Dong-Hyun Jung, Gyeongrae Im, Joon-Gyu Ryu, Seungkeun Park, Heejung Yu, Junil Choi
©SHUTTERSTOCK/PHONLAMAI PHOTO
Recently, megaconstellations with a massive number of low Earth orbit (LEO) satellites are being considered as a possible solution for providing global coverage due to relatively low latency and high throughput compared to geosynchronous orbit (GEO) satellites. However, as the number of satellites and operators participating in the LEO constellation increases, inter-satellite interference will become more severe, which may yield marginal improvement or even decrement in network throughput. In this article, we introduce the concept of satellite clusters that can enhance network performance through satellites’ cooperative transmissions. The characteristics, formation types, and transmission schemes for the satellite clusters are highlighted. Simulation results evaluate the impact of clustering from coverage and capacity perspectives, showing that when the number of satellites is large, the performance of clustered networks outperforms the unclustered ones. The viable network architectures of the satellite cluster are proposed based on the 3rd Generation Partnership Project (3GPP) standard. Finally, the future applications of clustered satellite networks are discussed.
Satellite communications have been recently employed to provide global Internet services exploiting the large coverage of satellites. The beam size of GEO satellites is typically 200–3,500 km, while that of non-GEO satellites, such as medium Earth orbits and LEOs is 100–1,000 km [1]. Although the LEO satellites’ coverage is small compared to that of GEO satellites, the LEO satellites have recently received great attention because of relatively low latency and high throughput due to their low altitudes. In addition, the LEO satellites are becoming more miniaturized in size, integrated, and light-weighted, which reduces manufacturing time and launch costs [2].
Typically, as the number of satellites in a constellation increases, the throughput of the satellite network is also enhanced. However, when a sufficiently large number of satellites has already been launched, adding more satellites in orbit may increase inter-satellite interference, resulting in the marginal enhancement of network throughput. Motivated by this, the concept of satellite cluster (i.e., a set of closely-located small satellites) has been investigated to further enhance the performance through cooperation [3].
A satellite cluster is a group of multiple satellites placed nearby where the satellites cooperatively transmit and receive signals as if they were a single multiantenna satellite, as shown in Figure 1. The satellites in a cluster act as either “leader” or “follower,” according to the preassigned roles where the cluster consists of one leader and multiple follower satellites. The leader satellite manages the entire cluster and plays a key role in cooperative transmissions by exchanging information and control signals with the follower satellites through inter-satellite links (ISLs). Thus, powerful on-board processors for inter-satellite communications and routing are required at the leader’s payload. The follower satellites passively operate as directed by their leader and serve as cooperative nodes to improve the performance of the cluster.
Figure 1 Concept of the satellite clusters.
In this article, we investigate satellite clusters to enhance the network capacity by cooperative transmissions among multiple satellites. We describe the characteristics, formation types, and transmission schemes of the satellite clusters and evaluate the network performance from coverage and capacity perspectives by stochastic geometry-based simulations. A key finding is that when the number of satellites is large, the performance of the clustered satellite networks is better than that of the unclustered networks. We also propose and compare 3GPP-based architectures for the satellite clusters and address possible challenges. The applications of the satellite clusters are also discussed.
In this section, we discuss the satellite cluster’s characteristics, formation, and cooperative transmission schemes. We also evaluate the network performance in terms of capacity and coverage probability based on the stochastic geometry.
Different from the unclustered satellite constellations, satellite clusters have the following advantages.
The satellites in a cluster shape a certain formation and move in groups. Thus, the relative distances among the satellites are very close, e.g., several to hundreds of kilometers according to the cluster size. Thanks to this geographical proximity, accurate beam pointing between satellites, which is generally considered as a main challenge of conventional ISLs, becomes more straightforward because the distance between the leader and follower is much shorter than that of the conventional ISLs. Moreover, as the orbits of followers are determined by the leader’s orbit, adding more followers in a cluster requires negligible orbital resources.
To cope with increased user requirements, satellites should have more complex functionalities. This in turn causes larger payloads and higher production and launch costs. The satellite cluster can solve this problem by dividing functionalities of one complex leader satellite into multiple smaller follower satellites. Thus, the follower satellites may have as few functionalities as possible and take a portion of the leader’s load. Moreover, as the follower satellites cannot operate without the leader but are used only for coverage extension and throughput increase, the full protocol stack is not a requirement for the followers. These make it easy to reduce the payload of the leader as well as scale up the cluster.
The LEO satellites do not always have a direct path to a ground terminal due to their low altitudes, since the probability that the satellites experience a line-of-sight (LOS) channel varies according to the elevation angle [4]. Transmissions from multiple satellites in the cluster may increase the LOS probability and overcome the performance degradation by blockages. Although a dominant LOS path between one satellite and a ground terminal exists, the channels from the satellites in a cluster to the terminal may not be highly correlated because the distance between satellites is much farther than the wavelength of the carrier frequency, e.g., several to hundreds of kilometers. Therefore, spatial diversity can be achieved, and this would enhance the network throughput [5].
