Building 5G Fingerprint Datasets for Accurate Indoor Positioning

Abstract

The fifth generation (5G) of mobile communication technology has developed rapidly in recent years. Millimeter wave (mmWave) communication, multi-input-multi-output (MIMO) techniques and beamforming technologies are widely considered for the 5G communication systems. The deployment of 5G networks in most countries is still sparse and real-world 5G signal acquisition is yet difficult and expensive. Therefore, simulation of the 5G environment and signal becomes a critical and vital approach for the research and development in various aspects of 5G wireless networks. The challenge is even more serious in the research of this domain where access to reliable datasets or regenerating simulated data to develop or improve solutions are sometimes extremely difficult processes or impossible. In this paper, we address this gap in the literature by developing a simulator for a 5G environment which considers the design of any urban area and generates beamformed MIMO air interface signals. This simulator is a key step to generate near-realistic data samples (i.e., dataset) which can be further used for various research topics on the 5G. As an example, we use this simulated data for the training of the machine learning models for an indoor positioning use-case scenario. The deterministic three-dimensional raytracing techniques are used to build the simulation model via a commercial software Wireless Insite. This paper describes the structure of the simulator, explains the details of generating and collecting the data samples, and interprets the obtained datasets for indoor localization, as a use-case example. The main goal here is to provide sufficient information and resources to regenerate this dataset for future research works on similar topics.