Intelligent handling system design and integration

Challenge

This project was funded by the SBRI (Small Business Research Initiative) to develop intelligent automated solutions for the sorting and segregation of nuclear waste. Currently waste from the decommissioning of nuclear facilities has to be sorted manually with the hazards associated and inefficient sorting can result in costly storage solutions with inefficiently packaged waste.

The aim was to develop a prototype system for autonomously sorting radioactive waste with a robot handling system using AI (Artificial Intelligence) vision systems to identify waste objects, and a measurement and sorting system to stream the objects to a set of output waste containers. This will enable the sorting of waste more safely with no/minimal human intervention and create a more efficient packaging process, reducing waste storage capacity requirements.

Solution

• MTC developed an AI based multi-camera computer vision software package with a custom user interface.
• Grasp planning algorithms for vacuum and parallel gripper types were developed and integrated with the robot to pick the detected waste items.
• Object measurement was also performed through the vision system, generating important information for the waste sorting decision model.
• A bin packing algorithm was developed for optimising use of output bin space, reducing the number of container swaps required.

Outcome

• An intelligent system that identifies individual waste items using its vision system. Multiple views are fused into a 3D model for input to the grasping algorithms.
• Waste objects are picked up with an appropriate gripper for radiological and chemical analysis which drives output stream sorting.
• The vision system provides a traceable data record with every sorted object.
• The efficient packing algorithm maximises use of the output containers for safe processing or storage.

Impact

• Reduction or elimination of human manual sorting tasks which must currently be conducted in a high risk, dangerous and dirty environment.
• Trainable AI system will enhance performance though machine learning during use.
• Each object receives traceable record including images, measurement, radiological and chemical data, building a digital data set for life.
• The developed solution will allow improved sorting and segregation of nuclear waste ensuring waste streams are appropriately classified and managed in a safe manner, and reduce waste storage costs.

Mark Wickens, Robotics Programme Director at Atkins, said:

MTC worked brilliantly in our collaborative team to deliver a pragmatic solution to a cutting-edge challenge. They used their experience and expertise to develop a state-of-the-art vision system solution based on proven off-the-shelf hardware and software.

Tom Cockerill, Chief Engineer at MTC, said:

The flexibility of the system has proven capability to utilise intelligent automation in a challenging environment. This offers significant advantages in making the workplace safer for humans and enables increased operating capacity.