Discriminator Augmented Model-Based Reinforcement Learning

TL;DR

This paper introduces a discriminator-augmented approach to model-based reinforcement learning that improves planning accuracy by correcting for learned model inaccuracies and minimizing value estimation variance, leading to better performance.

Contribution

It proposes an importance sampling framework and a new objective for fitting dynamics models, enhancing planning accuracy in MBRL.

Findings

Improved performance on stochastic control tasks.

Theoretical justification for the variance-minimizing objective.

Enhanced robustness to model inaccuracies.

Abstract

By planning through a learned dynamics model, model-based reinforcement learning (MBRL) offers the prospect of good performance with little environment interaction. However, it is common in practice for the learned model to be inaccurate, impairing planning and leading to poor performance. This paper aims to improve planning with an importance sampling framework that accounts and corrects for discrepancy between the true and learned dynamics. This framework also motivates an alternative objective for fitting the dynamics model: to minimize the variance of value estimation during planning. We derive and implement this objective, which encourages better prediction on trajectories with larger returns. We observe empirically that our approach improves the performance of current MBRL algorithms on two stochastic control problems, and provide a theoretical basis for our method.