Pith. sign in

REVIEW 4 major objections 5 minor 47 references

Learning more with the same effort: how randomization improves the robustness of a robotic deep reinforcement learning agent

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Randomized camera views during training make a progressive-network robot policy about 25 percent more accurate on unseen camera positions, at the same 70-million-step training budget.

desk verdict Clean within-simulation robustness comparison that overclaims a sim-to-real efficiency result: no PNN is actually trained, and the 25% headline lacks seed variance. read the letter →

arxiv 2501.14443 v1 pith:WKBUTSCE submitted 2025-01-24 cs.RO cs.AI

Lucía Güitta-López , Jaime Boal , Álvaro J. López-López This is my paper · ORCID
classification cs.ROcs.AI
keywords DeepReinforcementLearningSim-to-RealTransferProgressiveNeuralNetworksDomainRandomizationRoboticManipulationCameraViewpointRobustnessA3CVirtualTestBench
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 randomizing the camera viewpoint during the simulated training phase makes a progressive-neural-network (PNN) robot controller substantially more robust to camera changes before any real experience is added. The authors train two agents with identical 70-million-step budgets: a baseline fixed-camera PNN agent and a domain-randomized variant whose camera moves within a bounded range. Evaluated on a grid of 153 camera positions with 1,000 episodes each, the randomized agent keeps high accuracy over a much wider viewpoint region and beats the baseline by roughly 25 percent on average, with gains above 80 percent in the positions where the baseline collapses. The practical point is that if robustness after the virtual phase predicts how quickly a real column can adapt, then the same synthetic effort buys a smaller real-experience requirement.

What carries the argument

The load-bearing object is the progressive neural network architecture: a virtual column trained on synthetic experience, connected by lateral links to light real-world columns that are later trained on real data. The paper's instrument is a virtual test bench that sweeps 153 camera positions (z from 140 to 220 degrees, y from -50 to -10 degrees, in 5-degree steps) and labels each episode a success or failure, turning robustness into an accuracy map. The independent variable is the camera randomization interval used during training, and the evaluation grid deliberately extends beyond that interval so the agent must both interpolate and extrapolate. The key identity is that average accuracy over this grid is treated as a measure of how much real experience the PNN's real column will need after the virtual phase.

What would settle it

Deploy the same baseline and randomized policies as real columns on a physical arm with a camera placed at one of the grid positions and measure episodes-to-success for both; if the randomized model does not reach the baseline's success rate with less real experience, the proxy fails. A cheaper check: train a third agent randomizing lighting and textures instead of camera position; if its accuracy gain on the camera grid does not predict its robustness to those perturbations, viewpoint robustness is not a general sim-to-real readiness measure.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that a PNN-style agent trained with a randomized camera position (z from 160 to 200 degrees, y from -40 to -20 degrees) is more robust to camera viewpoint changes than the same agent trained with a fixed camera at (180, -30) degrees, under the same total training effort and with only about one extra hour of compute. The randomized model sustains accuracy above 90 percent across roughly 155 to 210 degrees in z and -40 to -25 degrees in y, achieves 100 percent accuracy at 11 camera positions, and handles shadow-induced failures that break the baseline. The authors interpret the gain as evidence that domain randomization should complement the PNN sim-to-real pipeline, reducing the amount of real experience the real column needs for the same final performance.

Load-bearing premise

The whole practical conclusion rests on treating camera-viewpoint robustness in the simulator as a stand-in for the sim-to-real gap and for how much real experience a real column would need; the paper never runs a real robot or a real transfer experiment, so if other discrepancies dominate, the claimed reduction in real experience is unsupported.

Editorial extensions

If this is right

  • With the same 70 million training steps, a robot policy trained with randomized camera positions reaches above 90 percent accuracy over a much wider range of viewpoints than the fixed-camera policy.
  • The largest gains, exceeding 80 percent in some cells, occur exactly where the baseline policy fails, so the improvement is not just a uniform shift.
  • The randomized model needs no additional experience, only about one more hour of training, meaning the robustness gain is nearly free in sample terms.
  • The accuracy difference after the virtual phase can be read as a decrease in the real experience the PNN real column would require, since a policy already invariant to viewpoint disturbances needs less adaptation.
  • Real experience should still improve the agent even when the virtual data is high quality, so domain randomization is a complement to, not a replacement for, real training.

