Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Received XX Month, XXXX; revised XX Month, XXXX; accepted XX Month, XXXX; Date of publication XX Month, XXXX; date of
current version XX Month, XXXX.
Satellite-MEC Integration for 6G
Internet of Things: Minimal Structures,
Advances, and Prospects
Yueshan Lin1, Wei Feng1(Senior Member, IEEE), Yanmin Wang2, Yunfei Chen3(Senior
Member, IEEE), Yongxu Zhu4(Senior Member, IEEE), Ximu Zhang1, Ning Ge1(Member,
IEEE), and Yue Gao5(Fellow, IEEE)
1Department of Electronic Engineering, Tsinghua University, Beijing 100084, China
2School of Information Engineering, Minzu University of China, Beijing 100041, China
3Department of Engineering, University of Durham, Durham DH1 3LE, U.K.
4National Mobile Communications Research Laboratory, Southeast University, Nanjing 210096, China
5School of Computer Science, Fudan University, Shanghai 200433, China
CORRESPONDING AUTHOR: Wei Feng (e-mail: fengwei@tsinghua.edu.cn).
This work was supported in part by the National Key Research and Development Program of China under Grant 2020YFA0711301, in part
by the National Natural Science Foundation of China under Grant 62341110 and Grant U22A2002, and in part by the Suzhou Science and
Technology Project.
ABSTRACT The sixth-generation (6G) network is envisioned to shift its focus from the service
requirements of human beings to those of Internet-of-Things (IoT) devices. Satellite communications are
indispensable in 6G to support IoT devices operating in rural or disaster areas. However, satellite networks
face the inherent challenges of low data rate and large latency, which may not support computation-
intensive and delay-sensitive IoT applications. Mobile Edge Computing (MEC) is a burgeoning paradigm
by extending cloud computing capabilities to the network edge. Using MEC technologies, the resource-
limited IoT devices can access abundant computation resources with low latency, which enables the highly
demanding applications while meeting strict delay requirements. Therefore, an integration of satellite
communications and MEC technologies is necessary to better enable 6G IoT. In this survey, we provide
a holistic overview of satellite-MEC integration. We first categorize the related studies based on three
minimal structures and summarize current advances. For each minimal structure, we discuss the lessons
learned and possible future directions. We also summarize studies considering the combination of minimal
structures. Finally, we outline potential research issues to envision a more intelligent, more secure, and
greener integrated satellite-MEC network.
INDEX TERMS Computation offloading, Internet of Things (IoT), mobile edge computing (MEC), satellite
communications, satellite-MEC integration.
I. INTRODUCTION
The past few years have witnessed the proliferation of
intelligent Internet-of-Things (IoT) devices, such as wire-
less sensors, industrial robots and intelligent vehicles. With
connection to the Internet, these IoT devices can enable a
myriad of emerging applications (e.g., autonomous driving).
The number of connected IoT devices will reach 30 billion
by the end of 2025 [1] [2]. As a consequence, future sixth-
generation (6G) networks will focus mainly on serving these
intelligent IoT devices instead of human beings. Providing
IoT devices with satisfactory services raises challenges for
the design of wireless systems. One major challenge is
that a considerable part of the IoT devices are deployed
in remote areas, such as oceans, deserts and forests, for
environmental monitoring and resource exploitation [3]. The
harsh geographical conditions in these areas make it difficult
or expensive to construct traditional terrestrial infrastructures
in fifth-generation (5G) networks [4] [5]. In addition, some
IoT devices are required in disastrous areas, where terrestrial
infrastructures may suffer from serious damage [6]. To
address this, a non-terrestrial network via satellites and un-
manned aerial vehicles (UAVs) may be used to complement
the terrestrial network and fill the coverage gap [7] [8].
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
VOLUME , 1
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Yueshan Lin et al.: Satellite-MEC Integration for 6G Internet of Things: Minimal Structures, Advances, and Prospects
Satellite communications are considered a promising so-
lution to providing ubiquitous broadband Internet access at
low cost [9] [10]. Geostationary earth orbit (GEO) satellite
networks, which are traditionally used for satellite commu-
nications, have experienced rapid developments in terms of
providing high speed services to global users [11]. Moreover,
low earth orbit (LEO) constellation networks have attracted
great attention due to their lower propagation latency and
higher transmission rate [12]. Several commercial projects of
LEO satellite communication, such as OneWeb, Telesat, and
Starlink, have been launched. Despite the many advantages,
satellite networks also face their inherent challenges. Com-
pared with terrestrial networks, satellite networks typically
have lower data rate and larger latency. In computation-
intensive applications, IoT devices may need to offload their
data for cloud computing, due to their limited computing
resources. Offloading through satellites can lead to a long
delay, which is unacceptable for IoT devices that require
delay-sensitive services.
To address these challenges, one promising direction is
to enable edge intelligence to replace traditional cloud com-
puting [13]. The basic idea is to extend cloud computing
capabilities to the network edge to enable artificial intel-
ligence (AI) applications, allowing the IoT devices to be
endowed with low-latency data processing and decision-
making capabilities. Mobile edge computing (MEC) tech-
nologies play an important role in the edge intelligence
paradigm. In current 5G networks, MEC technologies have
been used to enhanced the service quality for human beings.
We envision that integrating satellite networks and MEC can
better support IoT applications in remote or disastrous areas
as well [14].
Preliminary attempts have been made to integrate satel-
lite communications and MEC technologies. For instance,
Hewlett Packard Enterprise (HPE) partnered with National
Aeronautics and Space Administration (NASA) to first
launch computers to the International Space Station, namely
the HPE SpaceBorne Computer, which managed to operate
during its full time aboard. In addition, the cloud service
providers (e.g., Amazon, Microsoft and Google) have ex-
plored cloud-based ground stations which directly connect
satellites with ground data centers.
However, the design of integrated satellite-MEC networks
also face several challenges. First, the temporal and spatial
distributions of the IoT devices’ service requirements are
sparse and heterogeneous, and they could vary significantly
over time in terms of service number, service type, etc.
Besides, the communication and computing resources of the
integrated satellite-MEC network are limited. On the one
hand, satellite communications are inherently limited in data
rate and latency, and the limited orbit resources restrict the
number of operating satellites. On the other hand, the MEC
servers deployed on satellites or UAVs are restricted in terms
of size, weight and energy. Moreover, it is challenging to
configure the appropriate network resources to match the
service requirements to achieve higher resource efficiency.
There exist hierarchical network resources in the network
such as the communication links of different features and
multiple layers of edge servers. Meanwhile, the resources
could change dynamically due to the mobility of UAVs and
LEO satellites, making this problem complicated. Last but
not least, the harsh space environment renders deploying
MEC servers on satellites difficult. Space radiation is one
of the most important factors. It not only causes cumulative
effects that could influence the operational parameters of
on-board devices, but also triggers single-event effects that
lead to operational errors or even device damages [15]. In
addition to radiation, the low temperature and vacuum in
space could also cause damage to electronic devices [16].
To overcome these difficulties, certain mitigation measures
need to be taken, which could incur additional costs and
bring new challenges to integrated satellite-MEC networks’
system design.
With the above-mentioned challenges, the design of an
integrated satellite-MEC network is still an open issue, and
a number of studies have discussed this problem. There are
a few relevant survey papers on this subject. For instance,
[17] summarized the advances on satellite communication
networks and reviewed studies that consider enabling edge
computing on satellites. In addition, [18] reviewed the ex-
isting studies that consider a three-layer network, where the
air- and space-layer infrastructures are equipped with MEC
servers to provide services for users on the ground layer.
Although these surveys have made great contributions, they
focus on a special scenario of satellite-MEC integration. The
absence of a comprehensive review on integrated satellite-
MEC networks underlines the motivation for this article.
Referring to [19], a complex integrated satellite-MEC
network could be considered as an orchestration of three
minimal network structures, which are regarded as the basic
elements of the integrated satellite-MEC network. These
minimal structures have unique network properties, and
therefore differ in the enabled applications as well as the
design challenges. After the literature review, we believe that
all existing works are either based on one of the minimal
structures or a combination of multiple minimal structures.
Therefore, the outline of this survey is presented as follows.
•First, we summarize the studies that are based on
the computing-in-forward-link (CIF) structure, where
MEC servers are deployed on aerial platforms (APs)
connected to the satellite, as shown in Fig. 1(a). By
deploying MEC servers in proximity to IoT devices, the
CIF structure is suitable for local-area applications with
ultra-low latency requirements. However, the central
processing unit (CPU) capability of an on-board MEC
server is restricted since APs are limited in size and
energy, which should be considered in the system
design.
•Then we discuss the related studies considering
the computing-on-orbit (COO) structure, where MEC
2 VOLUME ,
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
(a) (b) (c)
Aerial platformAerial platform SatelliteSatellite Gateway stationGateway station Edge serverEdge server
FIGURE 1. Illustration of minimal structures of integrated satellite-MEC networks (a) CIF structure (b) COO structure (c) CAF structure [19].
servers are deployed directly on satellites, as shown in
Fig. 1(b). Since satellites provide much larger coverage
than APs, the COO structure can well support wide-area
applications with low latency requirements. However,
the minimal structure system design faces multiple
challenges due to the harsh space environment and the
limitations of satellites in terms of size, weight and
energy.
•The third category of studies are based on the
computing-after-feeder-link (CAF) structure, where
MEC servers are deployed at the gateway station, as
shown in Fig. 1(c). In this minimal structure configu-
ration, the MEC servers are endowed with enhanced
CPU capabilities. However, this also results in in-
creased latency for IoT devices accessing these servers.
Therefore, the CAF structure is suitable for wide-area
computation-intensive but delay-tolerant applications.
•Finally, we summarize the studies that consider the
combination of different minimal structures.
The rest of this paper is organized as follows. In Sections
II, III, IV, we summarize the related studies that are based
on the three minimal structures respectively. We further
discuss the learned lessons and possible future directions for
each minimal structure. In Section V, we review the studies
that investigate the combination of the minimal structures.
Finally, Section VI outlines open issues, and Section VII
draws the conclusion.
II. COMPUTING-IN-FORWARD-LINK STRUCTURE
The first minimal structure is the CIF structure. As shown
in Fig. 1(a), this minimal integrating structure consists of an
AP equipped with an MEC server, a satellite, a gateway
and multiple IoT devices. The AP can be a UAV or a
high-altitude platform (HAP). This minimal structure can be
extended by considering multiple APs in the network, as
shown in Fig. 2. Existing studies based on the CIF structure
primarily focus on two key areas: computation offloading
and content delivery.
A. COMPUTATION OFFLOADING
Computation task offloading is a basic service provided by
integrated satellite-MEC networks. For the CIF structure, an
important problem is to properly allocate different users’
computation tasks to MEC servers. To solve the problem, it
is necessary to comprehensively consider the characteristics
of computation tasks (e.g., delay requirement, input data
size), as well as the heterogeneous communication and
computation resources in the network.
