Skip to main content Skip to main navigation

More efficient recycling with AI: Joint project develops automated sorting solution for bulky waste

| Press release | Environment & Energy | Autonomous Systems | Image Recognition & Understanding | Machine Learning & Deep Learning | Robotics | Sensors & Networks | Robotics Innovation Center | Bremen

Currently, large pieces of waste, such as bulky or construction waste, can only be recycled with a great deal of manual and technical effort. The SmartRecycling-Up project has developed an innovative, AI-based concept that automates the sorting process using a crane or excavator for the first time. A consortium led by the German Research Center for Artificial Intelligence (DFKI) developed technologies combining robotics, artificial intelligence, and intelligent sensors to create an integrated functional system.

Real-world deployment: The automated crane operation was successfully tested at the ASO GmbH waste sorting facility in Osterholz-Scharmbeck.

Automated waste sorting is a key component of a sustainable circular economy. While smaller waste fractions, such as packaging and paper, can be mechanically separated in conveyor-guided sorting systems, large pieces of waste, such as bulky and construction waste, still pose a challenge. Currently, these materials must be laboriously shredded before they can be sorted in conventional plants. This process is inefficient, costly, and energy-intensive.

These challenges provided the impetus for the SmartRecycling-Up project, which was funded by the Federal Ministry for the Environment, Nature Conservation, Nuclear Safety and Consumer Protection (BMUV) as part of the "AI Lighthouses for the Environment, Climate, Nature and Resources" funding line. The project’s goal was to develop an AI-based solution for sorting large waste objects without the need for mechanical shredding. A particular focus was placed on the automated recovery of valuable materials such as wood, plastic, and metal – components that are often difficult to separate from mixed waste streams. The aim: to significantly increase recycling efficiency, conserve resources, and raise overall recycling rates.

AI-controlled complete solution for waste sorting

To achieve this, SmartRecycling-Up developed a novel technical approach that combines advanced AI-supported sensor technology, machine learning, and automated control systems into an intelligent, integrated concept. The system features a hydraulic crane, empowered by AI-based control, which autonomously performs the complex tasks involved in waste handling and sorting. These tasks include transferring and mixing the waste mixture, targeted feeding of a shredder, detection and removal of contaminants, and separate sorting of recyclable materials.

At the core of the system lies a sophisticated sensor platform, composed of multispectral cameras, depth sensors, and AI-driven evaluation processes that enable precise material classification and three-dimensional localization. Detected materials are analyzed in real time, with handling decisions automatically transmitted to the crane's control system. This allows the system not only to distinguish recyclable materials from contaminants but also to sort different fractions by material type.

An AI-powered sensor system detects and classifies materials in real time.

DFKI: Autonomous control through adaptive systems

As part of SmartRecycling-Up, DFKI has developed an AI-based control framework that can fully automate the movement and process control of hydraulic heavy machinery such as cranes or excavators for recycling. The framework consists of various modules that cover different functionalities for automation.

The "SmartMotionController" module uses deep reinforcement learning to learn the machine's motion sequences. To do this, it continuously analyzes how the machine reacts to certain control commands and adapts the motion planning accordingly. This is based on data from the SmartStateEstimator module. This processes perspective data captured by external camera and lidar systems during the movements of the machine and its joints. To do this, it works with generative AI for image processing and other advanced methods to continuously estimate the state of the robot. This enables the controller to establish the relationship between control commands and machine movements and autonomously control the machine along a desired trajectory. The "SmartProcessController" coordinates the entire process from object recognition and position estimation to collision-free trajectory planning and execution.

Successful tests in the laboratory and under real conditions

The technologies were first developed and tested in the laboratory using the ARTER excavator robot. This was followed by use under real conditions: Automated operation with a hydraulic crane was successfully implemented in a waste sorting plant and various aspects of the solution were validated. The systems were able to reliably identify and sort pre-defined recyclable materials. The tests showed that the combination of AI, sensor technology and robotics significantly increases efficiency. At the same time, the mechanical stress on the valuable waste materials and the human workload are reduced - thus improving the quality of the entire recycling process.

The autonomous walking excavator ARTER served as the robotic test platform for the AI-based control system.

Project partners and funding

SmartRecycling-Up was carried out under the leadership of DFKI together with the Institute for Energy and Circular Economy at Bremen University of Applied Sciences GmbH, the Smart Systems Research and Transfer Center at HAW Hamburg, Baljer & Zembrod GmbH & Co KG, KreisAbfallVerwertungsGesellschaft mbH Minden-Lübbecke, Karl Siedenburg GmbH & Co KG and ASO Abfall-Service Osterholz GmbH. The project was funded by the BMUV under the funding code 67KI21013 and supervised by Zukunft - Umwelt - Gesellschaft (ZUG) gGmbH.

 

Share this post:

Contact:

Ajish Babu, M.Sc.
Dr.-Ing. Thomas Vögele

Press contact:

Communications & Media DFKI Bremen