Integration of 5G and Motion Sensors for Vehicular Positioning: A Loosely-Coupled Approach

Abstract

Autonomous vehicles (AVs) are poised to revolutionize the transportation industry by enhancing traffic efficiency and road safety. However, achieving optimal vehicular autonomy demands an uninterrupted and precise positioning solution, especially in deep urban environments. 5G mmWave holds immense potential to provide such a service due to its accurate range and angle measurements. Yet, as mmWave signals are prone to signal blockage, severe positioning errors will occur. Most of the 5G positioning literature relies on constant motion models to bridge such 5G outages, which do not capture the true dynamics of the vehicle. Few proposed methodologies rely on inertial measurement units (IMUs) to bridge such gaps, where they predominantly use tightly coupled (TC) integration schemes, introducing a nonlinear 5G measurement model. Such approaches, which rely on Kalman filtering, necessitate the linearization of the measurement model, leading to pronounced positioning errors. In this paper, however, we propose a loosely coupled (LC) sensor fusion scheme to integrate 5G, IMUs, and odometers to mitigate linearization errors. Additionally, we propose a novel method to design the process covariance matrix of the extended Kalman filter (EKF). Moreover, we propose enhancements to the mechanization of the IMU data to enhance the standalone IMU solution. The proposed methodologies were tested using a novel setup comprising 5G measurements from Siradel's S_5G simulation tool and real IMU and odometer measurements from an hour-long trajectory. The proposed method resulted in 14 cm of error for 95% of the time compared to 1 m provided by the traditional constant velocity model approach.