In [20], the CIF network consisted of a satellite and
a multi-antenna access point with MEC capabilities. The
access point worked in full-duplex mode, and thus the
computation results could be transmitted back to users in
real time. To improve the offloading data rate, the authors
investigated users’ task offloading decision and resource
allocation jointly. The authors of [21] considered utilizing
satellite and multiple UAVs in the network, where the UAVs
were equipped with MEC servers to provide computation
services. In this paper, the task offloading decision was
jointly designed with the allocation of user power, bandwidth
and computing resources, and the aim was to minimize the
total energy cost in the system. The authors proposed an
algorithm based on double deep Q-learning as a solution.
In [22], multiple UAVs and multiple ground base stations,
all equipped with MEC servers, were used to provide edge
computing services. An LEO satellite was connected to both
the UAVs and the base stations for backhaul transmission. A
joint UAV placement and task offloading decision problem
was considered in the network to maximize the overall profit
of the MEC service provider, which was determined by
the number of completed computation tasks and the energy
consumption of the MEC servers. The authors provided a
two-stage algorithm to solve this problem.
VOLUME , 3
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Yueshan Lin et al.: Satellite-MEC Integration for 6G Internet of Things: Minimal Structures, Advances, and Prospects
TABLE 1. Summary of Advances on the CIF structure
Theme Ref. Network architecture Design objective Proposed solution
Computation
offloading
[20] A satellite and a ground
base station
Task offloading
rate
Design users’ task offloading decision and resource allocation, and
provide a solution by decomposing the problem into two sub-problems.
[21] A satellite and multiple
UAVs
Energy Jointly optimize the users’ task offloading decision and resource allocation
and propose a scheme based on double deep Q-learning.
[22] A satellite, multiple
UAVs and multiple
ground base stations
Profit of MEC
service provider
Jointly optimize the users’ task offloading decision and UAV placement
and propose a two-stage algorithm.
Content
delivery
[23] MEC-enabled radio ac-
cess network (RAN) with
satellite backhaul
/ Investigate two use cases for popular and personalized content delivery.
[24] MEC-enabled RAN with
satellite and terrestrial
backhaul
/ Propose a content delivery strategy to achieve optimal traffic distribution
among satellite and terrestrial backhaul links.
[25] MEC-enabled RAN with
satellite backhaul
/ Propose a SR-based adaptive video streaming scheme.
Ground
base station
UAV
HAP
Edge
server
Satellite
Gateway
station
Feeder
link
FIGURE 2. Illustration of one possible extension of the CIF structure.
B. CONTENT DELIVERY
In addition to computation offloading, another important
service considered in the CIF structure is the delivery
of bandwidth-demanding application data, such as high-
resolution video streaming. Specifically, the broadcast-
ing/multicasting capability of satellite communication en-
ables content delivery to multiple network locations, where
the data can be stored in MEC servers in proximity to users.
However, the system design still faces multiple challenges
due to the limited network resources and long latency, and
thus multiple studies have been conducted toward efficient
content delivery.
The authors of [23] proposed a network architecture where
a CIF satellite-MEC network was utilized to support mobile
video delivery. In this network, the authors investigated two
use cases to enhance the users’ Quality of Experience (QoE).
One use case used utilizing satellite communications to pre-
populate video content to MEC servers at different locations
based on the predictive content popularity. The other use case
pre-fetched video content segments to the MEC servers, in
order to overcome the long propagation latency of satellite
links. In [24], a similar CIF structure was considered, ex-
cept that both terrestrial and satellite backhaul links were
included. The MEC server selected a backhaul link for
each enhancement layer of the video, based on the playout
buffer size. The authors proposed a content delivery strategy
to achieve optimal traffic distribution among the backhaul
links. The authors of [25] proposed a super-resolution-
based (SR-based) adaptive video streaming scheme in a CIF
satellite-MEC network. Specifically, this SR-based method
transmitted low-resolution images through the satellite links
to overcome the limited transmission rate. The MEC server
provided the computation resources necessary to run a deep
neural network to reconstruct low-resolution images to high-
resolution images. Table 1 gives a summary of all these
works.
C. LESSONS LEARNED AND FUTURE DIRECTIONS
Existing works based on the CIF structure mainly focus on
two aspects. The first is the task offloading decision problem
in computation offloading, and the second is content delivery.
Potential new research avenues are listed as follows.
Initially, new design objectives may be contemplated
to enhance system performance. For instance, while prior
studies on task offloading decisions have considered energy
cost or other metrics as design objectives, the CIF structure
mainly emphasized delay-sensitive services. Future research
could integrate task latency as a key design objective.
Besides, there are also new topics that warrant discussion.
For instance, a critical issue is determining the on-board
MEC capability for HAPs or UAVs. In addressing this issue,
the number of service requirements and the energy consump-
tion are important factors to be considered. Since the flight
duration of an HAP can extend to months, configuring the
MEC capability on HAPs is a large-time scale problem. In
4 VOLUME ,
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
this case, the design of the MEC capability may be based
on the average service requirement number, which can be
estimated by analyzing historical data. For the UAV case,
however, the duration of one flight is only a few hours.
In this case, the MEC capability configured for UAVs can
be further optimized based on more specific information.
For instance, the large-scale channel state information (CSI)
during the UAV’s flight can be conveniently acquired from a
pre-established database, referred to as a radio map [26]. This
external information can assist the on-board MEC capability
configuration. Consequently, a new framework for medium-
timescale network adjustment based on external information
needs to be introduced—an area ripe for further exploration.
Additionally, the task pre-processing problem can be con-
sidered in this basic structure. Specifically, for computation
tasks that have a huge input data size, the MEC servers can
pre-process the task to compress the data size before offload-
ing it to the cloud. This aspect of network management also
warrants additional discussion.
III. COMPUTING-ON-ORBIT STRUCTURE
The second minimal structure is the COO structure. This
minimal integrating structure is composed of a satellite
equipped with an MEC server, a gateway and multiple IoT
devices, as shown in Fig. 1(b). There are several variants
of this minimal structure. For instance, the space segment
can be a constellation of LEO satellites. In addition, a GEO
satellite and an LEO constellation can coordinately provide
edge computing services, as shown in Fig. 3. The existing
studies based on the COO structure mainly focus on three
topics, namely MEC server placement, service placement
and computation offloading.
A. MEC SERVER PLACEMENT
Since there can be multiple satellites on different orbits in
space, the first problem of the COO structure is to determine
on which satellites the MEC servers are placed. On the
one hand, because of the harsh space environments such
as the severe radiation, placing MEC servers on satellites
requires hardening measures for the servers, which could in-
GEO
GEO-LEO
ISLs
LEO-LEO
ISLs
LEO
constellation Feeder
link
Ground
access point
Edge server
FIGURE 3. Illustration of one possible extension of the COO structure.
cur additional costs. On the other hand, the satellites without
MEC server equipment may need better inter-satellite links
(ISLs) to offload their tasks. This leads to a tradeoff that
requires careful consideration. Moreover, the temporal and
spatial distributions of the service requirements should also
be considered, which makes the problem more complicated.
In this context, both [27] and [28] explored the prob-
lem of server placement in a COO network with an LEO
constellation in space. By modeling the LEO constellation
as a two-dimensional torus network, the authors of [27]
aimed to place a minimum number of servers so that every
satellite can access a server within a threshold distance. To
achieve optimal server placement, an algorithm based on
the d-hops placement method was proposed. On the other
hand, the authors of [28] focused on computation latency
and considered two server placement problems. The first
problem aimed to minimize the task response delay at a given
snapshot, while the second aimed to minimize the average
response delay for an entire time period. A heuristic scheme
based on the genetic algorithm was proposed to solve both
problems. The proposed scheme yielded a performance gain
over traditional schemes, as it took into account the temporal
and spatial characteristics of LEO satellite networks.
B. SERVICE PLACEMENT
The execution of a computation task requires not only
computation resources but also a set of codes and related
libraries/databases. The MEC server can store the code and
databases of certain services, which is referred to as service
placement. Therefore, the next problem of the COO structure
is the service placement decision of the satellite-based MEC
servers. We note that the service placement problem is dif-
ferent from the above-mentioned server placement problem,
where the former decides how to deploy the software and
database of different applications on satellite-based MEC
servers, while the latter decides whether or not the satellites
should be equipped with MEC servers. For the service
placement problem, it is of great importance to consider
how different types of service requirements are distributed
spatially.
In [29], the authors considered a COO system, where a
constellation of satellites each equipped with an MEC server
provided computing services. The service placement problem
was investigated to maximize the robustness aware service
coverage of the system. Specifically, the problem aimed to
increase the user request number that can access the service,
as well as the user request number that can access more
than one service copy deployed on different satellite-based
servers. The authors proposed an online service placement
algorithm based on Lyapunov optimization and Gibbs sam-
pling to give a near-optimal solution. The authors of [30]
further extended the system in [29] considering ISLs among
LEO satellites. The joint service placement and service
request scheduling problem was investigated, which aimed
to reduce unsatisfactory service requests while minimizing
VOLUME , 5
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Yueshan Lin et al.: Satellite-MEC Integration for 6G Internet of Things: Minimal Structures, Advances, and Prospects
ISL transmission cost. The authors modeled it as a mixed-
integer linear programming problem and provided a solution,
which showed better performance than greedy methods.
C. COMPUTATION OFFLOADING
For the COO structure, in addition to MEC server placement
and service placement, another important problem is to
properly offload users’ computation tasks to MEC servers.
To solve this problem, it is necessary to comprehensively
consider the characteristics of computation tasks (e.g., delay
requirement, input data size), as well as the heterogeneous
communication and computation resources in the network.
In [31], the authors considered utilizing a single LEO
satellite with MEC equipment for computation offloading.
Specifically, ground users process their task data locally
or offload the data to the MEC-enabled satellite, where
the to-be-processed data wait in task queues. The joint
offloading decision and communication and computing re-
source allocation problem was investigated to minimize the
long-term power cost. To solve the problem, the Lyapunov
optimization was employed for problem decomposition and
an online algorithm combining deep reinforcement learning
and conventional optimization algorithms was proposed to
solve the sub-problems.
To implement computation offloading in the COO net-
work, many existing works considered utilizing an LEO con-
stellation in space, where each LEO satellite was equipped
with an MEC server. In [32], [33], and [34], each user
was associated with at most one satellite. In [32], the task
offloading decision problem was considered to achieve a
minimum energy consumption of the local and edge com-
puting. A distributed algorithm based on the multiplier alter-
nating direction method was proposed, which approximated
the optimal solution with low computational complexity.
Adopting the same user association method, [33] jointly
optimized the task offloading decision and the bandwidth and
computation resource allocation. To minimize the weighted
sum of the energy consumption and task delay costs, the
authors proposed an algorithm based on problem decompo-
sition. Since service placement is preliminary to the compu-
tation offloading process, the authors of [34] considered the
task offloading problem jointly with the service placement
problem. For the minimization of task execution delays,
the authors jointly optimized the service placement, task
offloading decision and resource allocation of the system.
