Latent Diffusion Planning for Imitation Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xie, Amber, Rybkin, Oleh, Sadigh, Dorsa, Finn, Chelsea
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909590315073536
author Xie, Amber
Rybkin, Oleh
Sadigh, Dorsa
Finn, Chelsea
author_facet Xie, Amber
Rybkin, Oleh
Sadigh, Dorsa
Finn, Chelsea
contents Recent progress in imitation learning has been enabled by policy architectures that scale to complex visuomotor tasks, multimodal distributions, and large datasets. However, these methods often rely on learning from large amount of expert demonstrations. To address these shortcomings, we propose Latent Diffusion Planning (LDP), a modular approach consisting of a planner which can leverage action-free demonstrations, and an inverse dynamics model which can leverage suboptimal data, that both operate over a learned latent space. First, we learn a compact latent space through a variational autoencoder, enabling effective forecasting of future states in image-based domains. Then, we train a planner and an inverse dynamics model with diffusion objectives. By separating planning from action prediction, LDP can benefit from the denser supervision signals of suboptimal and action-free data. On simulated visual robotic manipulation tasks, LDP outperforms state-of-the-art imitation learning approaches, as they cannot leverage such additional data.
format Preprint
id arxiv_https___arxiv_org_abs_2504_16925
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Latent Diffusion Planning for Imitation Learning
Xie, Amber
Rybkin, Oleh
Sadigh, Dorsa
Finn, Chelsea
Robotics
Artificial Intelligence
Recent progress in imitation learning has been enabled by policy architectures that scale to complex visuomotor tasks, multimodal distributions, and large datasets. However, these methods often rely on learning from large amount of expert demonstrations. To address these shortcomings, we propose Latent Diffusion Planning (LDP), a modular approach consisting of a planner which can leverage action-free demonstrations, and an inverse dynamics model which can leverage suboptimal data, that both operate over a learned latent space. First, we learn a compact latent space through a variational autoencoder, enabling effective forecasting of future states in image-based domains. Then, we train a planner and an inverse dynamics model with diffusion objectives. By separating planning from action prediction, LDP can benefit from the denser supervision signals of suboptimal and action-free data. On simulated visual robotic manipulation tasks, LDP outperforms state-of-the-art imitation learning approaches, as they cannot leverage such additional data.
title Latent Diffusion Planning for Imitation Learning
topic Robotics
Artificial Intelligence
url https://arxiv.org/abs/2504.16925