Vector Quantized Models for Planning

TL;DR

This paper introduces a novel planning method using vector quantized models that effectively handles stochastic and partially observable environments, outperforming existing methods like MuZero in complex scenarios.

Contribution

The paper presents a discrete autoencoder-based approach combined with stochastic Monte Carlo tree search for improved planning in uncertain environments.

Findings

Outperforms offline MuZero on stochastic chess variants

Scales to complex 3D environments like DeepMind Lab

Effectively models environment responses with discrete latent variables

Abstract

Recent developments in the field of model-based RL have proven successful in a range of environments, especially ones where planning is essential. However, such successes have been limited to deterministic fully-observed environments. We present a new approach that handles stochastic and partially-observable environments. Our key insight is to use discrete autoencoders to capture the multiple possible effects of an action in a stochastic environment. We use a stochastic variant of Monte Carlo tree search to plan over both the agent's actions and the discrete latent variables representing the environment's response. Our approach significantly outperforms an offline version of MuZero on a stochastic interpretation of chess where the opponent is considered part of the environment. We also show that our approach scales to DeepMind Lab, a first-person 3D environment with large visual observations and partial observability.