A Lagrange dual decomposition (LDD)-based algorithm was
proposed to obtain the closed-form optimal solution, and a
heuristic algorithm was also proposed to find an efficient
solution with low complexity.
In [35], [36], and [37], the users could offload their
computation tasks to multiple satellites simultaneously. The
authors of [35] optimized the offloading decision to minimize
the weighted sum of the average task response time and the
average task energy consumption. They proposed a game-
theoretic approach to solve this problem, which reached the
Nash equilibrium in an iterative manner. In [36], joint opti-
mization of task offloading decision and resource allocation
was considered in the system. The aim was to minimize
the total energy consumption of local and edge computing.
The authors proposed a novel algorithm which decomposes
the problem into two sub-problems and solves them re-
spectively. In [37], a special system model was considered
where the computation task data were generated from source
satellites (e.g., Earth observation satellites) and offloaded
to satellites with MEC equipment for edge computing.
For energy consumption minimization, the task offloading
decision and the communication and computation resource
allocation were jointly optimized. The authors divided the
original optimization problem into two sub-problems and
applied successive convex approximation method to design
an iterative algorithm.
Some studies further included ISLs in their considered sys-
tem model, where users’ computation tasks are first offloaded
to an access satellite and could further be forwarded to other
satellites for execution. The authors of [38] proposed a novel
task allocation algorithm based on the greedy strategy to
optimize the task offloading decision. The algorithm also
focused on average delay and energy consumption reduction,
and it showed a performance gain over the double edge
computation offloading algorithm. In [39], the joint task
admission and task scheduling problem was investigated,
aimed at jointly minimizing the delay and energy consump-
tion. Utilizing the delayed online learning method based on
the Lyapunov framework, the authors developed a practical
online distributed algorithm to solve the problem, which
could achieve close-to-optimal performance.
Additionally, some studies have considered a more com-
plicated double-layer architecture involving LEO and GEO
satellites in space. In [40], each LEO satellite was equipped
with an MEC server and executed the offloaded tasks, while
the GEO satellites managed and coordinated the satellite
MEC resources. To achieve task delay minimization, a
scheduling algorithm based on dynamic priority queue was
proposed to solve the task offloading decision problem.
In [41], the computation tasks could be executed at LEO
satellites or GEO satellites. With both latency and energy
costs considered, the authors jointly optimized the task of-
floading decision and communication resource allocation. An
improved two-sided many-to-one matching game algorithm
was proposed to solve the problem.
Moreover, the combination of the COO structure with
ground or aerial networks was investigated. The authors
of [42] considered a combined terrestrial-MEC and satellite-
MEC network, where an LEO satellite provided edge ser-
vices in space. In the system, the task offloading strategy
and the resource allocation of the satellite were jointly con-
sidered, aimed at maximizing the profit of the MEC service
provider. The proposed algorithm decomposed the problem
into two sub-problems and produced a solution. Focused
also on terrestrial-MEC and satellite-MEC combinations, the
6 VOLUME ,
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
TABLE 2. Summary of Advances on the COO Structure
Theme Ref. Network architecture Design objective Proposed solution
MEC
server
placement
[27] LEO constellation with ISLs Number of placed
servers
Design the MEC server placement method by modeling the LEO
constellation as a 2D torus network and proposing an algorithm
based on the d-hops placement method.
[28] LEO constellation with ISLs Latency Design the MEC server placement and the association of satellites
by a heuristic scheme based on the genetic algorithm.
Service
placement
[29] LEO constellation Service coverage
and robustness
Optimize the service placement by proposing a Lyapunov
optimization-based online service placement scheme.
[30] LEO constellation with ISLs Service satisfactory
rate and ISL costs
Jointly optimize the service placement and service request schedul-
ing scheme through mixed-integer linear programming.
Computation
offloading
[31] An LEO satellite Energy Propose a joint offloading decision and resource allocation scheme
based on Lyapunov optimization and deep reinforcement learning.
[32] LEO constellation Energy Design the task offloading scheme by proposing a distributed
algorithm based on the alternating direction method of multipliers.
[33] LEO constellation Latency and energy Jointly optimize the task offloading decision and the bandwidth and
computation resource allocation through problem decomposition.
[34] LEO constellation Latency Jointly optimize the service placement, offloading decision and
resource allocation by an LDD-based algorithm.
[35] LEO constellation Latency and energy Optimize the task offloading, propose a game-theoretic approach
to solve the problem and prove that the Nash equilibrium exists.
[36] LEO constellation Energy Jointly optimize the task offloading decision and the power, band-
width and computation resource allocation, and solve the problem
through problem decomposition.
[37] LEO constellation Latency and energy Develop a delayed online learning method under the Lyapunov
framework for the joint task admission and scheduling problem.
[38] LEO constellation with ISLs Latency and energy Propose a novel task allocation algorithm based on greedy strategy.
[39] LEO constellation with ISLs Energy Jointly optimize the offloading decision of source satellites and
communication and computing resource allocation through problem
decomposition and successive convex optimization.
[40] LEO constellation and GEO
with ISLs
Latency Propose a task scheduling algorithm based on dynamic priority
queue.
[41] LEO constellation and GEO
with ISLs
Latency and energy Jointly optimize the task offloading decision and communication
resource allocation, and propose an improved two-sided many-to-
one matching game algorithm to solve the problem.
[42] An LEO satellite and a
ground base station
Profit of MEC ser-
vice provider
Jointly optimize the task offloading decision and the communica-
tion and computation resource allocation, and solve the problem
through problem decomposition.
[43] An LEO satellite and multi-
ple ground base stations
Energy Jointly optimize the task offloading decision and the computing
resource allocation, decompose the problem and solve the sub-
problems using iterative algorithms.
[44] LEO constellation and a
UAV
Latency Design the task offloading decision by proposing a curriculum
learning-multi-agent deep deterministic policy gradient approach.
[45] An LEO satellite and multi-
ple UAVs
Latency and energy Jointly design the task offloading decision and UAV trajectory by
proposing a multi-agent reinforcement learning based algorithm.
[46] An LEO satellite and multi-
ple UAVs
Latency, energy and
resource cost
Design a hierarchical distributed iterative algorithm to achieve the
Stackelberg equilibrium of users’ task offloading decisions.
[47] LEO constellation and mul-
tiple UAVs
Energy Solve the task offloading decision problem through a deep rein-
forcement learning algorithm.
VOLUME , 7
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Yueshan Lin et al.: Satellite-MEC Integration for 6G Internet of Things: Minimal Structures, Advances, and Prospects
authors of [43] further considered a system model with
multiple base stations. To minimize the total energy con-
sumption, the task offloading decision was jointly optimized
with the computing resource allocation. The authors adopted
the classic alternating optimization method for decomposing
the original problem and then solved each sub-problem using
low-complexity iterative algorithms.
The authors of [44] considered combining aerial-MEC and
satellite-MEC, where users could offload computation tasks
to the LEO satellite or to a UAV flying on a predetermined
trajectory. The optimization of task offloading decision was
conducted to lower the time-averaged task execution latency.
To learn the near-optimal offloading strategy, a curriculum
learning-multi-agent deep deterministic policy gradient ap-
proach was proposed. In [45], the scenario involved multiple
UAVs and an LEO satellite, each equipped with an MEC
server. The authors jointly optimized the task offloading de-
cision and UAV trajectory to for latency and energy cost min-
imization. A multi-agent reinforcement learning based task
offloading algorithm was proposed to solve the problem. [46]
considered a similar network architecture but also included
the peer-to-peer (P2P) communication between ground users,
allowing computation tasks to be offloaded to peer users
for execution. The authors aimed to simultaneously improve
the task latency, energy consumption and resource costs by
optimizing users’ offloading decisions. The problem was
modeled as a multi-leader and multi-follower Stackelberg
game, and a hierarchical distributed iterative algorithm was
designed to achieve the Stackelberg equilibrium. The authors
of [47] further considered a system model with multiple LEO
satellites and UAVs, equipped with MEC servers, to process
or cache users’ tasks. The task offloading decision problem
was investigated to minimize the energy consumption for
task execution. The authors employed a constrained Markov
decision process to formulate the task offloading decision
problem and further devised a deep reinforcement learning-
based algorithm to solve the problem. Table 2 summarizes
and compares these works.
D. LESSONS LEARNED AND FUTURE DIRECTIONS
We first compare the MEC server placement problem and the
service placement problem in terms of design challenges and
solutions. In the former topic, the main difficulty lies in de-
ploying the minimum amount of computing resources while
considering the harsh space environment, the limited energy
supply, as well as the available ISLs. For service placement,
it is challenging to efficiently distribute the service codes and
databases under strict constraints on computing and storage
resources. Besides, adapting efficiently to the temporal and
spatial distribution of user requests is a major challenge
for both problems. In terms of potential solutions, heuristic
schemes and graph theory-based schemes are widely con-
sidered in the MEC server placement problem, while in the
service placement problem optimization schemes are often
adopted.
Then we present the potential research gaps for the COO
structure. First, more realistic scenarios and environments
should be considered. For instance, severe electromagnetic
radiation in space can influence satellite-based server perfor-
mance and even cause damage. Therefore, the servers need
to be radiation-hardened, which impacts the computation
performance. Few existing works have considered this factor.
Besides, the energy supply of satellites is heavily dependent
on solar power, which can be inconsistent. This also influ-
ences satellite-based servers’ performance. In future work,
these factors should be taken into consideration to obtain
more persuasive results.
In addition, a more complicated system model can be
investigated. For instance, for the MEC server placement
problem, MEC servers can also be placed on GEO satellites
in addition to LEO satellites. With their inherently large
coverage, the MEC-enabled GEO satellites can not only
provide edge computing services for ground users, but also
orchestrate the communication and computing resources for
LEO satellites, which enables better coordination in the
system. This idea has been mentioned in [40] and can be
further investigated.
Another direction of research is to choose proper design
objectives. We take the service placement problem as an
example. In existing studies, the design objective of [30]
is service coverage and service robustness, while [29] min-
imizes the service satisfactory rate and ISL costs. In the
future, novel design objectives should be considered to better
describe the performance of service placement.
Moreover, some new research topics can be explored. An
important instance is the MEC server activation problem.
Due to the limited energy on the satellites, adopting a full-on
mode for MEC servers may be impractical. Therefore, it is
important to decide which of the servers should be activated,
in order to satisfy the service requirements and save the
energy costs. Different from MEC server placement which is
adjusted at similar timescales as infrastructure changes (e.g.,
months), MEC server activation is often adjusted every few
hours or minutes. Network adjustments on this timescale
have yet to be investigated. Therefore, a novel network
architecture that enables on-demand network adjustments at
such a medium timescale needs to be considered [19].
For computation offloading, there are also new research
topics to be considered. For instance, the scenario of mul-
tiple MEC servers executing a single complicated task can
be considered. To achieve this, multiple satellites need to
provide edge computing services effectively. The authors
of [48] proposed an on-orbit federated learning system,
where LEO satellites serve as local servers and a medium
earth orbit (MEO) satellite serves as the global server.