Reading between the lines

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

  • If viewpoint perturbation is a proxy for the broader sim-to-real gap, the same randomization logic should extend to lighting, texture, and dynamics; the paper only varies camera position, so this is an extrapolation.
  • There is likely an optimal randomization width: the randomized model loses a little accuracy at the extreme -10 degree y-axis viewpoint compared to the baseline, suggesting that too much diversity can dilute precision even as it widens coverage.
  • The 153-position accuracy grid could serve as a diagnostic tool for deciding where real-world fine-tuning should concentrate: cells where the randomized model still underperforms are the ones whose visual features are least represented in training.
  • A direct real-robot test with the same PNN columns would settle whether the measured robustness gain translates into real sample savings; the paper sets up the benchmark but does not run that test.
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

4 major / 5 minor

Summary. This paper studies whether randomizing the camera viewpoint during simulation-based training improves the robustness of a deep RL agent to camera-position shifts, using that robustness as a proxy for behavior in a progressive-neural-network (PNN) sim-to-real pipeline. The authors train two A3C agents on a simulated IRB120 reaching task for 70 million steps each: a Baseline Model trained with a fixed camera at (180°, −30°) and a Domain Randomization Model trained with camera positions sampled from [160°, 200°] × [−40°, −20°]. Both policies are evaluated on a held-out grid of 153 camera positions spanning [140°, 220°] × [−50°, −10°], with 1,000 episodes per position. The DRM reports roughly 25 percentage points higher average success accuracy over the grid, with gains above 80% in some peripheral configurations. The authors interpret this as evidence that randomization would improve a PNN's virtual column and reduce the real experience needed for transfer.

Significance. The experimental benchmark is carefully designed: identical step budgets and hyperparameters, a test grid wider than the training distribution, and 1,000 evaluation episodes per camera pose. If the viewpoint-robustness result is reproducible across seeds and environments, it is a useful quantitative demonstration that domain randomization can widen the high-accuracy operating envelope of an image-based robotic controller at no additional training cost. The main advertised implication for real-world PNN transfer, however, goes beyond the evidence presented.

major comments (4)
  1. [Section 7 and Abstract] The conclusion that DR 'reduce[s] the amount of training and experience required when transferring this knowledge to the real world' is not supported by the experiments, because no progressive neural network is trained. The evaluated agent (Appendix A) is a standalone A3C with convolutional layers and an LSTM; there are no lateral connections, no frozen virtual column, and no real-column training. Section 5 explicitly measures robustness 'right before including real experience into the lateral connections and real columns of the PNN architecture,' so the paper tests zero-shot viewpoint robustness of a single policy, not the transfer speed of a PNN. This is a load-bearing extrapolation; either an actual PNN transfer experiment or a substantially weakened claim is needed.
  2. [Section 6, Figs. 8-11 and Table 3] Each model is trained only once, so the headline 'around 25%' average improvement has no training-seed variance. The shaded bands in Figures 8-11 reflect standard deviation across evaluation episodes, not uncertainty across training runs; with a single seed, the 0-5% gains near the training point and the >80% gains at the periphery cannot be separated from run-to-run stochasticity. Please provide multiple training seeds or present the claims with an explicit single-run caveat.
  3. [Sections 6.1 and 6.2] The DRM is not uniformly more robust by every metric: the maximum failure distance increases from 48 cm for the BM (at (140°, −25°) and (220°, −50°)) to 81 cm for the DRM (at (140°, −10°)). Since the paper defines robustness through several metrics (Table 3), the average-accuracy gain should be accompanied by a discussion of this worse worst-case behavior before claiming that DR 'substantially mitigates' the robustness problem.
  4. [Section 5] The proxy assumption—'only the camera position is used to represent discrepancies between the virtual and the real worlds'—is a reasonable choice for a controlled experiment, but it is an untested assumption that camera-viewpoint shifts dominate the sim-to-real gap. Dynamics, lighting, textures, and joint friction are not perturbed, so the results do not directly establish that the DRM would transfer better to a physical robot; this limits the practical claim in the abstract and conclusions.
minor comments (5)
  1. [Table 2] The target position entries 'U(−0,3,0.3) cm' and 'U(−0.3,0.3)' are typographically inconsistent; the intended interval is presumably U(−0.3, 0.3) cm.
  2. [Figure 5 and Figure 8 captions] The captions use 'shadowed region' and 'shading' inconsistently; use 'shaded' consistently for the gray training-area marker.
  3. [Figure 7 caption and Table 3] The phrase 'steady-stable regime' is used without a formal definition; define the criterion or replace it with 'steady state'.
  4. [Appendix A, Table A1] The text states that M1 uses a logarithmic action set, but Table A1 does not explicitly mark which action set is logarithmic; clarify the correspondence between the rows and the description.
  5. [Section 3] The related-work discussion would benefit from more recent sim-to-real and domain-randomization references, since several cited works are from the 2017-2020 period and the field has subsequently developed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the BM vs DRM robustness comparison is an empirical held-out evaluation, and the sim-to-real experience claim is an acknowledged extrapolation, not a derivation from fitted inputs.

