Design of a Method to Improve 5G Indoor Positioning Accuracy Using Sensor Fusion with an IMU and Floor Map Information

Sammanfattning

Indoor positioning is gaining interest with the surge of Industry 4.0 and the Internet of Things. This thesis focuses on sensor fusion techniques to improve the accuracy of radio-based indoor positioning systems. The main aim of this work is to increase the accuracy of a commercial 5G indoor positioning system by fusing information from an IMU and the indoor floor map. This is a challenging problem that implies taking into consideration several factors. Some of them are the way the position measurements are obtained, the effect of multipath propagation, the errors, biases, and drifts in each of the sensors, and the time synchronization and sampling frequency of all the devices involved. The proposed sensor fusion algorithm is a particle filter adapted to use the information of the sensors. The indoor positioning system’s measurements will be loosely coupled, while the IMU’s measurements will be tightly coupled. The IMU information is used to displace the particle filter’s particles according to the movement of the robot. The floor map is used to identify and filter out position measurement outliers, as well as to adapt the measurement model used in the particle filter. The performance was measured by comparing it to a ground truth obtained through a lidar-based SLAM algorithm. The results obtained showed a significant improvement in accuracy with respect to the position measurements from the commercial system. The particle filter offered better performance than an extended Kalman filter with access to IMU and floor map information. This is assumed to be caused by the non-Gaussian distribution of positioning errors caused by multipath propagation.