43% reductions are in-sample and cannot support the transfer claim. Please specify the split, the number of transitions per trajectory, and how the one-minute budget is allocated.
  2. [Tables I, II, IV] No statistical significance tests, confidence intervals, or per-trial results are reported. For example, in Table II, VertiAKD has MSE 0.144±0.150 while MAML has 0.330±0.390; the standard deviations overlap substantially. The 94.43% relative reduction, and even some of the smaller differences in Table I, may be within noise. The paper should report the number of evaluation rollouts/segments, per-seed or per-trial results, and paired bootstrap or significance tests. This is necessary to support the quantitative headline claims.
  3. [Sec. III-D, Eq. (2), and Sec. V-B] The weighted coefficient prior alpha_prior = sum_{k in N_new} w_k alpha*_k assumes that the novel vehicle's optimal coefficients lie in the span of the training fleet's descriptor coefficient vectors. Because the basis functions are frozen, the prior cannot represent dynamics outside this span. The physical experiment with the Four-Wheeled with Trailer platform is precisely a case where the new vehicle may excite coupled vehicle-trailer dynamics absent from the training fleet, yet the paper does not analyze whether or why the span assumption holds. I recommend adding a leave-one-vehicle-out analysis or an explicit out-of-span diagnostic (e.g., compare alpha_prior against coefficients estimated from a large new-vehicle dataset) to demonstrate when the prior is valid and when it is not.
minor comments (5)
  1. [Fig. 2] Typo: 'Funtion Encoder' should be 'Function Encoder'.
  2. [Sec. V-A] The paper alternates between 'one minute', 'three trajectories', and '400 trajectories' without defining trajectory length. Please clarify the data budget and make the units consistent throughout the abstract and experiments.
  3. [Tables I and II] The number of evaluation trajectories/rollouts used to compute MSE±Std is not stated. Please add this information, as it is essential for interpreting the variance.
  4. [Algorithm 1] The notation P_{t|t-B} and P_t is confusing: line 6 defines P_{t|t-B} but later updates P_t without an explicit recursion on B. Please clarify the indexing or add an explanatory sentence.
  5. [Related Work] The Related Work section is heavily self-referential (e.g., refs. [13], [18], [25]). Consider adding more discussion of recent cross-embodiment and few-shot dynamics methods beyond MAML and AnyCar to situate the contribution more broadly.

Circularity Check

0 steps flagged · score 1.0 of 10

No definitional circularity: the coefficient prior is a weighted combination of training-fleet descriptor coefficients; new-vehicle data only sets descriptor-relevance weights.

full rationale

VertiAKD's central prediction chain is not circular by construction. The coefficient prior for a novel vehicle is αprior = Σ_{k∈N_new} w_k α*_k (Sec. III-D), where each α*_k is computed via ridge-regularized least squares (Eqn. 2) on example transitions belonging to a training-fleet mobility descriptor. The novel vehicle's limited data are used only to compute an embedding distribution Z_new and the SWD-based relevance weights w_k (Secs. III-C and V-A); they are not used to regress the α coefficients against the new vehicle's transition labels. Thus the long-horizon prediction error is not minimized on the new-vehicle data by definition. The mobility encoder is trained with triplet losses (Sec. III-B), but this is an empirical representation-learning step, not a renaming of the target. The paper does build on prior function-encoder work by the same authors (Refs. [13], [25]) and uses self-developed benchmarks ([18], [30]); these are self-citations, but the method is described and trained in the paper and is evaluated against external baselines (MAML, AnyCar) plus physical experiments, so the citations are not load-bearing uniqueness arguments. The manuscript does not explicitly state whether the three trajectories used for descriptor identification are disjoint from the 64-step evaluation windows; that is a missing-support/experimental-design concern for the 'one minute' data-efficiency claim, but it is not a circularity because the prior is not fitted to the evaluation targets. Overall, no equation reduces to its own input and no fitted parameter is renamed as a prediction; the correct finding is no significant circularity.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The ledger lists the hand-set hyperparameters and load-bearing domain assumptions. The most important free parameters are the triplet margin, loss weights, ridge regularization, RLS forgetting factor and noise (values unreported), the 0.9 descriptor-retention threshold, the SWD^{-2} weighting exponent, and the silhouette-selected cluster count K. The central unstated premise is convex-hull coverage: a new vehicle's dynamics must be expressible as a weighted blend of training descriptor coefficients. No new physical entities are introduced; 'mobility descriptors' are a learned clustering, not an invented physical quantity.