Further research can be conducted on this topic. Besides,
the handover problem of satellite-based MEC servers can
be considered. After the edge server finishes computation,
the results need to be transmitted back to the user. This can
be difficult due to the mobility of the LEO satellites. Many
8 VOLUME ,
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Fiber
link
Feeder
link
Fiber
link
Feeder
link
Satellite
Edge
server
Gateway
station
FIGURE 4. Illustration of one possible extension of the CAF structure.
existing works tackle this problem by setting a computation
time constraint. However, this might not work when the
offloading task is computationally intensive. In that case, the
handover of computation results through ISLs is necessary,
which can be further investigated.
IV. COMPUTING-AFTER-FEEDER-LINK STRUCTURE
The third minimal structure is the CAF structure. As shown
in Fig. 1(c), this minimal integrating structure consists of
a satellite, a gateway equipped with an MEC server and
multiple IoT devices. The minimal structure can be extended
by considering multiple gateways, as shown in Fig. 4.
In this minimal structure, the MEC servers have higher
CPU capability, but the time delay for the IoT devices to ac-
cess the servers is also higher. Therefore, the CAF structure
is suitable for wide-area computation-intensive but delay-
tolerant applications. Existing studies based on the CAF
structure have mainly focused on computation offloading.
A. COMPUTATION OFFLOADING
The authors of [49] considered a system consisting of mul-
tiple LEO satellites and a gateway station equipped with an
MEC server. To achieve fast and energy efficient offloading,
the bandwidth and power resources of users were jointly
allocated. The authors introduced a multi-agent architecture
in which each LEO satellite made their own allocation poli-
cies based on historical policies, as well as users’ workload
situation provided by an information center. Based on that,
a novel multi-agent information broadcasting and judging
algorithm was proposed to allocate resources in a collabora-
tive manner. In [50], a similar system with multiple satellites
and a MEC-enabled ground station was considered. The
authors considered multiple-input-multiple-output (MIMO)
transmission between users and LEO satellites. The user
association, offloading decision, MIMO transmit precoding
and computing resource utilization are jointly optimized to
minimize the long-term average energy consumption. The
problem was solved based on Lyapunov optimization the-
ory and problem decomposition, where quadratic transform
based fractional programming methods were utilized to solve
certain sub-problems.
B. LESSONS LEARNED AND FUTURE DIRECTIONS
Only a few studies investigate the CAF structure, which
mainly focus on resource allocation in computation offload-
ing.
For the CAF structure, one possible future direction is the
problem of deciding the MEC capability configured for the
gateway stations. Since gateways often have sufficient energy
provision, the main focus of this problem is to satisfy users’
service requirements. This problem can be difficult since that
we need to consider not only the service requirement number
of the gateways’ neighboring areas, but also farther areas that
satellites may cover.
In terms of the computation offloading, existing studies
[49] only investigated the resource allocation problem. In
fact, the offloading decision is also important in this struc-
ture, which decides which gateway station the tasks should
be offloaded. Offloading may involve ISL transmissions,
which makes this problem even more complicated.
V. COMBINATION OF MINIMAL STRUCTURES
After discussing the three minimal structures, possible com-
binations of the minimal structures and the relevant problems
will be discussed in this section.
A. COMBINATION OF CIF AND COO
In Fig. 5(a), one possible combination of the CIF structure
and the COO structure is shown. In this section, we review
(a) (b) (c)
FIGURE 5. Illustration of combinations of minimal structures (a) combination of CIF and COO structure (b) combination of COO and CAF structure (c)
combination of CIF and CAF structure [19].
VOLUME , 9
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Yueshan Lin et al.: Satellite-MEC Integration for 6G Internet of Things: Minimal Structures, Advances, and Prospects
TABLE 3. Summary of Advances on the Combination of CIF and COO Structure
Theme Ref. Network architecture Design objective Proposed solution
Computation
offloading
[51] An LEO satellite and a
UAV
Energy Jointly optimize the task offloading decision and UAV trajectory
by proposing an alternating algorithm based on successive convex
approximation.
[52] An LEO satellite and a
UAV
Energy Jointly design the task allocation and the UAV trajectory in three
scenarios with different satellite availability.
[53] An LEO satellite and mul-
tiple UAVs
Latency Jointly design the offloading decision and resource allocation, and
propose a solution to the problem based on problem decomposition
and the block coordinate descent method.
[54] An LEO satellite and mul-
tiple UAVs
Energy Jointly optimize the task offloading decision and resource allocation
by proposing a low-complexity algorithm based on successive
convex optimization.
[55] An LEO satellite and mul-
tiple HAPs
Energy Jointly design the task offloading decision, network resource allo-
cation and MIMO transmit precoding, and propose an algorithm to
decompose the problem and solve the sub-problems iteratively.
[56] An LEO satellite and mul-
tiple UAVs with P2P links
Latency Jointly design the user association and task offloading decision
based on the block successive upper-bound minimization method.
[57] LEO constellation with
ISLs and a ground base
station
Latency/Energy Optimize the task offloading decision by proposing a double
edge computation offloading algorithm based on the Hungarian
algorithm.
[58] LEO constellation and an
HAP
Energy Jointly optimize the task offloading decision and the communica-
tion and computing resource allocation by problem decomposition
and an intelligent heuristic algorithm.
[59] LEO constellation and a
UAV
Latency Optimize the task offloading decision by proposing a deep rein-
forcement learning based algorithm.
[60] LEO constellation and
multiple UAVs
Number of completed
tasks
Model the offloading decision problem as a stochastic game and
propose a learning-based orbital edge offloading approach to solve
the problem.
Service place-
ment
[61] An LEO satellite and mul-
tiple ground base stations
Latency, resource uti-
lization and service
caching ratio
Jointly design the service placement strategy, offloading decision
and resource allocation by introduced the non-dominated sorting
genetic algorithm II.
Content deliv-
ery
[62] An LEO satellite and mul-
tiple ground base stations
/ Propose a novel cooperative multicast-unicast transmission scheme
to handle both popular requests and personalized requests.
the existing studies focused on the combined CIF and COO
structure.
A major part of the existing works focused on the com-
putation offloading problem. In [51], the authors considered
a system consisting of a UAV and an LEO satellite, both
of which were equipped with an MEC server. The joint
task offloading decision and UAV trajectory design problem
was investigated to minimize the total energy consumption.
The authors proposed an alternating algorithm based on
the successive convex approximation approach to solve the
problem. The authors of [52] considered a similar system
with an MEC-enabled UAV and an LEO satellite, but further
discussed three different scenarios according to the avail-
ability of satellite communication. In each scenario, the task
allocation was jointly designed with the UAV trajectory to
minimize the total energy consumption based on successive
convex approximation strategies.
Other works considered a more complicated system, con-
sisting of a satellite and multiple UAVs or HAPs with MEC
server equipment. The authors of [53] considered a latency-
oriented joint offloading decision and resource allocation
problem in the network. The authors proposed a solution
to the problem by decomposing the problem and utilizing
the block coordinate descent method. In [54], the authors
also investigated the task offloading decision and resource
allocation in this network, but turned to minimize the total
power consumption of the satellite, UAVs and users. A low-
complexity algorithm based on successive convex optimiza-
tion was proposed to solve the problem. In [55], the system
consisted of an LEO satellite and multiple HAPs. Spe-
cially, the user-HAP and HAP-LEO communication links all
10 VOLUME ,
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
adopted multiple-input-multiple-output (MIMO) techniques.
In this paper, the task offloading decision and the network
resource allocation were jointly designed with the MIMO
transmit precoding. The aim was to minimize the total
energy consumption of communication and computation in
the system. The authors proposed an algorithm to decompose
the problem and solve the sub-problems iteratively. In [56],
the authors considered that the UAVs could communicate
with each other, so that users’ computation tasks could be
offloaded among the UAVs for better execution. The user
association and offloading decision were jointly optimized
to minimize the sum of users’ task latency. The problem
was non-convex, and thus the block successive upper-bound
minimization method was proposed as a solution.
In addition, some studies considered the scenario of utiliz-
ing multiple satellites in space. In the system model of [57],
users offloaded their task data to a ground base station for
edge computing. The data could further be offloaded to an
access satellite and transmitted to other satellites through
ISLs for data processing. To allocate users’ tasks, the authors
proposed a double edge computation offloading algorithm
based on the Hungarian algorithm. This proposed algorithm
could minimize respectively the average task latency and
the average energy consumption of edge servers. In [58], an
HAP collected and processed users’ task data. Distinguishing
from [57], the HAP could further offload the task data to
multiple satellites simultaneously for edge computing. The
task offloading decision and the communication and com-
puting resource allocation were jointly optimized to achieve
energy-minimization in the system. The authors decoupled
the problem and proposed an intelligent heuristic algorithm
for solution. Moreover, [59] jointly considered the CIF and
COO combined network with the terrestrial MEC network.
Specifically, a UAV with MEC equipment collected users’
computation tasks. The tasks could be executed at the UAV-
based server, or offloaded to ground-based or satellite-based
servers. The optimization of the offloading decision was
performed to minimize the average execution latency. The
problem was formulated into a Markov decision process,
which was solved by a deep reinforcement learning based
algorithm. In [60], the authors further considered a network
consisting of multiple MEC-enabled UAVs as well as MEC-
enabled LEO satellites. Ground users offload their tasks to
the associated UAV, and the UAV process the task data or
further offload them to certain satellites for processing. The
authors aimed to maximize the number of tasks that were
completed before the deadline by designing the task offload-
ing decision. The problem was modeled as a stochastic game
and a learning-based orbital edge offloading approach was
proposed to solve the problem.
The authors of [61] took a step further to jointly consider
service placement and computation offloading in a combined
CIF and COO network. Specifically, some users in the
system offloaded not only their task data but also the corre-
sponding execution codes. The execution codes were cached
in ground-based or satellite-based servers, which could then
handle offloaded tasks of the same service type. The service
placement strategy, offloading decision and resource alloca-
tion were jointly optimized in the network. The aim was
to minimize the system cost, which was a weighted sum
of task latency, computation resource utilization, bandwidth
utilization and cache ratio. The authors introduced the non-
dominated sorting genetic algorithm II to solve the problem.
The authors of [62], on the other hand, focused on the
content delivery problem. They considered a network where
MEC servers were placed on the satellite and ground base
stations. A novel cooperative multicast-unicast transmission
scheme was proposed to handle both popular requests and
personalized requests. Table 3 gives a summary of all these
works.
B. COMBINATION OF COO AND CAF
In Fig. 5(b), one possible combination of the COO structure
and the CAF structure is shown. In this section, we review
the existing works considering a combined COO and CAF
structure.
We first summarize the studies focused on the computa-
tion offloading problem. Some studies considered a simple
system model which consisted of an LEO satellite and a
ground gateway station, both equipped with an MEC server.
