Channel Charting-Based Vehicle Position Estimation in Real-World Coordinates of Lanes

The development of vehicle-to-everything (V2X) technology enables the real-time sharing of various information, leading to the availability of future automated driving applications such as cooperative driving. To this end, vehicle position information is being fundamental since various autonomous functions rely on it. For the position information, the global positioning system (GPS) is generally considered. However, GPS is easily contaminated by environmental factors, which implies the occurrence of GPS shadow areas including canyons, highdensity urban areas, and so on. Therefore, V2X-based localization methods are proposed in various literature. However, previous localization algorithms often rely on prerequisites, such as strict synchronization or the availability of a large number of ground-truth positions for supervised learning, which may not always be feasible in practical scenarios. From this perspective, a channel charting-based approach can be an adequate solution, but its feasibility and accuracy in the non-line-of-sight (NLoS) outdoor environments and high-speed mobility conditions have not been verified. Therefore, in this paper, we propose channel charting-based vehicle position estimation under urban driving scenarios. The results demonstrate the feasibility of channel charting-based localization considering urban driving scenarios while reducing data overhead.