free parameters (7)
  • Triplet margin delta = 2.0
    Manual hyperparameter in the terrain-aware triplet loss (Sec. IV-B) that controls latent-space separation; chosen by hand, not fitted to a benchmark.
  • Triplet loss weights w_u, w_c, w_v = 1.0, 1.0, 0.1
    Weights of unconditional, conditional, and variance losses in the combined objective; chosen by hand.
  • Ridge regularization lambda (Eqn. 2) = 1e-3
    Regularization for closed-form coefficient estimation in the function encoder.
  • RLS forgetting factor gamma and noise matrix Q = not reported
    Algorithm 1 depends on gamma in (0,1] and noise Q, but no numerical values are given; replication requires contacting authors.
  • Descriptor retention threshold = 0.9
    Cumulative normalized weight threshold for selecting top mobility descriptors (Sec. III-C).
  • Sliced Wasserstein distance exponent = -2
    Relevance weights are defined as SWD^{-2} and then normalized; the power is chosen by hand without an ablation.
  • Number of mobility descriptors K = selected by silhouette score, value not reported
    K-means cluster count chosen to maximize silhouette; the resulting K is not reported, making the prior construction hard to reproduce.
assumptions (5)
  • domain assumption Forward kinodynamics can be represented exactly enough as a linear combination of a fixed set of neural ODE basis functions with a per-condition coefficient vector.
    Adopted from function-encoder literature (refs [11],[12],[13]); Sec. III-D states the model Delta x_t = sum_i alpha_i G_i(x_t,u_t,e_t; theta_i).
  • ad hoc to paper The kinodynamics of any novel vehicle on unseen terrain lies close to the convex hull of the training fleet's mobility-descriptor coefficient vectors.
    Sec. III-D constructs alpha_prior = sum_{k in N_new} w_k alpha*_k with nonnegative weights summing to 1. The paper provides no mechanism for dynamics outside this span.
  • domain assumption Elevation and RGB semantic patches encoded by reconstruction-only convolutional autoencoders retain the information needed for kinodynamic prediction.
    Sec. IV-A trains autoencoders with MSE reconstruction loss only; no dynamics-aware loss is used, so task-relevant features are assumed to survive.
  • domain assumption The mobility latent space learned from the training fleet transfers to novel vehicles and unseen terrain, so Sliced Wasserstein distance in that space ranks transferable kinodynamic knowledge.
    Sec. III-C uses SWD between Z_new and descriptor distributions to set weights; this assumes distribution distance correlates with dynamics transferability.
  • standard math Standard numerical and statistical machinery (RK4 integration, ridge regression, RLS, PCA, K-means, Sliced Wasserstein distance) is used correctly.
    Used throughout Secs. III and IV without proof; standard methods.

how reviews work

0 comments
Cite this review

Pith. "Pith review of VertiAKD: Adaptive Off-Road Kinodynamics on Vertically Challenging Terrain." pith.science (2026). https://pith.science/paper/RUFEIKRC

@misc{pith2026260800945,
  author       = {Pith},
  title        = {Pith review of: VertiAKD: Adaptive Off-Road Kinodynamics on Vertically Challenging Terrain},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RUFEIKRC}},
  note         = {Machine review of arXiv:2608.00945}
}
read the original abstract

Off-road mobility requires autonomous mobile robots to generalize across heterogeneous vehicle fleets and continuously changing terrain conditions. Existing cross-vehicle adaptation approaches generally assume flat terrain, while terrain-aware kinodynamic models often require platform-specific data collection and retraining. To this end, we propose VertiAKD, a unified framework for transferring and adapting off-road kinodynamic knowledge across diverse vehicles on geometrically and semantically complex terrain simultaneously. VertiAKD learns a shared mobility representation that jointly encodes vehicle configurations, trajectory transitions, and local elevation and semantic terrain features. Given limited data from a novel vehicle operating on unseen terrain, VertiAKD identifies the most relevant mobility descriptors and transfers their knowledge to initialize a terrain-aware kinodynamic model via function encoders, which is then periodically refined online from streaming observations without gradient-based retraining. We evaluate VertiAKD in the Verti-Bench simulator, built on the Chrono multi-physics engine, and on five physical configurations of the Verti-4-Wheeler platform. With only one minute of new trajectory data and associated terrain features, VertiAKD reduces long-horizon prediction error by up to 34.52% over direct mobility descriptor transfer across diverse unseen vehicle configurations and 94.43% over competing baselines. We further demonstrate robust closed-loop trajectory tracking in both simulation and physical experiments, highlighting the effectiveness of terrain-aware cross-vehicle knowledge transfer for accurate modeling and reliable off-road navigation.

Figures

Figures reproduced from arXiv: 2608.00945 by the authors.

