Learning from Pixels with Expert Observations
This addresses the problem of costly expert action acquisition in robot manipulation for RL researchers, though it appears incremental as it builds on existing goal-conditioned RL methods.
The paper tackles the challenge of sparse rewards in reinforcement learning for robot manipulation tasks by using readily available expert observations as intermediate visual goals instead of costly expert actions. Their approach significantly improves performance of state-of-the-art agents while requiring 4-20 times fewer expert actions during training.
In reinforcement learning (RL), sparse rewards can present a significant challenge. Fortunately, expert actions can be utilized to overcome this issue. However, acquiring explicit expert actions can be costly, and expert observations are often more readily available. This paper presents a new approach that uses expert observations for learning in robot manipulation tasks with sparse rewards from pixel observations. Specifically, our technique involves using expert observations as intermediate visual goals for a goal-conditioned RL agent, enabling it to complete a task by successively reaching a series of goals. We demonstrate the efficacy of our method in five challenging block construction tasks in simulation and show that when combined with two state-of-the-art agents, our approach can significantly improve their performance while requiring 4-20 times fewer expert actions during training. Moreover, our method is also superior to a hierarchical baseline.