Abstract
Nowadays, Global Navigation Satellite Systems (GNSS) enable us to determine our position whenever desired in outdoor areas. In spite of the global availability of GNSS, their signals cannot be used in indoor areas for positioning. A widely used indoor positioning method has therefore emerged in the last decade: fingerprinting.
Fingerprinting consists of two phases. There is an offline phase, where a radio map is recorded by measuring the signal strength of signals received in indoor areas (such as Bluetooth low energy or WLAN signals) at reference points. There is also an online positioning phase, where the same signals are measured and compared to the signals in the radio map to estimate a position. As the offline phase is time-consuming, there is a need to optimize fingerprinting.
This thesis focuses on radio fingerprinting optimization tailored to vehicles in parking garages. It evaluates deterministic and probabilistic fingerprinting methods regarding their optimization potential and investigates approaches to the algorithmic and economic optimization as well as an integrated solution with vehicle sensor data.
The algorithms developed were tested in a parking garage near Graz. The algorithmic optimization showed that the best results are achieved using deterministic Weighted K Nearest Neighbour (WKNN) fingerprinting with distance metrics based on the L1 norm. The economic optimization revealed that the achievable accuracy does not decrease when fingerprints are only recorded in areas where the vehicle is allowed to drive. The integration with vehicle sensor data lowered the Root Mean Square Error (RMSE) of the trajectories to less than 3 m.
Fingerprinting consists of two phases. There is an offline phase, where a radio map is recorded by measuring the signal strength of signals received in indoor areas (such as Bluetooth low energy or WLAN signals) at reference points. There is also an online positioning phase, where the same signals are measured and compared to the signals in the radio map to estimate a position. As the offline phase is time-consuming, there is a need to optimize fingerprinting.
This thesis focuses on radio fingerprinting optimization tailored to vehicles in parking garages. It evaluates deterministic and probabilistic fingerprinting methods regarding their optimization potential and investigates approaches to the algorithmic and economic optimization as well as an integrated solution with vehicle sensor data.
The algorithms developed were tested in a parking garage near Graz. The algorithmic optimization showed that the best results are achieved using deterministic Weighted K Nearest Neighbour (WKNN) fingerprinting with distance metrics based on the L1 norm. The economic optimization revealed that the achievable accuracy does not decrease when fingerprints are only recorded in areas where the vehicle is allowed to drive. The integration with vehicle sensor data lowered the Root Mean Square Error (RMSE) of the trajectories to less than 3 m.
| Original language | English |
|---|---|
| Qualification | Master of Science |
| Awarding Institution |
|
| Supervisors/Advisors |
|
| Publication status | Published - 2017 |