Learning Temporally-Consistent Representations for Data-Efficient Reinforcement Learning
This addresses the representation learning bottleneck in RL for agents in high-dimensional state spaces, leading to improved sample efficiency and generalization, though it is an incremental improvement over existing methods.
The paper tackles the problem of poor sample efficiency in deep reinforcement learning from high-dimensional states by introducing a new representation learning method called KSL, which enforces temporal consistency through a self-supervised auxiliary task, resulting in state-of-the-art data efficiency and asymptotic performance on the PlaNet benchmark.
Deep reinforcement learning (RL) agents that exist in high-dimensional state spaces, such as those composed of images, have interconnected learning burdens. Agents must learn an action-selection policy that completes their given task, which requires them to learn a representation of the state space that discerns between useful and useless information. The reward function is the only supervised feedback that RL agents receive, which causes a representation learning bottleneck that can manifest in poor sample efficiency. We present $k$-Step Latent (KSL), a new representation learning method that enforces temporal consistency of representations via a self-supervised auxiliary task wherein agents learn to recurrently predict action-conditioned representations of the state space. The state encoder learned by KSL produces low-dimensional representations that make optimization of the RL task more sample efficient. Altogether, KSL produces state-of-the-art results in both data efficiency and asymptotic performance in the popular PlaNet benchmark suite. Our analyses show that KSL produces encoders that generalize better to new tasks unseen during training, and its representations are more strongly tied to reward, are more invariant to perturbations in the state space, and move more smoothly through the temporal axis of the RL problem than other methods such as DrQ, RAD, CURL, and SAC-AE.