This thesis addresses indoor positioning using wireless networks. Since satellite
based systems are not sufficiently reliable inside buildings due to signal attenuation,
the main approaches and algorithms used for indoor positioning are presented,
together with their advantages and limitations. The WiFi and Bluetooth Low
Energy technologies are discussed in greater detail, including their use for position
estimation, along with commercial solutions and privacy issues related to the
detection of wireless devices. In the practical part, a system for determining the
location of mobile phones based on WiFi fingerprinting is developed. The system
consists of four Raspberry Pi receiver stations that passively capture WiFi frames
transmitted by known devices and use the measured signal strengths to construct
fingerprints. Based on these fingerprints, the system determines the spatial zone and
estimates the position of a device within the room. The system design and software
architecture, collection of training fingerprints, model training, the effect of phone
transmission frequency, and long-term tracking of three devices are described. The
estimated positions are also used to visualize the occupancy patterns of individual
users in the form of heatmaps. On unseen reference points, zone classification
accuracy was 63.8%, and the median position error was 1.51 m. Finally, the
limitations of the developed system and possibilities for its further development are
discussed.
|
