# DART: Noise Injection for Robust Imitation Learning

arXiv: https://arxiv.org/abs/1703.09327 · HTML: https://mlsanity.com/paper/1703.09327/

## Key facts

- Cited by 191 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 27 Mar 2017.

## Built on by (most cited first, 30 of 191)

- [Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware](https://mlsanity.com/paper/2304.13705/) (2023)
- [Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion](https://mlsanity.com/paper/2407.01392/) (2024)
- [BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learning](https://mlsanity.com/paper/2202.02005/) (2022)
- [RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation](https://mlsanity.com/paper/2506.18088/) (2025)
- [π*0.6: a VLA That Learns From Experience](https://mlsanity.com/paper/2511.14759/) (2025)
- [Causal Confusion in Imitation Learning](https://mlsanity.com/paper/1905.11979/) (2019)
- [Imitation Learning as f-Divergence Minimization](https://mlsanity.com/paper/1905.12888/) (2019)
- [Urban Driver: Learning to Drive from Real-world Demonstrations Using Policy Gradients](https://mlsanity.com/paper/2109.13333/) (2021)
- [Learning to drive from a world on rails](https://mlsanity.com/paper/2105.00636/) (2021)
- [Human-in-the-Loop Imitation Learning using Remote Teleoperation](https://mlsanity.com/paper/2012.06733/) (2020)
- [ThriftyDAgger: Budget-Aware Novelty and Risk Gating for Interactive Imitation Learning](https://mlsanity.com/paper/2109.08273/) (2021)
- [S4RL: Surprisingly Simple Self-Supervision for Offline Reinforcement Learning](https://mlsanity.com/paper/2103.06326/) (2021)
- [Data Quality in Imitation Learning](https://mlsanity.com/paper/2306.02437/) (2023)
- [Safe Imitation Learning via Fast Bayesian Reward Inference from Preferences](https://mlsanity.com/paper/2002.09089/) (2020)
- [Fighting Copycat Agents in Behavioral Cloning from Observation Histories](https://mlsanity.com/paper/2010.14876/) (2020)
- [Symphony: Learning Realistic and Diverse Agents for Autonomous Driving Simulation](https://mlsanity.com/paper/2205.03195/) (2022)
- [Behavior Retrieval: Few-Shot Imitation Learning by Querying Unlabeled Datasets](https://mlsanity.com/paper/2304.08742/) (2023)
- [NeRF in the Palm of Your Hand: Corrective Augmentation for Robotics via Novel-View Synthesis](https://mlsanity.com/paper/2301.08556/) (2023)
- [Motion Policy Networks](https://mlsanity.com/paper/2210.12209/) (2022)
- [HYDRA: Hybrid Robot Actions for Imitation Learning](https://mlsanity.com/paper/2306.17237/) (2023)
- [Bidirectional Decoding: Improving Action Chunking via Guided Test-Time Sampling](https://mlsanity.com/paper/2408.17355/) (2024)
- [LazyDAgger: Reducing Context Switching in Interactive Imitation Learning](https://mlsanity.com/paper/2104.00053/) (2021)
- [DiffuseLoco: Real-Time Legged Locomotion Control with Diffusion from Offline Datasets](https://mlsanity.com/paper/2404.19264/) (2024)
- [Re-Mix: Optimizing Data Mixtures for Large Scale Imitation Learning](https://mlsanity.com/paper/2408.14037/) (2024)
- [Overcoming Model Bias for Robust Offline Deep Reinforcement Learning](https://mlsanity.com/paper/2008.05533/) (2020)
- [Grasping with Chopsticks: Combating Covariate Shift in Model-free Imitation Learning for Fine Manipulation](https://mlsanity.com/paper/2011.06719/) (2020)
- [On the Sample Complexity of Stability Constrained Imitation Learning](https://mlsanity.com/paper/2102.09161/) (2021)
- [JUICER: Data-Efficient Imitation Learning for Robotic Assembly](https://mlsanity.com/paper/2404.03729/) (2024)
- [Efficient Data Collection for Robotic Manipulation via Compositional Generalization](https://mlsanity.com/paper/2403.05110/) (2024)
- [Much Ado About Noising: Dispelling the Myths of Generative Robotic Control](https://mlsanity.com/paper/2512.01809/) (2025)

## Abstract (arXiv)

One approach to Imitation Learning is Behavior Cloning, in which a robot observes a supervisor and infers a control policy. A known problem with this "off-policy" approach is that the robot's errors compound when drifting away from the supervisor's demonstrations. On-policy, techniques alleviate this by iteratively collecting corrective actions for the current robot policy. However, these techniques can be tedious for human supervisors, add significant computation burden, and may visit dangerous states during training. We propose an off-policy approach that injects noise into the supervisor's policy while demonstrating. This forces the supervisor to demonstrate how to recover from errors. We propose a new algorithm, DART (Disturbances for Augmenting Robot Trajectories), that collects demonstrations with injected noise, and optimizes the noise level to approximate the error of the robot's trained policy during data collection. We compare DART with DAgger and Behavior Cloning in two domains: in simulation with an algorithmic supervisor on the MuJoCo tasks (Walker, Humanoid, Hopper, Half-Cheetah) and in physical experiments with human supervisors training a Toyota HSR robot to perform grasping in clutter. For high dimensional tasks like Humanoid, DART can be up to 3x faster in computation time and only decreases the supervisor's cumulative reward by 5% during training, whereas DAgger executes policies that have 80% less cumulative reward than the supervisor. On the grasping in clutter task, DART obtains on average a 62% performance increase over Behavior Cloning.