Disentangling Controllable Object through Video Prediction Improves Visual Reinforcement Learning
This work addresses sample efficiency and performance in visual reinforcement learning for domains like video games and robotics, representing an incremental improvement over existing methods.
The paper tackled the problem of improving vision-based reinforcement learning by disentangling controllable objects from visual observations using action-conditioned video prediction, resulting in enhanced sample efficiency and game performance with normalized scores increasing from 222.8% to 261.4% in Atari games.
In many vision-based reinforcement learning (RL) problems, the agent controls a movable object in its visual field, e.g., the player's avatar in video games and the robotic arm in visual grasping and manipulation. Leveraging action-conditioned video prediction, we propose an end-to-end learning framework to disentangle the controllable object from the observation signal. The disentangled representation is shown to be useful for RL as additional observation channels to the agent. Experiments on a set of Atari games with the popular Double DQN algorithm demonstrate improved sample efficiency and game performance (from 222.8% to 261.4% measured in normalized game scores, with prediction bonus reward).