Edited by: Chie Takahashi, University of Cambridge, United Kingdom
Reviewed by: C. James Taylor, Lancaster University, United Kingdom; Sean Andersson, Boston University, United States
This article was submitted to Sensor Fusion and Machine Perception, a section of the journal Frontiers in Robotics and AI
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
The use of a robotic arm manipulator as a platform for coincident radiation mapping and laser profiling of radioactive sources on a flat surface is investigated in this work. A combined scanning head, integrating a micro-gamma spectrometer and Time of Flight (ToF) sensor were moved in a raster scan pattern across the surface, autonomously undertaken by the robot arm over a 600 × 260 mm survey area. A series of radioactive sources of different emission intensities were scanned in different configurations to test the accuracy and sensitivity of the system. We demonstrate that in each test configuration the system was able to generate a centimeter accurate 3D model complete with an overlaid radiation map detailing the emitted radiation intensity and the corrected surface dose rate.
The global nuclear industry is facing significant challenges in decommissioning and nuclear waste management owing to an ever-increasing amount of nuclear waste awaiting to be processed and prepared for long-term storage. In 2018,
An alternative approach, which is only now becoming possible, is in the use of robotic manipulators equipped with micro gamma-spectrometers to scan waste packages in a more dynamic way. Such solid-state detection units are usually very compact with detector crystals of 1–30 cm3 and fast counting rates (typically 20,000 cps) able to discern different gamma-emitting radioisotopes based on their differing decay energies. Conceptually, they enable radiation scans to be performed robotically at much smaller stand-off distances (<10 cm) than segmented gamma scanning, yielding a much higher spatial resolution and sensitivity. However, in order to conduct such close-proximity scanning, a method of determining the sensor stand-off distance must also be integrated. To touch the waste material could potentially contaminate or damage the detector and hence this needs to be prevented at all costs.
There are therefore numerous robotic technologies and sensors which are capable of being combined to achieve this target of combined gamma scanning and 3D profiling of nuclear waste objects—yet to our knowledge this has not previously been reported. The use of robotics for gamma inspection in the nuclear industry is not new, but has typically been employed for plant inspection and not waste assay. Tsitsimpelis et al. (
The use of point cloud data in the formation of 3D models is another recent innovation that the field of robotics has been quick to adopt. Various techniques are routinely being applied to generate point cloud data, which in turn can be transformed into a 3D model. Within nuclear robotics, LiDAR scanning is already an established technique used in 3D environment reconstruction. Aerial radiation mapping routinely relates recorded radiometric data to a 3D model collected by either LiDAR or photogrammetry to produce a combined 3D representation (Connor et al.,
The research in the current manuscript adopts a similar process to Martin et al. (
Robotic manipulation and instrument scanning must be well-synchronized to provide effective integrated measurements. For radiation scanning a Kromek
A figure showing the 3D model and overlaid radiation map generated when the robot scan was completed at a 1 cm standoff above 2 Cs-137 sources 30 cm apart. Left to right the source activity is 7.5 and 10 μSvh−1, respectively. The color represents the gamma radiation counts in counts per second for a given 2 × 2 mm pixel.
A figure showing the 3D model and overlaid radiation map generated when the robot scan was completed at a 1 cm standoff above 2 Cs-137 sources directly adjacent. Left to right the source activity is 7.5 and 10 μSvh−1, respectively. The color represents the gamma radiation counts in counts per second for a given 2 × 2 mm pixel.
A figure showing the 3D model and overlaid radiation map generated when the robot scan was completed at a 1 cm standoff above 2 Cs-137 sources directly adjacent, from a second angle to aid visual clarity of the distinguished pucks. Left to right the source activity is 7.5 and 10 μSvh−1 respectively. The color represents the gamma radiation counts in counts per second for a given 2 × 2 mm pixel.
A figure showing the 3D model and overlaid radiation map generated when the robot scan was completed at a 1 cm standoff above 2 Cs-137 sources 1 cm apart. Left to right the source activity is 7.5 and 10 μSvh−1, respectively. The color represents the gamma radiation counts in counts per second for a given 2 × 2 mm pixel.
A figure showing the 3D model and overlaid radiation map generated when the robot scan was completed at a 1 cm standoff above 2 Cs-137 sources 30 cm apart. Left to right the source activity is 7.5 and 10 μSvh−1, respectively. The color represents the radiation dose rate in μSvh−1 for a given 2 × 2 mm pixel.