When the satellites are in operation after settling in orbit, some satellites may break down due to unexpected hardware or software problems. In the unclustered constellations, if a satellite malfunctions or does not work for a certain time, its coverage area can be served by spare satellites in the same or another orbital plane. However, replacing a satellite from one in the same or another orbital plane may take a long time, e.g., several days to months. In contrast, when a follower satellite in a cluster gets out of order, the coverage may remain by the other satellites in the cluster.
A possible candidate to enable formation flying of satellites is the projected circular orbit (PCO). In PCO-based formation flying, the leader satellite in the cluster follows a reference orbit, while the remaining follower satellites travel along the PCOs, which have slightly different inclinations and eccentricities from the reference orbit, as shown in Figure 2. With this small change of the orbital configuration, the follower satellites tend to circularly orbit the leader satellite when viewed from the Earth. This PCO-based formation flying was successfully demonstrated in the Canadian nanosatellite program called the CanX-4&5 mission [6]. By simply extending this concept to multiple satellites, the leader satellite surrounded by more than two follower satellites could be implemented [7]. In this article, we consider two PCO-based cluster formations: circular and uniform clusters, as shown in Figure 3.
Figure 2 (a) Orbital configuration and (b) view from the Earth for the PCO-based satellite cluster with three satellites, where the gray sphere is the leader satellite and the blue and green spheres are the followers. From ${t} = {t}_{i}$ to ${t} = {t}_{{i} + {1}},{i} = {\{}{1},{2},{3}{\}},$ the followers rotate 90° when viewed from the Earth [6].
Figure 3 Circular and uniform cluster distributions. The shaded area is the spherical-cap-shaped region where the satellites in a cluster can be distributed. The altitude, the radius of the Earth, and the polar angle of the spherical cap are denoted by a, ${r}_{e},$ and ${\theta}_{c},$ respectively.
A circular cluster is a cluster in which the follower satellites are equidistant from the leader satellite, forming a circular swarm. With the circular formation, the distances of the ISLs are comparable, which is a big advantage for synchronizing transmission timing across the satellites in the cluster. In addition, when the follower satellites are equally distant from the adjacent followers, it is expected that the antenna or lens alignment for ISL communications can be simple. However, when the number of satellites in clusters is large, there exists a space limit to deploy the follower satellites over a ring-shaped track.
In a uniform cluster, the leader satellite is in the middle of the spherical cap where the follower satellites are uniformly distributed. This formation may be made up of multilayered PCOs with a different distance to the leader satellite [7]. In contrast to the circular formation, the uniform cluster has unequal distances of the ISLs, which yields difficulty in timing synchronization of the satellites and transceiver alignment of the ISLs. This formation, however, is appropriate for dense deployment of satellites, since the uniform distribution efficiently uses the distributed area. In other words, the uniform formation can accommodate more satellites than the circular formation, when the two formations have the same minimum distance among the satellites.
The concept of cooperative multipoint (CoMP) was proposed to enhance the throughput of cell-edge users by mitigating inter-cell interference from multiple transmission points. Several scenarios for intra-evolved Node B (eNB) CoMP, which uses multiple remote radio heads (RRHs) to perform CoMP at a single eNB, were considered in release 11, assuming ideal backhauling, while release 12 focused on inter-eNB CoMP, i.e., CoMP involving multiple eNBs, with nonideal backhauling. The 3GPP has considered several transmission schemes for downlink (DL) CoMP, such as joint transmission (JT) and dynamic point selection (DPS), which can be applied to the satellite clusters.
The JT uses multiple transmission points in a cell (intracell JT) or in different cells (inter-cell JT) to transmit signals to a user. The maximum ratio transmission (MRT) can be applicable to the JT of the satellite clusters. However, as the amplitude and phase of the channel coefficients are required for the MRT, which should be first estimated and then distributed to all of the follower satellites, strict requirements of the ISL would be necessary. In addition, the satellites may have a problem in amplifying signals because the satellites usually operate near the saturation level of power amplifier to compensate for the large path loss. This results in performance degradation due to nonlinear distortion of the desired signals. Instead of the MRT, the equal gain transmission can be used as an alternative for the JT, equally allocating transmit power to the satellites in the cluster. This scheme only requires the phase of channel coefficients, which allows the inexpensive power amplifiers to be mounted at the satellite’s payload. This can be a great merit for the satellite clusters because the power amplifier of the satellite payload (e.g., traveling wave tube amplifier) is a key component to compensate for the substantial power loss of the received signals. However, as the JT relies on precise cooperation among satellites in the cluster, strict requirements on the ISL capacity and synchronization are necessary, especially with a large number of followers.
