1 Goal-Oriented Semantic Communications for 6G Networks

2025-04-30 0 0 487.48KB 7 页 10玖币

侵权投诉

Goal-Oriented Semantic Communications

for 6G Networks

Hui Zhou, Yansha Deng, Xiaonan Liu, Nikolaos Pappas, and Arumugam Nallanathan

Abstract—Upon the arrival of emerging devices, including

Extended Reality (XR) and Unmanned Aerial Vehicles (UAVs),

the traditional communication framework is approaching Shan-

non’s physical capacity limit and fails to guarantee the massive

amount of transmission within latency requirements. By jointly

exploiting the context of data and its importance to the task, an

emerging communication paradigm shift to semantic level and

effectiveness level is envisioned to be a key revolution in Sixth

Generation (6G) networks. However, an explicit and systematic

communication framework incorporating both semantic level and

effectiveness level has not been proposed yet. In this article,

we propose a generic goal-oriented semantic communication

framework for various tasks with diverse data types, which incor-

porates both semantic level information and effectiveness-aware

performance metrics. We ﬁrst analyze the unique characteristics

of all data types, and summarise the semantic information,

along with corresponding extraction methods. We then propose

a detailed goal-oriented semantic communication framework

for different time-critical and non-critical tasks. In the goal-

oriented semantic communication framework, we present the

goal-oriented semantic information, extraction methods, recovery

methods, and effectiveness-aware performance metrics. Last but

not least, we present a goal-oriented semantic communication

framework tailored for Unmanned Aerial Vehicle (UAV) control

task to validate the effectiveness of the proposed goal-oriented

semantic communication framework.

Index Terms—6G, Task-oriented and semantics-aware commu-

nication, information extraction, effectiveness layer, performance

metric, data importance.

I. INTRODUCTION

Inspired by Shannon’s classic information theory, Weaver

and Shannon proposed a more general deﬁnition of a commu-

nication system involving three different levels of problems,

namely, (i) the bits conveying information should be reliably

transmitted to the recipient (the technical problem); (ii) the

context conveyed by the transmitted bits should accurately

reﬂect the intentions of the sender (the semantic problem);

and (iii) the conduct or action of the system in response

to communications should be effective in accomplishing a

desired task (the effectiveness problem) [1]. The ﬁrst level

of communication, which is the transmission of bits, has

been well studied and realized in conventional communication

systems based on Shannon’s technical framework. However,

Hui Zhou, and Yansha Deng are with the Department of Engineering, King’s

College London, London, U.K. (e-mail: {hui.zhou, yansha.deng}@kcl.ac.uk)

(Corresponding author: Yansha Deng).

Xiaonan Liu, and Arumugam Nallanathan are with the School of Electronic

Engineering and Computer Science, Queen Mary University of London,

London, U.K. (e-mail: {x.l.liu, a.nallanathan}@qmul.ac.uk).

Nikolaos Pappas is with the Department of Computer and Information

Science, Link¨

oping University, Sweden (email:nikolaos.pappas@liu.se).

with the massive deployment of emerging devices, includ-

ing Extended Reality (XR) and Unmanned Aerial Vehicles

(UAVs), diverse tasks with stringent requirements pose critical

challenges to traditional communications, which are already

approaching the Shannon physical capacity limit. This imposes

the Sixth Generation (6G) network towards a communication

paradigm shift to semantic level and effectiveness level by

exploiting the context of data and its importance to the task.

It is noted that the signiﬁcance and importance of information

evaluates the importance of extracted semantic information in

accomplishing a speciﬁc task and is closely coupled with the

considered task.

Initial works on “semantic communications” have mainly

focused on identifying the content of the traditional text and

speech [2], and the information freshness, i.e., age of informa-

tion (AoI) [3] as a semantic metric that captures the timeliness

of the information. However, these cannot capture the data

importance sufﬁciently of achieving a speciﬁc task. In [4], a

joint design of information generation, transmission, and re-

construction was proposed. Although the authors explored the

beneﬁts of including the effectiveness level in [5], an explicit

and systematic communication framework incorporating both

semantic level and effectiveness level has not been proposed

yet. There is an urgent need for a uniﬁed communication

framework aiming at task-oriented performances for diverse

data types.

Motivated by this, in this paper, we propose a generic goal-

oriented semantic communication framework, which jointly

considers the semantic level information about the data con-

text and effectiveness-aware performance metrics about data

importance for different tasks with various data types. The

main contributions of this paper are:

1) We ﬁrst present the existing semantics for traditional

text, speech, image, and video data types. More impor-

tantly, we analyze the unique characteristics of emerging

data types, including 360◦video, sensor, haptic, and

machine learning models, and propose corresponding se-

mantics deﬁnition and extraction methods in Section II.

2) We then propose a generic goal-oriented semantic com-

munication framework for typical time-critical and non-

critical tasks, where semantic level and effectiveness

level are jointly considered. Speciﬁcally, by exploiting

the unique characteristics of different tasks, we present

goal-oriented semantic information, their extraction and

recovery methods, and effectiveness-aware performance

metrics to guarantee the task requirements in Section III.

