Robotic Intelligence for Construction Waste Sorting

OBJECTIVES

To develop an automated, intelligent robotic system for real-time sorting of irregular construction and demolition waste

To build a deep learning model that can identify and classify construction waste using data from multiple sensors

To create a self-supervised robotic sorting system using digital twin simulation and adaptive learning

To test and refine the system’s accuracy, efficiency, and scalability in real-world conditions with industry partners

Team

Kevin Zhang

Lead Researcher

Lei Hou

Research Fellow

Ehsan Asadi

Research Fellow

Shanuka Kamesh Dodampegama

PhD

Partners

Shenzhen Yuezhong Green Building Sc-Tech Development Co Ltd

Progress

SUMMARY

This project is building a robotic waste-sorting system using AI and sensor-driven technology to tackle the challenge of mixed, irregular construction waste. A domain-adaptive deep learning model has been developed to identify construction waste and transfer learning between systems. Key equipment for the robotic sorting lab has been purchased, including a conveyor system. Initial testing is underway using ROS2 with a UR5e robotic arm and Robotiq gripper. A custom tool is in development to manage image annotation, prediction, correction, and future real-time visualization and control of the full system. The project team also shared its work internationally at the Chinese Materials Conference 2024.

OPPORTUNITIES

Access to industry research funding for AI and robotics in sustainable construction

Strong collaboration opportunities with government and green building sectors

Real-world testing and commercialisation of a scalable robotics solution

Potential to change the way construction waste is sorted and processed on building sites

Platform to train students and researchers in cutting-edge AI and automation for the circular economy

AS SEEN AT

Chinese Material Conference 2024 in Guangzhou, China – Keynote speech