full rationale

No circular step found. The central comparison is empirical: BM and DRM are trained with identical hyperparameters, action space, reward, and the same 70-million-step budget; only the camera randomization differs (Table 2). Robustness is evaluated on a fixed grid of 153 camera poses (Fig. 5) with 1,000 episodes per pose, and the DR training region is only a subset of the test grid, so accuracy at held-out poses is a genuine prediction rather than a reconstruction of a fitted quantity. The reported 25% average improvement and the heat maps are measured outcomes, not derived from the inputs by construction. The PNN framework is adopted from Rusu et al. [5], an external citation, and no uniqueness theorem or self-citation is used to force the choice. The conclusion that DR would reduce real-world training experience is an extrapolation, explicitly deferred to future work ('We leave for future work the quantification of the actual reduction of the total amount of real experience required'), and the evaluated agent is a standalone A3C without lateral connections or real columns; this is a scope/validity limitation, not a circular reduction. Under the hard rule that only construction-level reductions count as circularity, the appropriate score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

No new physical entities or theoretical constructs are introduced. The load-bearing assumptions are practical: the simulation is an adequate testbed, camera viewpoint is the relevant perturbation, and the chosen success criterion and training distributions are valid. The main unvalidated extension is interpreting viewpoint robustness as predictive of real-experience savings in an actual PNN transfer.

free parameters (3)
  • DR training camera ranges = z in [160,200] deg, y in [-40,-20] deg
    The randomization intervals are chosen by the authors, not derived from any analysis of the test grid or from measured sim-to-real discrepancies. The test grid extends to [140,220] and [-50,-10], so the training range covers the central region; the choice affects the magnitude of the reported gain.
  • Evaluation success distance = 10 cm (training success at 5 cm)
    The evaluation tolerance was deliberately increased by 5 cm relative to training, 'trusting that training the agent in more challenging conditions than required in evaluation would lead to a better performance.' This choice affects all accuracy numbers and the headline 25 percent gain.
  • Episode length / reward design (M1) = 50 steps; reward 70 or -(2*dist)^2
    Selected among seven MDP variants (Appendix A, Table A1) based on average accuracy, max failure distance, and learning time. The chosen design is a hand-selected configuration, not a first-principles choice, and the robustness result may depend on it.
assumptions (4)
  • domain assumption MuJoCo simulation of IRB120 dynamics is an adequate proxy for the robotic task.
    The entire study is in simulation; the paper relies on MuJoCo as ground truth and on simple Matplotlib rendering without rich textures. Invoked in Sections 4 and 5. This is the core proxy assumption for the claimed sim-to-real relevance.
  • domain assumption A3C with the Rusu et al. architecture is an appropriate and state-of-the-art DRL backbone.
    The paper adopts the A3C network from [5] without comparing to more recent algorithms. Invoked in Section 5 and Appendix A. A different algorithm could have different robustness behavior under randomization.
  • ad hoc to paper Camera-position perturbation alone captures the sim-to-real discrepancy.
    Stated explicitly in Section 5: 'only the camera position is used to represent discrepancies between the virtual and the real worlds.' This is a modeling choice that the conclusion about reducing real experience depends on.
  • standard math Standard RL math: MDP formulation, discounted return, A3C policy gradient with LSTM.
    Background assumptions from Sutton and Barto and Mnih et al., used in Sections 2 and 5, are standard and uncontroversial.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning more with the same effort: how randomization improves the robustness of a robotic deep reinforcement learning agent." pith.science (2026). https://pith.science/paper/WKBUTSCE