The authors of [63] jointly optimized the task offloading
decision and the bandwidth allocation of user-satellite and
satellite-gateway links, to minimize the weighted sum of
task execution latency and energy consumption. The authors
proposed a deep reinforcement learning-based algorithm to
solve the problem, which could achieve near-optimal of-
floading cost performance with low computation complexity.
In [64], joint optimization of task offloading decision and
resource allocation was performed to minimize the time-
averaged task execution latency. The authors leveraged the
framework of Lyapunov optimization to convert the problem
into multiple sub-problems, which were then solved in an
iterative manner.
Further studies were conducted which focused on utilizing
multiple satellites in space and a single gateway station.
Assuming each user was associated with a single satellite,
the authors of [65] jointly optimized the offloading decision
and bandwidth allocation, aimed at both latency and energy
costs. A distributed deep learning algorithm was introduced
to solve the problem in two stages. Adopting the same
user association scheme, [66] investigated the computation
offloading of two types of computation tasks, namely edgy-
cloud and cloudy-edge. Joint optimization of offloading de-
cision and computation resource allocation were considered
for each user to reduce the system costs. Specifically, the
system costs took delay, energy consumption, and resource
utilization into consideration. The authors provided a game-
based perspective on the problem and proposed a hybrid
particle swarm optimization-based algorithm to achieve the
Nash equilibrium. [67] also investigated a system consisting
VOLUME , 11
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Yueshan Lin et al.: Satellite-MEC Integration for 6G Internet of Things: Minimal Structures, Advances, and Prospects
TABLE 4. Summary of Advances on the Combination of COO and CAF Structure
Theme Ref. Network architecture Design objective Proposed solution
Computation
offloading
[63] An LEO satellite and a
gateway station
Latency and energy Jointly optimize the task offloading decision and the bandwidth
allocation by proposing a deep reinforcement learning-based algo-
rithm.
[64] An LEO satellite and a
gateway station
Latency Jointly optimize the task offloading decision and allocation, and
solve the problem by leveraging the framework of Lyapunov
optimization to convert the problem into multiple sub-problems.
[65] LEO constellation and a
gateway station
Latency and energy Jointly optimize the task offloading decision and bandwidth allo-
cation, and solve the problem in two stages by a distributed deep
learning algorithm.
[66] LEO constellation and a
gateway station
Latency, energy and
resource utilization
Jointly optimize the task offloading decision and computation
resource allocation by adopting a game-based perspective, and
propose a hybrid particle swarm optimization-based algorithm to
achieve the Nash equilibrium.
[67] LEO constellation and a
gateway station
Energy Jointly optimize the user association, offloading decision and com-
munication resource utilization and solve the problem by problem
decomposition.
[68] Three LEO satellites with
ISLs and a gateway station
Energy Jointly optimize the task offloading decision and computation
resource allocation based on the improved non-dominated sorting
genetic algorithm II.
[69] LEO constellation with
ISLs and a gateway station
Latency and energy jointly optimized the task offloading decision and computation
resource utilization by designing a deep reinforcement learning
method based on proximal policy optimization.
[70] LEO constellation with
ISLs and a gateway station
Energy Optimize the inter-satellite routing scheme jointly with the task
offloading decision and transmission power, and solve the problem
by a two-stage algorithm.
[71] LEO constellation with
ISLs and a gateway station
Latency and resource
utilization
Jointly optimize the task offloading decision and the communica-
tion resource utilization, and solve the problem by decomposing it
into two sub-problems.
[72] LEO constellation and
multiple gateway stations
Latency and energy Jointly design the task offloading decision and resource allocation
by proposing a solution based on deep reinforcement learning.
Computation
offloading
and content
delivery
[73] LEO constellation and
GEO/MEO with ISLs, and
a gateway station
Latency and resource
utilization
Jointly Design the task offloading decision and caching decision by
proposing a deep imitation learning-driven offloading and caching
algorithm to achieve real-time decision making.
of multiple MEC-enabled LEO satellites and an MEC-
enabled gateway station, but in the system users could of-
fload their task data to multiple satellites. The authors jointly
optimized the user association, task offloading decision and
communication resource utilization, aimed at minimizing the
overall energy consumption. The problem was decomposed
into four sub-problems and solved based on relaxation trans-
formation and fractional programming. Some studies further
included ISLs in the system model, where task data could
be transmitted between satellites through ISLs. In [68], the
authors considered a simple scenario with an access satellite
and two nearby satellites. Joint task offloading decision and
computation resource allocation were conducted for energy
consumption minimization. The authors provided a solution
to the problem based on the improved non-dominated sorting
genetic algorithm II. A more complicated system model
which utilized an LEO constellation with ISLs in space was
further explored [69]. The authors jointly optimized the task
offloading decision and computation resource utilization,
aimed at lowering both the task execution latency and energy
costs. A deep reinforcement learning method based on prox-
imal policy optimization was designed to approximate the
optimal solution. Considering a similar network, the authors
of [70] optimized the inter-satellite routing scheme, jointly
with the task offloading decision and the transmission power.
The objective was to minimize the energy consumption at
the satellites while fulfilling latency constraints. The authors
proposed an algorithm which decomposed the problem and
12 VOLUME ,
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
solved it in two stages. In [71], a computation task was
modeled as a directed graph consisting of multiple virtual
network functions. These virtual network functions could be
uploaded to different satellites through user-satellite links or
ISLs for execution. Joint optimization of offloading decision
and communication resource utilization was conducted for
bandwidth and delay cost minimization. A distributed algo-
rithm based on multi-agent systems was proposed, which
achieved better system performance than the Viterbi and
game theory algorithms.
Moreover, a more complicated system could be consid-
ered, which consisted of multiple LEO satellites and multiple
gateway stations, each with an attached MEC server [72].
The authors investigated the joint task offloading decision
and resource allocation problem in the network, with the
aim of improving system latency and on-orbit computing
energy consumption. A solution based on deep reinforcement
learning was proposed for this problem.
On the other hand, the authors of [73] jointly considered
the computation offloading and content delivery problem.
Specifically, the computation tasks could be executed on
satellites or at the gateway, and the results could be cached on
satellite-based servers for further reuse. Joint optimization of
task offloading decision and caching decision was performed,
aiming to improve both the latency performance and the
resource utilization in the system. To this end, the authors
proposed a deep imitation learning-driven offloading and
caching algorithm which could achieve real-time decision
making. Table 4 summarizes and compares these works.
C. COMBINATION OF CIF AND CAF
In Fig. 5(c), one possible combination of the CIF structure
and the CAF structure is shown. Very few studies have
focused on the combination of the CIF structure and the
CAF structure. The authors of [74] considered a system
with multiple ground base stations, a satellite and a gateway,
where MEC servers were deployed at the base stations and
the gateway. The users’ task data could be offloaded to an
associated base station, or further to the distant gateway
with stronger computing capability through satellite commu-
nication. The task offloading decision was jointly optimized
with the communication and computation resource allocation
to minimize the task execution delay. The authors divided
the problem into to sub-problems, where the task offloading
decision subproblem was solved with theoretical analysis and
mathematical derivation, and the resource allocation problem
was solved by utilizing the particle swarm optimization
algorithm. In [75], the authors considered a system consisting
of an LEO satellite that connects to the computing server at
the gateway station and multiple MEC-enabled UAVs. The
UAV trajectory and communication resource allocation were
jointly optimized for energy minimization, and the problem
was solved leveraging an iterative algorithm.
D. LESSONS LEARNED AND FUTURE DIRECTIONS
By deploying multiple layers of MEC servers, the com-
bined structures could incorporate the advantages of basic
structures and therefore support more demanding service
requirements. However, for the combined network structures,
new system design challenges will arise since the network
architecture is more complicated. We select two of the
combined structures as examples to discuss their advantages
and challenges compared to basic structures.
The combined CIF and COO structure deploys MEC
servers on both APs and satellites. Compared with the CIF
structure, the task data of IoT devices could not only be
offloaded to UAVs/HAPs for local applications, but also
aggregated with other users’ data at the satellite-based server
for execution, which further support wide-area applications
with relatively low latency. Besides, compared to the COO
structure, the combined structure allows the task data to be
pre-processed at the APs to reduce the data size by extracting
key information. This enables more efficient utilization of
communication and computing resources, as well as reduce
the task latency. Nevertheless, this combined structure also
faces new challenges. For instance, the multi-tier computing
resources render the offloading decision problem more diffi-
cult. Besides, both UAVs and LEO satellites could be highly
mobile, leading to a dynamic network architecture. This also
adds to the difficulty of system design.
For the combined COO and CAF structure, MEC servers
are deployed on satellites as well as at gateway sta-
tions. This combined structure could usually empower more
computation-intensive applications than the COO structure,
since satellites are often strictly limited in computing re-
sources. Compared with the CAF structure, on the other
hand, it enables that the application data generated on satel-
lites (e.g., remote sensing data) could be offloaded through
ISLs for low-latency processing. Such combination also
raises new challenges, such as how to efficiently distributing
task data among satellites and gateways through ISLs and
high-speed feeder links.
Furthermore, we note that for these structures, some
existing works study a simple network architecture by con-
sidering a single satellite, while other works further consider
multiple satellites in the system. Extending a single satellite
to multiple satellites raises new system design challenges.
First of all, the offloading decision problem naturally be-
comes more complicated since more than one satellite is
equipped with an MEC server. In addition, ISLs among
satellites need to be considered in a satellite constellation,
which have different characteristics compared with other
communication links in the network, such as ground-to-space
links and feeder links. This renders the joint orchestration
of heterogeneous communication and computing resources
more difficult. Besides, handover among satellites as well
as coordination among satellites for task data processing are
also important challenges in the multi-satellite architecture.
VOLUME , 13
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Yueshan Lin et al.: Satellite-MEC Integration for 6G Internet of Things: Minimal Structures, Advances, and Prospects
MEC server placement
MEC server activation
Service placement
Computation offloading
Task offloading decision
Cooperative computation
Handoff of MEC servers
Computation offloading
Task offloading decision
Content delivery
Decision on On-board MEC
capability of UAVs
Task pre-processing at edge
Decision on MEC capability
of gateway stations
Computing-in-Forward-Link Computing-on-Orbit Computing-after-Feeder-Link
Computation offloading
Resource allocation
Task offloading decision
Decision on On-board MEC
capability of HAPs
Large timescale
(e.g., months/years)
Medium timescale
(e.g., minutes/hours)
Small timescale
(e.g., milliseconds)
FIGURE 6. Existing and future research directions of the three basic structures at different timescales. (The blue-colored topics are yet to be
discussed.)
VI. OPEN RESEARCH ISSUES
This section outlines a few open research issues in the
integration of satellite and MEC.
