Simplistic Machine Learning-Based Air-to-Ground Path Loss Modeling in an Urban Environment

Abstract

Unmanned aerial vehicles (UAVs) are being broadly employed lately in different domains because of their unique features such as ease of mobility and feasibility. A high fidelity communication link is the basis for guaranteeing the robustness of the UAV network between its ends. To offer reliable models for utilization in designing UAV communication systems, in addition to the processes of planning, deploying, and operating these systems, accurate estimation of the prevailing radio channel framework parameters is required. In this work, we suggest and present a strategy for constructing an empirical path loss (PL) model for air-to-ground radio frequency channels relying on machine learning (ML). ML regression algorithms including K-nearest-neighbors (kNN), Regression Trees (RT) and Artificial Neural Networks (ANN) are utilized in our versatile three-dimensional (3D) technique. To that end, we investigate the use of GPS coordinates (i.e., latitude, longitude, and altitude.) of both of the UAV transmitter and ground receiver, in addition to humidity, temperature and atmospheric pressure as features into the ML algorithm to predict the link PL. Hence, all environment parameters, and the corresponding implicit relationships are incorporated in the learning phase, and the subsequent prediction of the PL. The validity of our model and approach is verified through numerical results.