7ConclusionThe motivation of this thesis, the analysis of state-of-the-art localization algorithmsbased on the prediction of Wi-Fi signal distributions has lead to the design of alocalization framework providing a combined controll of the raytracing system, thelocalization system and the evaluation tools. The implemented system was em-ployed to build a representative evaluation corpus for the localization problem and atraining corpus for the Wi-Fi prediction problem. The evaluation of the predictioncapabilities of the used PHOTON raytracer and the accuracy of the localizationalgorithm over both data sets has brought up the following conclusions:The achievable prediction accuracy of the radio propagation models is about 4 dBmcompared to real world measurements by investigating only the basic components,i.e. reflection and attenuation, of the PHOTON raytracer and using 3D scene ge-ometry that distinguishes at least between two material types. This result has beenobserved over 22 deployed APs on the investigated 8000 m3UMIC scene. Fur-thermore, the propagation models were able to generalize over two device classes.This means, the localization accuracy was not affected by using two different typesof client devices and without a specific device adaptation process although theyprovided significant different RSSI measurements.The primary goal of this work, the evaluation of different localization algorithms ona complex scene and a large evaluation corpus under the constraints of the accuracyinduced by the raytracer model, has been reached. The evaluation of the HMM, PFand LMSE localization algorithms has shown, that the PHOTON driven models canbe used for predictions with an accuracy in the range of 2- 3 m under the constraintsof the described environment settings.The evaluation of the location accuracy has shown that the RSSI based system hasa large variance for the localization error. The primary source for the variabilityin the observed accuracy is the constitution of the collected evaluation corpus. Forexample, some paths have complex sections that lead along stairways with only asparse AP coverage. On the simpler pathway zones of the building, that are wellcovered by about 6-8 APs, the system can reach localization error rates of between