Pith. sign in

REVIEW 3 major objections 6 minor 36 references

Predictive Training with Latent Imagination for Visual Quadruped Navigation

T0 review · 3 major / 6 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a training-only predictive branch supervising the policy's own recurrent hidden state can make a reactive quadruped navigation policy anticipate moving obstacles, cutting collisions from 14.0% to 3.8% in dynamic scenes

desk verdict Useful training-only predictive supervision idea, but the missing SIGReg-only control leaves the causal role of the predictor unverified. read the letter →

arxiv 2607.17574 v1 pith:64NN237N submitted 2026-07-20 cs.RO cs.AI

Yancheng Zhu , Wanli Ma , Chen Han , Irvin Haozhe Zhan , Bingfeng Qin , Yixin Xu This is my paper · ORCID
classification cs.ROcs.AI
keywords predictivetraininglatentimaginationquadrupednavigationdynamicobstaclesrecurrentpolicyJEPASIGRegregularizationsim-to-realtransfer
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

This paper tries to establish that a purely training-time predictive loss — an MLP that predicts the recurrent navigation policy's next hidden state, plus a variance/decorrelation regularizer to stop collapse — can make a reactive LSTM-SRU controller anticipate moving obstacles at deployment, with the predictor discarded and no extra runtime cost. On a dynamic obstacle simulation task the predictive branch lifts success from 86.0% to 96.2% and cuts collisions from 14.0% to 3.8% compared with the same backbone trained without it, and the resulting policy transfers zero-shot to a physical Unitree Go2 quadruped in outdoor scenes. A sympathetic reader would care because it suggests anticipatory behavior can be shaped into a reactive controller for free during training, sidestepping the inference-time burden of world models and planners.

What carries the argument

The load-bearing object is Lpred = (1/d)‖gθ(ht, at) − sg(ht+1)‖², a one-step prediction loss in which a lightweight MLP gθ is trained to predict the policy's own deterministic hidden state ht+1 from (ht, at), with a stop-gradient on the target to prevent a trivial constant solution. SIGReg adds variance and decorrelation penalties on ht to prevent representational collapse. The predictive branch operates in latent space, detached from the observation encoder, and is entirely discarded at inference, leaving the deployed controller identical to the reactive LSTM-SRU baseline.

What would settle it

Train SRU-WM and the SRU baseline to convergence on identical dynamic scenes, freeze both, and fit a linear probe that decodes next-frame obstacle positions from each hidden state. If probe accuracy for SRU-WM is not clearly higher, the collision drop is not evidence of anticipatory encoding; it would more likely come from a smoothed or better-conditioned representation.

Watch

Extended reading notes

Core claim

The central claim is that predictive supervision of the recurrent hidden state — not a separate world model or planner — is sufficient to give a reactive navigation policy anticipatory behavior in dynamic scenes. The predictor gθ learns the mapping (ht, at) ↦ ht+1 during training; the resulting gradients reshape the LSTM-SRU memory such that, once the predictor is removed, the policy begins evasive steering while obstacles are still outside immediate collision range. The strongest evidence is the DynObs comparison: the same backbone with and without the predictive branch scores 96.2% vs 86.0% success and 3.8% vs 14.0% collisions, a gap the paper attributes to the predictive signal rather tha

Load-bearing premise

The load-bearing premise is that minimizing self-prediction of the policy's own next hidden state forces that state to represent short-horizon obstacle motion; the loss could be minimized instead by making the hidden state trivially predictable — compressed, low-dimensional, or smooth — without representing moving obstacles at all.

Editorial extensions

If this is right

  • Recurrent RL policies can be improved by an auxiliary self-prediction loss on their own hidden states, with no change to the deployed architecture.
  • Dynamic obstacle avoidance does not require inference-time imagination or planning; training-time latent prediction suffices to shape the controller's behavior.
  • SIGReg-style regularization is what keeps the hidden state well-conditioned during predictive training; without it, the same predictive branch makes the policy worse than no prediction at all (86.0% vs 90.7% success).
  • Predictive training on one navigation task transfers to a different dynamic-obstacle task, suggesting the learned hidden-state structure is not a brittle task-specific hack.

