Object-Detection-Aided GNSS and Its Integration With Lidar in Highly Urbanized Areas

Abstract

Positioning is a key function for autonomous vehicles that requires globally referenced localization information. Lidarbased mapping, which refers to simultaneous localization and mapping (SLAM), provides continuous positioning in diverse scenarios. However, SLAM error can accumulate through time. Besides, only relative positioning is provided by SLAM. The Global Navigation Satellite System (GNSS) receiver is one of the significant sensors for providing globally referenced localization, and it is usually integrated with lidar in autonomous driving. However, the performance of the GNSS is severely challenged due to the reflection and blockage caused by buildings in superurbanized cities, including Hong Kong, China; Tokyo; and New York, resulting in the notorious non-line-of-sight (NLOS) receptions. Moreover, the uncertainty of the GNSS positioning is ambiguous, leading to the incorrect tuning of its weight during GNSS?lidar integration. This article innovatively employs lidar to identify the NLOS measurement of the GNSS receiver using point-cloud-based object detection. Measurements from satellites suffering from NLOS reception will be excluded based on the proposed fault detection and exclusion (FDE) algorithm. Then, GNSS-weight least-square positioning is conducted based on the surviving measurements from FDE. The noise covariance of the GNSS positioning is calculated by considering the potential location errors caused by the NLOS and the remaining LOS measurements. The improved GNSS result and its corresponding noise covariance are integrated with lidar through a graph-based SLAM-integration framework. Experimental results indicate that the proposed GNSS?lidar integration can obtain improved positioning accuracy in a highly urbanized area in Hong Kong.