Advancing Location Accuracy via Collimated Nuclear Assay for Decommissioning Robotic Applications (ALACANDRA)

Radioactivity is all around us but it is usually dispersed such that it poses little risk to human health. However, past industrial activities associated with nuclear weapons production, the manufacture of fuel for nuclear power stations and the management of radioactive waste from these activities have resulted in a significant number of highly contaminated facilities. The level of contamination can be so great that people cannot enter because the radiation level is too high. Further, because we do not understand the long-term risks associated with low-level radiation exposures, entry to place contaminated less is often discouraged to minimise any risk that there might be. Matters are complicated further because difficulty getting inside complicates our ability to understand exactly what needs to be done to make these places safe.

Some of these facilities are not safe because they are old and were not designed to last this long. It is important to make them safe now to ensure radioactivity does not get out, and because the longer this takes the more difficult and expensive it becomes as new problems arise. However, this will take a long time to complete: at Sellafield, the time needed to complete this is forecast to be 120 years. This means that if they are not dealt with effectively now, these problems will fall to future generations; hence, from an ethical standpoint, the imperative is to prevent this by action now.

One way to understand these radiological hazards is to send in a robot. Great advances have been made in this regard as a result of recent research, done in part by the people leading this proposal. However, simply transporting a radiation detector into a place and trying to determine where it detects the most radiation does not work for two important reasons: Firstly, radioactivity in these places is often dispersed, meaning that it is not concentrated in one place that might be dealt with easily and quickly. Instead, contamination arises from leaks, splashes, tide marks in vessels and it migrates into porous materials, yielding a 3D distribution in space. Radiation detector systems and imagers have difficulty with this because they often provide an assessment from a particular perspective that may not tell us everything we need to know. Secondly, contaminated places are often cluttered with process equipment, detritus and construction materials. These can cause the radiation to scatter and also absorb it. This influences the 'picture' and can influence how much radioactivity is thought to be present.

With a human 'in the loop' - in the space with the contamination - they could improvise by moving to different vantage points, moving debris out of the way and by inferring what is involved from what they see. This not being possible, the use of a commercial robotic platform constitutes a way by which this might be replicated. For example, by assessments from a number of complementary vantage points and fusing the data obtained from this variety of perspectives. However, it is important to maintain human oversight of these operations by driving the robot rather than affording it full autonomy in case difficulties arise in recovering it etc. This raises the question: How can we interpret robot-derived information from a variety of perspectives, from a cluttered space contaminated with dispersed radioactivity, to help us understand what hazards may exist, quickly and effectively? Our research appeals directly to this requirement: we suspect that a detector's response is related to a relatively simple combination of sub-responses, as if the contamination were comprised of pixels of contamination. By advancing our interpretation of the combined influence of these on a radiation detector system configured by a robot, we hope to connect what we observe with nature of the radioactivity that is present, hence enabling robots to assist in the clean-up of these spaces more efficiently.