S-TRAVERSE: GNSS-Free Vehicle Trajectory Reconstruction Using Smartphone IMU and Vehicle Speed Sensors

Technologies based on the Global Navigation Satellite System (GNSS), such as vehicle navigation systems, often experience reduced accuracy in urban environments because of signal blockages, delays, or physical obstructions. While sensor fusion is the predominant strategy for addressing these challenges, the literature shows that such methods often face two critical limitations: a strong dependency on external data such as GPS and maps, and the inevitable accumulation of errors due to inherent sensor drift. To provide a more robust solution, this study proposes a novel system, S-TRAVERSE, for reconstructing and tracking vehicle trajectories with high accuracy in GNSS-denied environments. This system integrates IMU data from the smartphone’s onboard sensor with speed data from vehicle sources, such as the Controller Area Network (CAN), using a Transformer-based model. We conducted extensive experiments involving multiple drivers, vehicle models, and varied driving environments to capture the complex temporal relationships in sensor data and validate S-TRAVERSE’s performance. In our evaluation, the system demonstrated high accuracy and outperformed both traditional filtering methods and a representative deep learning baseline. By demonstrating robust, long-duration performance without reliance on maps or continuous GPS, this study advances location tracking and offers a reliable, versatile solution for vehicle operation and safety management in challenging GNSS-denied environments.