Robust and Precise Vehicle Localization Based on Multi-Sensor Fusion in Diverse City Scenes

Abstract

We present a robust and precise localization system that achieves centimeter-level localization accuracy in disparate city scenes. Our system adaptively uses information from complementary sensors such as GNSS, LiDAR, and IMU to achieve high localization accuracy and resilience in challenging scenes, such as urban downtown, highways, and tunnels. Rather than relying only on LiDAR intensity or 3D geometry, we make innovative use of LiDAR intensity and altitude cues to significantly improve localization system accuracy and robustness. Our GNSS RTK module utilizes the help of the multi-sensor fusion framework and achieves a better ambiguity resolution success rate. An error-state Kalman filter is applied to fuse the localization measurements from different sources with novel uncertainty estimation. We validate, in detail, the effectiveness of our approaches, achieving 5-10cm RMS accuracy and outperforming previous state-of-the-art systems. Importantly, our system, while deployed in a large autonomous driving fleet, made our vehicles fully autonomous in crowded city streets despite road construction that occurred from time to time. A dataset including more than 60 km real traffic driving in various urban roads is used to comprehensively test our system.

Monocular Localization with Vector HD Map (MLVHM): A Low-Cost Method for Commercial IVs

“…As for the vehicle localization, two main methods are used in localization parameter estimations: filtering and optimization. Some common filtering methods are the Kalman Filter [3], the Extended Kalman Filter [36] and the particle filter [33,37]-Monte Carlo localization algorithm-which models the probability of sensor readings in order to predict and update the camera pose. In the works pertaining to optimization method [24], the pose estimation problem is viewed as a maximum likelihood (ML) formula where the sensor readings are compared with the map re-projection results to produce constraints, against which the camera pose is estimated iteratively.…”

“…Besides fusing the absolute localization information such as GNSS, it can also exploit the inter-frame motion information for smoothing. For example, IMU, vehicle dynamic constraints, and wheel odometry have all been integrated in the localization system as map-matching supplements [3,39]. These additional inter-frame constraints can be easily incorporated into the prediction step in the filtering framework.…”

“…Precise knowledge of the localization and orientation of a vehicle is critical in realizing many map-based IV applications [1,2], such as decision-making, cooperative driving, map updating, etc. The state-of-the-art integrated inertial navigation system based on the global navigation satellite system (GNSS) real-time kinematic (RTK) module, high-quality IMU, and LiDAR-based positioning technology is well-known to deliver high-precision localization [3]; however, due to its high cost, it remains mostly in the research stage and is seldom seen in mass-produced vehicles.…”

IMU-Aided High-Frequency Lidar Odometry for Autonomous Driving

“…All these lidar odometry algorithms can only output localization results at a frequency of less than 10 Hz, which might not be fast enough. Some recent approaches have started to fuse the lidar odometry with the IMU using either the Kalman Filter framework [2][3][4][5]17,18] or the factor graph representation [6,7].…”

Low-Cost GPS-Aided LiDAR State Estimation and Map Building

“…For example, Kümmerle et al [9] proposed a graphbased optimization approach, which treats the robot state as a node and the constraint gained from measurement as an edge. Wang et al [27] proposed a method to fuse the data collected by a multiple-sensor system that consists of RTK modules, a LiDAR sensor, and IMUs. However, the performance of such a system is subject to GPS single point positioning.…”