Highly Automated Vehicle Absolute Positioning Using LiDAR Unique Signatures

Abstract

The safe operation of Highly Automated Vehicles (HAV) depends on the availability of accurate, high integrity on-board measurements of the vehicle dynamic states such as position and velocity. Augmented GNSS-based solutions have been shown to meet availability, accuracy, and integrity requirements under certain conditions, such as unimpeded open sky with appropriate receiver and antenna designs that have access to high accuracy correction services (e.g. RTK or PPP/RTK). Integrated GNSS/IMU solutions provide continuity during brief GNSS signal disruption when traveling through a short tunnel, under an overpass, or on a road with trees intermittently covering short sections along the vehicle’s path. However, in the event of more challenging conditions resulting in prolonged GNSS signal disruption, manipulation, or solution degradation on the order of multiple minutes or more, additional sensors are needed to provide the dynamic state measurements or provide periodic updates to keep IMU errors sufficiently bounded. In an open sky environment, measurements from additional onboard sensors can enhance the integrity of the HAV position and velocity estimates by detecting and possibly mitigating the impact of intentional jamming and spoofing of GNSS signals. Additionally, crowdsourcing of such measurements from multiple HAVs operating in the same geographic region can potentially be used to identify the area of impact and localize the source of thespoofing signal. In recent years, LiDAR has emerged as one of the key localization-assist sensors for automated vehicles. The accurate, high-resolution range and angular measurements offered by current LiDAR technology coupled with high-definition maps of the HAV operation environment and camera data processed through object recognition algorithms can provide absolute localization information, as well as IMU updates in a GNSS denied or degraded environment. However, these types of solutions rely on the HAV operating in a feature rich environment with on-board access to high-definition maps and the presence of processing memory-intensive object recognition algorithms. These algorithms must be applied in real time to high throughput images from multiple cameras for a full 360o object recognition and coupling with LiDAR measurements. In this paper, we examine an alternative solution based on the design and deployment of geo-located LiDAR-recognizable targets that allow for absolute position measurements in a GNSS challenged or denied environment. This solution precludes the need to fuse LiDAR data with the output of resource intensive real-time image recognition algorithms or high-definition maps of the HAV operation environment and instead requires a significantly smaller database with mapping between LiDAR target signatures and absolute geographic locations. This work will demonstrate vehicle navigation performance when using periodic updates of absolute positions obtained from a LiDAR when GNSS measurements are unavailable or degraded. The vehicle position will be obtained through the use of presurveyed LiDAR targets along the test route and each target will have a unique LiDAR signature to identify it with its pre-surveyed location. The paper will discuss the LiDAR target design and processing algorithm to allow unique LiDAR identification of individual targets and assess performance for a HAV dynamic scenario. An integrated GNSS, IMU, and wheel pulse transducer odometer reference system will be used for truth absolute position and velocity information. The platform utilized for this work is one of USDOT Volpe Center’s test vehicles, which for this effort is a 4WD Mercedes Sprinter van outfitted with multiple antenna mounts, equipment racks and instrumentation power (inverter, storage batteries, solar). Integrated into this mobile laboratory are GNSS, IMU, wheel pulse transducer, LiDAR, and camera sensors that will be leveraged in this work. The significance of this work is to show navigation potential when using sparse LiDAR targets in areas where GNSS reception can be interrupted (e.g. dense urban, parking garage, jamming) or when the GNSS position solutions have been intentionally manipulated such as with GNSS spoofing.