自动驾驶的传感器及数据融合
摘要:利用传感器采集数据,然后对不同数据源得到的数据进行融合,再把
生成的数据输入神经网络等算法进行处理,最终的结果用于指导机器的动作。这
是现代人工智能一般会采取的方式,可将其概括地分为感知层、控制层和执行层。
自动驾驶是人工智能的一大应用领域,离不开这三层的相互配合。对于感知层,
摄像头、雷达等传感器获取图像、距离、速度等信息,扮演眼睛、耳朵的角色。
所以,研究自动驾驶感知层的设备和实现方式——传感器应用情况和数据融合算
法——有着重要的意义。人工智能时代,智能传感器和基于 AI 的数据融合将成
为主流。本文将着重在自动驾驶系统领域中,探讨传感器技术和数据融合算法的
现状。
关键词:自动驾驶;传感器;融合算法;人工智能;
Sensors and data fusion for autonomous driving
Abstract: Use sensors to collect data, and then fuse the data from different data
sources, and then input the generated data into neural networks and other
algorithms for processing, the final result is used to guide the machine action.This is
the general approach of modern artificial intelligence, which can be broadly divided
into a perception layer, a control layer and an executive layer.Autonomous driving is
a major application field of artificial intelligence, which cannot be separated from
the coordination of these three layers.For the perception layer, sensors such as
cameras and radar acquire information such as images, distance and speed, acting as
eyes and ears.Therefore, it is of great significance to study the devices and
implementation methods of automatic driving awareness layer -- sensor application
and data fusion algorithm.In the era of artificial intelligence, smart sensors and
AI-based data fusion will become the mainstream.This paper will focus on the status
quo of sensor technology and data fusion algorithm in the field of autonomous
driving system.
Keyword: automatic driving;The sensor;Fusion algorithm;Artificial intelligence;
一、 传感器
1.传感器简述
传感器是数据的采集入口,是物联网、智能工业、智能设备、无人驾驶等的
必要组成部分[1]。传感器应用领域广泛,且在很多领域嵌入程度深刻。下表是
主要领域里面传感器使用情况,以及主要分类。
表一 六大主要领域里面传感器分类