THE CONFERENCE
FOR AUTHORS
FOR PARTICIPANTS

Timezone: »

 
Talk
Robust Adversarial Reinforcement Learning
Lerrel Pinto · James Davidson · Rahul Sukthankar · Abhinav Gupta

Sun Aug 06 05:30 PM -- 05:48 PM (PDT) @ C4.5
in Reinforcement learning 1 »

Deep neural networks coupled with fast simulation and improved computational speeds have led to recent successes in the field of reinforcement learning (RL). However, most current RL-based approaches fail to generalize since: (a) the gap between simulation and real world is so large that policy-learning approaches fail to transfer; (b) even if policy learning is done in real world, the data scarcity leads to failed generalization from training to test scenarios (e.g., due to different friction or object masses). Inspired from H-infinity control methods, we note that both modeling errors and differences in training and test scenarios can just be viewed as extra forces/disturbances in the system. This paper proposes the idea of robust adversarial reinforcement learning (RARL), where we train an agent to operate in the presence of a destabilizing adversary that applies disturbance forces to the system. The jointly trained adversary is reinforced -- that is, it learns an optimal destabilization policy. We formulate the policy learning as a zero-sum, minimax objective function. Extensive experiments in multiple environments (InvertedPendulum, HalfCheetah, Swimmer, Hopper, Walker2d and Ant) conclusively demonstrate that our method (a) improves training stability; (b) is robust to differences in training/test conditions; and c) outperform the baseline even in the absence of the adversary.

[ PDF]  [ ss Summary/Notes]  [ Video] 

Author Information

Lerrel Pinto (Carnegie Mellon University)
James Davidson (Google Brain)
Rahul Sukthankar (Google Research)
Abhinav Gupta (Carnegie Mellon University)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors