Wave Propagation Computations with Parabolic Equations: 3D, forest height and GPU

Authors:

  • Eric Corrigan
  • Aleksander Blagoiev
  • Gunnar Eriksson
  • Erik Axell

Publish date: 2024-10-15

Report number: FOI-R--5650--SE

Pages: 19

Written in: Swedish

Keywords:

  • radio wave propagation
  • parabolic equations
  • GPU

Abstract

Existing radio channel models used at FOI model the radio channel between transmitter and receiver in two dimensions, vertically and horizontally in the direction of wave propagation. With access to new map data with higher resolution and different categorization of terrain than in the past, there is an increasing need for modelling of the radio channel in three dimensions and with respect to the terrain. The main reason for this is that the radio signal can travel around an object in the horizontal direction, which is not considered in a two-dimensional model. Computational complexity is already a major challenge with existing models in two dimensions and with lower map resolution. Radio channel modeling in an additional dimension, and with higher resolution, increases the computational complexity significantly. Therefore, the development of methods for faster calculation of radio channel models is also required. The aim of this report is to increase knowledge of radio channel modelling, with a focus on using higher resolution terrain information well. This report presents research results about PE modelling in the following areas: full 3D modelling, simplified 3D modelling, analysis of the impact of forest height on wave propagation and faster wave propagation calculations with graphics processing units (GPU). Several existing 3D models have been evaluated and the need to develop a new 3D PE model has been identified. The effect forest height has on wave propagation is complicated and depends on frequency, antenna heights and dominant propagation modes. By parallelizing and performing calculations with the GPU, the computation time was reduced between one and 30 times, depending on system size and solution method. Overall, there is a need for further development of wave propagation models to utilize available map data for increased computational accuracy and efficiency.