3.2. Positioning and Tracking 373.2 Positioning and TrackingThe following sections will present literature for the major algorithmic approachesto the localization problem under the restriction of AP generated RSSI readings.An exception has been made for the related work about Particle Filters in 3.2.2.Their indoor localization approaches seems to derive of use-cases from the fieldof robot navigation and have therefore been historically adapted to other sensortypes. The presented radio propagation dependent localization systems use the RSSIinformation in an active way. This means, a mobile device is actively sensing forthe information. Recent research has also been conducted on the passive approachthat is given by measuring fluctuations in the radio propagation with stationarydevice. Such a method was proposed in[32] and has the advantage to drop the needfor special equipment at the mobile localization target in the form of some Wi-Fi-receiver. With exception of the presented system in[13], described in section 3.2.3,all RSSI based systems use a single device for measurements. In[13], three devicesare used synchronously.3.2.1 Hidden Markov Models In this section, three HMM based localization systems and their evaluation result arepresented. The focus is placed on the following model properties and environmentconfigurations that are used for the evaluation aspect:1. What radio propagation model was employed? Are different materials param-eters distinguished if it is an analytical one?2. How are is the PDFs of the emission probabilities p( x| s) modelled and bywhich means are they obtained from the radio propagation model?3. How are the discrete PDFs of the transition probabilities p( s| s) modelled?4. What is the rasterization ratio of the environment that determines the area orvolume that is covered by a state?5. How many predecessor states are evaluated and finally stored for backtrackingto the best state sequence s1T?6. How was the radio propagation model configured for evaluation? How manymaterial parameters were used. How was the model obtained if it is an empir-ical one?7. What is the size of the scene, and how are the LOS/NLOS conditions. Howmany APs were used?8. What was the evaluation corpus of tracked paths and what is the reportederror.The first HMM based pedestrian localization system, that is analysed here, is re-ported in[25]. For the radio propagation model an empirical model is used that