mlsanityData to 7 Oct 2026

MEGA-DAgger: Imitation Learning with Multiple Imperfect Experts

arXiv 2303.00638 ↗First posted 1 Mar 2023cs.LG cs.ROcited by 17

On the map: Offline RL and imitation learning from data → Imitation learning methods

Abstract (arXiv)

Imitation learning has been widely applied to various autonomous systems thanks to recent development in interactive algorithms that address covariate shift and compounding errors induced by traditional approaches like behavior cloning. However, existing interactive imitation learning methods assume access to one perfect expert. Whereas in reality, it is more likely to have multiple imperfect experts instead. In this paper, we propose MEGA-DAgger, a new DAgger variant that is suitable for interactive learning with multiple imperfect experts. First, unsafe demonstrations are filtered while aggregating the training data, so the imperfect demonstrations have little influence when training the novice policy. Next, experts are evaluated and compared on scenarios-specific metrics to resolve the conflicted labels among experts. Through experiments in autonomous racing scenarios, we demonstrate that policy learned using MEGA-DAgger can outperform both experts and policies learned using the state-of-the-art interactive imitation learning algorithms such as Human-Gated DAgger. The supplementary video can be found at urlhttps://youtu.be/wPCht31MHrw.

Key facts. Cited by 17 papers in the map (data through 7 Oct 2026). Placed in Imitation learning methods, one of 13 directions in Offline RL and imitation learning from data. Builds on 16 of its 41 references that are in the map. First posted on arXiv 1 Mar 2023.

Builds on

16 of 41 references.

Closest ancestors · 4

Same direction, newest first

  1. 2021ThriftyDAgger: Budget-Aware Novelty and Risk Gating for Interactive Imitation Learning17 Sep
  2. 2019Imitation Learning from Imperfect Demonstration27 Jan
  3. 2018HG-DAgger: Interactive Imitation Learning with Human Experts5 Oct
  4. 2018EnsembleDAgger: A Bayesian Approach to Safe Imitation Learning22 Jul

Other references · 12

  1. 2016End to End Learning for Self-Driving Carscited by 1,000
  2. 2016Control Barrier Function Based Quadratic Programs for Safety Critical Systemscited by 261
  3. 2019Extrapolating Beyond Suboptimal Demonstrations via Inverse Reinforcement Learning from Observationscited by 256
  4. 2016Query-Efficient Imitation Learning for End-to-End Autonomous Drivingcited by 83
  5. 2022Autonomous Vehicles on the Edge: A Survey on Autonomous Vehicle Racingcited by 56
  6. 2021LazyDAgger: Reducing Context Switching in Interactive Imitation Learningcited by 50
  7. 2017Deep Learning in Robotics: A Review of Recent Researchcited by 42
  8. 2021Learn-to-Race: A Multimodal Control Environment for Autonomous Racingcited by 23
  9. 2021Exploring Imitation Learning for Autonomous Driving with Feedback Synthesizer and Differentiable Rasterizationcited by 22
  10. 2022A Benchmark Comparison of Imitation Learning-based Control Policies for Autonomous Racingcited by 11
  11. 2022Indy Autonomous Challenge -- Autonomous Race Cars at the Handling Limitscited by 10
  12. 2021Enhancing Feasibility and Safety of Nonlinear Model Predictive Control with Discrete-Time Control Barrier Functionscited by 8

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 2; 2024: 5; 2025: 6; 2026: 4.

17 papers.

Sort:

By area:

  1. 2023Provable Guarantees for Generative Behavior Cloning: Bridging Low-Level Stability and High-Level Behaviorcited by 33
  2. 2025The Pitfalls of Imitation Learning when Actions are Continuouscited by 25
  3. 2024Do We Really Need a Complex Agent System? Distill Embodied Agent into a Single Modelcited by 19
  4. 2024Autonomous Navigation of Unmanned Vehicle Through Deep Reinforcement Learningcited by 12
  5. 2025Dynamic Rank Adjustment in Diffusion Policies for Efficient and Flexible Trainingcited by 6
  6. 2025PRISM-DP: Spatial Pose-based Observations for Diffusion-Policies via Segmentation, Mesh Generation, and Pose Trackingcited by 5
  7. 2026DexHiL: A Human-in-the-Loop Framework for Vision-Language-Action Model Post-Training in Dexterous Manipulationcited by 5
  8. 2023IIFL: Implicit Interactive Fleet Learning from Heterogeneous Human Supervisorscited by 4
  9. 2024Developing Path Planning with Behavioral Cloning and Proximal Policy Optimization for Path-Tracking and Static Obstacle Nudgingcited by 3
  10. 2024Optimizing Active Perception for Learning Simultaneous Viewpoint Selection and Manipulation with Diffusion Policycited by 3
  11. 2025Interactive Imitation Learning for Dexterous Robotic Manipulation: Challenges and Perspectives -- A Surveycited by 3
  12. 2025Arnold: A multi-task, multi-embodiment muscle transformer policycited by 2
5 more
  1. 2024NeRF-Aug: Data Augmentation for Robotics with Neural Radiance Fieldscited by 1
  2. 2025GrOMP: Grasped Object Manifold Projection for Multimodal Imitation Learning of Manipulationcited by 1
  3. 2026Imitation Learning for Combinatorial Optimisation under Uncertaintycited by 0
  4. 2026Oracle-Guided Soft Shielding for Safe Move Prediction in Chesscited by 0
  5. 2026Artificial Intelligence for Modeling and Simulation of Mixed Automated and Human Trafficcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2023RLIF: Interactive Imitation Learning as Reinforcement Learning
  2. 2024Efficient Active Imitation Learning with Random Network Distillation
  3. 2023Agnostic Interactive Imitation Learning: New Theory and Practical Algorithms
  4. 2022Imitation Learning by Estimating Expertise of Demonstrators
  5. 2024Interactive and Hybrid Imitation Learning: Provably Beating Behavior Cloning
  6. 2025Beyond-Expert Performance with Limited Demonstrations: Efficient Imitation Learning with Double Exploration
  7. 2021Imitation Learning: Progress, Taxonomies and Challenges
  8. 2023A Survey of Imitation Learning: Algorithms, Recent Developments, and Challenges