Reward-free World Models for Online Imitation Learning

TL;DR

This paper introduces a reward-free world model approach for online imitation learning that models environment dynamics in latent space, improving stability and performance in complex high-dimensional tasks.

Contribution

It proposes a novel method using latent space dynamics and inverse soft-Q learning to enhance stability and effectiveness in online imitation learning for complex environments.

Findings

Achieves stable, expert-level performance in high-dimensional tasks

Outperforms existing methods on benchmarks like DMControl, MyoSuite, and ManiSkill2

Demonstrates the effectiveness of reward-free latent dynamics modeling

Abstract

Imitation learning (IL) enables agents to acquire skills directly from expert demonstrations, providing a compelling alternative to reinforcement learning. However, prior online IL approaches struggle with complex tasks characterized by high-dimensional inputs and complex dynamics. In this work, we propose a novel approach to online imitation learning that leverages reward-free world models. Our method learns environmental dynamics entirely in latent spaces without reconstruction, enabling efficient and accurate modeling. We adopt the inverse soft-Q learning objective, reformulating the optimization process in the Q-policy space to mitigate the instability associated with traditional optimization in the reward-policy space. By employing a learned latent dynamics model and planning for control, our approach consistently achieves stable, expert-level performance in tasks with high-dimensional observation or action spaces and intricate dynamics. We evaluate our method on a diverse set of benchmarks, including DMControl, MyoSuite, and ManiSkill2, demonstrating superior empirical performance compared to existing approaches.