Current waste processing methods for the decommissioning of nuclear facilities involve manual sorting tasks which are potentially hazardous for the workers involved. The MTC’s robotic model uses an AI-based vision system and radiological sensors to identify radioactive waste objects and sort them into appropriate waste streams. This minimises the risk to operators and increases the productivity of the process.
As radioactive waste comes in a variety of different shapes and sizes, the robot’s vision model learns to identify individual items. The multi-camera computer vision software gathers multiple views of the waste objects which are fused into a 3D model for input into algorithms. The algorithms inform the robot of the object’s shape and location, and combined with sensor data on radioactivity level, drive appropriate waste stream sorting. Every object sorted provides new data points for the vision system, allowing it to use machine learning to become more autonomous with each new item it sorts.
By efficiently sorting nuclear waste into separate streams, items can be directed to the appropriate output. This allows certain, low-risk waste to be recycled instead of placed into storage along with higher risk items, reducing storage costs and environmental impact.