Reading between the lines

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

  • The recipe is generic to any recurrent policy, so comparable gains are plausible for wheeled robots, drones, or manipulators, though the paper does not test these morphologies.
  • A linear probe from ht to next-frame obstacle positions would directly test whether the collision reduction is caused by encoding obstacle motion; the paper does not include such an analysis.
  • Rolling out gθ for multiple steps during training is a natural extension for faster or more erratic obstacles; the paper explicitly leaves this to future work, so its benefit is an open question.
  • The near-constant SIGReg loss around 2.06 suggests the representation is stable but somewhat redundant across dimensions; reducing this floor might extract further gains, an inference not examined in the paper.
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 / 6 minor

Summary. The paper proposes a training-only predictive supervision method for visual quadruped navigation. A lightweight MLP predictor is added to an LSTM-SRU recurrent policy during training; it is trained to predict the policy's own next hidden state (with stop-gradient) under a SIGReg anti-collapse regularizer, and is discarded at inference. Experiments in Isaac Lab compare a static single-goal task (Nav) and a dynamic-obstacle task (DynObs) against NavRL and NavDP baselines and an internal SRU baseline, with a qualitative zero-shot deployment on a Unitree Go2. The central claim is that the predictive loss alone—not the regularizer or extra capacity—reduces collisions in dynamic scenes, with the sharpest quantitative evidence in Table 4 (3.8% vs 14.0% collision rate).

Significance. The core idea is attractive: a training-only auxiliary predictor with zero inference overhead could be a cheap way to shape recurrent representations for dynamic navigation. The internal controlled baseline (SRU without the predictive branch), the matched reward/observation/training protocol, and the zero-shot hardware deployment are clear strengths. If the causal role of Lpred were established, the DynObs collision reduction would be a useful result for legged local navigation. However, the missing SIGReg-only control leaves open the possibility that the observed gains come entirely from the known variance/covariance regularizer; the proposed anticipatory mechanism is not directly evidenced; and the lack of error bars prevents assessment of effect size and significance. Conditional on additional ablations and variance reporting, the paper could be a solid contribution.

