Enabling High-Precision 5G mmWave-Based Positioning for Autonomous Vehicles in Dense Urban Environments
5G , Autonomous Vehicles , Multipath , Positioning , Sensor Fusion
Autonomous vehicles (AVs) have the potential to transform the transportation industry by altering conventional modes of travel, enhancing road safety measures, and mitigating traffic congestion and greenhouse gas emissions. An accurate, continuous, and robust positioning solution is required for AVs to operate safely in dense urban environments. In this thesis, we tackle the problem of AV positioning by utilizing the emerging 5G New Radio (NR) millimeter Wave (mmWave) technology along with AV's OBMS, including accelerometers, gyroscopes, and speed measurements. To achieve this, three novel contributions are introduced. The first contribution introduces a high-precision standalone 5G mmWave-based positioning method that utilizes a novel non-line-of-sight (NLoS) detection scheme that can effectively mitigate the detrimental effects of multipath and shadowing, which can be significant in dense urban environments. The second contribution of this research is an ensemble-learning-based OoRI method, which is essential to facilitate robust multipath positioning in dense urban environments. The proposed classifier is crucial for the realization of many multipath positioning schemes which have been limited to working with single-bounce reflections (SBRs). Lastly, an integrated positioning solution based on an unscented Kalman filter as a multi-system fusion engine is developed to integrate the 5G line-of-sight (LoS) and multipath signals with the AV’s OBMS to achieve an uninterrupted positioning solution at high precision. The methodology also features a measurement exclusion scheme and an additional validation stage for NLoS measurements using motion constraints. To validate the proposed methodologies, quasi-real 5G measurements were collected using a commercially available ray-tracing tool that incorporates 3D map scans of downtown Toronto (ON, Canada), allowing for realistic road test scenarios in dense urban environments involving realistic multipath challenges. Additionally, for the same road tests, real OBMS data were collected from the test vehicle moving in downtown Toronto at various motion dynamics. The results of this work demonstrate that the proposed system is capable of maintaining a level of accuracy below 30 cm for approximately 97% of the time, which is superior to the accuracy level achieved when multipath signals and OBMS are not considered, which is only around 91% of the time.