The Study of Empirical Path Loss Models for Accurate Prediction of GSM Signal for Applications in Microcellular Environments

Abstract

Path loss models are used generally in GSM signal predictions, optimization of coverage and interference analysis. Peculiarity of these models gives rise to high errors in prediction when they are deployed in environments different from the one they are designed for. In this study, the fitness of six path loss models was evaluated using five metrics to gauge their performance. In order to achieve this, empirical prediction models were fitted with the measurement survey data performed on live signals transmission on microcellular service provider in an outdoor propagation in urban, suburban and rural environments of two states in the western part of Nigeria.

The results show that no single model consistently fits the path loss values along the routes. However, Cost-231, Erceg and SPM models provided better fitness in some monitored base stations with RMSE values slightly above 7dB for urban environment and less than 10 dB in some suburban and rural environments. The prediction errors distribution for Erceg model distributes nearly symmetrical around a mean error of 0.71dB for urban, -8.56 dB for suburban and -1.67 dB for rural environments and some of the model prediction error distributions fairly follow a normal distribution curve. 

The Gaussian error distribution within the windows of 0 to ±10 dominates the frequency counts. From the analysis, Erceg and Cost-231 presented the closest fit and least RMSE value. However, adjustment of the parameters of the Erceg and Cost-231 models is necessary to minimize the RMSE values to the acceptable ranges suitable for all the environments.