Dual-Arm Cable Manipulation on Untangling, Cutting & Tension-Aware Control

Institut

TUM Munich Institute of Robotics and Machine Intelligence (Institut)

Typ

Masterarbeit /

Inhalt

 

Beschreibung

Topic Overview
We investigate robotic manipulation of deformable cable-like objects using a dual-arm system. The focus is on untangling, trimming, and tension-aware pulling under realistic visual interference. The student will work with a loop-based scene, where the robot actively disturbs the cable, observes the propagation of deformation, infers the cable topology, and progressively identifies feasible manipulation points. Based on the inferred cable structure, the robot executes tension-aware pulling, collision-free and tangle-free motion, and cutting when necessary.

Research Components (depending on student background)

• Vision-based cable segmentation under occlusion

• Robot-in-the-loop active perturbation & optical flow analysis

• Online estimation of perturbation propagation & cable topology

• Dual-arm motion planning with collision/tanglement avoidance

• Tension-aware control for safe cable pulling

• Cutting and post-manipulation routines

Motivation
Cables and other Deformable Linear Objects (DLOs) are common in manufacturing, service robotics, electronics assembly, and recycling. Unlike rigid objects, cables deform continuously, occlude themselves, and exhibit tension-dependent dynamics. This project develops techniques that go beyond passive perception by integrating active exploration, physical interaction, and online estimation.

 

Voraussetzungen

Candidate Profile (we do not expect all)

• Strong interest in robotics, control, motion planning, or machine learning

• Solid programming skills in Python and/or C++

• Experience in at least one of: computer vision, dynamics & control, motion planning, RL/IL, or optimization

• Familiarity with Franka, ROS2, Isaac/MuJoCo, or segmentation models is a plus, not a requirement

What We Offer

• Hands-on access to dual-arm Franka Research 3 platform

• Integration with real vision pipelines and cable setups

• Possibility to publish results depending on outcome

• Supervision environment with regular feedback

• Opportunity to combine perception, estimation, and control in a realistic robotic task

How to apply

Please send application including your CV, transcript to Mr. Li, yu.li(at)tum.de and Dr. Sadeghian, hamid.sadeghian(at)tum.de.

Reference

O. Holešovský, R. Škoviera and V. Hlaváč, "Interactive Robotic Moving Cable Segmentation by Motion Correlation," in IEEE Robotics and Automation Letters, vol. 10, no. 7, pp. 7420-7427, July 2025, doi: 10.1109/LRA.2025.3574960.

Shivakumar K, Viswanath V, Gu A, et al. Sgtm 2.0: Autonomously untangling long cables using interactive perception[J]. arXiv preprint arXiv:2209.13706, 2022.

Möglicher Beginn

sofort

Kontakt

Yu Li, M.Sc.
yu.litum.de