UAV-UGV Teaming for Rapid Radiological Mapping

Abstract

This paper presents a novel configuration of UAV-UGV teams for rapid radiological mapping. The UGVs are equipped with low cost Geiger-Müller counters whose measurements are simulated using Poisson statistics. Gaussian Process Regression (GPR) is used to generate a model of the radiation field that includes uncertainty estimates. In the current work, the UAVs do not have sensors and only act as carrier drones for the UGVs equipped with sensors. The UAVs leverage information-driven path planning where the metric for information is the uncertainty in the GPR model. This information metric is used to determine regions to deploy the UGVs. The UGVs cover their given region using Boustrophedon cellular decomposition. Monte Carlo studies show that UAV-UGV teams using information theoretic path planning (ITPP) are able to lower the model error significantly faster relative to control experiments with UGV-only mapping or with UAV-UGV teams performing random sampling (RS). The model error decays exponentially for the UAV-UGV teams but only linearly for the UGV-only teams. These results illustrate a potential system concept for UAV-UGV teams performing radiation mapping and provide baseline results quantifying potential performance improvements over systems employing only mobile ground sensors.