| 1234567891011121314 |
- \cleardoublepage
- \vspace {2cm}
- \begin{center}
- \paragraph{Abstract}
- This thesis focuses on the localization problem adapted to the constraints of a raytracer simulated signal distribution for Wi-Fi capable mobile devices in indoor scenarios.
- The localization problem is defined as predicting the most probable locations for an observed sequence of Wi-Fi signal strength readings.
- An accurately performing solution is of high interest because Wi-Fi signals can be observed cheaply due to an already widespread deployment of Access Points.
- For an efficient analysis of the problem, a framework is implemented that combines the raytracing, the localization and evaluation components.
- Based on this framework, it is investigated whether the raytracing tool provides an effective basis for an accurate Wi-Fi localization system.
- Furthermore, the performance of a Hidden Markov Model, a Particle Filter and a Nearest Neighbour based localization approach are evaluated on automatically trained raytracer models.
- Therefore, a representative corpus of location annotated signal measurements is assembled and subsequently employed for a thorough investigation of the algorithm properties with respect to tracking the device in scenes of various complexity. The trained Wi-Fi signal strength predictions diverge in average by $4dBm$ from the real measurements. Under these predictions, the tracking algorithms reach a localization accuracy of about $1.5m$ on pathways and degrades up to $4m$ in complex scenes like stairways.
- \hrulefill
- \end{center}
|
