Indoor Localization Using Multi-operator Public Land Mobile Networks and Support Vector Machine Learning Algorithms

Given Public Land Mobile Networks (PLMN) ubiquitous infrastructure and excellent coverage, they can be used for providing location information for various location-based services, especially in indoor environments. This paper investigates indoor positioning solutions which utilize received signal strength measurements obtained by mobile device from PLMN cells belonging to multiple mobile network operators. Two indoor positioning methods, based on Support Vector Machine learning algorithms and space-partitioning principle, are proposed. The first method utilizes Support Vector Regression (SVR) algorithm, whilst the second one introduces space-partitioning principle and the combined use of Support Vector Classification and SVR algorithms. The proposed techniques are thoroughly investigated in a real indoor environment. In addition, several criteria for choosing relevant cells for the positioning purposes have been explored. Positioning with SVR has demonstrated good results, while utilizing space-partitioning principle has further reduced the average positioning error by 27%. Moreover, the proposed solution has outperformed positioning methods based on k Nearest Neighbours and Artificial Neural Networks, when implemented in the same verification test bed.