The DPS is a very simple beamforming scheme that selects only a single transmission point with the best channel condition. The DPS can mitigate the inter-cell interference because it mutes other transmission points that are not selected. With this advantage, the DPS would be well applicable to clustered satellite networks, especially with a massive number of satellites. Unless the leader is the best choice, the leader needs to inform only one follower, which has the best channel condition, through the intra-cluster link. In this regard, the DPS can significantly reduce the requirement for ISL capacity and timing synchronization among the satellites. However, the DPS leads to frequent changes between satellites in the cluster, so low-latency ISL switching is required.
Considering the discussion so far, we evaluate the DL performance of clustered satellite networks using stochastic geometry-based simulations [8]. Assume that the satellites are located at the altitude of 600 km and operated in S-band (2 GHz) with 30-MHz bandwidth. The free-space path-loss model is adapted with the path-loss exponent of 3, and the shadowed-Rician fading is assumed with average shadowing. For simplicity, we assume the satellites generate a single beam with the maximum transmit antenna gain of 30 dBi, the 3-dB beamwidth of 20°, and the beam pattern given in [9]. The satellites maintain their beam boresight in the direction of the subsatellite point. The terminal has an omnidirectional antenna with the gain of 0 dB. The effective isotropic radiated power density and noise spectral density are set to 34 dBW/Hz and –174 dBm/Hz, respectively. For the clustered networks, the polar angle of the spherical cap where the satellites can be distributed is set to 1°, and 10% of the total satellites are the leaders, while the remaining satellites are the followers. For example, when the number of satellites is 1,000, there are 100 clusters, each consisting of one leader and nine followers. For the unclustered networks, all satellites are independently distributed where each satellite works as an independent transmitter.
We use a homogeneous binomial point process (BPP) to model the cluster distribution [8]. The leaders are uniformly distributed according to the homogeneous BPP over a sphere. For the circular formation, the followers are spaced at the boundary of the spherical cap, as shown in Figure 3, maintaining relatively equal distances to the adjacent followers. For the uniform formation, the followers are distributed on the spherical cap according to the homogeneous BPP.
In Figure 4, the ergodic capacities of the unclustered and clustered networks are compared with various numbers of satellites, assuming the circular formation and the MRT-based JT for the clustered networks. When the number of satellites is small, the unclustered network achieves a higher ergodic capacity than the clustered networks. In contrast, for large numbers of satellites, the satellite cluster achieves higher performance by cooperative transmissions. This proves that the satellite cluster can play an important role to enable megaconstellations as more satellites exist in the space. With an excessively large number of satellites, however, the performance of both unclustered and clustered networks decreases due to the higher inter-satellite interference. This explains that the constellation design with an appropriate number of satellites is very important in terms of the network performance. For different beamwidths, this tendency still holds, but the number of satellites at which the performance starts to degrade would be increased with a smaller beamwidth and vice versa. The JT has a greater capacity than the DPS for small numbers of satellites due to the optimality of the MRT with negligible inter-satellite interference. In contrast, the DPS outperforms the JT when the number of satellites becomes large because the DPS significantly reduce the inter-satellite interference.
Figure 4 Ergodic capacity versus the number of satellites.
Figure 5 shows the coverage probabilities, i.e., the probabilities that the signal-to-interference-plus-noise ratio (SINR) is higher than a threshold. As expected, the clustering has benefits in terms of coverage probability. It is shown that the satellite clusters with the circular formation have better performance than that with the uniform formation. Thus, with the small cluster size, the circular formation is preferable due to low requirements for ISLs and accuracy of position control, while uniform formation may be suitable with the large cluster size because the satellites can be deployed efficiently.
Figure 5 Coverage probability versus SINR threshold when the total number of satellites is 10,000.
In this section, we propose the 3GPP-based network architectures for the satellite clusters considering functional split options presented in [10]. We assume that the leader has all of the network functions of next-generation (NG) Node B (gNB) in 5G new radio (NR) and is connected to the 5G core (5GC) network through a gateway using the NG interface logically and the satellite radio interface physically. We discuss feasible functional split options that can be applicable to the clustered satellite networks and address several technical challenges to enable the proposed architectures for the satellite clusters.
A centralized or cloud radio access network (C-RAN) architecture has been proposed to increase the installation efficiency of base stations in a cost-effective manner. The C-RAN physically separates the base station into the RRH and baseband unit (BBU). The RRHs are deployed and distributed at each cell site while the BBUs are colocated and centralized. The common public radio interface is used as the fronthaul interface between the RRH and BBU. With such a C-RAN architecture, the centralized BBU is able to reduce the rental cost and electricity rate by performing baseband processing for the multiple distributed RRHs.
