Smartphone-based Indoor Pedestrian Tracking via Transformer

In a GPS-blocked environment, pedestrians often lose directions and struggle with finding the correct path. We propose an indoor pedestrian tracking method based on smartphone sensors, including the IMU (Inertial Measurement Unit) and the barometer. The system tracks pedestrians’ indoor walking trajectories by using a single smartphone. Without prerequisite infrastructures, the system utilizes Transformer for speed inference and integrates the gyroscope to estimate real-time bearing. During the trajectory estimation process, the system combines the speed and the bearing to track the trajectory. To enable multi-floor tracking, we propose a stair detection algorithm based on the barometer and the accelerometer. We also implement backtracking based on prior trajectories, which allows the pedestrian to get back to the starting position. We evaluate our system in an underground parking lot and multi-story indoor buildings. The results show our advantages.