Domain Adaptation In Reinforcement Learning Via Latent Unified State Representation

TL;DR

This paper introduces a two-stage reinforcement learning method that learns a unified latent state representation to enable zero-shot domain adaptation across different visual environments, improving generalization in complex tasks.

Contribution

The paper proposes a novel two-stage RL approach that learns a cross-domain consistent latent state representation, enabling effective zero-shot transfer without additional training.

Findings

Achieves state-of-the-art domain adaptation in CarRacing variants.

Outperforms prior latent-representation and image translation methods.

Demonstrates effectiveness in complex autonomous driving simulations.

Abstract

Despite the recent success of deep reinforcement learning (RL), domain adaptation remains an open problem. Although the generalization ability of RL agents is critical for the real-world applicability of Deep RL, zero-shot policy transfer is still a challenging problem since even minor visual changes could make the trained agent completely fail in the new task. To address this issue, we propose a two-stage RL agent that first learns a latent unified state representation (LUSR) which is consistent across multiple domains in the first stage, and then do RL training in one source domain based on LUSR in the second stage. The cross-domain consistency of LUSR allows the policy acquired from the source domain to generalize to other target domains without extra training. We first demonstrate our approach in variants of CarRacing games with customized manipulations, and then verify it in CARLA, an autonomous driving simulator with more complex and realistic visual observations. Our results show that this approach can achieve state-of-the-art domain adaptation performance in related RL tasks and outperforms prior approaches based on latent-representation based RL and image-to-image translation.