TimeRewarder: Learning Dense Reward from Passive Videos via Frame-wise Temporal Distance
This addresses the scalability and effort issues in reward design for robotics, offering a solution that can leverage diverse video sources like demonstrations and human videos.
The paper tackles the problem of manually designing dense rewards in reinforcement learning for robotics by introducing TimeRewarder, a method that learns reward signals from passive videos via frame-wise temporal distance modeling. It achieved nearly perfect success in 9 out of 10 Meta-World tasks with only 200,000 interactions per task, outperforming previous methods and even manually designed rewards.
Designing dense rewards is crucial for reinforcement learning (RL), yet in robotics it often demands extensive manual effort and lacks scalability. One promising solution is to view task progress as a dense reward signal, as it quantifies the degree to which actions advance the system toward task completion over time. We present TimeRewarder, a simple yet effective reward learning method that derives progress estimation signals from passive videos, including robot demonstrations and human videos, by modeling temporal distances between frame pairs. We then demonstrate how TimeRewarder can supply step-wise proxy rewards to guide reinforcement learning. In our comprehensive experiments on ten challenging Meta-World tasks, we show that TimeRewarder dramatically improves RL for sparse-reward tasks, achieving nearly perfect success in 9/10 tasks with only 200,000 interactions per task with the environment. This approach outperformed previous methods and even the manually designed environment dense reward on both the final success rate and sample efficiency. Moreover, we show that TimeRewarder pretraining can exploit real-world human videos, highlighting its potential as a scalable approach path to rich reward signals from diverse video sources.