major comments (3)
  1. [§5.2, Table 2; Eq. (14)] The paper's central claim—that predictive supervision Lpred, rather than the SIGReg regularizer, is responsible for the collision reduction—is not isolated by the reported ablations. The full objective in Eq. (14) is LRL + α_pred Lpred + α_sigreg Lsigreg. In Table 2, the variant 'SRU-WM (no SIGReg)' (α_sigreg=0) obtains 86.0% SR / 12.1% CR, worse than the SRU baseline (90.7% / 5.6%), whereas the full 'SRU-WM (MLP)' reaches 94.4% / 5.6%. Since there is no control with α_pred=0 and α_sigreg=0.003, the entire improvement over the baseline could be due to SIGReg alone. This is load-bearing for the abstract claim that the 'predictive training signal alone' improves navigation. Please add a SIGReg-only run, and also report an internal ablation on the DynObs task, where Table 4 currently has no internal ablation at all.
  2. [§3.3, Eq. (13); §5.5] The predictive target is the policy's own next hidden state, sg(ht+1), not an external observation of obstacle motion. Minimizing ||gθ(ht, at) − sg(ht+1)||² can be achieved by making ht+1 trivial to predict (e.g., low-dimensional or slowly varying), without encoding obstacle trajectories. The paper infers from the collision-rate drop that 'the learned recurrent representation encodes short-horizon obstacle motion' (§5.5), but no direct evidence is provided. To substantiate the anticipatory mechanism, add a probe or decoding experiment (e.g., decode obstacle velocity or position from ht) or a counterfactual (e.g., replace dynamic obstacles with static ones while keeping the policy otherwise identical) and show that the predictive loss's effect disappears. Without such evidence, the mechanism claim is not supported.
  3. [§5.2–5.3, Tables 2–4] All evaluation tables report single point estimates with no standard deviations, seeds, or confidence intervals. The headline differences—+3.7 percentage points SR on Nav and −10.2 percentage points CR on DynObs—could be within run-to-run noise, especially because metrics are averaged over the final 10% of training steps and the number of evaluation episodes is not stated for the ablation. Please report mean ± std over at least three independent training seeds (or, at minimum, episode-level confidence intervals) for all variants and tasks. This is necessary to assess the reliability of the quantitative claims.
minor comments (6)
  1. [§3.2, Eqs. (4)–(10)] The equations mix a GRU-style reset gate (rt in Eqs. 4–5) with LSTM cell equations (input, forget, output gates). Please clarify the exact LSTM-SRU cell update; as written, the role of rt in the LSTM variant is undefined.
  2. [§5.3, Tables 3–4] The NavRL/NavDP comparison uses official checkpoints adapted to a different simulator, sensor configuration, and action space, and the text acknowledges this. Given these confounds, the highlighted margins (e.g., '12× lower') should be softened or clearly labeled as system-level rather than method-level comparisons.
  3. [§5.4, Figures 8–9] The real-robot evidence is qualitative (four-frame sequences). Please report the number of trials, collision counts, or a quantitative success metric; otherwise the claim of 'zero-shot sim-to-real transfer' should be presented as a feasibility demonstration.
  4. [§5.5, Table 4] The phrase 'same predictor trained on navigation episodes' is ambiguous. Was the DynObs model trained with dynamic obstacles, or is this a zero-shot transfer from the static Nav task? Please clarify the training protocol for each row of Table 4.
  5. [§3.3, Remark on multi-step rollout] The paper honestly states that multi-step rollout is not used. This limitation is relevant to the title's 'latent imagination' wording, since no imagination or rollout is performed at training or inference. Consider rephrasing the title or carefully defining the term.
  6. [§4.1] The claim that gains are 'not due to additional model capacity' is not demonstrated: adding the predictor increases training-time parameters. The inference-time cost is indeed unchanged; please restrict the claim to inference-time capacity.

Circularity Check

0 steps flagged · score 1.0 of 10

No formal circularity: the auxiliary predictive loss is self-referential in target, but the reported success/collision metrics are external and the core derivation is not reduced to its inputs.

full rationale

I walked the paper's claimed derivation chain. The predictive branch (Eq. 12-13) is an auxiliary JEPA-style objective whose target is the policy's own next hidden state, sg(h_{t+1}); in that narrow sense the 'prediction' is self-referential rather than tied to external ground-truth obstacle trajectories. However, the paper's central quantitative claims (Tables 2-4) are held-out navigation success/collision/timeout rates, which are external to the fitted auxiliary loss; success is not defined in terms of L_pred or L_sigreg. The method builds on cited prior work (SRU [35], SIGReg [24], TD-MPC2 [18], etc.) with no evident author overlap and no load-bearing self-citation chain. The main weakness is experimental rather than circular: the SIGReg-only control (alpha_pred=0, alpha_sigreg>0) is missing, and the no-SIGReg ablation is worse than the SRU baseline, so the DynObs improvement cannot be unambiguously attributed to the predictive loss rather than the regularizer. That is a confound in causal attribution, not an equivalence by construction. No equation in the paper defines the claimed outcome in terms of the fitted predictor, so no formal circularity is established.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on two calibrated loss weights (α_pred = 0.03, α_sigreg = 0.003), a manually designed reward, and the prior assumptions that SRU aligns memory, SIGReg prevents collapse, one-step latent prediction suffices, and the simulator transfers to the real robot. No new physical entities are introduced.