A. HIERARCHICAL ORCHESTRATION OF MINIMAL
STRUCTURES IN THE INTEGRATED SATELLITE-MEC
NETWORK
The scale of a practical integrated satellite-MEC network is
often huge, and the network consists of a massive amount
of minimal structures. These minimal structures are often
strongly coupled in terms of resource utilization (e.g., com-
munication bandwidth) and task division, which renders the
system design complicated. Inspired by the structure of
proteins, we believe that adopting a hierarchical minimal
structure orchestration framework could be a promising solu-
tion. Specifically, minimal structures can be simply orches-
trated into secondary structures (amino acids orchestrated
into peptides), which further form more complicated tertiary
structures (larger peptides), and so forth. Eventually, these
structures form highly functional integrated satellite-MEC
networks (proteins). In this framework, the system design
of a tertiary structure, for instance, could directly utilize
the secondary structures as basic elements for network or-
chestration, without going into the details of the lower-level
minimal structures. Therefore, the computation complexity
of network orchestration could be significantly reduced. Fur-
ther researches into this hierarchical orchestration framework
could be considered.
B. HIERARCHICAL-TIMESCALE NETWORK
ADJUSTMENTS IN THE INTEGRATED SATELLITE-MEC
NETWORK
The current network is mainly adjusted or optimized at two
timescales. Network planning and network architecture ad-
justment often take place at a large timescale (e.g., months or
years). Specific communication and computation parameters
are adjusted at a small timescale (e.g., milliseconds). For
the integrated satellite-MEC network, service requirements’
number, service type and spatial distribution change dy-
namically at a timescale in between. Thus, the traditional
network adjustment is unable to match the service require-
ments, resulting in degraded resource efficiency. Therefore,
a hierarchical-timescale network adjustment framework is
of interest. We categorize in Fig. 6 the existing and future
research topics, based on a hierarchical-timescale framework.
It can be observed that for each minimal structure there exist
many problems of medium-timescale network adjustment to
be explored. Moreover, since the medium timescale is usu-
ally much larger than the channel coherence time, a process-
oriented optimization framework could be considered for the
design of network adjustments [76].
C. AI-BASED INTEGRATED SATELLITE-MEC NETWORK
AI-based tools and methods have been widely used recently.
AI-based methods can be applied to the integrated satellite-
MEC network in two aspects. First, the integration of satellite
and MEC raises new problems, some of which may be hard
to model. In this context, AI-based methods can be utilized
to provide a feasible solution. In fact, many learning-based
schemes have been utilized in existing studies. For instance,
14 VOLUME ,
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
in [37] a delayed online learning method was developed
for task admission and scheduling. Besides, many existing
studies considered adopting a deep reinforcement learning-
based algorithm [47] [59] [69] [72]. On the other hand, the
widely distributed MEC servers with close proximity to users
can support AI-based applications in return. In fact, [48]
considered utilizing MEO and LEO satellites to implement
a federated learning network. Further investigations into both
aspects can be considered.
D. SECURITY ISSUES IN THE INTEGRATED
SATELLITE-MEC NETWORK
Security is an important issue for the integrated satellite-
MEC network. Satellite networks provide coverage for a
wide geographical area. The openness of electromagnetic en-
vironment makes the network susceptible to cyber-attacks of
different types, such as eavesdropping and jamming. Besides,
the sophisticated integration of satellite and MEC recalls
novel system design methods, which may also raise new
security risks. To address these problems, new security mea-
sures need to be designed and implemented in the network.
One possible solution is to combine the integrated satellite-
MEC network with the blockchain technique, where each
MEC server can work as a node in the blockchain network.
However, this may require massive data transmission for data
synchronization, which can be difficult for the integrated
satellite-MEC network. This tradeoff between security and
resource utilization can be further investigated.
E. INTEGRATED SATELLITE-MEC NETWORK
COORDINATED WITH REMOTE SENSING
In addition to ground IoT devices, the satellites for remote
sensing also generated a great amount of data that needs to be
processed. By deploying integrated satellite-MEC networks,
the remote sensing task data could be computed at the
network edge, which reduces the task latency and saves the
communication resources for offloading to the remote cloud.
In fact, several existing studies have investigated this issue.
The work [28] has considered a simple coordination scenario
where the computation tasks generated by the satellite itself
and offloaded by ground users are processed together. In
[37], the offloading decisions of source satellites are jointly
designed with the satellite-MEC network’s communication
and computing resources. Nevertheless, there still exist many
research gaps that require further consideration, such as dis-
tributing the task data efficiently through ISLs in the satellite
constellation. Moreover, the coordination could be further
investigated to achieve functional cooperation in an efficient
manner, where new applications that require joint sensing-
communication-computation capabilities can be enabled.
F. GREEN INTEGRATED SATELLITE-MEC NETWORK
In the integrated satellite-MEC network, a huge number of
MEC servers will be deployed in a hierarchical manner to
provide services, which leads to massive energy consump-
tion. Besides, UAVs, HAPs and other vehicles (automated or
manual) are widely adopted in the integrated satellite-MEC
network, which also leads to substantial energy consumption
and carbon emissions. Therefore, it is important to develop a
green integrated satellite-MEC network. Traditional methods
used in the terrestrial networks may not be applicable,
because servers and vehicles in the integrated network are
highly mobile and distributed sparsely and heterogeneously
in wide area. Novel techniques for a greener network can be
interesting.
G. SUPPORTING
SENSING-COMMUNICATION-COMPUTING-CONTROL
CLOSED-LOOP DESIGN WITH THE INTEGRATED
SATELLITE-MEC NETWORK
Certain mission-critical IoT applications require the tasks ex-
ecuted in a sensing-communication-computing-control (SC3)
closed-loop manner [77]. Specifically, the sensors collect
information on the field and transmit the sensing data to
the computing server. The server processes the sensing data
to make decisions, which are further transmitted to the
field robots for execution. Since these applications could
take place in remote or rural areas while require low-
latency closed-loop control, supporting them with integrated
satellite-MEC networks is a potential solution. To this end,
the data scheduling and resource orchestration in the in-
tegrated satellite-MEC network needs to be designed for
control-oriented optimization, which requires further discus-
sion.
VII. CONCLUSIONS
In this paper, we have captured the latest technical advances
in satellite-MEC integration. Specifically, we have classified
the existing studies based on three minimal structures. For
each minimal structure, we have presented a comprehensive
literature review based on the research topics, and have
further discussed the gaps and research directions. In ad-
dition, we have also reviewed the studies that focus on
the combination of minimal structures. Finally, we have
outlined the open issues for satellite-MEC integration, such
as introducing a hierarchical-timescale network adjustment
framework to improve resource efficiency, and combining the
integrated network with AI-based techniques, blockchain-
based security measures, as well as sensing and navigation
functions.
REFERENCES
[1] IoT Analytics. (2020). State of the IoT 2020: 12 Billion IoT Con-
nections, Surpassing Non-IoT for the First Time. [Online]. Available:
https://iot-analytics.coms
[2] Z. Lin, M. Lin, B. Champagne, W. -P. Zhu, and N. Al-Dhahir, “Secrecy-
energy efficient hybrid beamforming for satellite-terrestrial integrated
networks,” IEEE Trans. Commun., vol. 69, no. 9, pp. 6345-6360, Sept.
2021.
[3] R. Liu, K. Guo, X. Li, et al., “RIS-empowered satellite-aerial-terrestrial
networks With PD-NOMA,” IEEE Commun. Surveys Tuts.,early access,
2024.
VOLUME , 15
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Yueshan Lin et al.: Satellite-MEC Integration for 6G Internet of Things: Minimal Structures, Advances, and Prospects
[4] W. Feng, Y. Lin, Y. Wang, et al., “Radio map-based cognitive satellite-
UAV networks towards 6G on-demand coverage,” IEEE Trans. Cogn.
Commun. Netw., early access, 2023.
[5] Z. Lin, M. Lin, T. de Cola, J. -B. Wang, W. -P. Zhu, and J. Cheng,
“Supporting IoT with rate-splitting multiple access in satellite and
aerial-integrated networks,” IEEE Internet Things J., vol. 8, no. 14,
pp. 11123-11134, Jul., 2021.
[6] M. Wu, K. Guo, X. Li, et al., “Deep reinforcement learning-based
energy efficiency optimization for RIS-aided integrated satellite-aerial-
terrestrial relay networks,” IEEE Trans. Commun., early access, 2024.
[7] W. Feng, Y. Wang, Y. Chen, N. Ge, and C. -X. Wang, “Structured
satellite-UAV-terrestrial networks for 6G Internet of Things,” IEEE
Netw., early access, 2024.
[8] K. Guo, R. Liu, X. Li, L. Yang, K. An, and Y. Huang, “Outage
performance of RIS-assisted cognitive non-terrestrial network with
NOMA,” IEEE Trans. Veh. Technol., vol. 73, no. 4, pp. 5953-5958,
Apr. 2024.
[9] O. Kodheli, E. Lagunas, N. Maturo, et al., “Satellite communications
in the new space era: A survey and future challenges,” IEEE Commun.
Surveys Tuts., vol. 23, no. 1, pp. 70-109, 1st Quart., 2021.
[10] K. Guo, M. Wu, X. Li, H. Song, and N. Kumar, “Deep reinforcement
learning and NOMA-based multi-objective RIS-assisted IS-UAV-TNs:
Trajectory optimization and beamforming design,” IEEE Trans. Intell.
Transp. Syst., vol. 24, no. 9, pp. 10197-10210, Sept. 2023.
[11] G. Giambene, S. Kota, and P. Pillai, “Satellite-5G integration: A
network perspective,” IEEE Netw., vol. 32, no. 5, pp. 25-31, Sep./Oct.
2018.
[12] Y. Su, Y. Liu, Y. Zhou, J. Yuan, H. Cao, and J. Shi, “Broadband LEO
satellite communications: Architectures and key technologies,” IEEE
Wireless Commun., vol. 26, no. 2, pp. 55-61, Apr. 2019.
[13] P. Porambage, J. Okwuibe, M. Liyanage, M. Ylianttila, and T. Taleb,
“Survey on multi-access edge computing for Internet of Things realiza-
tion,” IEEE Commun. Surveys Tuts., vol. 20, no. 4, pp. 2961-2991, 4th
Quart., 2018.
[14] Y. Yang, “Multi-tier computing networks for intelligent IoT,” Nat.
Electron., vol. 2, no. 1, pp. 4–5, Jan. 2019.
[15] R. H. Maurer, M. E. Fraeman, M. N. Martin, and D. R. Roth, “Harsh
environments: Space radiation environment, effects, and mitigation,”
Johns Hopkins APL Tech. Dig., vol. 28, no. 1, pp. 17-29, 2008.
[16] A. D. George and C. M. Wilson, “Onboard processing with hybrid and
reconfigurable computing on small satellites,” Proc. IEEE, vol. 106, no.
3, pp. 458-470, Mar. 2018.
[17] S. Wang and Q. Li, “Satellite computing: Vision and challenges,” IEEE
Internet Things J., vol. 10, no. 24, pp. 22514-22529, Dec. 2023.
[18] Q. Zhang, Y. Luo, H. Jiang, and K. Zhang, “Aerial edge computing:
A survey,” IEEE Internet Things J., vol. 10, no. 16, pp. 14357-14374,
Aug. 2023.