Figure 1
Figure 1. A heterogeneous vehicle fleet with diverse physical [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Verti-Arena comprises a variety of off-road terrain, [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

30 extracted references · 19 canonical work pages

  1. [1]

    Motion planning and control for mobile robot navigation using machine learning: a survey,

    X. Xiao, B. Liu, G. Warnell, and P. Stone, “Motion planning and control for mobile robot navigation using machine learning: a survey,” Autonomous Robots, vol. 46, no. 5, pp. 569–597, 2022

    2022

  2. [2]

    A survey on path planning for autonomous ground vehicles in unstructured environ- ments,

    N. Wang, X. Li, K. Zhang, J. Wang, and D. Xie, “A survey on path planning for autonomous ground vehicles in unstructured environ- ments,”Machines, vol. 12, no. 1, p. 31, 2024

    work page 2024

  3. [3]

    A survey on terrain traversability analysis for autonomous ground vehicles: Methods, sensors, and challenges,

    P. V . Borges, T. Peynot, S. Liang, B. Arain, M. Wildie, M. G. Minareci, S. Lichman, G. Samvedi, I. Sa, N. Hudsonet al., “A survey on terrain traversability analysis for autonomous ground vehicles: Methods, sensors, and challenges,”Field Robotics, vol. 2, pp. 1567– 1627, 2022

    2022

  4. [4]

    Information-theoretic model predictive control: Theory and applica- tions to autonomous driving,

    G. Williams, P. Drews, B. Goldfain, J. M. Rehg, and E. A. Theodorou, “Information-theoretic model predictive control: Theory and applica- tions to autonomous driving,”IEEE Transactions on Robotics, vol. 34, no. 6, pp. 1603–1622, 2018

    2018

  5. [5]

    Randomized kinodynamic plan- ning,

    S. M. LaValle and J. J. Kuffner Jr, “Randomized kinodynamic plan- ning,”The international journal of robotics research, vol. 20, no. 5, pp. 378–400, 2001

    2001

  6. [6]

    Toward wheeled mobility on vertically challenging terrain: Platforms, datasets, and algorithms,

    A. Datar, C. Pan, M. Nazeri, and X. Xiao, “Toward wheeled mobility on vertically challenging terrain: Platforms, datasets, and algorithms,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 16 322–16 329

    work page 2024

  7. [7]

    Learning inverse kinodynamics for accurate high-speed off-road navigation on unstructured terrain,

    X. Xiao, J. Biswas, and P. Stone, “Learning inverse kinodynamics for accurate high-speed off-road navigation on unstructured terrain,”IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 6054–6060, 2021

    2021

  8. [8]

    Terrain- attentive learning for efficient 6-dof kinodynamic modeling on verti- cally challenging terrain,

    A. Datar, C. Pan, M. Nazeri, A. Pokhrel, and X. Xiao, “Terrain- attentive learning for efficient 6-dof kinodynamic modeling on verti- cally challenging terrain,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 5438– 5443

    work page 2024

Show all 30 references
  1. [9]

    Pietra: Physics-informed evidential learning for traversing out-of-distribution terrain,

    X. Cai, J. Queeney, T. Xu, A. Datar, C. Pan, M. Miller, A. Flather, P. R. Osteen, N. Roy, X. Xiaoet al., “Pietra: Physics-informed evidential learning for traversing out-of-distribution terrain,”IEEE Robotics and Automation Letters, vol. 10, no. 3, pp. 2359–2366, 2025

    2025

  2. [10]

    Physord: a neuro-symbolic approach for physics-infused motion prediction in off-road driving,

    Z. Zhao, B. Li, Y . Du, T. Fu, and C. Wang, “Physord: a neuro-symbolic approach for physics-infused motion prediction in off-road driving,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 11 670–11 677

    2024

  3. [11]

    Online adaptation of terrain-aware dynamics for planning in unstruc- tured environments,

    W. Ward, S. Etter, T. Ingebrand, C. Ellis, A. J. Thorpe, and U. Topcu, “Online adaptation of terrain-aware dynamics for planning in unstruc- tured environments,”arXiv preprint arXiv:2506.04484, 2025

    2025 arXiv

  4. [12]

    Zero-shot transfer of neural odes,

    T. Ingebrand, A. J. Thorpe, and U. Topcu, “Zero-shot transfer of neural odes,”Advances in Neural Information Processing Systems, vol. 37, pp. 67 604–67 626, 2024

    2024

  5. [13]

    Vertiadaptor: Online kin- odynamics adaptation for vertically challenging terrain,

    T. Xu, C. Pan, A. Datar, and X. Xiao, “Vertiadaptor: Online kin- odynamics adaptation for vertically challenging terrain,” in2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2026

    2026

  6. [14]

    Neu- ral ordinary differential equations,

    R. T. Chen, Y . Rubanova, J. Bettencourt, and D. K. Duvenaud, “Neu- ral ordinary differential equations,”Advances in neural information processing systems, vol. 31, 2018

    2018

  7. [15]

    Pushing the limits of cross-embodiment learning for manipulation and navigation,

    J. Yang, C. Glossop, A. Bhorkar, D. Shah, Q. Vuong, C. Finn, D. Sadigh, and S. Levine, “Pushing the limits of cross-embodiment learning for manipulation and navigation,” inProceedings of Robotics: Science and Systems, 2024