A figure showing the 3D model and overlaid radiation map generated when the robot scan was completed at a 1 cm standoff above 2 Cs-137 sources 1 cm apart and 2 Pitchblende sources 30 cm apart. Left to right the source activity is 4.5 μSvh−1 Pitchblende, 7.5 μSvh−1 Cs-137, 10 μSvh−1 Cs-137 and 4.3 μSvh−1, respectively. The color represents the gamma radiation counts in counts per second for a given 2 × 2 mm pixel.
A figure showing the 3D model and overlaid radiation map generated when the robot scan was completed at a 1 cm standoff above 2 Cs-137 sources 1 cm apart and 2 Pitchblende sources 30 cm apart. Left to right the source activity is 4.5 μSvh−1 Pitchblende, 7.5 μSvh−1 Cs-137, 10 μSvh−1 Cs-137, and 4.3 μSvh−1, respectively. The color represents the radiation dose rate in μSvh−1 for a given 2 × 2 mm pixel.
A flowchart showing how the system was integrated.
The data collected by the arm contained a full spectrum of 4096 energy bins ranging from 50 keV to 2 MeV. This allows for the distinct Cs-137 peak at 662 keV to be identified. A python script was written to calculate the counts identified within the peak range and subtract from that the baseline reading to correct for background. A multiplication factor was applied to convert the raw count value within the energy range to a dose rate in μSvh−1, using the method described in Connor et al. (
For scanning tests the robot was programmed to perform a basic raster pattern scan of the “scan surface”, which was a 0.6 × 0.26 m area. The raster scan had a step length of 1 cm and the speed of the scan could be varied, depending on the activity of the test sources, less active sources require longer counting times (slower scans) to achieve adequate detection. Sealed radioactive sources containing caesium-137 (Cs-137), one of 7.5 μSvh−1 contact dose and the other at 10 μSvh−1 contact dose were used, alongside naturally occurring uranium (pitchblende) sealed sources of 4.3 and 4.5 μSvh−1 contact dose rate to test the radiation response of the system. The robot arm was set up to move at a consistent speed in a continuous linear motion of 10 mm per second, with a single scan taking approximately 30 min. Scan time could be reduced if higher activity sources were used, but as a proof of concept where timing is not restricted, greater scan time is able to yield a higher resolution radiation map, with a longer exposure for each collection interval. A photograph of the scanning system is shown in
A photograph of the scanning system in action.
To test the system, several different scanning scenarios were set up using the Cs-137 and Pitchblende sources available. The first used 2 Cs-137 sources separated apart by a distance of 30 cm, center to center. The scanning head was programmed to trace the raster scan path designed at a rate 10 mm per second. The resulting data can be seen in
This result demonstrates the capability of the 3D model generation using the ToF sensor, as it generates an identifiable geometric representation of the source pucks. The data is displayed with the counts recorded by the detector for a given 100 ms exposure. Following this experiment, the two sources were placed directly next to each other. This served as a test of the ToF mapping procedure. The resulting figure is shown in
From this we can clearly identify the radioactive hot spot. In addition the physical 3D separation of the sources can be comprehended. It is easier to distinguish on software which enables the rotation of the generated model. In order to aid the visual clarity in distinguishing the pucks,
To further monitor this 3D modeling routine, the sources were next placed 1 cm apart from each other. The result of this test is shown in
Here the centimeter gap is visible and both sources may be physically distinguished. This concept works and is sufficient for identifying radiation hotspots present in given scan and sort scenarios. However it does not provide any dose rate information. The data was subsequently processed to give an estimate of the dose rate of the given sources. The sources comprised of Cs-137 containing moss samples collected from Fukushima and were hence not perfect point-source emitters. Instead we approximated each source as a point emitter where the origin of the point was 3 cm beneath the puck surface, which corresponds to the thickness of the perspex and a small air gap, as shown in
A diagram explaining the point source modeling used to invoke the inverse square law.
For our calculations we assume that the perspex is gamma transparent due to its low density and Z number. Hence we may apply a correction factor to the dose rate method described in the above section, to display the data in surface dose rate format. The two Cs-137 sources positioned at 30 cm apart may be seen on a dose rate map in
The dose rates recorded at the 1 cm standoff are in good agreement with the actual recorded contact dose rates measured for each source: 7.5 and 10 μSvh−1. This is very promising in demonstrating that the system sensitivity is sufficient to discriminate radioactive objects that classify at the Very Low Level Waste (VLLW) to Low Level Waste (LLW) threshold. In the UK VLLW is normally regarded as material with a specific activity up to 100 Bqg−1 (RWM,
An ability to measure the full gamma radiation spectrum enables different radioisotopes to be identified.
