COMPARISON OF POSITIONING ACCURACY OF GRID AND PATH LOSS-BASED MOBILE POSITIONING METHODS USING RECEIVED SIGNAL STRENGTHS Elina Laitinen, Jukka Talvitie, Elena-Simona Lohan and Markku Renfors Department of Communications Engineering Tampere University of Technology P. O. Box 553, FIN-33101 Tampere, Finland E-mail: {elina.laitinen, jukka.talvitie, elena-simona.lohan, markku.renfors}@tut.fi ABSTRACT In this paper, positioning accuracy of two basic mobile po- sitioning methods based on Received Signal Strength (RSS) measurements is compared via comprehensive simulations. The first method is a path loss-based technique using gen- eral free-space loss model and brute-force algorithm. The second method is a grid (fingerprint)-based positioning ap- proach, with averaging over several nearest neighbor grid points. In the analysis, these two methods are compared in various outdoor scenarios using Matlab simulations. It is shown that the grid-based method performs better when fast fading is present and/or shadowing variance is 10 dB or more, even if path loss parameters are perfectly estimated. In addition, for the grid-based method, the optimal choice of the number of neighbor points for averaging is shown. Keywords: Positioning, Received Signal Strength, Grid Point, Fingerprints, Path loss, Nearest neighbors. 1. INTRODUCTION The number of different positioning applications has been enormously increasing in many fields, ranging from busi- ness and leisure time to surveying and health care. The re- ceiver unit for Global Positioning System (GPS) has been included in many new mobile phones, and also applications like Google Maps are available in many new models. How- ever, there is still a demand especially for low cost position- ing techniques, e.g., for cheaper mobile phones without GPS or other satellite-based positioning systems. GPS also has performance limitations in certain circumstances, such as ur- ban canyons, and relatively high power consumption when implemented into a cell phone. Several different cellular mobile positioning techniques for outdoor situation based only on Received Signal Strength (RSS) have been proposed over the years. One approach is a path loss method, where position of the mobile station (MS) is estimated based on some signal propagation model. RSS from several Base Stations (BS) is measured, and the posi- tion estimate is then calculated via trilateration using esti- mated distances between BSs and MS [1, 2]. In order to cal- culate a unique location estimate for the mobile, distances from at least three BSs are needed. In addition, BS locations are needed as well. [2] In the grid-based or fingerprint po- sitioning methods, RSS values are compared to a database with a number of pre-measured signal strength samples with known locations. The best hit, i.e., the most similar situa- tion is then decided to be the estimated location of MS [3]. The main disadvantage of the grid-based method is database generation and maintenance requirements. However, it has to be taken into account that some pre-measured samples are needed in the path loss approach as well for estimating the path loss channel characteristics. Even though different RSS based positioning techniques are widely known, a fair comparison of path loss-based and grid-based approaches in terms of positioning accuracy for outdoor case cannot be found in the literature so far to the best of the Authors’ knowledge. Most publications related to the subject are evaluating the performance for only one of these methods, such as [4, 5] for fingerprints and [6, 7] for path loss approaches, or are limited to indoor circum- stance only (such as [8, 9, 10]). Some related works are also utilizing Wi-Fi networks for outdoor positioning, e.g., [1]. Therefore a comprehensive comparison of these two ba- sic positioning methods for outdoor case with varying fading phenomenon is very interesting and important indeed when an RSS based positioning technique is intended to be im- plemented for mobile phones or other receiver equipment. Moreover, a small enhancement of grid-based positioning based on averaging over a certain number of neighbor mea- surements is proposed here. Brute-force algorithm with path loss model parameter estimates is used for the path loss approach when estimat- ing user location. In the positioning phase of the grid-based method applied in this paper, also averaging over N neigh near- est neighbor grid points is used, starting from some ideas found in [3] and [11] and developed further here with respect to the optimum neighbor points. The differences between RSS levels of the user and RSS values in the database are calculated, and the estimated position for mobile is then cal- culated as mean of known locations of N neigh grid points with the minimum difference. The analysis is performed in var- ious scenarios, i.e., with/without shadowing, with/without fading and for both sectorized and omni-directional anten- nas, using Matlab simulations. Both cell radius and num- ber of hearable BSs are varied. Traditional Cell-ID (CID) method is included as benchmark. In addition, the optimal value of N neigh neighbor points for averaging is searched and results are shown here, using both simulated data and real data measurements. Proceedings of SPAMEC 2011, Cluj-Napoca, Romania 1 © EURASIP 2011