Semi-Supervised Offline Reinforcement Learning with Action-Free Trajectories

TL;DR

This paper introduces a semi-supervised offline reinforcement learning framework where unlabelled trajectories are augmented with proxy labels generated via an inverse dynamics model, enabling effective learning even with limited labelled data.

Contribution

It proposes a simple meta-algorithmic pipeline combining inverse dynamics models with offline RL, demonstrating strong empirical results on benchmarks with minimal labelled data.

Findings

High success in matching fully labelled performance with only 10% labelled data.

The pipeline is flexible and effective across various offline RL algorithms.

Empirical study reveals key data and algorithmic factors influencing performance.

Abstract

Natural agents can effectively learn from multiple data sources that differ in size, quality, and types of measurements. We study this heterogeneity in the context of offline reinforcement learning (RL) by introducing a new, practically motivated semi-supervised setting. Here, an agent has access to two sets of trajectories: labelled trajectories containing state, action and reward triplets at every timestep, along with unlabelled trajectories that contain only state and reward information. For this setting, we develop and study a simple meta-algorithmic pipeline that learns an inverse dynamics model on the labelled data to obtain proxy-labels for the unlabelled data, followed by the use of any offline RL algorithm on the true and proxy-labelled trajectories. Empirically, we find this simple pipeline to be highly successful -- on several D4RL benchmarks~\cite{fu2020d4rl}, certain offline RL algorithms can match the performance of variants trained on a fully labelled dataset even when we label only 10\% of trajectories which are highly suboptimal. To strengthen our understanding, we perform a large-scale controlled empirical study investigating the interplay of data-centric properties of the labelled and unlabelled datasets, with algorithmic design choices (e.g., choice of inverse dynamics, offline RL algorithm) to identify general trends and best practices for training RL agents on semi-supervised offline datasets.