부산대학교 도서관

Attaining centimeter-level vehicle positioning is a fundamental requirement for lane-level autonomous driving. Pursuing this objective from the perspective of direction-of-arrival (DOA) estimation is promising, which has emerged as a prominent research topic. In order to simultaneously meet the demands of low complexity and high accuracy, it is crucial for DOA-based solutions to address pressing challenges, including the model mismatch problem and reliable positioning in scenarios with limited samples. This article explores a novel vehicle positioning scheme employing DOAs obtained from collaborative base stations (BSs) or roadside unit (RSU). To cope with the actual propagation scenarios and avoid nonrandom systematic error, the exact spatial geometry (ESG) for DOA estimation is adopted. Under the ESG model, a two-stage unitary approximate message passing (UAMP)-based DOA estimation method is proposed. With DOAs estimated at multiple collaborative BSs/RSUs, the locations of vehicles are finally obtained with cross-localization criterion. Numerical simulations are provided to show that the proposed method is effective and delivers competitive performance. Furthermore, inspired by intriguing simulation results, we design a DOA subset selection mechanism that enhances the reliability of positioning performance.