Augmenting large scale propagation models using field measurements

Abstract

There is a need to use the available radio spectrum more efficiently because of the current spectrum scarcity and the ongoing increase in demand for usable bands of the radio spectrum. One way to use the radio spectrum more efficiently is to share the licensed band when and where it is not used by the licensed user. Spectrum sharing approaches like Dynamic Spectrum Access (DSA) typically makes use of a Geolocation Database (GLDB) to determine where secondary users can reuse the licensed spectrum band without interfering with the licensed primary users. Television White Space (TVWS) networks operate on this principle. Radio wave propagation models are fundamental to geolocation databases and are used to predict the coverage areas of transmitters, and so also, determine where the spectrum can be shared. Modelling the attenuation of the signal power of a radio wave in a wireless channel is no trivial task and all propagation models have a degree of inaccuracy. Large scale radio propagation models suffer from inaccuracies because there are numerous parameters with variability to account for. This research study aims to develop and implement a method to augment large scale radio propagation models with field measurements in order to reduce the mean error between propagation predictions and field strength measurements. Field measurements includes field strength measurements, or signal power, and land cover data. Transmitters in the radio band (88 - 108 MHz) were available and selected for this study. A measurement study spanning a period of three months, was conducted to determine the average received power from the various transmitters at a single stationary receiver. The effective heights of the land cover data were varied and added to the Digital Elevation Model (DEM), hence affecting the path obstruction-loss calculations. The heights of the land cover data modelled the effect of the land cover data on the attenuation of the signal and not the physical height of the objects associated with the land cover type. The gradient descent method was used to adjust the heights associated with the different land cover classes, by minimising the Root Mean Square Error (RMSE) between the propagation predictions and the field strength measurements. The standard deviation and Pearson correlation coefficient were also considered when evaluating the augmented propagation model. The Longley-Rice irregular terrain model (ITM) and the irregular terrain model with obstructions (ITWOM) were chosen for this study. The results of the study showed that it is possible to augment large scale radio propagation models with field measurements and reduce the RMSE between predictions and field strength measurements by more than 7 dBm. The results of the augmented ITM follows. The RMSE was reduced from 12.16 dBm to 4.44 dBm. That is reduction of 7.72 dBm. The standard deviation increased from 5.87 dBm to 8 dBm. That is an increase of 2.13 dBm. The Pearson correlation coefficient reduced from 0.88 to 0.78 which is a reduction of 0.1. The results of the augmented ITWOM follows. The RMSE was reduced from 10.95 dBm to 3.12 dBm, which is a reduction of 7.83 dBm. The standard deviation increased from 5.87 dBm to 6.1 dBm. That is only an increase of 0.23 dB. The Pearson correlation coefficient increased from 0.88 to 0.9, which is an increase of 0.02. The augmented large scale propagation models require initial training but provide more accurate propagation predictions. This approach can be used to improve on the propagation models used in network planning, spectrum monitoring and spectrum management. A better informed decision regarding the use of spectrum can be made, to ultimately use the available spectrum more efficiently.