Mobile position estimation with RSS Based techniques in an urban city with multiple multi-storied structures

Accurate position estimation is the vital issue in Location Based Services (LBS) of 3G communications. The quality of LBS depends on the positioning technologies used and the network topology. In this paper, the performance of four prominent self-position fixing methods namely i) RSS Statistics, ii) Least squares (LS) iii) Weighted LS (WLS) and iv) Constrained WLS (CWLS) are compared. The estimations are based on RSS measurements from surrounding base stations (BS) in GSM network. In statistics of Received Signal Strength (RSS) method, 2-D location of MS is estimated based on PDF of propagation distances derived from RSS data using Minimum Mean Square Error criterion. In LS estimation techniques, correlation that exists between propagation distances and RSS measurements is exploited. The performance of WLS method is better than LS method with the inclusion of weight matrix. In CWLS, Lagrangian multiplier is introduced to include the radial distance constraint to improve reliability and integrity of estimates. Experiments are conducted to acquire RSS data by using Test Equipment for Mobile Systems (TEMS) investigation tool and GSM/GPRS 900/1800MHz dual frequency modem for both stationary and kinematic modes of mobile. RSS data acquired in the coverage area of Tirumalgiri BS (17° 28' 11", 78° 30' 55", 560m) Hyderabad city, India. BS antenna is at the height of 33m. It transmits power at an effective isotropic radiated power (EIRP) of 43 dBm on 950.6 MHz frequency. The acquired field strength values at various points are ranging from -80dBm to -35dBm. For a typical urban GSM cell structure (σ
_shadowing
= 6 to 8dB) simulated position estimates are obtained. Experimentally observed results are verified with GPS measurements and found that accuracy is limited by number of hearable BSs involved in the estimation. Comparative error analysis indicates that the accuracy of the location estimation greatly depends on the cell size. Error is least when the BSs involved are well distributed in all the directions around and nearer to the MS which has least Geometric Dilution of Precision (GDOP). It is found that, CWLS method performs well in a densely populated urban city with Multiple Multi-storied structures. Experimental results indicate that CWLS offer less Circular Error Probability (86.7m), RMSE (2.58m), least GDOP (2.0) and error a (1.54m) with 7 BSs. The result will be an important input for designing location assisted handoffs required for reliable mobile links and in deployment of BSs at proper locations to achieve uninterrupted coverage.