free parameters (4)
  • α_pred = 0.03
    Weight of the auxiliary prediction loss in Eq. (14); described as 'calibrated' to shape the representation without disrupting policy optimization.
  • α_sigreg = 0.003
    Weight of the SIGReg variance-decorrelation loss in Eq. (14); chosen to prevent representational collapse.
  • Reward shaping weights (Table 1) = multiple coefficients (e.g., -50 failure penalty, +2.5 tight goal kernel, +0.5 soft goal, -0.1 lateral/action penalties)
    The multi-term shaped reward is manually designed; the failure penalty of -50 is large enough to dominate short episodes and drive collision avoidance.
  • Prediction horizon = 1 step
    The choice of one-step prediction instead of multi-step rollout is central to the claimed short-horizon mechanism; multi-step is explicitly deferred to future work.
assumptions (5)
  • domain assumption LSTM-SRU spatial modulation keeps the hidden state registered to the changing egocentric frame, as claimed in [35].
    The entire memory mechanism in §3.2 inherits this from prior work; if SRU's spatial alignment fails, the predictive-loss interpretation collapses.
  • domain assumption SIGReg (from [24]) prevents representational collapse of deterministic recurrent states without a stochastic bottleneck.
    Used in §3.3 to justify the variance/decorrelation penalty; the paper relies on this prior result rather than proving it here.
  • ad hoc to paper One-step prediction is sufficient to encode short-horizon obstacle motion.
    §3.3 explicitly defers multi-step rollout; the claimed collision-avoidance gains are attributed to one-step latent prediction without evidence about the required horizon.
  • domain assumption Isaac Sim/PhysX simulation plus the listed domain randomizations is a faithful enough proxy for the Unitree Go2 real world to support zero-shot transfer.
    §4.3 and §5.4 assume the sim-to-real gap is covered by randomizations; the real-robot validation is qualitative.
  • standard math MDPO is a valid trust-region policy optimizer for this setting.
    The paper cites [33]; no derivation is provided, but this is a standard background method.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Predictive Training with Latent Imagination for Visual Quadruped Navigation." pith.science (2026). https://pith.science/paper/64NN237N

@misc{pith2026260717574,
  author       = {Pith},
  title        = {Pith review of: Predictive Training with Latent Imagination for Visual Quadruped Navigation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/64NN237N}},
  note         = {Machine review of arXiv:2607.17574}
}
read the original abstract

Reinforcement-learning navigation policies for legged robots select actions reactively from current observations and short-term memory, with limited capacity to anticipate how moving obstacles will evolve in the near future. In dynamic environments, this reactivity causes the robot to respond too late because collision risk depends on short-horizon scene structure rather than on current obstacle positions alone. Lightweight predictive supervision applied to the policy's recurrent state during training can encode anticipatory obstacle dynamics without modifying the inference-time controller. We augment a reactive LSTM-SRU navigation backbone with an auxiliary JEPA-style predictor and SIGReg regularization: during training, the predictor supervises the deterministic hidden state to anticipate its own next state; at inference, it is fully discarded, incurring zero additional computational cost. On simulated and real-world navigation benchmarks with dynamic obstacles, our method substantially improves navigation success while reducing collision rates through the predictive training signal alone, without additional inference-time parameters. Real-robot deployment on a Unitree Go2 demonstrates zero-shot sim-to-real transfer: the controller navigates cluttered indoor and dynamic outdoor environments without fine-tuning, with evasive behavior consistent with the collision reduction observed in simulation.

Figures

Figures reproduced from arXiv: 2607.17574 by the authors.