    2024

  8. [16]

    Scaling cross-embodied learning: One policy for manipulation, navigation, locomotion and aviation,

    R. Doshi, H. R. Walke, O. Mees, S. Dasari, and S. Levine, “Scaling cross-embodied learning: One policy for manipulation, navigation, locomotion and aviation,” inConference on Robot Learning. PMLR, 2025, pp. 496–512

    2025

  9. [17]

    Anycar to anywhere: Learning universal dynamics model for agile and adaptive mobility,

    W. Xiao, H. Xue, T. Tao, D. Kalaria, J. M. Dolan, and G. Shi, “Anycar to anywhere: Learning universal dynamics model for agile and adaptive mobility,” in2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2025, pp. 8819–8825

    2025

  10. [18]

    Verti-bench: A general and scalable off-road mobility benchmark for vertically challenging terrain,

    T. Xu, C. Pan, M. B. Rao, A. Datar, A. Pokhrel, Y . Lu, and X. Xiao, “Verti-bench: A general and scalable off-road mobility benchmark for vertically challenging terrain,” inRobotics: Science and Systems (RSS) 2025, 2025

    2025

  11. [19]

    J. Y . Wong,Theory of ground vehicles. John Wiley & Sons, 2022

    2022

  12. [20]

    High-speed accurate robot control using learned forward kinody- namics and non-linear least squares optimization,

    P. Atreya, H. Karnan, K. S. Sikand, X. Xiao, S. Rabiee, and J. Biswas, “High-speed accurate robot control using learned forward kinody- namics and non-linear least squares optimization,” in2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2...

    2022

  13. [21]

    Learning to model and plan for wheeled mobility on vertically challenging terrain,

    A. Datar, C. Pan, and X. Xiao, “Learning to model and plan for wheeled mobility on vertically challenging terrain,”IEEE Robotics and Automation Letters, vol. 10, no. 2, pp. 1505–1512, 2024

    2024

  14. [22]

    Model-agnostic meta-learning for fast adaptation of deep networks,

    C. Finn, P. Abbeel, and S. Levine, “Model-agnostic meta-learning for fast adaptation of deep networks,” inInternational conference on machine learning. PMLR, 2017, pp. 1126–1135

    2017

  15. [23]

    Hyperdynamics: Meta-learning object and agent dynamics with hy- pernetworks,

    Z. Xian, S. Lal, H.-Y . Tung, E. A. Platanios, and K. Fragkiadaki, “Hyperdynamics: Meta-learning object and agent dynamics with hy- pernetworks,”arXiv preprint arXiv:2103.09439, 2021

    2021 arXiv

  16. [24]

    Rma: Rapid motor adaptation for legged robots,

    A. Kumar, Z. Fu, D. Pathak, and J. Malik, “Rma: Rapid motor adaptation for legged robots,”arXiv preprint arXiv:2107.04034, 2021

    2021 arXiv

  17. [25]

    Car: Cross-vehicle kinodynamics adap- tation via mobility representation,

    T. Xu, C. Pan, and X. Xiao, “Car: Cross-vehicle kinodynamics adap- tation via mobility representation,”arXiv preprint arXiv:2603.06866, 2026

    2026 arXiv

  18. [26]

    Online learning of dynamical systems using low-rank updates to physics-informed kernel distribution embeddings,

    K. Ortiz, R. DiPirro, A. J. Thorpe, and M. Oishi, “Online learning of dynamical systems using low-rank updates to physics-informed kernel distribution embeddings,” in2024 IEEE 63rd Conference on Decision and Control (CDC). IEEE, 2024, pp. 7548–7555

    2024

  19. [27]

    Gaussian process dynamical models,

    J. Wang, A. Hertzmann, and D. J. Fleet, “Gaussian process dynamical models,”Advances in neural information processing systems, vol. 18, 2005

    2005

  20. [28]

    Scalable diffusion models with transformers,

    W. Peebles and S. Xie, “Scalable diffusion models with transformers,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 4195–4205

    2023

  21. [29]

    Gen- eralized sliced wasserstein distances,

    S. Kolouri, K. Nadjahi, U. Simsekli, R. Badeau, and G. Rohde, “Gen- eralized sliced wasserstein distances,”Advances in neural information processing systems, vol. 32, 2019

    2019

  22. [30]

    Verti-arena: A controllable and standardized indoor testbed for multi-terrain off-road autonomy,

    H. Chen, A. Datar, T. Xu, F. Cancelliere, H. Rangwala, M. B. Rao, D. Song, D. Eichinger, and X. Xiao, “Verti-arena: A controllable and standardized indoor testbed for multi-terrain off-road autonomy,” 2026, pp. 133–138

    2026

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.