Adaptive localization in wireless networks

Abstract

Indoor positioning approaches based on communication systems typically use the received signal strength (RSS) as measurements. To work properly, such a system often requires many calibration points before its start. Based on theoretical-propagation models (RF planning) and on self-organizing maps (SOM) an adaptive approach for Simultaneous Localization and Learning (SLL) has been developed. The algorithm extracts out of online measurements the true model of the RF propagation. Applying SLL, a self-calibrating RSS-based positioning system with high accuracies can be realized without the need of cost intensive calibration measurements during system installation or maintenance. The main aspects of SLL are addressed as well as convergence and statistical properties. Results for real-world DECT and WLAN setups are given, showing that the localization starts with a basic performance slightly better than Cell-ID, finally reaching the accuracy of pattern matching using calibration points. © 2008 Springer US.