부산대학교 도서관

We propose a radio-based approach for height classification of mobile devices in cellular networks for the purpose of enabling the network infrastructure to distinguish between ground users and aerial devices like drones. The classifier is based on learning the properties of the reference signal received power (RSRP) values that each device obtains from base stations. Scenario-based simulations using the Vienna 5G System Level Simulator with adopted base station antenna patterns demonstrate the feasibility of decision tree classifiers with an average misclassification rate at about one percent with three height levels. It is shown that decision trees outperform other classification algorithms in this context.