Using digital solutions to track and position personnel and vehicles in underground mines is vital to creating a safe working environment. Epiroc Mining Intelligence provides positioning tags capable of tracking and positioning personnel and vehicles in the form of Android applications. While the personal tags are positioned by using stationary reference points, the vehicle tags have access to vehicle sensors, such as accelerometers and gyroscopes, providing a more accurate position. Due to the different positioning techniques, a personal tag in a vehicle might not report the same position as the vehicle. This leads to discrepancies in positioning, which is problematic in dangerous situations, such as during blasting or emergencies. An approach to minimize these discrepancies is to allow personal tags to inherit the position of the vehicle. This thesis will address one core component of such a solution, identifying which personal tags are inside a vehicle.

By utilizing Bluetooth Low Energy, a tag can identify which devices are in close proximity and establish a connection. Once the connection has been established, the personal tag and vehicle tag samples accelerometer data for a set duration and compares them using dynamic time warping. The results from dynamic time warping and the variance of the accelerometer time series are used to classify whether the personal tag is inside the same vehicle as the vehicle tag or not. 

Test cases with different combinations of personal tag position and movement of vehicle/personnel were evaluated. The tests conducted show promising results for both the proximity detection and in-vehicle classification component of the system. Bluetooth Low Energy provides a robust proximity detection solution, with capabilities exceeding the requirements for this thesis. Combining the results from dynamic time warping and variance, it is possible to determine whether a personal tag is inside a vehicle or not with an accuracy of approximately 90%. The system also shows an increased power consumption of 22,17% with all the features enabled and in-vehicle classification performed once every 30s.

Future work can be done to further optimize the system, especially the proximity detection component, as it currently picks up devices that are in the far distance. This system is one part of a comprehensive positioning solution, and only provides detection and in-vehicle classification of nearby tags. To report and determine the position of the tags, this system needs to be integrated with the existing positioning solutions.