Network Design for Accurate Vehicle Localization

Abstract

Network-based localization plays a key role on the introduction of emerging road applications, such as connected autonomous driving. These applications demand unprecedented precise, reliable and secure positioning, with localization requirements below one meter. This stringent demand is pushing for the use of road-side units (RSUs) from fifth generation (5G) and vehicular networks for accurate vehicle localization. However, these networks are not typically designed for positioning but for data communication purposes, which follow a different paradigm for the network deployment and operation. This paper provides design guidelines on dedicated network deployments with the aim of achieving accurate vehicle-to-infrastructure (V2I) positioning in road scenarios. First, the network layout or site placement is assessed with a geometrical metric. Then, the minimum density of RSUs along the road is bounded with line-of-sight (LoS) probability models for urban street and highway scenarios. Finally, the Cramer-Rao bound (CRB) for joint time-of-arrival (ToA) and angle-of-arrival (AoA) localization is used to maximize the distance between RSUs along the road, by exploiting multi-antenna deployments. According to the simulation results, the network sites are recommended to be located at alternate sides of the road, with a maximum distance between RSUs of 40 m and 230 m for urban and rural environments, respectively. Following these design guidelines, there is also the need to exploit antenna arrays to combine uplink ToA and AoA estimates, in order to ensure a vehicle location accuracy below one meter on the 95% of the cases. The use of eight antenna elements at the RSUs is found to reduce nearly one and a half times the minimum network density in highway localization deployments.