mlsanityData to 14 Oct 2026

Generalization Guarantees for Imitation Learning

arXiv 2008.01913 ↗First posted 5 Aug 2020cs.RO cs.LG cs.SYcited by 2

On the map: Offline RL and imitation learning from data → Robot manipulation imitation learning

Abstract (arXiv)

Control policies from imitation learning can often fail to generalize to novel environments due to imperfect demonstrations or the inability of imitation learning algorithms to accurately infer the expert's policies. In this paper, we present rigorous generalization guarantees for imitation learning by leveraging the Probably Approximately Correct (PAC)-Bayes framework to provide upper bounds on the expected cost of policies in novel environments. We propose a two-stage training method where a latent policy distribution is first embedded with multi-modal expert behavior using a conditional variational autoencoder, and then "fine-tuned" in new training environments to explicitly optimize the generalization bound. We demonstrate strong generalization bounds and their tightness relative to empirical performance in simulation for (i) grasping diverse mugs, (ii) planar pushing with visual feedback, and (iii) vision-based indoor navigation, as well as through hardware experiments for the two manipulation tasks.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in Robot manipulation imitation learning, one of 13 directions in Offline RL and imitation learning from data. Builds on 25 of its 39 references that are in the map. First posted on arXiv 5 Aug 2020.

Builds on

25 of 39 references.

Closest ancestors · 4

Same direction, newest first

  1. 2020Learning to Generalize Across Long-Horizon Tasks from Human Demonstrations13 Mar
  2. 2019Relay Policy Learning: Solving Long-Horizon Tasks via Imitation and Reinforcement Learning25 Oct
  3. 2019Self-Supervised Correspondence in Visuomotor Policy Learning16 Sep
  4. 2018Reinforcement and Imitation Learning for Diverse Visuomotor Skills26 Feb

Other references · 21

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2015ShapeNet: An Information-Rich 3D Model Repositorycited by 2,897
  3. 2017Evolution Strategies as a Scalable Alternative to Reinforcement Learningcited by 574
  4. 2017Computing Nonvacuous Generalization Bounds for Deep (Stochastic) Neural Networks with Many More Parameters than Training Datacited by 533
  5. 2014Semi-Supervised Learning with Deep Generative Modelscited by 496
  6. 2017End-to-end Driving via Conditional Imitation Learningcited by 485
  7. 2018An Algorithmic Perspective on Imitation Learningcited by 459
  8. 2018ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worstcited by 397
  9. 2017A PAC-Bayesian Approach to Spectrally-Normalized Margin Bounds for Neural Networkscited by 314
  10. 2018Stronger generalization bounds for deep nets via a compression approachcited by 307
  11. 2017Deep Imitation Learning for Complex Manipulation Tasks from Virtual Reality Teleoperationcited by 289
  12. 2017DART: Noise Injection for Robust Imitation Learningcited by 191
9 more
  1. 2017Vision-Based Multi-Task Manipulation for Inexpensive Robots Using End-To-End Learning from Demonstrationcited by 100
  2. 2019Interactive Gibson Benchmark (iGibson 0.5): A Benchmark for Interactive Navigation in Cluttered Environmentscited by 90
  3. 2017Robust Imitation of Diverse Behaviorscited by 86
  4. 2018Reinforcement Learning from Imperfect Demonstrationscited by 75
  5. 2017Multi-Modal Imitation Learning from Unstructured Demonstrations using Generative Adversarial Netscited by 64
  6. 2018PAC-Bayes Control: Learning Policies that Provably Generalize to Novel Environmentscited by 22
  7. 2017Simultaneous Policy Learning and Latent State Inference for Imitating Driver Behaviorcited by 16
  8. 2020Probably Approximately Correct Vision-Based Planning using Motion Primitivescited by 16
  9. 2019Learning a Multi-Modal Policy via Imitating Demonstrations with Mixed Behaviorscited by 9

Built on by

Cited by year

Citing papers per year: 2021: 1; 2022: 0; 2023: 0; 2024: 1.

2 papers.

  1. 2024Deep Generative Models for Offline Policy Learning: Tutorial, Survey, and Perspectives on Future Directionscited by 20
  2. 2021Contrastively Learning Visual Attention as Affordance Cues from Demonstrations for Robotic Graspingcited by 7

Similar, not cited

Close in wording, no citation either way

  1. 2025Generalization Capability for Imitation Learning
  2. 2023Decomposing the Generalization Gap in Imitation Learning for Visual Robotic Manipulation
  3. 2021Bayesian Disturbance Injection: Robust Imitation Learning of Flexible Policies
  4. 2023Generalizable Imitation Learning Through Pre-Trained Representations
  5. 2026Beyond the Majority: Long-tail Imitation Learning for Robotic Manipulation
  6. 2026An Empirical Study on What Matters for Viewpoint-Generalizable Policies in Visual Imitation Learning
  7. 2025Improving Generalization Ability of Robotic Imitation Learning by Resolving Causal Confusion in Observations
  8. 2023Agnostic Interactive Imitation Learning: New Theory and Practical Algorithms