Learning Predictive Representations for Deformable Objects Using Contrastive Estimation

TL;DR

This paper introduces a contrastive learning framework for deformable object manipulation that jointly optimizes visual and dynamics models, enabling effective simulation-to-real transfer and improved task performance.

Contribution

It presents a novel joint optimization approach for visual and dynamics models using contrastive estimation, specifically tailored for deformable object manipulation.

Findings

Substantial performance improvements over standard methods.

Successful transfer of policies from simulation to real robot.

Effective manipulation of ropes and cloths in experiments.

Abstract

Using visual model-based learning for deformable object manipulation is challenging due to difficulties in learning plannable visual representations along with complex dynamic models. In this work, we propose a new learning framework that jointly optimizes both the visual representation model and the dynamics model using contrastive estimation. Using simulation data collected by randomly perturbing deformable objects on a table, we learn latent dynamics models for these objects in an offline fashion. Then, using the learned models, we use simple model-based planning to solve challenging deformable object manipulation tasks such as spreading ropes and cloths. Experimentally, we show substantial improvements in performance over standard model-based learning techniques across our rope and cloth manipulation suite. Finally, we transfer our visual manipulation policies trained on data purely collected in simulation to a real PR2 robot through domain randomization.