V2X-Based Cooperative Positioning for Connected Vehicles in GNSS-Denied Environments

An accurate cooperative positioning system is essential for autonomous driving and traffic dispatching, especially in the GNSS-denied environment. The existing V2V-based positioning strategies still rely on the GNSS signal, while most V2I-based approaches could not provide the relative position with surrounding vehicles. Some perception-based methods are limited in real-world field tests. To address these issues, this paper proposed a novel vehicle-infrastructure cooperative positioning (VICP) solution for connected vehicles in GNSS-denied environments. A cooperation mechanism is suggested from roadside perception to V2X communication and multi-source data fusion. The presented cooperative positioning method considers the roadside perception and ego vehicle information, which mainly contains data preprocessing, multi-object matching and tracking, object position prediction, and ego confidence prediction. In particular, to distinguish the ego vehicle and the surrounding vehicles, an adaptive weighting strategy is adopted to predict the ego vehicle confidence based on the coefficient of variation statistic. Furthermore, a connected vehicle and two sets of roadside infrastructures are utilized for real-world experiments in our test field. The road test demonstrates that the proposed method can maintain continuous and accurate positioning in the real V2X scene.