TY - GEN
T1 - State-Only Imitation Learning for Dexterous Manipulation
AU - Radosavovic, Ilija
AU - Wang, Xiaolong
AU - Pinto, Lerrel
AU - Malik, Jitendra
N1 - Publisher Copyright:
© 2021 IEEE.
PY - 2021
Y1 - 2021
N2 - Modern model-free reinforcement learning methods have recently demonstrated impressive results on a number of problems. However, complex domains like dexterous manipulation remain a challenge due to the high sample complexity. To address this, current approaches employ expert demonstrations in the form of state-action pairs, which are difficult to obtain for real-world settings such as learning from videos. In this paper, we move toward a more realistic setting and explore state-only imitation learning. To tackle this setting, we train an inverse dynamics model and use it to predict actions for state-only demonstrations. The inverse dynamics model and the policy are trained jointly. Our method performs on par with state-action approaches and considerably outperforms RL alone. By not relying on expert actions, we are able to learn from demonstrations with different dynamics, morphologies, and objects. Videos available on the { text{project page}}.
AB - Modern model-free reinforcement learning methods have recently demonstrated impressive results on a number of problems. However, complex domains like dexterous manipulation remain a challenge due to the high sample complexity. To address this, current approaches employ expert demonstrations in the form of state-action pairs, which are difficult to obtain for real-world settings such as learning from videos. In this paper, we move toward a more realistic setting and explore state-only imitation learning. To tackle this setting, we train an inverse dynamics model and use it to predict actions for state-only demonstrations. The inverse dynamics model and the policy are trained jointly. Our method performs on par with state-action approaches and considerably outperforms RL alone. By not relying on expert actions, we are able to learn from demonstrations with different dynamics, morphologies, and objects. Videos available on the { text{project page}}.
UR - http://www.scopus.com/inward/record.url?scp=85122173330&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85122173330&partnerID=8YFLogxK
U2 - 10.1109/IROS51168.2021.9636557
DO - 10.1109/IROS51168.2021.9636557
M3 - Conference contribution
AN - SCOPUS:85122173330
T3 - IEEE International Conference on Intelligent Robots and Systems
SP - 7865
EP - 7871
BT - IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2021
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2021
Y2 - 27 September 2021 through 1 October 2021
ER -