WOMBET: World Model-based Experience Transfer for Robust and Sample-efficient Reinforcement Learning

TL;DR

WOMBET introduces a world model-based framework that generates and filters prior data for efficient transfer in reinforcement learning, improving sample efficiency and performance in robotics tasks.

Contribution

It jointly learns a world model and generates offline data with uncertainty-aware planning, enabling stable transfer and adaptation in RL.

Findings

WOMBET outperforms strong baselines on continuous control benchmarks.

Uncertainty-penalized planning provides a lower bound on true return.

Joint data generation and transfer improve sample efficiency and final performance.

Abstract

Reinforcement learning (RL) in robotics is often limited by the cost and risk of data collection, motivating experience transfer from a source task to a target task. Offline-to-online RL leverages prior data but typically assumes a given fixed dataset and does not address how to generate reliable data for transfer. We propose \textit{World Model-based Experience Transfer} (WOMBET), a framework that jointly generates and utilizes prior data. WOMBET learns a world model in the source task and generates offline data via uncertainty-penalized planning, followed by filtering trajectories with high return and low epistemic uncertainty. It then performs online fine-tuning in the target task using adaptive sampling between offline and online data, enabling a stable transition from prior-driven initialization to task-specific adaptation. We show that the uncertainty-penalized objective provides a lower bound on the true return and derive a finite-sample error decomposition capturing distribution mismatch and approximation error. Empirically, WOMBET improves sample efficiency and final performance over strong baselines on continuous control benchmarks, demonstrating the benefit of jointly optimizing data generation and transfer.