Offline Reinforcement Learning as Anti-Exploration

TL;DR

This paper introduces a novel offline reinforcement learning method that uses an anti-exploration bonus to keep the policy close to the dataset, improving performance on continuous control tasks.

Contribution

It proposes a new approach of subtracting a prediction-based exploration bonus to prevent the policy from venturing into unseen actions, connecting it to regularization techniques.

Findings

Competitive performance on continuous control tasks

Effective in maintaining dataset support

Outperforms some existing offline RL methods

Abstract

Offline Reinforcement Learning (RL) aims at learning an optimal control from a fixed dataset, without interactions with the system. An agent in this setting should avoid selecting actions whose consequences cannot be predicted from the data. This is the converse of exploration in RL, which favors such actions. We thus take inspiration from the literature on bonus-based exploration to design a new offline RL agent. The core idea is to subtract a prediction-based exploration bonus from the reward, instead of adding it for exploration. This allows the policy to stay close to the support of the dataset. We connect this approach to a more common regularization of the learned policy towards the data. Instantiated with a bonus based on the prediction error of a variational autoencoder, we show that our agent is competitive with the state of the art on a set of continuous control locomotion and manipulation tasks.