Motion Perception in Reinforcement Learning with Dynamic Objects
This addresses a specific bottleneck in reinforcement learning for dynamic scenarios, offering incremental improvements for tasks like robotic control.
The paper tackles the problem of reinforcement learning agents failing to explicitly account for motion in dynamic environments, and shows that learning an explicit motion representation improves controller quality across several continuous control benchmarks and robotic tasks, with image difference inputs outperforming temporal frame stacking.
In dynamic environments, learned controllers are supposed to take motion into account when selecting the action to be taken. However, in existing reinforcement learning works motion is rarely treated explicitly; it is rather assumed that the controller learns the necessary motion representation from temporal stacks of frames implicitly. In this paper, we show that for continuous control tasks learning an explicit representation of motion improves the quality of the learned controller in dynamic scenarios. We demonstrate this on common benchmark tasks (Walker, Swimmer, Hopper), on target reaching and ball catching tasks with simulated robotic arms, and on a dynamic single ball juggling task. Moreover, we find that when equipped with an appropriate network architecture, the agent can, on some tasks, learn motion features also with pure reinforcement learning, without additional supervision. Further we find that using an image difference between the current and the previous frame as an additional input leads to better results than a temporal stack of frames.