3) To demonstrate the effectiveness of our proposed

goal-oriented semantic communication framework, we

arXiv:2210.09372v3 [eess.SY] 6 Apr 2024

present the goal-oriented semantic communication solu-

tion tailored for Unmanned Aerial Vehicle (UAV) control

and analyze the results in Section IV.

II. SEMANTIC INFORMATION EXTRACTION

To exploit the context of the data for transmission, the

challenge lies in designing the algorithm to identify and then

extract the semantic information from each data type based

on its unique characteristics. It is noted that each data type

needs a customized semantic information extraction algorithm

to exploit its characteristics fully. Therefore, in this section, we

focus on analyzing the characteristics of both traditional and

emerging data types in the 6G network and summarizing the

semantic information deﬁnition with corresponding extraction

methods, as shown in Table I.

TABLE I

SEMANTIC INFORMATION EXTRACTION OF DIFFERENT DATA TYPES

Data Type Semantic Information Semantic Information

Extraction Method

Text Embedding BERT

Speech Embedding BERT

Image Edge, Corner, Blob, Ridge SIFT, CNN

Video Temporal Correlation CNN

360◦Video FoV Biological Information

Compression

Haptic Data JND Web’s Law

Sensor and

Control Data Freshness AoI

A. Speech and Text

For one-dimensional speech signal, the speech-to-text con-

version can be ﬁrst performed by speech recognition. With

the extracted text information, the famous embedding extrac-

tion method, i.e., Bidirectional Encoder Representations from

Transformers (BERT) can be applied to extract embedding

as typical semantic information, which represents the words,

phrases, or text as a low-dimensional vector [2]. However,

during the speech-to-text conversion process, the timbre and

emotion conveyed in the speech may be lost.

B. Image and Video

As a two-dimensional data type, the image geometric struc-

tures, including edges, corners, blobs, and ridges, can be

identiﬁed as typical semantic information, where the Convo-

lutional Neural Network (CNN) has shown stronger capability

to extract complex geometric structures with its matrix kernel

[6], [7]. Since video combines two-dimensional images with

an extra time dimension, the temporal correlation between

adjacent frames can be identiﬁed as important semantic in-

formation, where the static background can be ignored during

transmission.

C. 360◦Video

The 360◦rendered video is a new data type in emerging

XR applications. The most important semantic information

is identiﬁed as human ﬁeld-of-view (FoV), which occupies

around one-third of the 360◦video and only has the highest

resolution requirement at the center [8]. In this case, biological

information, such as retinal foveation and ballistic saccadic

eye movements can be leveraged for semantic information ex-

traction. Therefore, biological information compression meth-

ods have been utilized to extract the semantic information,

where retinal foveation and ballistic saccadic eye movements

are jointly considered to optimize the semantic information

extraction process.

D. Haptic Data

Haptic data consists of two submodalities, which are tactile

information and kinesthetic [9], [10]. For tactile information,

ﬁve major dimensions can be identiﬁed, which are fric-

tion, hardness perception, warmth conductivity, macroscopic

roughness, and microscopic roughness. Kinesthetic informa-

tion refers to the position/orientation of human body parts

and external forces/torques applied to them. To reduce the

redundant raw haptic data, Just Noticeable Difference (JND) is

identiﬁed as valuable semantic information to ﬁlter the haptic

signal that cannot be perceived by the human, where Weber’s

law serves as an important semantic information extraction

criterion.

E. Sensor and Control data

Sensors are usually deployed to monitor the physical char-

acteristics of the environment (e.g., temperature, humidity,

or trafﬁc) in a geographical area. The acquisition of data is

transformed to status updates that are transmitted through a

network to the destination nodes. Then, these data are pro-

cessed to extract useful information, such as control commands

or remote source reconstruction, that can be further utilized

to predict the evolution of the initial source. The accuracy

of the reconstructed data, either in control commands or in

predicting the evolution, is directly related to the relevance

or the semantic value of the data measurements. Thus, an

important aspect is the generation of trafﬁc and how it can be

affected in order to ﬁlter only the most important samples so

the redundant or less useful data will be eliminated to reduce

potential congestion inside the network.

The AoI also has a critical role in dynamic control sys-

tems since it was shown that non-linear AoI and Value of

Information (VoI) are paradigm shifts, and they can improve

the performance of such systems. Furthermore, we have seen

in early studies that the semantics of information (beyond

timeliness) can provide further gains by reducing the amount

of information that is generated and transmitted without de-

grading the performance.

F. Machine Learning Model

With the massive deployment of machine learning algo-

rithms, machine learning-related model has been regarded as

another important data type.

•Federated Learning (FL) Model: The FL framework

has been considered as a promising approach to preserve

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

1Goal-OrientedSemanticCommunicationsfor6GNetworksHuiZhou,YanshaDeng,XiaonanLiu,NikolaosPappas,andArumugamNallanathanAbstract—Uponthearrivalofemergingdevices,includingExtendedReality(XR)andUnmannedAerialVehicles(UAVs),thetraditionalcommunicationframeworkisapproachingShan-non’sphysicalcapacitylimitand...

展开>> 收起<<

1 Goal-Oriented Semantic Communications for 6G Networks.pdf

共7页,预览2页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

1 Goal-Oriented Semantic Communications for 6G Networks

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: