Related Experiment Video
Updated: Feb 24, 2026
03:55
Author Spotlight: Enhancing Grasping Abilities for Hemiplegic Patients with Flexible Robotic Limbs
Published on: October 27, 2023
2.9K
Deep learning-based robotic cloth manipulation applications: systematic review, challenges and opportunities for
Ningquan Gu1, Mitsuhiro Hayashibe1, Kyo Kutsuzawa2
1Neuro-Robotics Lab, Department of Robotics, Graduate School of Engineering, Tohoku University, Sendai, Japan.
Frontiers in Robotics and AI
|February 23, 2026
Summary
This review summarizes deep learning for robotic cloth manipulation, highlighting progress and challenges in unfolding and folding tasks. Future work needs better data, simulators, and multi-modal sensors for physical AI advancements.
Area of Science:
- Robotics
- Artificial Intelligence
- Computer Vision
Background:
- Robotic cloth manipulation, specifically unfolding and folding, is crucial for physical AI.
- Recent deep learning advances have accelerated progress in this field.
Purpose of the Study:
- To systematically review and summarize deep learning-based methods for cloth unfolding and folding.
- To identify current challenges and future research directions.
Main Methods:
- A systematic literature review following PRISMA guidelines.
- Analysis of 41 papers published between 2019 and 2024.
- Categorization of learning methods into six paradigms.
Main Results:
- Current methods show promise but face challenges with irregular cloth sizes and diverse initial states.
- Need for improved real-world data, realistic simulators, and addressing the Sim2Real gap.
- Multi-modal sensors and novel primitive actions enhance performance; consistent metrics and failure mode strategies are needed.
Conclusions:
- The field of deep learning for robotic cloth manipulation is rapidly advancing.
- Addressing data limitations, simulation realism, sensor integration, and algorithmic paradigms is key for future progress in physical AI.
