Recognising Known Configurations of Garments For Dual-Arm Robotic Flattening

TL;DR

This paper presents a method for robots to recognize known garment configurations to facilitate efficient flattening, reducing computational complexity in deformable-object manipulation.

Contribution

It introduces a learning approach for recognizing garment states, enabling pre-designed manipulation plans for improved robotic flattening of garments.

Findings

Robots can accurately identify garment configurations.

Pre-designed plans improve flattening efficiency.

Recognition reduces computational time.

Abstract

Robotic deformable-object manipulation is a challenge in the robotic industry because deformable objects have complicated and various object states. Predicting those object states and updating manipulation planning is time-consuming and computationally expensive. In this paper, we propose learning known configurations of garments to allow a robot to recognise garment states and choose a pre-designed manipulation plan for garment flattening.