A Statistical Analysis of Feature Transformation for Efficient Localisation in Urban Environments

In this paper, we perform a transformation-based statistical analysis with an eye to designing a robust and efficient localisation scheme. To this end, we evaluate the coefficient of determination (COD) also denoted as $R^{2}$ on simulated electromagnetic wave propagation models in an urban environment. Our transformation-based statistical models show that two measurable network parameters, namely the received power (RP) and the time of arrival (ToA), present a strong correlation with a mobile user's given location. By transforming the network parameters we were able to achieve COD of 0.577 for the RP and 0.549 for ToA using the Modulus transformation. We believe that by exploiting the high correlation of parameter transformation, there is potential to design fast and robust machine learning localisation schemes through which the future location of a mobile user can be accurately and reliably predicted.