Nuclear detectionScavenger hunt for simulated nuclear materials

Published 16 July 2018

Competing in a fictitious high-stakes scenario, a group of scientists at the Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) bested two dozen other teams in a months-long, data-driven scavenger hunt for simulated radioactive materials in a virtual urban environment. The competition platform was also built and managed by Lab researchers.

Competing in a fictitious high-stakes scenario, a group of scientists at the Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) bested two dozen other teams in a months-long, data-driven scavenger hunt for simulated radioactive materials in a virtual urban environment.

The goal of this hackathon-styled event was both to improve the detection methods that could be applied to actual threats involving nuclear materials, and to create a platform to virtually vet out these methods. The computer-programming challenge ran from 22 January to 14 May and featured 66 researchers on 25 teams based at DOE national laboratories and other U.S. government research labs. The teams received data that simulated radiation sources scanned by a vehicle-mounted detector system traveling along urban streets.

Teams were scored on the success of the algorithm or algorithms they developed in sleuthing the time slots marking the detection of the simulated radioactive hotspots likely associated with potentially hazardous nuclear materials – and in ruling out common sources of radiation. Government agencies have previously conducted many field tests and performed comparisons to study how to find radioactive materials in different settings, but this was the first online competition.

LBL says that Tenzing Joshi, an applied nuclear physicist in Berkeley Lab’s Nuclear Science Division, led the winning team in this Urban Radiological Search Competition created by the DOE’s National Nuclear Security Administration. His teammates included Mark Bandstra, a senior scientific engineering associate, and UC Berkeley graduate student Kyle Bilton.

Berkeley Lab served as the host institution for the competition, which was also administered by Los Alamos National Laboratory and Oak Ridge National Laboratory. A group of researchers in Berkeley Lab’s Computational Research, Nuclear Science, and Information Technology divisions teamed with other researchers in the Nuclear Science Division to develop the competition platform.

Berkeley Lab organizers leveraged the Lab’s capabilities in support of big data science to develop and roll out the competition platform, and they hope it can be used in other data analytics challenges. The Lab’s Applied Nuclear Physics program has been developing a variety of mobile and portable detector systems to quickly identify radiation sources.