@misc{pith2026250114443,
  author       = {Pith},
  title        = {Pith review of: Learning more with the same effort: how randomization improves the robustness of a robotic deep reinforcement learning agent},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WKBUTSCE}},
  note         = {Machine review of arXiv:2501.14443}
}
read the original abstract

The industrial application of Deep Reinforcement Learning (DRL) is frequently slowed down because of the inability to generate the experience required to train the models. Collecting data often involves considerable time and economic effort that is unaffordable in most cases. Fortunately, devices like robots can be trained with synthetic experience thanks to virtual environments. With this approach, the sample efficiency problems of artificial agents are mitigated, but another issue arises: the need for efficiently transferring the synthetic experience into the real world (sim-to-real). This paper analyzes the robustness of a state-of-the-art sim-to-real technique known as progressive neural networks (PNNs) and studies how adding diversity to the synthetic experience can complement it. To better understand the drivers that lead to a lack of robustness, the robotic agent is still tested in a virtual environment to ensure total control on the divergence between the simulated and real models. The results show that a PNN-like agent exhibits a substantial decrease in its robustness at the beginning of the real training phase. Randomizing certain variables during simulation-based training significantly mitigates this issue. On average, the increase in the model's accuracy is around 25% when diversity is introduced in the training process. This improvement can be translated into a decrease in the required real experience for the same final robustness performance. Notwithstanding, adding real experience to agents should still be beneficial regardless of the quality of the virtual experience fed into the agent.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 27 canonical work pages

  1. [1]

    MIT Press, Cambridge (Massachusetts), London (England) (2018)

    Sutton, R.S., Barto, A.G.: Reinforcement Learning: an Introduction. MIT Press, Cambridge (Massachusetts), London (England) (2018)

    work page 2018

  2. [2]

    Computing Research Repository (CoRR) (2017)

    Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., Lillicrap, T., Simonyan, K., Hassabis, D.: Mastering chess and shogi by self-play with a general reinforcement learning algorithm. Computing Research Repository (CoRR) (2017)

    work page 2017

  3. [3]

    Benchmarking Reinforcement Learning Algorithms on Real-World Robots

    Mahmood, A.R., Korenkevych, D., Vasan, G., Ma, W., Bergstra, J.: Benchmarking reinforcement learning algorithms on real-world robots. In: Proceedings of the Second Conference on Robot Learning (CoRL) (2018). http://arxiv.org/abs/1809.07731

    work page Pith review arXiv 2018

  4. [4]

    Foundations and Trends in Machine Learning 11(3-4), 219–354 (2018)

    Fran¸ cois-Lavet, V., Henderson, P., Islam, R., Bellemare, M.G., Pineau, J.: An introduction to deep reinforcement learning. Foundations and Trends in Machine Learning 11(3-4), 219–354 (2018). https://doi.org/10.1561/ 2200000071

    work page 2018

  5. [5]

    In: 1st Conference on Robot Learning (CoRL) (2017)

    Rusu, A.A., Veˇ cer ´ ık, M., Roth¨ orl, T., Heess, N., Pascanu, R., Hadsell, R.: Sim-to-real robot learning from pixels with progres- sive nets. In: 1st Conference on Robot Learning (CoRL) (2017). https://arxiv.org/abs/1610.04286

    arXiv 2017

  6. [6]

    In: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp

    Tobin, J., Fong, R., Ray, A., Schneider, J., Zaremba, W., Abbeel, P.: Domain randomization for transferring deep neural networks from simulation to the real world. In: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 23–30 (2017). http://arxiv.org/abs/1703.06907

    arXiv 2017

  7. [7]

    Journal of Mathematics and Mechanics, 679–684 (1957)

    Bellman, R.: A markovian decision process. Journal of Mathematics and Mechanics, 679–684 (1957)

    work page 1957

  8. [8]

    Management Science 28(1), 1–16 (1982)

