From Waste to Value: How TECNALIA’s Robotic technologies developed in RECONMATIC are reinventing Construction Waste Sorting

RECONMATIC
Jul 24
3 min read

Automation is unlocking a new era for circular construction. and RECONMATIC is leading the way. With the development of intelligent robotic systems for off-site sorting of construction and demolition waste (CDW), the project is paving the way for a safer, smarter, and more sustainable future for the built environment.

The Challenge: Complexity in Construction Waste Streams

Construction and demolition sites generate highly heterogeneous waste. Concrete, ceramics, wood, plasterboard, plastic, and contaminant materials like gypsum or asbestos often end up in a mixed stream, making automated separation particularly challenging. Traditional sorting systems rely on manual labour or simple mechanical processes, which limit efficiency, safety, and material quality in recycling.

To meet the challenge of producing high-purity, high-quality recycled materials, RECONMATIC partner TECNALIA’s robotics team set out to develop a solution that combines multisensory integration, AI-driven material recognition, and robotic sorting.

Developing an AI-driven detection and classification of the waste material

One of the project's early breakthroughs was the integration, geometric calibration, and spatial synchronisation of multiple sensors (RGB, 3D laser, and NIR hyperspectral) into a unique multispectral hypercube. This hypercube is the primary input used for creating a dataset of CDW fragments (semi-automatically segmented and annotated), which is then used to train AI deep learning models capable of recognising the different materials of the CDW waste items.

During operation, the hypercube is generated in real-time and fed to the AI deep learning models to infer the material category of each pixel of the previously unknown items, detect the object shape and compute the optimal grasping point for the robot to pick the undesired contaminant out of the main waste stream on the conveyor belt.

The Robotic Cell

The core of the solution is an automated robotic cell that includes:

The perception sensors (RGB, 3D laser and NIR)
The deep-learning model trained for real-time classification of materials on the sorting belt, achieving high accuracy.
A custom-designed multi-purpose gripper that handles both heavy and lightweight materials using a combination of fingers and vacuum suction.
A sophisticated pick-and-throw control algorithm allowing the Omron Quattro robot to throw the waste item instead of placing it in the target box, achieving higher performance.

This hybrid approach enables the robot to act as a "final sorter", particularly targeting the removal of gypsum, plastic, wood, and potentially glass from the main conveyor belt, and allowing to increase the purity of the final product.

Real-World Integration: Pilot at the RECSO Facility

The system is currently under development at TECNALIA premises. It will be demonstrated at the RECSO plant, and its performance analysed.

Once productized, TECNALIA’s WasteSorting robot co-developed in RECONMATIC, will find its place in the manual final sorting cabin where operators currently perform this tedious task, moving multiple tons of material every day. The AI-driven sorting robot will contribute to reducing injuries and improving the final product quality and market acceptance by reducing chemical and visual contaminants.

Stay tuned as we continue to push the boundaries of automation and sustainability. Robotics is no longer just about productivity - it’s about creating value from waste, one intelligent pick at a time.

This article was written by Ibon Merino and Damien Sallé from TECNALIA. In the RECONMATIC project, TECNALIA leads efforts to bring construction and demolition waste materials back into the market by developing smart reuse strategies. They also support the project in other areas, helping create digital tools for better planning before demolition and building a robot that can sort materials more efficiently in off-site recycling facilities.