Direct Learning Localization in the Presence of Multiple Access Interference

This article demonstrates that the performance degradation in wireless communication localization can result from multiple access interference (MAI). To address this issue, we introduce a new supervised machine-learning technique named direct learning localization (DLL). This approach determines user locations by analyzing the statistical model of received signal strength (RSS) at base stations. Our method utilizes the probability density function (PDF) of RSS rather than its instantaneous values. It operates independently of the prior knowledge of channel statistics and MAI profiles, as the RSS PDF is derived from direct observations. Consequently, our method obviates the need for interference cancellation algorithms and incorporates interference profiles as a part of the localization fingerprint. DLL calculates the Kullback-Leibler (KL) divergence between the RSS PDFs of training and test users as a measure of dissimilarity. Utilizing these dissimilarities, it then predicts user locations with a logistic regression model. We demonstrate that DLL significantly enhances localization accuracy in the presence of MAI and outperforms other methods in the literature. Additionally, we present an analysis of the algorithm's performance to prove its reliability.