Provable RL with Exogenous Distractors via Multistep Inverse Dynamics

TL;DR

This paper introduces a new model and algorithm for reinforcement learning in environments with high-dimensional observations and exogenous noise, providing theoretical guarantees and empirical validation for effective latent state discovery.

Contribution

The paper proposes the EX-BMDP model and the PPE algorithm, addressing the challenge of exogenous noise in representation learning for RL, with provable efficiency and robustness.

Findings

PPE is sample and computationally efficient in EX-BMDPs.

Prior approaches fail in environments with temporally correlated noise.

Empirical results demonstrate PPE's effectiveness in challenging exploration tasks.

Abstract

Many real-world applications of reinforcement learning (RL) require the agent to deal with high-dimensional observations such as those generated from a megapixel camera. Prior work has addressed such problems with representation learning, through which the agent can provably extract endogenous, latent state information from raw observations and subsequently plan efficiently. However, such approaches can fail in the presence of temporally correlated noise in the observations, a phenomenon that is common in practice. We initiate the formal study of latent state discovery in the presence of such exogenous noise sources by proposing a new model, the Exogenous Block MDP (EX-BMDP), for rich observation RL. We start by establishing several negative results, by highlighting failure cases of prior representation learning based approaches. Then, we introduce the Predictive Path Elimination (PPE) algorithm, that learns a generalization of inverse dynamics and is provably sample and computationally efficient in EX-BMDPs when the endogenous state dynamics are near deterministic. The sample complexity of PPE depends polynomially on the size of the latent endogenous state space while not directly depending on the size of the observation space, nor the exogenous state space. We provide experiments on challenging exploration problems which show that our approach works empirically.