[19] Y. Lin, W. Feng, T. Zhou, Y. Wang, Y. Chen, N. Ge, and C. -X. Wang,
“Integrating satellites and mobile edge computing for 6G wide-area
edge intelligence: Minimal structures and systematic thinking,” IEEE
Netw., vol. 37, no. 2, pp. 14-21, Mar./Apr. 2023.
[20] J. Fu, J. Hua, J. Wen, K. Zhou, J. Li, and B. Sheng, “Optimization of
achievable rate in the multiuser satellite IoT system with SWIPT and
MEC,” IEEE Trans. Ind. Informat., vol. 17, no. 3, pp. 2072-2080, Mar.
2021.
[21] N. Waqar, S. A. Hassan, A. Mahmood, K. Dev, D. -T. Do, and M.
Gidlund, “Computation offloading and resource allocation in MEC-
enabled integrated aerial-terrestrial vehicular networks: A reinforcement
learning approach,” IEEE Trans. Intell. Transp. Syst., vol. 23, no. 11,
pp. 21478-21491, Nov. 2022.
[22] Y. -H. Chao, C. -H. Chung, C. -H. Hsu, Y. Chiang, H. -Y. Wei, and
C. -T. Chou, “Satellite-UAV-MEC collaborative architecture for task
offloading in vehicular networks,” in Proc. IEEE Globecom Workshops
(GC Wkshps), Taipei, Taiwan, 2020, pp. 1-6.
[23] N. Wang, N. Nouwell, C. Ge, et al., “Satellite support for enhanced
mobile broadband content delivery in 5G,” in Proc. IEEE Int. Symp.
Broadband Multimedia Syst. Broadcast. (BMSB), Valencia, Spain, 2018,
pp. 1-6.
[24] S. Kumar, N. Wang, C. Ge, and B. Evans, “Optimising layered
video content delivery based on satellite and terrestrial integrated 5G
networks,” in Proc. Eur. Conf. Networks Commun. (EuCNC), Valencia,
Spain, 2019, pp. 161-166.
[25] H. Cai, W. Jing, X. Wen, Z. Lu, and Z. Wang, “MEC-based QoE
optimization for adaptive video streaming via satellite backhaul,” in
Proc. IEEE Int. Conf. Commun. Workshops (ICC Workshops), 2021,
pp. 1-7.
[26] X. Li, W. Feng, J. Wang, Y. Chen, N. Ge, and C.-X. Wang, “Enabling
5G on the ocean: A hybrid satellite-UAV-terrestrial network solution,”
IEEE Wireless Commun., vol. 27, no. 6, pp. 116-121, Dec. 2020.
[27] T. Pfandzelter and D. Bermbach, “QoS-aware resource placement for
LEO satellite edge computing,” in Proc. IEEE Int. Conf. Fog Edge
Comput. (ICFEC), Messina, Italy, 2022, pp. 66-72.
[28] Z. Yan, T. d. Cola, K. Zhao, W. Li, S. Du, and H. Yang, “Exploiting
edge computing in Internet of Space Things networks: Dynamic and
static server placement,” in Proc. IEEE Veh. Technol. Conf. (VTC2021-
Fall), Norman, OK, USA, 2021, pp. 1-6.
[29] Q. Li, S. Wang, X. Ma, et al., “Service coverage for satellite edge
computing,” IEEE Internet Things J., vol. 9, no. 1, pp. 695-705, Jan.
2022.
[30] Y. Zhang, Y. Tang, and W. Wang, “Service deployment and service
request optimization scheduling in MEC enabled LEO networks,” in
Proc. Int. Conf. Comput. Commun. Networks (ICCCN), Athens, Greece,
2021, pp. 1-6.
[31] Q. Tang, Z. Fei, B. Li, et al., “Stochastic computation offloading for
LEO satellite edge computing networks: A learning-based approach,”
IEEE Internet Things J., vol. 11, no. 4, pp. 5638-5652, Feb. 2024.
[32] Q. Tang, Z. Fei, B. Li, and Z. Han, “Computation offloading in
LEO satellite networks with hybrid cloud and edge computing,” IEEE
Internet Things J., vol. 8, no. 11, pp. 9164-9176, Jun. 2021.
[33] B. Wang, J. Xie, D. Huang, and X. Xie, “A computation offloading
strategy for LEO satellite mobile edge computing system,” in Proc. Int.
Conf. Commun. Softw. Networks (ICCSN), Chongqing, China, 2022, pp.
75-80.
[34] Y. Hao, Z. Song, Z. Zheng, Q. Zhang, and Z. Miao, “Joint commu-
nication, computing, and caching resource allocation in LEO satellite
MEC networks,” IEEE Access, vol. 11, pp. 6708-6716, 2023.
[35] Y. Wang, J. Yang, X. Guo, and Z. Qu, “A game-theoretic approach to
computation offloading in satellite edge computing,” IEEE Access, vol.
8, pp. 12510-12520, 2020.
[36] Z. Song, Y. Hao, Y. Liu, and X. Sun, “Energy-efficient multiaccess
edge computing for terrestrial-satellite Internet of Things,” IEEE Inter-
net Things J., vol. 8, no. 18, pp. 14202-14218, Sept. 2021.
[37] R. Wang, W. Zhu, G. Liu, et al., “Collaborative computation offloading
and resource allocation in satellite edge computing,” in Proc. IEEE
Glob. Commun. Conf. (GLOBECOM), Rio de Janeiro, Brazil, 2022, pp.
5625-5630.
[38] Y. Zhang, C. Chen, L. Liu, D. Lan, H. Jiang, and S. Wan, “Aerial
edge computing on orbit: A task offloading and allocation scheme,”
IEEE Trans. Netw. Sci. Eng., early access, 2022.
[39] X. Zhang, J. Liu, R. Zhang, et al., “Energy-efficient computation peer
offloading in satellite edge computing networks,” IEEE Trans. Mob.
Comput., vol. 23, no. 4, pp. 3077-3091, Apr. 2024.
[40] J. Han, H. Wang, S. Wu, J. Wei, and L. Yan, “Task scheduling of high
dynamic edge cluster in satellite edge computing,” in Proc. IEEE World
Congr. Serv. (SERVICES), Beijing, China, 2020, pp. 287-293.
[41] H. Fang, Y. Jia, Y. Wang, Y. Zhao, Y. Gao, and X. Yang, “Matching
game based task offloading and resource allocation algorithm for
satellite edge computing networks,” in Proc. Int. Symp. Networks,
Comput. and Commun. (ISNCC), Shenzhen, China, 2022, pp. 1-5.
[42] B. Wang, X. Li, D. Huang, and J. Xie, “A profit maximization
strategy of MEC resource provider in the satellite-terrestrial double
edge computing system,” in Proc. Int. Conf. Commun. Technol. (ICCT),
Tianjin, China, 2021, pp. 906-912.
[43] X. Cao, B. Yang, Y. Shen, et al., “Edge-assisted multi-layer offloading
optimization of LEO satellite-terrestrial integrated networks,” IEEE J.
Sel. Areas Commun., vol. 41, no. 2, pp. 381-398, Feb. 2023.
[44] Z. Wang, H. Yu, S. Zhu, and B. Yang, “Curriculum reinforcement
learning-based computation offloading approach in space-air-ground
integrated network,” in Proc. Int. Conf. Wirel. Commun. Signal Process.
(WCSP), Changsha, China, 2021, pp. 1-6.
[45] K. Yu, Q. Cui, Z. Zhang, X. Huang, X. Zhang, and X. Tao, “Efficient
UAV/satellite-assisted IoT task offloading: A multi-agent reinforcement
learning solution,” in Proc. Asia-Pacific Conf. Commun. (APCC), Jeju
Island, Korea, Republic of, 2022, pp. 83-88.
[46] X. Lin, A. Liu, C. Han, X. Liang, K. Pan, and Z. Gao, “LEO satellite
and UAVs assisted mobile edge computing for tactical ad-hoc network:
A game theory approach,” IEEE Internet Things J., vol. 10, no. 23, pp.
20560-20573, Dec. 2023.
16 VOLUME ,
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
[47] Y. Liu, L. Jiang, Q. Qi, and S. Xie, “Energy-efficient space–air–ground
integrated edge computing for Internet of Remote Things: A federated
DRL approach,” IEEE Internet Things J., vol. 10, no. 6, pp. 4845-4856,
Mar. 2023.
[48] Y. Jing, C. Jiang, N. Ge, and L. Kuang, “Resource optimization for
signal recognition in satellite MEC with federated learning,” in Proc.
Int. Conf. Wirel. Commun. Signal Process. (WCSP), Changsha, China,
2021, pp. 1-5.
[49] Y. Lyu, Z. Liu, R. Fan, C. Zhan, H. Hu, and J. An, “Optimal
computation offloading in collaborative LEO-IoT enabled MEC: A
multi-agent deep reinforcement learning approach,” IEEE Trans. Green
Commun. Netw., early access, 2022.
[50] C. Ding, J. -B. Wang, M. Cheng, M. Lin, and J. Cheng, “Dynamic
transmission and computation resource optimization for dense LEO
satellite assisted mobile-edge computing,” IEEE Trans. Commun., vol.
71, no. 5, pp. 3087-3102, May 2023.
[51] D. Kim and S. Jeong, “Joint optimization of offloading scheduling
and path planning for space-air-ground integrated edge computing
systems,” in Proc. Int. Conf. ICT Convergence (ICTC), Jeju Island,
Korea, Republic of, 2022, pp. 230-232.
[52] S. Jung, S. Jeong, J. Kang, and J. Kang, “Marine IoT systems
with space–air–sea integrated networks: Hybrid LEO and UAV edge
computing,” IEEE Internet of Things Journal, vol. 10, no. 23, pp. 20498-
20510, Dec. 2023.
[53] N. N. Ei, J. S. Yoon, and C. S. Hong, “Energy-aware task offloading
and resource allocation in space-aerial-integrated MEC system,” in
Proc. Asia-Pacific Netw. Oper. Manag. Symp. (APNOMS), Takamatsu,
Japan, 2022, pp. 1-6.
[54] B. Chen, N. Li, Y. Li, X. Tao, and G. Sun, “Energy efficient hybrid
offloading in space-air-ground integrated networks,” in Proc. IEEE
Wireless Commun. Networking Conf. (WCNC), Austin, TX, USA, 2022,
pp. 1319-1324.
[55] C. Ding, J. -B. Wang, H. Zhang, M. Lin, and G. Y. Li, “Joint
optimization of transmission and computation resources for satellite and
high altitude platform assisted edge computing,” IEEE Trans. Wireless
Commun., vol. 21, no. 2, pp. 1362-1377, Feb. 2022.
[56] Y. K. Tun, K. T. Kim, L. Zou, Z. Han, G. D´
an, and C. S. Hong, “Col-
laborative computing services at ground, air, and space: An optimization
approach,” IEEE Trans. Veh. Technol., vol. 73, no. 1, pp. 1491-1496,
Jan. 2024.