Figure 1
Figure 1. Model architecture of the SRU-WM framework. The reactive backbone (black) fuses depth, proprioception, and [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. LSTM-SRU recurrent cell with spatial modulation. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. World-model training process. The LSTM-SRU back [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Simulated training environments spanning diverse [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Training loss curves for the predictive branch. Shaded [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Qualitative trajectory comparison on static obstacle [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Dynamic obstacle navigation environment. Pedestri [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 9
Figure 9. Figure 9: Outdoor deployment. Top: static obstacle avoidance [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 3 linked inside Pith

  1. [1]

    So- cial lstm: Human trajectory prediction in crowded spaces

    Alexandre Alahi, Kratarth Goel, Vignesh Ramanathan, Alexandre Robicquet, Li Fei-Fei, and Silvio Savarese. So- cial lstm: Human trajectory prediction in crowded spaces. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016

    2016

  2. [2]

    V-jepa 2: Self-supervised video models enable understanding, prediction and planning.arXiv preprint arXiv:2506.09985, 2025

    Mido Assran, Adrien Bardes, David Fan, Quentin Garrido, et al. V-jepa 2: Self-supervised video models enable understanding, prediction and planning.arXiv preprint arXiv:2506.09985, 2025

    arXiv 2025

  3. [3]

    Self-supervised learning from im- ages with a joint-embedding predictive architecture

    Mido Assran, Quentin Duval, Ishan Misra, Piotr Bo- janowski, Pascal Vincent, Michael Rabbat, Yann LeCun, and Nicolas Ballas. Self-supervised learning from im- ages with a joint-embedding predictive architecture. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023

    2023

  4. [4]

    Navigation world models

    Amir Bar, Gaoyue Zhou, Danny Tran, Trevor Darrell, and Yann LeCun. Navigation world models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 15791–15801, 2025

    2025

  5. [5]

    VICReg: Variance-invariance-covariance regularization for self- supervised learning.International Conference on Learn- ing Representations (ICLR), 2022

    Adrien Bardes, Jean Ponce, and Yann LeCun. VICReg: Variance-invariance-covariance regularization for self- supervised learning.International Conference on Learn- ing Representations (ICLR), 2022

    2022

  6. [6]

    The vector field histogram—fast obstacle avoidance for mobile robots

    Johann Borenstein and Yoram Koren. The vector field histogram—fast obstacle avoidance for mobile robots. IEEE Transactions on Robotics and Automation, 7(3):278– 288, 1991

    1991

  7. [7]

    NavDP: Learning sim-to-real navigation diffusion policy with privileged information guidance

    Wenzhe Cai, Jiaqi Peng, Yuqiang Yang, Yujian Zhang, Meng Wei, Hanqing Wang, Yilun Chen, Tai Wang, and Jiangmiao Pang. NavDP: Learning sim-to-real navigation diffusion policy with privileged information guidance. arXiv preprint arXiv:2505.08712, 2025

    arXiv 2025

  8. [8]

    SEA-Nav: Efficient policy learning for safe and agile quadruped navigation in cluttered environments

    Shiyi Chen, Mingye Yang, Haiyan Mao, Jiaqi Zhang, Haiyi Liu, Shuheng He, Debing Zhang, Zihao Qiu, and Chun Zhang. SEA-Nav: Efficient policy learning for safe and agile quadruped navigation in cluttered environments. arXiv preprint arXiv:2603.09460, 2026

    arXiv 2026

Show all 36 references
  1. [9]

    Exploring simple siamese representation learning

    Xinlei Chen and Kaiming He. Exploring simple siamese representation learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021

    2021

  2. [10]

    Learning phrase representations using rnn encoder–decoder for statistical machine translation

    Kyunghyun Cho, Bart van Merriënboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. Learning phrase representations using rnn encoder–decoder for statistical machine translation. In Proceedings of the 2014 Conference on Empirical Meth- od...

