厘米级城市多传感器融合定位系统：Apollo论文详解

1星需积分: 50 140 浏览量更新于2024-09-05 4 收藏 4.13MB PDF 举报

本文档标题"apollo传感器融合论文.pdf"聚焦于自动驾驶领域中关键的传感器融合技术，特别是针对多场景下的车辆精确定位。作者Guowei Wan、Xiaolong Yang等人提出了一种创新且稳健的系统，旨在实现厘米级的定位精度，即使在城市市区、高速公路和隧道等复杂环境中也能保持高精度和鲁棒性。论文的核心内容强调了系统如何有效地利用互补传感器数据，如全球导航卫星系统（GNSS）、激光雷达（LiDAR）和惯性测量单元（IMU）。传统上，LiDAR主要用于提供三维几何信息，但论文中提出了一种创新方法，即不仅依赖于LiDAR强度和三维数据，还利用LiDAR的强度和高度信息，显著提升了定位系统的准确性和抗干扰能力。特别地，论文中的GNSS实时差分（RTK）模块利用了多传感器融合框架的优势，提高了模糊度解算的成功率。为了整合来自不同源的定位测量，文中采用了误差状态卡尔曼滤波器，其对不确定性进行了新颖的估计。通过详细的实验证明，该方法在各种场景下都能达到5-10厘米的根均方误差（RMS），明显优于现有的同类技术。这篇论文深入探讨了自动驾驶系统中传感器融合的重要性，展示了如何通过巧妙地结合多种传感器的数据，克服环境挑战，提升车辆定位的精度和鲁棒性。这对于自动驾驶技术的发展具有重要意义，因为它确保了车辆在复杂环境下的稳定运行和安全性。此外，论文的实施方法和技术对于其他自动驾驶平台优化定位性能也有很高的参考价值。

Robust and Precise Vehicle Localization based on Multi-sensor Fusion

in Diverse City Scenes

Guowei Wan, Xiaolong Yang, Renlan Cai, Hao Li, Yao Zhou, Hao Wang, Shiyu Song

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.

I. INTRODUCTION

Vehicle localization is one of the fundamental tasks in

autonomous driving. The single-point positioning accuracy

of the global navigation satellite system (GNSS) is about

10m due to satellite orbit and clock errors, together with

tropospheric and ionospheric delays. These errors can be

calibrated out with observations from a surveyed reference

station. The carrier-phase based differential GNSS tech-

nique, known as Real Time Kinematic (RTK), can provide

centimeter positioning accuracy [1]. The most signiﬁcant

advantage of RTK is that it provides almost all-weather

availability. However, its disadvantage is equally obvious

that it’s highly vulnerable to signal blockage, multi-path

because it relies on the precision carrier-phase positioning

techniques. Intuitively, LiDAR is a promising sensor for

precise localization. Failure during harsh weather conditions

and road construction still is an important issue of LiDAR-

based methods, although related works have shown good

progress in solving these problems, for example, light rain [2]

and snow [3]. Furthermore, LiDAR and RTK are two sensors

those are complementary in terms of applicable scenes.

LiDAR works well when the environment is full of 3D or

*This work is supported by Baidu Autonomous Driving Business Unit

in conjunction with the Apollo Project.

The authors are with Baidu Autonomous Driving Business Unit,

{wanguowei, yangxiaolong02, cairenlan, lihao30,

zhouyao, wanghao29, songshiyu}@baidu.com.

Author to whom correspondence should be addressed, E-mail:

songshiyu@baidu.com

Fig. 1: Our autonomous vehicle is equipped with a Velodyne LiDAR HDL-

64E. An integrated navigation system, NovAtel ProPak6 plus NovAtel IMU-

IGM-A1, is installed for raw sensor data collection, such as GNSS pseudo

range and carrier wave, IMU speciﬁc force and rotation rate. The built-in

tightly integrated inertial and satellite navigation solution was not used. A

computing platform equipped with Dual Xeon E5-2658 v3 12 cores, and a

Xilinx KU115 FPGA chip with 55% utilization for LiDAR localization.

texture features, while RTK performs excellently in open

space. An inertial measurement unit (IMU), including the

gyroscopes and the accelerometers, continuously calculate

the position, orientation, and velocity via the technology

that is commonly referred to as dead reckoning. It’s self-

contained navigation method, that is immune to jamming

and deception. But it suffers badly from integration drift.

Thus, each sensor has its own unique characteristics and its

working conditions. Here, we propose a robust and precise

localization system using multi-sensor fusion designed for

autonomous vehicles driving in complex urban and highway

scenes. More precisely, we adaptively fuse different local-

ization methods based on sensors such as LiDAR, RTK, and

IMU. The sensor conﬁguration of our system is shown in

Figure 1. Our system provides stable, resilient and precise lo-

calization service to other modules in an autonomous vehicle,

which has the capability of driving in several complex scenes,

such as downtown, tunnels, tree-lined roads, parking garages,

and highways. We demonstrate large-scale localization using

over 60 km of data in dynamic urban and highway scenes.

In Figure 2, we show the architecture of our multi-sensor

fusion framework.

To summarize, our main contributions are:

• A joint framework for vehicle localization that adap-

tively fuses different sensors including LiDAR, RTK,

and IMU. It effectively leverages their advantages and

shields our system from their failure in various scenes

下载后可阅读完整内容，剩余7页未读，立即下载

LEO爱排骨

粉丝: 14

厘米级城市多传感器融合定位系统：Apollo论文详解

传感器融合论文

Apollo规划论文

多传感器融合Multi-Sensor Fusion

技术文档 _ Apollo传感器标定指南.pdf

Apollo公开课丨ASU（Apollo 传感器单元）硬件详解.pdf

Apollo公开课丨Apollo Fuel 技术分享.pdf

Apollo公开课丨Apollo决策技术分享.pdf

技术文档 _ Apollo的坐标系.pdf

微服务配置中心Apollo架构和实践.pdf

技术文档丨Apollo3.5 感知模块.pdf

最新资源