[57] Y. Wang, J. Zhang, X. Zhang, P. Wang, and L. Liu, “A computation
offloading strategy in satellite terrestrial networks with double edge
computing,” in Proc. IEEE Int. Conf. Commun. Syst. (ICCS), Chengdu,
China, 2018, pp. 450-455.
[58] C. Mei, C. Gao, Y. Xing, X. Bian, and B. Hu, “An energy consumption
minimization optimization scheme for HAP-satellites edge computing,”
in Proc. Int. Conf. Commun. Technol. (ICCT), Nanjing, China, 2022,
pp. 857-862.
[59] C. Zhou, W. Wu, H. He, et al., “Deep reinforcement learning for
delay-oriented IoT task scheduling in SAGIN,” IEEE Trans. Wireless
Commun., vol. 20, no. 2, pp. 911-925, Feb. 2021.
[60] S. Zhang, A. Liu, C. Han, X. Liang, X. Xu, and G. Wang, “Multi-
agent reinforcement learning-based orbital edge offloading in SAGIN
supporting Internet of Remote Things,” IEEE Internet Things J., vol.
10, no. 23, pp. 20472-20483, Dec. 2023.
[61] Y. Song, X. Li, H. Ji, and H. Zhang, “Joint computing, caching and
communication resource allocation in the satellite-terrestrial integrated
network with UE cooperation,” in Proc. IEEE/CIC Int. Conf. Commun.
China (ICCC), Sanshui, Foshan, China, 2022, pp. 604-609.
[62] L. Liu, J. Zhang, X. Zhang, P. Wang, Y. Wang, and L. Ouyang, “Design
and analysis of cooperative multicast-unicast transmission scheme in
hybrid satellite-terrestrial networks,” in Proc. IEEE Int. Conf. Commun.
Syst. (ICCS), Chengdu, China, 2018, pp. 309-314.
[63] D. Zhu, H. Liu, T. Li, et al., “Deep reinforcement learning-based task
offloading in satellite-terrestrial edge computing networks,” in Proc.
IEEE Wireless Commun. Networking Conf. (WCNC), Nanjing, China,
2021, pp. 1-7.
[64] L. Cheng, G. Feng, Y. Sun, M. Liu, and S. Qin, “Dynamic computation
offloading in satellite edge computing,” in Proc. IEEE Int. Conf.
Commun. (ICC), Seoul, Korea, Republic of, 2022, pp. 4721-4726.
[65] H. Li, C. Chen, C. Li, L. Liu, and G. Gui, “Aerial computing of-
floading by distributed deep learning in collaborative satellite-terrestrial
networks,” in Proc. Int. Conf. Wirel. Commun. Signal Process. (WCSP),
Changsha, China, 2021, pp. 1-6.
[66] P. Li, Y. Wang, and Z. Wang, “A game-based joint task offloading
and computation resource allocation strategy for hybrid edgy-cloud and
cloudy-edge enabled LEO satellite networks,” in Proc. IEEE/CIC Int.
Conf. Commun. China (ICCC), Sanshui, Foshan, China, 2022, pp. 868-
873.
[67] Y. Zhang, H. Zhang, K. Sun, J. Huo, N. Wang, and V. C. M. Leung,
“Partial computation offloading in satellite-based three-tier cloud-edge
integration networks,” IEEE Trans. Wireless Commun., vol. 23, no. 2,
pp. 836-847, Feb. 2024.
[68] Y. Song, X. Li, H. Ji, and H. Zhang, “Energy-aware task offloading
and resource allocation in the intelligent LEO satellite network,” in
Proc. IEEE Int. Symp. Person. Indoor Mobile Radio Commun. (PIMRC),
Kyoto, Japan, 2022, pp. 481-486.
[69] H. Wu, X. Yang, and Z. Bu, “Deep reinforcement learning for
computation offloading and resource allocation in satellite-terrestrial
integrated networks,” in Proc. IEEE Veh. Technol. Conf. (VTC2022-
Spring), Helsinki, Finland, 2022, pp. 1-5.
[70] M. M. Gost, I. Leyva-Mayorga, A. P´
erez-Neira, M. ´
A. V´
azquez, B.
Soret, and M. Moretti, “Edge computing and communication for energy-
efficient earth surveillance with LEO satellites,” in Proc. IEEE Int.
Conf. Commun. Workshops (ICC Workshops), Seoul, Korea, Republic
of, 2022, pp. 556-561.
[71] X. Gao, R. Liu, A. Kaushik, and H. Zhang, “Dynamic resource
allocation for virtual network function placement in satellite edge
clouds,” IEEE Trans. Netw. Sci. Eng., vol. 9, no. 4, pp. 2252-2265,
Jul./Aug. 2022.
[72] G. Cui, X. Li, L. Xu, and W. Wang, “Latency and energy optimization
for MEC enhanced SAT-IoT networks,” IEEE Access, vol. 8, pp. 55915-
55926, 2020.
[73] S. Yu, X. Gong, Q. Shi, X. Wang, and X. Chen, “EC-SAGINs: Edge-
computing-enhanced space–air–ground-integrated networks for Internet
of Vehicles,” IEEE Internet Things J., vol. 9, no. 8, pp. 5742-5754,
April, 2022.
[74] X. Zhu and C. Jiang, “Delay optimization for cooperative multi-
tier computing in integrated satellite-terrestrial networks,” IEEE J. Sel.
Areas Commun., vol. 41, no. 2, pp. 366-380, Feb. 2023.
[75] Z. Hu, F. Zeng, Z. Xiao, et al., “Joint resources allocation and 3D
trajectory optimization for UAV-enabled space-air-ground integrated
networks,” IEEE Trans. Veh. Technol., vol. 72, no. 11, pp. 14214-14229,
Nov. 2023.
[76] Y. Wang, W. Feng, J. Wang, and T. Q. S. Quek, “Hybrid satellite-
UAV-terrestrial networks for 6G ubiquitous coverage: A maritime
communications perspective,” IEEE J. Sel. Areas Commun., vol. 39,
no. 11, pp. 3475-3490, Nov. 2021.
[77] C. Lei, W. Feng, J. Wang, S. Jin, and N. Ge, “Control-oriented
power allocation for integrated satellite-UAV networks,” IEEE Wireless
Commun. Lett., vol. 12, no. 5, pp. 883-887, May 2023.
Yueshan Lin received the B.S. degree from the
Department of Electronic Engineering, Tsinghua
University, Beijing, China, in 2021. He is currently
pursuing the Ph.D. degree with the Department
of Electronic Engineering, Tsinghua University,
Beijing. His research interests include UAV com-
munications, satellite communications and mobile
edge computing.
Wei Feng (Senior Member, IEEE) received the
B.S. and Ph.D. degrees from the Department of
Electronic Engineering, Tsinghua University, Bei-
jing, China, in 2005 and 2010, respectively. He
is currently a Professor with the Department of
Electronic Engineering, Tsinghua University. His
research interests include maritime communication
networks, large-scale distributed antenna systems,
and coordinated satellite-UAV-terrestrial networks.
He serves as the Assistant to the Editor-in-Chief
of CH INA COMMUNICATIONS and an Editor of
IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NE T-
WORKING.
VOLUME , 17
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Yueshan Lin et al.: Satellite-MEC Integration for 6G Internet of Things: Minimal Structures, Advances, and Prospects
Yanmin Wang received the B.S. degree from
Shandong University, China, in 2008, and the
Ph.D. degree from the Department of Electronic
Engineering, Tsinghua University, Beijing, China,
in 2013. She is currently an Associate Professor
at the School of Information Engineering, Minzu
University of China. Her research interests include
distributed antenna systems, satellite networks, and
coordinated satellite-UAV-terrestrial networks.
Yunfei Chen (Senior Member, IEEE) received
the B.E. and M.E. degrees in electronics engineer-
ing from Shanghai Jiaotong University, Shanghai,
China, in 1998 and 2001, respectively, and the
Ph.D. degree from the University of Alberta in
2006. He is currently a Professor with the De-
partment of Engineering, University of Durham,
U.K. His research interests include wireless com-
munications, cognitive radios, wireless relaying,
and energy harvesting.
Yongxu Zhu (Senior Member, IEEE) received
her Ph.D. degree in electrical engineering from
University College London in 2017. From 2017
to 2023, she was a Research Associate at Lough-
borough University, a Senior Lecturer at London
South Bank University, and an Assistant Professor
at Warwick University. Since 2023, she has been a
Professor at Southeast University. She also serves
as an Editor for IE EE WIRELESS COMMUNI-
CATIONS LETTE RS and IEE E TRANSACTIONS
ON WIRELESS COMMUNICATIONS. Her research
interests include B5G/6G, heterogeneous networks, non-terrestrial networks,
and collective intelligence networks.
Ximu Zhang received the M.Sc. degree and
Ph.D. degree in information and communication
engineering from the Harbin Institute of Tech-
nology (HIT), Harbin, China, in 2017 and 2022,
respectively. She is currently a Postdoctoral Fellow
with the Department of Electronics Engineering,
Tsinghua University. Her research interests fo-
cus on satellite networking, integrated terrestrial
satellite communications, and resource allocation
problems.
Ning Ge (Member, IEEE) received the B.S. and
Ph.D. degrees from Tsinghua University, Beijing,
China, in 1993 and 1997, respectively. From 1998
to 2000, he was with ADC Telecommunications,
Dallas, TX, USA, where he researched the de-
velopment of ATM switch fabric ASIC. Since
2000, he has been a Professor with the Department
of Electronics Engineering, Tsinghua University.
He has published over 100 papers. His current
research interests include communication ASIC
design, short range wireless communication, and
wireless communications. Dr. Ge is a senior member of the China Institute
of Communications and the Chinese Institute of Electronics.
Yue Gao (Fellow, IEEE) received a PhD from the
Queen Mary University of London (QMUL), U.K.,
in 2007. He is a chair professor at the School
of Computer Science, director of the Intelligent
Networking and Computing Research Centre at
Fudan University, China, and a visiting professor
at the University of Surrey, U.K. He worked as a
lecturer, a senior lecturer, a reader, and the chair
professor at QMUL and the University of Surrey,
respectively. He has published 200 peer-reviewed
journal and conference papers. His research in-
terests include sparse signal processing, smart antennas and cognitive
networks for mobile and satellite systems. He was a co-recipient of the
EU Horizon Prize Award on Collaborative Spectrum Sharing in 2016 and
an Engineering and Physical Sciences Research Council fellow in 2017. He
is a member of the Board of Governors and a Distinguished Speaker of the
IEEE Vehicular Technology Society (VTS), the chair of the IEEE ComSoc
Wireless Communication Technical Committee, and the past chair of the
IEEE ComSoc Technical Committee on Cognitive Networks. He has been
an editor of several IEEE Transactions and Journals, the Symposia Chair,
and the Track Chair. He has other roles in the organising committee of
several IEEE ComSoC, VTS, and other conferences.
18 VOLUME ,
This article has been accepted for publication in IEEE Open Journal of the Communications Society. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJCOMS.2024.3418860
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
