Teleoperation

DexPilot: Vision Based Teleoperation of Dexterous Robotic Hand-Arm System - Handa, Ankur, et al. (2020 ICRA)

3D Representation

• 3D Diffuser Actor: Multi-task 3D Robot Manipulation with Iterative Error Feedback. - OpenReview
• [IROS 2024] RISE: 3D Perception Makes Real-World Robot Imitation Simple and Effective. - Website
• [RSS 2024] Dexcap: Scalable and portable mocap data collection system for dexterous manipulation - Website
• [RSS 2024] 3d Diffusion Policy - Website

Imitation Learning Theory

1. On the Performance of Empirical Risk Minimization with Smoothed Data
2. On the Imitation of Non-Markovian Demonstrations: From Low-Level Stability to High Level Planning
3. Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning
4. How Generalizable Is My Behavior Cloning Policy? A Statistical Approach to Trustworthy Performance Evaluation
5. Transcendence: Generative Models Can Outperform The Experts That Train Them
6. Scaling Laws for Imitation Learning in Single-Agent Games
7. How to Leverage Diverse Demonstrations in Offline Imitation Learning

Data Quality

1. Behavioral Cloning from Noisy Demonstrations
2. Imitation Learning from Purified Demonstrations
3. DEMODICE: Offline Imitation Learning with Supplementary Imperfect Demonstrations
4. Data Quality in Imitation Learning
5. HYDRA: Hybrid Robot Actions for Imitation Learning
6. VIOLA: Imitation Learning for Vision-Based Manipulation with Object Proposal Priors

Imitation Learning Generalization

1. Efficient Data Collection for Robotic Manipulation via Compositional Generalization

Scale Imitation Learning

1. Re-Mix: Optimizing Data Mixtures for Large Scale Imitation Learning

Imitation Learning + RL

1. Imitation Bootstrapped Reinforcement Learning