Use of Gaussian process regression for radiation mapping of a nuclear reactor with a mobile robot

Abstract

Collection and interpolation of radiation observations is of vital importance to support routine operations in the nuclear sector globally, as well as for completing surveys during crisis response. To reduce exposure to ionizing radiation that human workers can be subjected to during such surveys, there is a strong desire to utilise robotic systems. Previous approaches to interpolate measurements taken from nuclear facilities to reconstruct radiological maps of an environment cannot be applied accurately to data collected from a robotic survey as they are unable to cope well with irregularly spaced, noisy, low count data. In this work, a novel approach to interpolating radiation measurements collected from a robot is proposed that overcomes the problems associated with sparse and noisy measurements. The proposed method integrates an appropriate kernel, benchmarked against the radiation transport code MCNP6, into the Gaussian Process Regression technique. The suitability of the proposed technique is demonstrated through its application to data collected from a bespoke robotic system used to conduct a survey of the Joz̆ef Stefan Institute TRIGA Mark II nuclear reactor during steady state operation, where it is shown to successfully reconstruct gamma dosimetry estimates in the reactor hall and aid in identifying sources of ionizing radiation.

Real-Time Avoidance of Ionising Radiation Using Layered Costmaps for Mobile Robots

“…However, in the current work, experiments are performed in both simulation and in an active environment with real ionising radiation sources. The robot modelled in simulation is a Clearpath Jackal, reflecting the hardware used in active deployments ( Bird et al, 2018 ; West et al, 2021b ; Tsitsimpelis et al, 2021 ), equipped with a single front-facing lidar sensor, utilising the GMapping 3 package for SLAM. The robot uses a four-wheel skid-steer drive for forward/reverse locomotion and yaw control.…”

“…The count rate was reported to the robot via ROS at a rate of 1 Hz, integrated over an energy range of 300–2,500 keV. More details regarding the platform and radiation sensing can be found in the work of Tsitsimpelis et al (2021 ) and West et al (2021b ). An annotated picture of the deployed platform is shown in Figures 6 , 2D .…”

“…The method of inverse distance weighting was chosen for interpolation, which relies on the principle that locations close to each other are more highly correlated than distant locations, and by using surrounding observations, a value at a new location can be inferred. The use of an inverse distance weighting approach has benefits over other approaches in that it can handle fluctuating, sparse, and repeat measurements, which many interpolation techniques cannot ( West et al, 2021b ), but is considerably quicker to compute than more accurate methods such as Gaussian process regression.…”

Stiffness estimation of planar spiral spring based on Gaussian process regression

“…The study verifies that the improved GPR can deal with high-dimensional input vectors with complex relationships because all the signal features are extracted from time-domain, frequency-domain, and wavelet-domain, respectively. West et al 14 utilise the capacity of interpolation and extrapolation of GPR with sparse, highly clustered data to produce smoother estimates of radiation intensity in the reconstruction of nuclear radiation map when provided with uncertain, irregularly spaced observations. The fields of the applications of GPR include but are not limited to environment 15 – 17 , materials 18 , engineering 19 , 20 , and finance 21 .…”

Sparse Gaussian Process Regression for Landslide Displacement Time-Series Forecasting

“…In recent years, Gaussian process regression (GPR) has demonstrated superior predictive strength in many regression tasks (He and Kusiak 2017;Schulz et al, 2018;Li et al, 2021a;Li et al, 2021b;Deringer et al, 2021;Jamei et al, 2021;West et al, 2021;Fuhg et al, 2022;Tamhidi et al, 2022). Generally, GPR is a nonlinear, nonparametric regression tool useful for interpolating data points scattered in a high-dimensional input space.…”