A Low-Complexity Path Loss Prediction Model for Multi-UAV Communication in Urban Environments

Ray tracing method (RT) is a computational method for modeling light propagation and interaction, which can be used to simulate communication signal propagation, path loss, etc. in unmanned aerial vehicles(UAV). However, the ray tracing method may face some challenges in real-time simulation of multi-UAVs. In this paper, we propose a low-complexity deterministic path loss prediction model based on RT and backpropagation neural network (BPNN) for online decision-making systems of multiple UAVs in urban environments. The dataset is built by calculating the path loss (PL) between each pair of receivers and transmitters in a deterministic urban scenario via RT. However, numerical results show the high complexity when using the path loss dataset directly to train the neural network. To address this, we simplify the PL dataset by incorporating communication system parameters (e.g., receiver sensitivity and transmission power) and categorizing the path loss severity into four levels: Low Path Loss, Moderate Path Loss, High Path Loss, and Severe Path Loss. The resulting low-complexity path loss prediction model, trained with BPNN and the simplified dataset, achieves efficient PL prediction while maintaining high accuracy. Compared to RT, the proposed model is computationally less time-consuming and can effectively predict the degree of path loss attenuation across various UAV distributions and operating frequencies. Furthermore, it provides a practical solution for assessing communication performance under dynamic urban propagation environments and varying signal conditions, including channel fading.