A Train-Once Approach for Location Estimation Using the Directional Propagation Model

Abstract

Location estimation that is based on the mobile phone network has drawn considerable attention in the field of wireless communications. Among the different mobile location estimation methods, the one that estimates a mobile station location with reference to a wave propagation model is shown to be effective and is applicable to different kinds of cellular networks, including Global System for Mobile Communications (GSM), cdmaOne, CDMA2000, and the Universal Mobile Telecommunications System. We have designed a train-once approach for location estimations using the directional propagation model (DPM). The DPM is an improved model that is based on the traditional free-space wave propagation model with the directional gain and environmental factors integrated in the estimation. The train-once approach works because we observe that different types of antennas are designed for different types of environments. Thus, a parameter estimation is related to the antenna type and, in turn, related to the environment. In this paper, we report our study of the train-once approach with the DPM for location estimations. We have tested our model with 192 177 sets of real-life data that have been collected from a major mobile phone operator in Hong Kong. Experimental results show that the train-once approach with the DPM is practical and outperforms the existing location estimation algorithms in terms of accuracy, stability among different types of terrains, and success rates.