Learning Markov State Abstractions for Deep Reinforcement Learning

TL;DR

This paper introduces a new method for learning Markovian state abstractions in deep reinforcement learning, improving sample efficiency and performance without relying solely on reward signals.

Contribution

It proposes a novel set of conditions and a practical training procedure combining inverse modeling and contrastive learning for Markov abstraction in RL.

Findings

Learned representations capture domain structure effectively.

Achieves better sample efficiency than existing methods.

Matches or exceeds performance of hand-designed states.

Abstract

A fundamental assumption of reinforcement learning in Markov decision processes (MDPs) is that the relevant decision process is, in fact, Markov. However, when MDPs have rich observations, agents typically learn by way of an abstract state representation, and such representations are not guaranteed to preserve the Markov property. We introduce a novel set of conditions and prove that they are sufficient for learning a Markov abstract state representation. We then describe a practical training procedure that combines inverse model estimation and temporal contrastive learning to learn an abstraction that approximately satisfies these conditions. Our novel training objective is compatible with both online and offline training: it does not require a reward signal, but agents can capitalize on reward information when available. We empirically evaluate our approach on a visual gridworld domain and a set of continuous control benchmarks. Our approach learns representations that capture the underlying structure of the domain and lead to improved sample efficiency over state-of-the-art deep reinforcement learning with visual features -- often matching or exceeding the performance achieved with hand-designed compact state information.