    Monahan, G.E.: Survey of partially observable markov decision processes - theory, models and algortihms. Management Science 28(1), 1–16 (1982). https://doi.org/10.1287/mnsc.28.1.1

    work page doi:10.1287/mnsc.28.1.1 1982

Show all 47 references
  1. [9]

    https://www

    Silver, D.: Lecture 2: Markov Decision Processes (2020). https://www. davidsilver.uk/wp-content/uploads/2020/03/MDP.pdf

    2020

  2. [10]

    IEEE Access 8, 176598–176623 (2020)

    Al-Masrur Khan, M.D., Khan, M.R.J., Tooshil, A., Sikder, N., Parvez Mahmud, M.A., Kouzani, A.Z., Nahid, A.A.: A systematic review on reinforcement learning-based robotics within the last decade. IEEE Access 8, 176598–176623 (2020). https://doi.org/10.1109/ACCESS.2020. 3027152 ...

    2020 doi

  3. [11]

    In: Proceedings of The 33rd International Conference on Machine Learning (PMLR), vol

    Mnih, V., Puigdom` enech Badia, A., Mirza, M., Graves, A., Harley, T., Lillicrap, T.P., Silver, D., Kavukcuoglu, K.: Asynchronous methods for deep reinforcement learning. In: Proceedings of The 33rd International Conference on Machine Learning (PMLR), vol. 48, pp. 1928–1937 (2...

    2016

  4. [12]

    Journal of Artificial Intelligence Research 69, 1421–1471 (2020)

    Lazaridis, A.: Deep reinforcement learning: A state-of-the-art walk- through. Journal of Artificial Intelligence Research 69, 1421–1471 (2020)

    2020

  5. [13]

    IEEE Signal Processing Magazine 34(6), 26–38 (2017)

    Arulkumaran, K., Deisenroth, M.P., Brundage, M., Bharath, A.A.: Deep reinforcement learning: A brief survey. IEEE Signal Processing Magazine 34(6), 26–38 (2017). https://doi.org/10.1109/MSP.2017.2743240

    2017

  6. [14]

    MIT Press, - (2016)

    Goodfellow, I., Bengio, Y., Courville, A.: Deep Learning. MIT Press, - (2016). http://www.deeplearningbook.org

    2016

  7. [15]

    https://doi.org/10.1038/nature14539

    Lecun, Y., Bengio, Y., Hinton, G.: Deep learning 521(7553), 436–444 (2015). https://doi.org/10.1038/nature14539

    2015 doi

  8. [16]

    International Journal of Robotics Research 32(11), 1238–1274 (2013)

    Kober, J., Bagnell, J.A., Peters, J.: Reinforcement learning in robotics: A survey. International Journal of Robotics Research 32(11), 1238–1274 (2013). https://doi.org/10.1177/0278364913495721

    2013 doi

  9. [17]

    Computing Research Repository (CoRR) (2019)

    Li, Y.: Reinforcement learning applications. Computing Research Repository (CoRR) (2019)

    2019

  10. [18]

    In: 2017 IEEE International Conference on Robotics and Automation, pp

    Zhu, Y., Mottaghi, R., Kolve, E., Lim, J.J., Gupta, A., Fei-Fei, L., Farhadi, A.: Target-driven visual navigation in indoor scenes using deep reinforcement learning. In: 2017 IEEE International Conference on Robotics and Automation, pp. 3357–3364 (2017). https://doi.org/10. 11...

    2017

  11. [19]

    In: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp

    Zhang, J., Springenberg, J.T., Boedecker, J., Burgard, W.: Deep reinforcement learning with successor features for navigation across similar environments. In: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 2371–2378 (2017). https: //doi.org/10....

    2017

  12. [20]

    IEEE Sensors Journal 21(2), 2230–2240 (2021)

    Ejaz, M.M., Tang, T.B., Lu, C.K.: Vision-based autonomous navigation approach for a tracked robot using deep reinforcement learning. IEEE Sensors Journal 21(2), 2230–2240 (2021). https://doi.org/10.1109/JSEN. 2020.3016299

    2021

  13. [21]

    IEEE Robotics and Automation Magazine 13(2), 99–110 (2006)

    Durrant-Whyte, H., Bailey, T.: Simultaneous localization and mapping: Part i history of the slam problem. IEEE Robotics and Automation Magazine 13(2), 99–110 (2006). https://doi.org/10.1109/MRA.2006. 1638022 Preprint sent to Applied Intelligence Learning more with the same effort 25