One can determine they are all radioactive, with different intensities, but not tell which source is which. By restricting the spectral window to only the 662 keV gamma photons, one generates, as a Cs-137 specific plot, as in
Using spectral gating like this the Pitchblende sources become invisible to the radiation scanning system. This is key for the nuclear industry, as it would enable radionuclide characterization and separation to be carried out autonomously.
This paper introduces a novel integration of technologies that facilitates the scanning of radioactive materials and waste-forms, creating a 3D model of the object or environment and adding an overlaid radiation map. The setup is comprised of certified commercial off-the-shelf (COTS) components that in consequence require little control performance verification. Each component is integrated in a modular manner system, allowing for a highly flexible system design. The nature of the integrated system means that a multitude of sensor packages and grippers could be added alongside or in place of the combined detector unit used in this paper. Even the robotic arm used for this project could be replaced with an alternate choice as there are numerous different robotic arm systems commercially available of different sizes, lift capabilities, reaches and radiation tolerances that could enable a range of different scale applications, from sorting bulk rubble or pipework to sifting sediments for hot micro-particles.
Our experiments demonstrate that it is possible for such a scanning system to make very accurate, high sensitivity, high spatial resolution radiation maps for resolving nuclear waste materials from each other on the basis of emitted gamma intensity. The result was also successful from a 3D modeling perspective, as it clearly identifies the sources as separate objects despite their close proximity. The spherically symmetric radiative flux which is emitted from the radioactive sources means that there are limitations on identifying which physical shape corresponds to which emission on the radiation map. This is something that could be improved by an algebraic reconstruction technique and a comprehensive understanding of the detectors response. This forms an important part of the future work this project will require. The radiation sources used in this work were relatively weak compared to real ILW and LLW. For real waste scenarios it would be expected that a smaller micro gamma spectrometer with greater peak dose measurement capability e.g., CZT or GaAs, could be utilized. Scan times would also be dramatically reduced with increasing radioactivity levels.
The next proposed step in development is to use the laser profiling to generate a volume for the object and then based on an assumed density, e.g., for concrete, the system could be developed to automatically threshold objects and assign them as either VLLW, LLW, or ILW based on their radioactivity and calculated mass. This thresholding programme would need to utilize an algorithm based on the inverse square law to calculate the intensity of the emitted radiation at the surface of the objects being scanned, accounting for the efficiency and solid angle of the radiation detector being used. It is crucial to get an accurate distance from the detector to the target object because contact dose calculations are based on the inverse square law. This means any discrepancy in reading will significantly affect the corrected dose calculation. Ultimately, the full gamma spectrometry capability of the system could be used to distinguish different gamma emitters too. This would add a further level of finesse for separating mixed nuclear wastes that might be expected to arise during nuclear decommissioning activities. The accuracy of the 3D model would facilitate a robotic manipulator for grasping in addition, fulfilling the requirements of the sort and segregation table.
The system's scanning methodology could also be substantially refined vs. the current simple raster scanning we have demonstrated. For example, an initial survey scan could conduct rapid raster scan of a scene to determine the degree of variability in scene topology and from that, determine automatically a more detailed scanning path that would maintain a safe but close scanning proximity to the waste objects. The initial survey scan would also locate any strong radiation emitters, which the adaptive path plan could deferentially focus on to provide a more detailed scan of that specific area of the scene.
This work demonstrates the use of a combined laser profiling and gamma-scanning sensor unit, mounted on a robot arm, to form an accurate 3D profile of a series of test objects on a scanning table, with a coincident overlay of the mapped radiation intensity. Radiation maps are successfully created by the system, which is able to correctly identify radioactive sources of different intensities on a flat scan surface. The generated 3D surface model reveals an accurate visualization of the tested scene and is accurate to within a centimeter. In addition to this an estimate of the surface dose rate produced by the radioactive emitters is made based on the scan data received to a good level of accuracy, correctly identifying the dose rate of two radioactive Cs sources to within 1 μSvh−1.
The raw (unprocessed) data that supports the findings from this study are available from Mendeley Data with the link: doi:
All authors contributed technically, both in the data collection and writing of this paper.
JD was employed by the company KUKA Systems UK Ltd. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The authors would like to thank John Jowsey and Alex Jenkins from Sellafield Ltd., for their valuable technical input. We would also like to thank KUKA AG., for providing technical support for the project and Jim Brooke, DRPS for the School of Physics for his assistance with the radiation sources used in this experiment.