AI-Enhanced Robotic Unfolding for Sustainable Textile Recycling – with a Focus on Energy Minimisation

About the Project

The textile industry is a significant contributor to global environmental waste, with millions of tons of discarded fabrics being sent to landfills annually [1]. Although recycling textiles can mitigate this issue, the current processes are labour-intensive, requiring human intervention for tasks such as sorting, unfolding, and classifying fabrics. This PhD project aims to develop a fully automated robotic system capable of unfolding textiles, using computer vision, artificial intelligence (AI), and energy-efficient control algorithms. The system will address the challenges of handling diverse textile types by leveraging RGB-D imaging for detailed analysis and optimised robotic motion planning.

The system will consist of a robotic arm over a conveyor belt, where textiles are fed for processing. Using RGB-D cameras, the system will capture both colour and depth data, allowing the identification of key textile features such as folds, edges, and textures. Convolutional Neural Networks (CNNs) will process these images to analyse the textile’s structure and identify manipulation points. Additionally, the system will classify textile types (e.g., cotton, polyester) to facilitate sorting for recycling purposes.

To ensure sustainability, the project will implement Model Predictive Control (MPC) to minimize energy consumption during the robotic unfolding process. MPC will calculate the most energy-efficient sequence of movements, continuously adjusting the robot’s actions based on real-time feedback from sensors. The system will reduce unnecessary motions, focusing on energy-efficient manipulation. This will lead to a more sustainable recycling process, as the energy footprint of handling and sorting textiles will be significantly reduced.

The main research objectives include: (1) designing a robotic system capable of automating the unfolding of textiles with minimal human intervention, (2) utilising AI-driven image processing techniques to identify key textile features and ensure efficient manipulation, (3) minimising energy consumption through optimised control algorithms, and (4) developing machine learning models for textile classification to assist in sorting and recycling.

The project builds upon recent advancements in robotics and AI, particularly in using CNNs for object recognition and depth-sensing technologies for 3D manipulation. While there has been significant research on robotic handling of flexible objects like cloth [2,3,4,5], there is still a gap in applying these technologies to textile recycling. This project aims to fill that gap by developing an integrated system specifically for textiles. Depth sensors will be used to capture folds and contours, while AI techniques will ensure accurate identification and manipulation. Furthermore, energy optimisation will be prioritized through algorithms like MPC, which have been shown to minimise energy usage in other robotic applications.

The system will be rigorously tested using a wide range of textile samples, varying in material type, thickness, and folding conditions. Performance metrics will include the accuracy of unfolding, energy consumption, and processing time. These tests will validate the system’s ability to handle different textile types efficiently while maintaining a low energy footprint.

Funding Notes

there is no funding for this project