    2006 doi

  14. [22]

    IEEE Robotics and Automation Magazine 13(3), 108–117 (2006)

    Bailey, T., Durrant-Whyte, H.: Simultaneous localization and mapping (slam): Part ii. IEEE Robotics and Automation Magazine 13(3), 108–117 (2006). https://doi.org/10.1109/MRA.2006.1678144

    2006 arXiv

  15. [23]

    Intelligent Service Robotics13(2), 263–272 (2020)

    Wen, S., Zhao, Y., Yuan, X., Wang, Z., Zhang, D., Manfredi, L.: Path planning for active slam based on deep reinforcement learning under unknown environments. Intelligent Service Robotics13(2), 263–272 (2020). https://doi.org/10.1007/s11370-019-00310-w

    2020 doi

  16. [24]

    Computing Research Repository (CoRR) (2016)

    Tai, L., Zhang, J., Liu, M., Boedecker, J., Burgard, W.: A survey of deep network solutions for learning control in robotics: From reinforcement to imitation. Computing Research Repository (CoRR) (2016)

    2016

  17. [25]

    2016-June, pp

    Mahler, J., Pokorny, F.T., Hou, B., Roderick, M., Laskey, M., Aubry, M., Kohlhoff, K., Kroger, T., Kuffner, J., Goldberg, K.: Dex-net 1.0: A cloud-based network of 3d objects for robust grasp planning using a multi- armed bandit model with correlated rewards, vol. 2016-June, p...

    2016

  18. [26]

    Robotics: Science and Systems (2017)

    Mahler, J., Liang, J., Niyaz, S., Laskey, M., Doan, R., Liu, X., Ojea, J.A., Goldberg, K.: Dex-net 2.0: Deep learning to plan robust grasps with synthetic point clouds and analytic grasp metrics. Robotics: Science and Systems (2017). https://doi.org/10.15607/RSS.2017.XIII.058

    2017 doi

  19. [27]

    1–8 (2018)

    Dex-net 3.0: Computing robust robot suction grasp targets in point clouds using a new analytic model and deep learning, pp. 1–8 (2018). http:// berkeleyautomation.github.io/dex-net.https://arxiv.org/abs/1709.06670

    2018 arXiv

  20. [28]

    Learning deep policies for robot bin picking by simulating robust grasping sequences, vol. 78, pp. 515–524 (2017). http://berkeleyautomation.github. io/dex-net.http://proceedings.mlr.press/v78/mahler17a.html

    2017

  21. [29]

    Science Robotics 4 (2019)

    Mahler, J., Matl, M., Satish, V., Danielczuk, M., Derose, B., Mckinley, S., Goldberg, K.: Learning ambidextrous robot grasping policies. Science Robotics 4 (2019). https://doi.org/10.1126/scirobotics.aau4984

    2019 doi

  22. [30]

    Kalashnikov, D., Irpan, A., Pastor, P., Ibarz, J., Herzog, A., Jang, E., Quillen, D., Holly, E., Kalakrishnan, M., Vanhoucke, V., Levine, S.: Scalable deep reinforcement learning for vision-based robotic manipulation, vol. 87, pp. 651–673 (2018). https://proceedings.mlr.press/...

    2018 arXiv

  23. [31]

    Applied Intelligence 50 (2020)

    Shoeleh, F., Asadpour, M.: Skill based transfer learning with domain adaptation for continuous reinforcement learning domains. Applied Intelligence 50 (2020). https://doi.org/10.1007/s10489-019-01527-z Preprint sent to Applied Intelligence 26 Learning more with the same effort

    2020 doi

  24. [32]

    27 (2014)

    Goodfellow, I.J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde- Farley, D., Ozair, S., Courville, A., Bengio, Y.: Generative adver- sarial nets, vol. 27 (2014). http://www.github.com/goodfeli/adversarial https://arxiv.org/abs/1406.2661

    2014 arXiv

  25. [33]

    Computing Research Repository (CoRR) (2016)

    Taigman, Y., Polyak, A., Wolf, L.: Unsupervised cross-domain image generation. Computing Research Repository (CoRR) (2016)

    2016

  26. [34]

    12619–12629 (2019)

    James, S., Wohlhart, P., Kalakrishnan, M., Kalashnikov, D., Irpan, A., Ibarz, J., Levine, S., Hadsell, R., Bousmalis, K.: Sim-to-real via sim-to-sim: Data-efficient robotic grasping via randomized-to-canonical adaptation networks, pp. 12619–12629 (2019). https://doi.org/10.110...