    2014

  3. [11]

    Facing off world model backbones: RNNs, transformers, and S4

    Fei Deng, Junyeong Park, and Sungjin Ahn. Facing off world model backbones: RNNs, transformers, and S4. InAdvances in Neural Information Processing Systems, 2023

    2023

  4. [12]

    The dynamic window approach to collision avoidance.IEEE Robotics & Automation Magazine, 4(1):23–33, 1997

    Dieter Fox, Wolfram Burgard, and Sebastian Thrun. The dynamic window approach to collision avoidance.IEEE Robotics & Automation Magazine, 4(1):23–33, 1997

    1997

  5. [13]

    Bootstrap your own latent: A new approach to self-supervised learning

    Jean-Bastien Grill, Florian Strub, Florent Altché, Corentin Tallec, Pierre Richemond, Elena Buchatskaya, Carl Do- ersch, Bernardo Pires, Zhaohan Daniel Guo, Moham- mad Gheshlaghi Azar, et al. Bootstrap your own latent: A new approach to self-supervised learning. InAdvances in ...

    2020

  6. [14]

    Social gan: Socially acceptable trajectories with generative adversarial networks

    Agrim Gupta, Justin Johnson, Li Fei-Fei, Silvio Savarese, and Alexandre Alahi. Social gan: Socially acceptable trajectories with generative adversarial networks. InPro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018

    2018

  7. [15]

    9 Learning latent dynamics for planning from pixels.Pro- ceedings of the 36th International Conference on Machine Learning, 2019

    Danijar Hafner, Timothy Lillicrap, Ian Fischer, Ruben Villegas, David Ha, Honglak Lee, and James Davidson. 9 Learning latent dynamics for planning from pixels.Pro- ceedings of the 36th International Conference on Machine Learning, 2019

    2019

  8. [16]

    Mastering atari with discrete world mod- els.International Conference on Learning Representa- tions, 2021

    Danijar Hafner, Timothy Lillicrap, Mohammad Norouzi, and Jimmy Ba. Mastering atari with discrete world mod- els.International Conference on Learning Representa- tions, 2021

    2021

  9. [17]

    Mastering diverse domains through world models

    Danijar Hafner, Jurgis Pasukonis, Jimmy Ba, and Timo- thy Lillicrap. Mastering diverse domains through world models. InInternational Conference on Learning Repre- sentations, 2024

    2024

  10. [18]

    Td-mpc2: Scalable, robust world models for continuous control

    Nicklas Hansen, Hao Su, and Xiaolong Wang. Td-mpc2: Scalable, robust world models for continuous control. In International Conference on Learning Representations, 2024

    2024

  11. [19]

    Social force model for pedestrian dynamics.Physical Review E, 51(5):4282– 4286, 1995

    Dirk Helbing and Péter Molnár. Social force model for pedestrian dynamics.Physical Review E, 51(5):4282– 4286, 1995

    1995

  12. [20]

    Long short- term memory.Neural Computation, 9(8):1735–1780, 1997

    Sepp Hochreiter and J"urgen Schmidhuber. Long short- term memory.Neural Computation, 9(8):1735–1780, 1997

    1997

  13. [21]

    Muon: An optimizer for hidden layers in neural networks, 2024

    Keller Jordan, Yuchen Jin, Vlado Boza, Jiacheng You, Franz Cesista, Laker Newhouse, and Jeremy Bernstein. Muon: An optimizer for hidden layers in neural networks, 2024

    2024

  14. [22]

    A path towards autonomous machine intel- ligence.OpenReview preprint, 2022

    Yann LeCun. A path towards autonomous machine intel- ligence.OpenReview preprint, 2022

    2022

  15. [23]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. InInternational Conference on Learning Representations, 2019

    2019

  16. [24]

    LeWorldModel: Stable end-to-end joint-embedding predictive architecture from pixels, 2026

    Lucas Maes, Quentin Le Lidec, Damien Scieur, Yann LeCun, and Randall Balestriero. LeWorldModel: Stable end-to-end joint-embedding predictive architecture from pixels, 2026

    2026

  17. [25]

    Learning robust perceptive locomotion for quadrupedal robots in the wild.Science Robotics, 7(62), 2022

