A recent arXiv preprint outlines a system that brings machine learning to radiation classification within the Robot Operating System (ROS). The study describes two ROS nodes built to differentiate seven radioactive isotopes, though the abstract provides limited detail on the overall architecture.
The first node, called the model training node, relies on Gradient Boosting, a popular ensemble learning method. The abstract cuts off after mentioning this node, so the exact role of the second node and the full training pipeline are not available from the source.
The work is notable for combining ROS, a standard robotics middleware, with AI-based classification of nuclear materials. This could be relevant for robots operating in environments where identifying radioactive sources is necessary, although the abstract does not discuss specific applications or performance metrics.