Assessing Transferability From Simulation to Reality for Reinforcement Learning
Open accessThe DOI matches its Crossref record; 19 of the 19 references it cites were found in a registry or an index.
Abstract
Learning robot control policies from physics simulations is of great interest to the robotics community as it may render the learning process faster, cheaper, and safer by alleviating the need for expensive real-world experiments. However, the direct transfer of learned behavior from simulation to reality is a major challenge. Optimizing a policy on a slightly faulty simulator can easily lead to the maximization of the 'Simulation Optimization Bias' (SOB). In this case, the optimizer exploits modeling errors of the simulator such that the resulting behavior can potentially damage the robot. We tackle this challenge by applying domain randomization, i.e., randomizing the parameters of the physics simulations during learning. We propose an algorithm called Simulation-based Policy Optimization with Transferability Assessment (SPOTA) which uses an estimator of the SOB to formulate a stopping criterion for training. The introduced estimator quantifies the over-fitting to the set of domains experienced while training. Our experimental results on two different second order nonlinear systems show that the new simulation-based policy search algorithm is able to learn a control policy exclusively from a randomized simulator, which can be applied directly to real systems without any additional training.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
Proximal Policy Optimization Algorithms
Schulman, J., Wolski, F., Dhariwal, P., Radford, A., & Klimov, O. (2017) · arXiv · doi:10.48550/arxiv.1707.06347 · Cited by 10,625
Domain randomization for transferring deep neural networks from simulation to the real world
Tobin, J., Fong, R., Ray, A., Schneider, J., Zaremba, W., & Abbeel, P. (2017) · 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2017.8202133 · Cited by 3,124
Bootstrap confidence intervals
DiCiccio, T. J., & Efron, B. (1996) · Statistical Science · doi:10.1214/ss/1032280214 · Cited by 2,286
Sim-to-Real Transfer of Robotic Control with Dynamics Randomization
Peng, X. B., Andrychowicz, M., Zaremba, W., & Abbeel, P. (2018) · 2018 IEEE International Conference on Robotics and Automation (ICRA) · doi:10.1109/icra.2018.8460528 · Cited by 1,309 · cited 4×
“From the listed approaches [10, 33, 15] were able to cross the reality gap without acquiring samples from the real world.”
BackgroundMethodInfluentialResilient Machines Through Continuous Self-Modeling
Bongard, J., Zykov, V., & Lipson, H. (2006) · Science · doi:10.1126/science.1133687 · Cited by 709
Noise and the reality gap: The use of simulation in evolutionary robotics
Jakobi, N., Husbands, P., & Harvey, I. (1995) · Lecture Notes in Computer Science · doi:10.1007/3-540-59496-5_337 · Cited by 631
CAD2RL: Real Single-Image Flight Without a Single Real Image
Sadeghi, F., & Levine, S. (2017) · Robotics: Science and Systems XIII · doi:10.15607/rss.2017.xiii.034 · Cited by 618
Model-Ensemble Trust-Region Policy Optimization
Kurutach, T., Clavera, I., Duan, Y., Tamar, A., & Abbeel, P. (2018) · arXiv · doi:10.48550/arxiv.1802.10592 · Cited by 537
Robust adversarial reinforcement learning
Pinto, L., Davidson, J., Sukthankar, R., & Gupta, A. (2017) · arXiv · doi:10.48550/arxiv.1703.02702 · Cited by 507 · cited 2×
“[29] introduced the idea of a second agent whose goal is to hinder the first agent from fulfilling its task.”
