mlsanityData to 7 Oct 2026

Efficient Training of Generalizable Visuomotor Policies via Control-Aware Augmentation

arXiv 2401.09258 ↗First posted 17 Jan 2024cs.RO cs.CVcited by 1

Not on the map. Why?

Abstract (arXiv)

Improving generalization is one key challenge in embodied AI, where obtaining large-scale datasets across diverse scenarios is costly. Traditional weak augmentations, such as cropping and flipping, are insufficient for improving a model's performance in new environments. Existing data augmentation methods often disrupt task-relevant information in images, potentially degrading performance. To overcome these challenges, we introduce EAGLE, an efficient training framework for generalizable visuomotor policies that improves upon existing methods by (1) enhancing generalization by applying augmentation only to control-related regions identified through a self-supervised control-aware mask and (2) improving training stability and efficiency by distilling knowledge from an expert to a visuomotor student policy, which is then deployed to unseen environments without further fine-tuning. Comprehensive experiments on three domains, including the DMControl Generalization Benchmark, the enhanced Robot Manipulation Distraction Benchmark, and a long-sequential drawer-opening task, validate the effectiveness of our method.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 48 of its 60 references that are in the map. First posted on arXiv 17 Jan 2024.

Builds on

48 of 60 references.

References · 48

  1. 2017Attention Is All You Needcited by 59,932
  2. 2023Segment Anythingcited by 8,158
  3. 2018Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actorcited by 4,635
  4. 2017CARLA: An Open Urban Driving Simulatorcited by 2,431
  5. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
  6. 2021Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learningcited by 882
  7. 2018QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulationcited by 730
  8. 2020Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixelscited by 573
  9. 2020Learning Quadrupedal Locomotion over Challenging Terraincited by 494
  10. 2019Leveraging Procedural Generation to Benchmark Reinforcement Learningcited by 478
  11. 2020Reinforcement Learning with Augmented Datacited by 452
  12. 2020Learning Invariant Representations for Reinforcement Learning without Reconstructioncited by 424
36 more
  1. 2019Learning by Cheatingcited by 361
  2. 2021Mastering Visual Continuous Control: Improved Data-Augmented Reinforcement Learningcited by 359
  3. 2018Quantifying Generalization in Reinforcement Learningcited by 345
  4. 2020Data-Efficient Reinforcement Learning with Self-Predictive Representationscited by 309
  5. 2019Improving Sample Efficiency in Model-Free Reinforcement Learning from Imagescited by 268
  6. 2017Revisiting the Arcade Learning Environment: Evaluation Protocols and Open Problems for General Agentscited by 267
  7. 2017Asymmetric Actor Critic for Image-Based Robot Learningcited by 202
  8. 2018A Study on Overfitting in Deep Reinforcement Learningcited by 170
  9. 2020RIDE: Rewarding Impact-Driven Exploration for Procedurally-Generated Environmentscited by 135
  10. 2018Assessing Generalization in Deep Reinforcement Learningcited by 132
  11. 2020Generalization in Reinforcement Learning by Soft Data Augmentationcited by 132
  12. 2019Network Randomization: A Simple Technique for Generalization in Deep Reinforcement Learningcited by 102
  13. 2019Generalization in Reinforcement Learning with Selective Noise Injection and Information Bottleneckcited by 101
  14. 2021Stabilizing Deep Q-Learning with ConvNets and Vision Transformers under Data Augmentationcited by 101
  15. 2021The Distracting Control Suite -- A Challenging Benchmark for Reinforcement Learning from Pixelscited by 97
  16. 2018A Dissection of Overfitting and Generalization in Continuous Reinforcement Learningcited by 86
  17. 2019Distilling Policy Distillationcited by 81
  18. 2018Illuminating Generalization in Deep Reinforcement Learning through Procedural Level Generationcited by 80
  19. 2017Towards Generalization and Simplicity in Continuous Controlcited by 65
  20. 2022Watch and Match: Supercharging Imitation with Regularized Optimal Transportcited by 64
  21. 2018Data augmentation instead of explicit regularizationcited by 62
  22. 2021Learning Task Informed Abstractionscited by 58
  23. 2022Pre-Trained Image Encoder for Generalizable Visual Reinforcement Learningcited by 58
  24. 2022VRL3: A Data-Driven Framework for Visual Deep Reinforcement Learningcited by 51
  25. 2021Towards Robust Bisimulation Metric Learningcited by 49
  26. 2022Look Closer: Bridging Egocentric and Third-Person Views with Transformers for Robotic Manipulationcited by 49
  27. 2021SECANT: Self-Expert Cloning for Zero-Shot Generalization of Visual Policiescited by 42
  28. 2021DRIBO: Robust Deep Reinforcement Learning via Multi-View Information Bottleneckcited by 31
  29. 2021Unsupervised Visual Attention and Invariance for Reinforcement Learningcited by 30
  30. 2019On the Generalization Gap in Reparameterizable Reinforcement Learningcited by 21
  31. 2022Don't Touch What Matters: Task-Aware Lipschitz Data Augmentation for Visual Reinforcement Learningcited by 21
  32. 2023MoVie: Visual Model-Based Policy Adaptation for View Generalizationcited by 15
  33. 2023Robust Representation Learning by Clustering with Bisimulation Metrics for Visual Reinforcement Learning with Distractionscited by 13
  34. 2023Normalization Enhances Generalization in Visual Reinforcement Learningcited by 13
  35. 2020Domain Adaptation Through Task Distillationcited by 11
  36. 2022Data augmentation for efficient learning from parametric expertscited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026See Less, Specify More: Visual Evidence Budgets for Generalizable VLAscited by 1