    Takahiro Miki, Jemin Lee, Jemin Hwangbo, Lorenz Well- hausen, Vladlen Koltun, and Marco Hutter. Learning robust perceptive locomotion for quadrupedal robots in the wild.Science Robotics, 7(62), 2022

    2022

  18. [26]

    Rep- resentation learning with contrastive predictive coding

    Aaron van den Oord, Yazhe Li, and Oriol Vinyals. Rep- resentation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748, 2018

    2018 arXiv

  19. [27]

    Mastering atari, go, chess and shogi by planning with a learned model.Nature, 588(7839):604–609, 2020

    Julian Schrittwieser, Ioannis Antonoglou, Thomas Hubert, Karen Simonyan, Laurent Sifre, Simon Schmitt, Arthur Guez, Edward Lockhart, Demis Hassabis, Thore Graepel, Timothy Lillicrap, and David Silver. Mastering atari, go, chess and shogi by planning with a learned model.Nature...

    2020

  20. [28]

    Proximal policy optimization algorithms.arXiv preprint arXiv:1707.06347, 2017

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms.arXiv preprint arXiv:1707.06347, 2017

    2017 arXiv

  21. [29]

    Vint: A foundation model for visual navigation.Proceedings of The 7th Conference on Robot Learning, 2023

    Dhruv Shah, Ajay Sridhar, Kunal Bhargava, et al. Vint: A foundation model for visual navigation.Proceedings of The 7th Conference on Robot Learning, 2023

    2023

  22. [30]

    Dreamernav: Learning-based autonomous nav- igation in dynamic indoor environments using world mod- els.Frontiers in Robotics and AI, 12, 2025

    Stuart Shanks, Jonathan Embley-Riches, Jianheng Liu, Andromachi Maria Delfaki, Carlo Ciliberto, and Dimitrios Kanoulas. Dreamernav: Learning-based autonomous nav- igation in dynamic indoor environments using world mod- els.Frontiers in Robotics and AI, 12, 2025

    2025

  23. [31]

    No- mad: Goal masked diffusion policies for navigation and exploration

    Ajay Sridhar, Dhruv Shah, Kunal Bhargava, et al. No- mad: Goal masked diffusion policies for navigation and exploration. InProceedings of the IEEE International Conference on Robotics and Automation, 2024

    2024

  24. [32]

    Lei Tai, Giuseppe Paolo, and Ming Liu. Virtual-to-real deep reinforcement learning: Continuous control of mo- bile robots for mapless navigation.2017 IEEE/RSJ Inter- national Conference on Intelligent Robots and Systems (IROS), 2017

    2017

  25. [33]

    Mirror descent policy optimization

    Manan Tomar, Lior Shani, Yonathan Efroni, and Moham- mad Ghavamzadeh. Mirror descent policy optimization. InInternational Conference on Learning Representations, 2022

    2022

  26. [34]

    Drl-vo: Learning to navigate dynamic environments with velocity obstacles

    Zhanteng Xie and Philip Dames. Drl-vo: Learning to navigate dynamic environments with velocity obstacles. IEEE Robotics and Automation Letters, 2023

    2023

  27. [35]

    Spatially-enhanced recurrent memory for long-range mapless navigation via end-to- end reinforcement learning.The International Journal of Robotics Research, 2025

    Fan Yang, Per Frivik, David Hoeller, Chen Wang, Cesar Cadena, and Marco Hutter. Spatially-enhanced recurrent memory for long-range mapless navigation via end-to- end reinforcement learning.The International Journal of Robotics Research, 2025

    2025

  28. [36]

    Barlow twins: Self-supervised learning via redundancy reduction.International Conference on Machine Learning (ICML), 2021

    Jure Zbontar, Li Jing, Ishan Misra, Yann LeCun, and Stéphane Deny. Barlow twins: Self-supervised learning via redundancy reduction.International Conference on Machine Learning (ICML), 2021. 10

    2021

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.