Deep visual foresight for planning robot motion
Abstract
A key challenge in scaling up robot learning to many skills and environments is removing the need for human supervision, so that robots can collect their own data and improve their own performance without being limited by the cost of requesting human feedback. Model-based reinforcement learning holds the promise of enabling an agent to learn to predict the effects of its actions, which could provide flexible predictive models for a wide range of tasks and environments, without detailed human supervision. We develop a method for combining deep action-conditioned video prediction models with model-predictive control that uses entirely unlabeled training data. Our approach does not require a calibrated camera, an instrumented training set-up, nor precise sensing and actuation. Our results show that our method enables a real robot to perform nonprehensile manipulation - pushing objects - and can handle novel objects not seen during training.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The generative motion planning approach aims to enhance generalization to unseen environments by generating intermediate visual sub-goals rather than explicit action sequences.
“The generative motion planning approach (Black et al., 2023; Du et al., 2024; Finn and Levine, 2017; Ko et al., 2024; Zhou et al., 2024) aims to enhance generalization to unseen environments by generating intermediate visual sub-goals rather than explicit action sequences.”
From Cosmos World Foundation Model Platform for Physical AI (2025) · cites this paper as Finn and Levine, 2017
Dreamer learns a model of the environment and improves its behavior by imagining future scenarios.
“Dreamer learns a model of the environment and improves its behavior by imagining future scenarios.”
From Mastering Diverse Domains through World Models (2023) · cites this paper as [15]
Cited by
Sorted by citations
On the Opportunities and Risks of Foundation Models
Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S., Bohg, J., Bosselut, A., Brunskill, E., et al. (2021) · arXiv · doi:10.48550/arxiv.2108.07258 · Cited by 2,676 · see citing works
Self-citationLearning Synergies Between Pushing and Grasping with Self-Supervised Deep Reinforcement Learning
Zeng, A., Song, S., Welker, S., Lee, J., Rodriguez, A., & Funkhouser, T. (2018) · 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2018.8593986 · Cited by 606 · see citing works
How to train your robot with deep reinforcement learning: Lessons we have learned
Ibarz, J., Tan, J., Finn, C., Kalakrishnan, M., Pastor, P., & Levine, S. (2021) · The International Journal of Robotics Research · doi:10.1177/0278364920987859 · Cited by 570 · see citing works
Self-citationLee, M. A., Zhu, Y., Srinivasan, K., Shah, P., Savarese, S., Fei-Fei, L., Garg, A., & Bohg, J. (2019) · 2019 International Conference on Robotics and Automation (ICRA) · doi:10.1109/icra.2019.8793485 · Cited by 349 · see citing works
RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control
Brohan, A., Brown, N., Carbajal, J., Chebotar, Y., Chen, X., Choromanski, K., Ding, T., Driess, D., Dubey, A., Finn, C., et al. (2023) · arXiv · doi:10.48550/arxiv.2307.15818 · Cited by 276 · see citing works
“We refer to this category of models as vision-language-action (VLA) models.”
MethodSelf-citationStochastic latent actor–critic: Deep reinforcement learning with a latent variable model
Lee, A. X., Nagabandi, A., Abbeel, P., & Levine, S. (2019) · arXiv · doi:10.48550/arxiv.1907.00953 · Cited by 159 · see citing works
Self-citationMastering diverse control tasks through world models
Hafner, D., Pasukonis, J., Ba, J., & Lillicrap, T. (2025) · Nature · doi:10.1038/s41586-025-08744-2 · Cited by 189 · see citing works
“Learning dynamics models of unknown environments and using them for reinforcement learning 15 has been explored in early algorithms, such as PILCO, E2C and Visual Foresight 16 .”
Grounded Human-Object Interaction Hotspots From Video
Nagarajan, T., Feichtenhofer, C., & Grauman, K. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00878 · Cited by 147 · see citing works
Octo: An Open-Source Generalist Robot Policy
Ghosh, D., Walke, H., Pertsch, K., Black, K., Mees, O., Dasari, S., Hejna, J., Kreiman, T., Xu, C., Luo, J., et al. (2024) · Robotics: Science and Systems XX · doi:10.15607/rss.2024.xx.090 · Cited by 129 · see citing works
Self-citationMastering Diverse Domains through World Models
Hafner, D., Pasukonis, J., Ba, J., & Lillicrap, T. (2023) · arXiv · doi:10.48550/arxiv.2301.04104 · Cited by 124 · see citing works
“The algorithm is based on the idea of learning a world model that equips the agent with rich perception and the ability to imagine the future 14,15,16 .”
