Mobile Device Positioning by Using Dynamic Weighted Centroid Model

GPS signals are used to locate a GPS-enabled mobile device. However, this method consumes a lot of energy on the device. An alternative way to overcome this problem is by using cellular signals instead of that of GPS. We propose the dynamic weighted centroid model to predict the device location. The method utilizes signal strengths (RSSI) to define the weights, and uses those weights to find the centroid of all detected cells. We apply two models based on how the weight is defined. The first model (model A) uses
$w_{i}=\frac{1}{\mathrm{R}\mathrm{S}\mathrm{S}\mathrm{I}_{\mathrm{i}}}$
and the second model (model B) uses
$w_{i}=\text{RXLEV}_{i}$
. To test the model, we build signal fingerprint from collected Network Measurement Reports (NMR) in the determined area. Based on the experimental results, model A has smaller minimum and maximum error, that are 0.7860 m and 290 m, respectively. As for model B, the results give smaller mean error, median and standard deviation, that are 107.5686 m, 99.5500 m and 54.7201 m, respectively. The results show that the error of model A spreads more than that of model B.