BackgroundSim-to-Real Robot Learning from Pixels with Progressive Nets
Rusu, A. A., Vecerik, M., Rothörl, T., Heess, N., Pascanu, R., & Hadsell, R. (2016) · arXiv · doi:10.48550/arxiv.1610.04286 · Cited by 385
Learning Dexterous In-Hand Manipulation
OpenAI, Andrychowicz, M., Baker, B., Chociej, M., Jozefowicz, R., McGrew, B., Pachocki, J., Petron, A., Plappert, M., Powell, G., et al. (2018) · arXiv · doi:10.48550/arxiv.1808.00177 · Cited by 248
Sim-to-real reinforcement learning for deformable object manipulation
Matas, J., James, S., & Davison, A. J. (2018) · arXiv · doi:10.48550/arxiv.1806.07851 · Cited by 241
Preparing for the unknown: Learning a universal policy with online system identification
Yu, W., Tan, J., Karen Liu, C., & Turk, G. (2017) · Robotics: Science and Systems XIII · doi:10.15607/rss.2017.xiii.048 · Cited by 237
Preparing for the Unknown: Learning a Universal Policy with Online System Identification
Yu, W., Tan, J., Liu, C. K., & Turk, G. (2017) · arXiv · doi:10.48550/arxiv.1702.02453 · Cited by 237 · cited 2×
“As done in [10, 11, 14, 15, 16] we use the domain parameter distribution as a prior which ensures the physical plausibility of each parameter.”
BackgroundMethodEnsemble-CIO: Full-body dynamic motion planning that transfers to physical humanoids
Mordatch, I., Lowrey, K., & Todorov, E. (2015) · 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2015.7354126 · Cited by 135
Adversarially Robust Policy Learning: Active construction of physically-plausible perturbations
Mandlekar, A., Zhu, Y., Garg, A., Fei-Fei, L., & Savarese, S. (2017) · 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2017.8206245 · Cited by 116
Towards Generalization and Simplicity in Continuous Control
Rajeswaran, A., Lowrey, K., Todorov, E., & Kakade, S. (2017) · arXiv · doi:10.48550/arxiv.1703.02660 · Cited by 94
Is optimization optimistically biased?
Hobbs, B. F., & Hepenstal, A. (1989) · Water Resources Research · doi:10.1029/wr025i002p00152 · Cited by 33
Cited by
Transferring policy of deep reinforcement learning from simulation to reality for robotics
Ju, H., Juan, R., Gomez, R., Nakamura, K., & Li, G. (2022) · Nature Machine Intelligence · doi:10.1038/s42256-022-00573-6 · Cited by 147
Robot Learning From Randomized Simulations: A Review
Muratore, F., Ramos, F., Turk, G., Yu, W., Gienger, M., & Peters, J. (2022) · Frontiers in Robotics and AI · doi:10.3389/frobt.2022.799893 · Cited by 128
“Moreover, this approach was employed as a baseline by Muratore et al. (2021b), who introduced a measure for the inter-domain transferability of controllers together with a risk-neutral randomization scheme.”
MethodSelf-citationData-Efficient Domain Randomization With Bayesian Optimization
Muratore, F., Eilers, C., Gienger, M., & Peters, J. (2021) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2021.3052391 · Cited by 77
“In this case, we observe a simulation optimization bias, which leads to an overestimation of the policy’s performance [2].”
BackgroundMethodInfluentialSelf-citationData-efficient Domain Randomization with Bayesian Optimization
Muratore, F., Eilers, C., Gienger, M., & Peters, J. (2020) · arXiv · doi:10.48550/arxiv.2003.02471 · Cited by 77
Self-citationProcedural Content Generation: Better Benchmarks for Transfer Reinforcement Learning
Muller-Brockhausen, M., Preuss, M., & Plaat, A. (2021) · 2021 IEEE Conference on Games (CoG) · doi:10.1109/cog52621.2021.9619000 · Cited by 11
“To narrow the gap between simulation and reality, in many papers, noise is applied for regularization or smoothing, e.g., Gaussian Noise [62], uniform random noise [63], the Simulation Optimization Bias (SOB) [64]).”
Background
Metadata
- Authors
- Muratore, F., Gienger, M., & Peters, J.
- Venue
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- Published
- 2021-04-01
- Volume
- 43
- Issue
- 4
- Pages
- 1172-1183
- ISSN
- 0162-8828 (ISSN-L)
- 2160-9292
- 1939-3539
- OpenAlex
- W2960705509
- Cited by
- 65
- References
- 19
- FWCI
- 2.88
- Type
- journal article
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.