Deformable Object Manipulation

A generalizable policy to fold clothes

In Progress

Andnet DeBoer, Conor Hayes, Robert Zhu, Praneeth Reddy Mallupalli

Northwestern University

LeRobot VLA Bimanual Manipulation
Project image 1

Overview

Deformable object manipulation remains an open challenge in robotics. This project addresses robotic garment folding as part of the LeHome Challenge 2026 at ICRA, aiming to develop policies that generalize across diverse clothing types using limited training data and simulation-based evaluation.

Problem Statement

Garment manipulation remains a fundamental challenge in robotics due to the high-dimensional state space and complex dynamics of deformable objects. The LeHome Challenge provides 600 teleoperation demonstrations across four garment categories (long/short-sleeved tops, long/short pants) to train policies that generalize to unseen garments.

Hardware: SO101 Dual 6DOF Robot Arms

Four garment categories in LeHome Challenge

Approach

Our pipeline consists of three key components:

  1. Perception
  2. Policy
  3. Training

Acknowledgments

This project is conducted at Northwestern University for the LeHome Challenge 2026 (ICRA).

← Back to Projects