90 6. EvaluationFigure 6.5 The two devices, the Iconia tablet and the Nexus smartphone, showcomparable error rates on a radio propagation model that is trained with measure-ments from both.this behaviour can be found by remembering the device classes of both models. TheIconia is a tablet device with a ten inch display, whereas the Nexus is of the smallersmartphone device class. Due to the size of its body, a possible interpretation is,that the Iconia tablet is equipped with a larger antenna which leads to smother RSSIreadings.Figure 6.6 Interestingly, the Nexus smartphone outperforms the Iconia tablet, withrespect to the localization accuracy, on a radio propagation model, that is onlytrained on the latter.In the last part of the evaluation it is investigated whether the assumption holds,that the radio propagation model generalizes over the two used devices. Therefore,an experiment with both devices is conducted on a radio propagation model that isonly trained with measurements of the Iconia device. The results of the employedmodel for the Iconia were already shown in figure 6.3. Using the model also for theNexus device leads to the results of figure 6.6.Comparing the results of both devices leads to the conclusion, that the Iconia trainedradio propagation model generalizes well for the Nexus specific localization problem.The reported error rates for the Nexus are even lower for most algorithmic combi-nations which has not been expected.The next step of the evaluation is conducted by adding an additional Device Adap-tation step as described in the previous section 6.2.3.1. Figure 6.7 reports the resultsof this procedure.
Thesis (Diplom)
Indoor Localization of Mobile Devices Based on Wi-Fi Signals Using Raytracing Supported Algorithms
Download single image
avaibable widths