自动驾驶车联网中通感算融合研究综述与展望.pdfVIP

工程科学学报，第

45

卷，第

1

期：137−149，2023

年

1

月

Chinese

Journal

of

Engineering,

Vol.

45,

No.

1:

137−149,

January

2023

/10.13374/j.issn2095-9389.2022.04.16.003;

自动驾驶车联网中通感算融合研究综述与展望

苣

马忠贵，李

卓，梁彦鹏

北京科技大学计算机与通信工程学院，北京

100083

苣通信作者，

E-mail:

163.com

摘

要

为了应对自动驾驶车联网极低的通信时延、极高的可靠性、更高的传输速率等极致性能需求，亟需破解现有车联网

中通信、感知、计算相互割裂与独立分治的问题，实现“云−边−端”一体化协同感知、协同传输和协同决策.

为此，急需对自动

驾驶车联网的通感算融合开展研究，实现三者的高效融合.

首先论述了目前在通信、感知、计算融合领域的研究进展，然后给

出了通感算融合网络的定义，论述了通感助算、通算助感以及感算助通的研究进展.

针对自动驾驶车联网的应用场景，创造

性地提出了“五层四面”通感算融合的网络架构，横向五层自下而上分别是：多元接入层、统一网络层、多域资源层、协同服

务层、管理与应用层；纵向四面分别是：通信面、感知面、算力面、智能融合面，通过五层四面的深度融合，进一步提升了自动

驾驶车联网中通感算融合网络的性能.

其次，提出了评价通感算融合网络的性能指标体系，最后针对目前研究存在的问题以

及未来发展方向给出了四点可行性建议.

关键词

车联网；通信−感知−计算融合；自动驾驶；边缘智能；智能交通

分类号

U463.67

Overview

and

prospect

of

communication-sensing-computing

integration

for

autonomous

driving

in

the

internet

of

vehicles

MAZhong-gui，LIZhuo苣，LIANGYan-peng

School

of

Computer

and

Communication

Engineering,

University

of

Science

and

Technology

Beijing,

Beijing

100083,

China

苣Corresponding

author,

E-mail:

163.com

ABSTRACT

To

meet

extreme

performance

requirements,

such

as

extremely

low

communication

delay,

extremely

high

reliability,

and

a

higher

transmission

rate,

for

autonomous

driving

in

the

Internet

of

vehicles

(IoV),

the

future

IoV

should

be

merged

into

a

united

framework

that

integrates

communication,

sensing,

and

computing.

At

the

same

time,

as

the

5G-Advanced

system

moves

toward

supporting

a

broader

toB

vertical

industry,

it

will

face

a

more

complex

and

heterogeneous

user

environment

and

multidimensional

digital

space,

which

requires

5G-Advanced

terminals

and

5G-Advanced

networks

to

have

stronger

environmental

sensing,

computing,

and

intelligence

capabilities.

To

realize

deep

integration

for

autonomous

driving

in

the

IoV,

the

sensing

of

IoV

depends

on

not

only

radar

positioning,

camera

imaging,

and

various

sensor

detections

but

also

communication,

which

can

collect

a

variety