    2019

  27. [35]

    2242–2251 (2017)

    Shrivastava, A., Pfister, T., Tuzel, O., Susskind, J., Wang, W., Webb, R.: Learning from simulated and unsupervised images through adversarial training, pp. 2242–2251 (2017). https://doi.org/10.1109/CVPR.2017.241. http://arxiv.org/abs/1612.07828

    2017 arXiv

  28. [36]

    2242–2251 (2017)

    Zhu, J.-Y., Park, T., Isola, P., Efros, A.A.: Unpaired image- to-image translation using cycle-consistent adversarial networks, pp. 2242–2251 (2017). https://doi.org/10.1109/ICCV.2017.244. http://arxiv.org/abs/1703.10593

    2017 arXiv

  29. [37]

    Computing Research Repository (CoRR) (2020)

    Gamrian, S., Goldberg, Y.: Transfer learning for related reinforcement learning tasks via image-to-image translation. Computing Research Repository (CoRR) (2020)

    2020

  30. [38]

    Zhu, Y., Wang, Z., Merel, J., Rusu, A., Erez, T., Cabi, S., Tunyasuvu- nakool, S., Kram´ ar, J., Hadsell, R., de Freitas, N., Heess, N.: Reinforce- ment and imitation learning for diverse visuomotor skills. (2018). https: //doi.org/10.15607/RSS.2018.XIV.009. http://arxiv.org/a...

    2018 arXiv

  31. [39]

    4572– 4580 (2016)

    Ho, J., Ermon, S.: Generative adversarial imitation learning, pp. 4572– 4580 (2016). http://arxiv.org/abs/1606.03476

    2016 arXiv

  32. [40]

    Computing Research Repository (CoRR) (2016)

    Rusu, A.A., Rabinowitz, N.C., Desjardins, G., Soyer, H., Kirkpatrick, J., Kavukcuoglu, K., Pascanu, R., Hadsell, R.: Progressive neural networks. Computing Research Repository (CoRR) (2016)

    2016

  33. [41]

    5026–5033 (2012)

    Todorov, E., Erez, T., Tassa, Y.: Mujoco: A physics engine for model- based control, pp. 5026–5033 (2012). https://doi.org/10.1109/IROS.2012. 6386109

    2012 doi

  34. [42]

    Simonyan, K., Zisserman, A.: Very deep convolutional networks for large- scale image recognition. (2015). http://arxiv.org/abs/1409.1556 Preprint sent to Applied Intelligence Learning more with the same effort 27

    2015 arXiv

  35. [43]

    704–720 (1995)

    Jakobi, N., Husbands, P., Harvey, I.: Noise and the reality gap: The use of simulation in evolutionary robotics, pp. 704–720 (1995). https://doi.org/ 10.5555/645300.648380. https://dl.acm.org/doi/10.5555/645300.648380

    1995

  36. [44]

    Computing Research Repository (CoRR) (2020)

    Mozifian, M., Zhang, A., Pineau, J., Meger, D.: Intervention design for effective sim2real transfer. Computing Research Repository (CoRR) (2020)

    2020

  37. [45]

    Chan, S.C.Y., Fishman, S., Canny, J., Korattikara, A., Guadarrama, S.: Measuring the reliability of reinforcement learning algorithms. (2020). http://arxiv.org/abs/1912.05663

    2020 arXiv

  38. [46]

    Jordan, S.M., Chandak, Y., Cohen, D., Zhang, M., Thomas, P.S.: Evaluating the performance of reinforcement learning algo- rithms. (2020). http://proceedings.mlr.press/v119/jordan20a/jordan20a.pdf https://arxiv.org/abs/2006.16958

    2020 arXiv

  39. [47]

    https://pytorch.org/assets/deep-learning/Deep-Learning- with-PyTorch.pdf

    Stevens, E., Antiga, L., Viehmann, T.: Deep Learning with Py- Torch, (2020). https://pytorch.org/assets/deep-learning/Deep-Learning- with-PyTorch.pdf

    2020

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.