MethodVisuoSpatial Foresight for Multi-Step, Multi-Task Fabric Manipulation
Hoque, R., Seita, D., Balakrishna, A., Ganapathi, A., Tanwani, A., Jamali, N., Yamane, K., Iba, S., & Goldberg, K. (2020) · Robotics: Science and Systems XVI · doi:10.15607/rss.2020.xvi.034 · Cited by 74 · see citing works
Understanding World or Predicting Future? A Comprehensive Survey of World Models
Ding, J., Zhang, Y., Shang, Y., Zhang, Y., Zong, Z., Feng, J., Yuan, Y., Su, H., Li, N., Sukiennik, N., et al. (2026) · ACM Computing Surveys · doi:10.1145/3746449 · Cited by 58 · see citing works
“This shift provides agents with richer, more varied training experiences, improving their adaptability and generalization in complex, unpredictable real-world situations.”
Yuan, W., Hang, K., Kragic, D., Wang, M. Y., & Stork, J. A. (2019) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2019.06.007 · Cited by 54 · see citing works
DayDreamer: World Models for Physical Robot Learning
Wu, P., Escontrela, A., Hafner, D., Goldberg, K., & Abbeel, P. (2022) · arXiv · doi:10.48550/arxiv.2206.14176 · Cited by 76 · see citing works
Visuospatial foresight for physical sequential fabric manipulation
Hoque, R., Seita, D., Balakrishna, A., Ganapathi, A., Tanwani, A. K., Jamali, N., Yamane, K., Iba, S., & Goldberg, K. (2022) · Autonomous Robots · doi:10.1007/s10514-021-10001-0 · Cited by 45 · see citing works
Kadi, H. A., & Terzić, K. (2023) · Sensors · doi:10.3390/s23052389 · Cited by 19 · see citing works
“The most direct way to reduce the compounding error in planning is to (1) use multi-step prediction in model training [171,172] and (2) have a short planning horizon .”
BackgroundCosmos World Foundation Model Platform for Physical AI
NVIDIA, :, Agarwal, N., Ali, A., Bala, M., Balaji, Y., Barker, E., Cai, T., Chattopadhyay, P., Chen, Y., et al. (2025) · arXiv · doi:10.48550/arxiv.2501.03575 · Cited by 13 · see citing works
Tung, H.-Y., Ding, M., Chen, Z., Bear, D., Gan, C., Tenenbaum, J., Yamins, D., Fan, J., & Smith, K. (2023) · Advances in Neural Information Processing Systems 36 · doi:10.52202/075280-2929 · Cited by 8 · see citing works
IRASim: A Fine-Grained World Model for Robot Manipulation
Zhu, F., Wu, H., Guo, S., Liu, Y., Cheang, C., & Kong, T. (2024) · arXiv · doi:10.48550/arxiv.2406.14540 · Cited by 6 · see citing works
V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning
Assran, M., Bardes, A., Fan, D., Garrido, Q., Howes, R., Mojtaba, Komeili, Muckley, M., Rizvi, A., Roberts, C., et al. (2025) · arXiv · doi:10.48550/arxiv.2506.09985 · Cited by 8 · see citing works
“Following many previous works in goal-conditioned robot manipulation (Finn and Levine, 2017; Lynch et al., 2020; Chebotar et al., 2021; Jang et al., 2022; Liu et al., 2022; Gupta et al., 2022), our current formulation of the optimization target assumes that we have access to visual goals.”
Procedural Generalization by Planning with Self-Supervised World Models
Anand, A., Walker, J., Li, Y., Vértes, E., Schrittwieser, J., Ozair, S., Weber, T., & Hamrick, J. B. (2021) · arXiv · doi:10.48550/arxiv.2111.01587 · Cited by 5 · see citing works
- Authors
- Finn, C., & Levine, S.
- Affiliations
- Google (United States) (US)
- Venue
- 2017 IEEE International Conference on Robotics and Automation (ICRA)
- Publisher
- IEEE
- Published
- 2017-05
- Pages
- 2786-2793
- ISSN
- 1050-4729 (ISSN-L)
- OpenAlex
- W2528489519
- Cited by
- 599
- References
- 0
- Citations by year
- 2026: 15 · 2025: 32 · 2024: 33 · 2023: 63 · 2022: 60 · 2021: 99 · 2020: 119 · 2019: 94 · 2018: 67 · 2017: 17
- FWCI
- 37.05
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- Proceedings - IEEE International Conference on Robotics and Automation/Proceedings (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Engineering › Control and Systems Engineering › Robot Manipulation and Learning
- Physical Sciences › Computer Science › Artificial Intelligence › AI-based Problem Solving and Planning
- Keywords
- robot motion planning, model-based reinforcement learning, model predictive control, unsupervised learning, non-prehensile manipulation, novel object generalization, robot learning
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