In 5G NR, however, the C-RAN structure brings another problem in that the fronthaul requires a much higher capacity to transmit the I/Q data sampled in the time domain. To mitigate such a problem, the open RAN structure has been proposed by selectively applying eight functional split options for the gNB [10]. Note that, in 3GPP, option 2, i.e., the package data convergence protocol/radio link control (PDCP/RLC) split, was recommended as the higher layer split, while the O-RAN Alliance selected option 7-2x, i.e., the low-physical layer (PHY) split between the resource element mapper and beamformer, as the lower layer split. Different from the C-RAN, the open RAN splits the functions of the gNB in higher layers into two units: a distributed unit (DU) and a centralized unit (CU).
Specifically, the leader is assumed to have full functionalities of the gNB (DU and CU), while the only gNB-DU is equipped at each follower to support the cluster’s cooperative transmissions. With this assumption, we propose the following three functional split options between the leader and follower:
The features and applicability of the three options are summarized in Table 1, where the required data rates are calculated with the following parameters: 100-MHz bandwidth, 256-quadrature amplitude modulation, 32 antenna ports, and eight multiple-input, multiple-output layers [11]. Satellite operators planning to offer 3GPP-based services with satellite clusters may select the best split option based on the following discussions.
Table 1 Candidate options for functional split.
As shown in Figure 6, in this split option, the follower has the low-PHY layer, including orthogonal frequency division multiplexing modulation/demodulation, and resource element mapping/demapping, and the radio frequency (RF) part, while the leader includes the other upper-layer functions. This intra-PHY split is similar to the open RAN structure, and it can be a suitable option to ensure compatibility and coordination with the terrestrial network. This option suggests a centralized architecture optimized for utilizing NR features, such as CoMP and carrier aggregation (CA). It also requires the simplest architecture among the three considered options for the follower satellite’s payload. This advantage significantly reduces the manufacturing and launch costs of the followers and saves the energy for signal processing, which consequently makes it easy to scale up the number of satellites to obtain more diversity gain. However, the data rate requirement in the ISL significantly increases and the latency requirement must be very low. Thus, this option can be applied only when the distance between the leader and follower is close enough to satisfy the latency requirements, and the ISL between leader and follower satellites is nearly perfect.
Figure 6 Cluster architecture with intra-PHY split. IP: Internet protocol; GTP-U: general packet radio service tunnelling protocol-user plane; SDAP: service data adaptation protocol; UDP: user datagram protocol; SRI: satellite radio interface.
This option splits the functions in the middle of MAC layer: that is, radio resource control (RRC), PDCP, RLC, and high-MAC layers are in the leader and low-MAC and PHY layers, as well as the RF part are in the follower. This intra-MAC split is aimed at reducing the latency of hybrid automatic repeat request (HARQ) protocol to ease the constraints of fronthaul capacity and increase reliability while ensuring data rate. Since HARQ is processed in DU, the followers handle their own HARQ processes. Hence, the leader can have much smaller buffer and less complex processor because the leader only handles its own HARQ. Still, some of CoMP functions and CA can be utilized at the leader. However, the interface between CU/DU becomes complex, and the scheduling operations over CU/DU should be defined additionally.
In this option, RRC and PDCP layers are in the leader and the RLC, MAC, and PHY layers and the RF part are in the follower. With this higher-layer split, the followers take a large portion of gNB’s processing, resulting in the low ISL capacity and latency requirements. The clustered satellite network with this split can utilize dual connectivity (DC) since it has a similar structure to the existing DC. When the cluster size is large, the satellites may have a large delay and performance degradation in the ISLs due to the long distance and antenna or lens misalignment. Thus, this option is well applicable to the satellite cluster especially when the satellites are far apart. In contrast, the network functions up to the RLC layer should be implemented at the followers, which requires complex payloads and reduces the scalability of the cluster. In addition, it is difficult to use cooperative transmission schemes such as CoMP, since the MAC layer is not centralized, and there may be security issues because the coordination of security configurations between different PDCP instances is required.
Maintaining the formation of satellites in a cluster is crucial to guarantee reliable inter-satellite communications. However, the accurate position control of the multiple satellites is challenging due to environmental disturbances, such as atmospheric drag and solar pressure [12]. This may require a complex ground control system, as well as high-performance sensors on board. For example, optical sensor-aided position control systems have been developed to enable proximate formation flying of spacecrafts at the Marshall Space Flight Center in the National Aeronautics and Space Administration (NASA), where the sensors are used to calculate the relative distance between two spacecrafts. Moreover, the Goddard enhanced onboard navigation system, also known as GEONS, is a software developed at the Goddard Space Flight Center in NASA, which uses standard GPS receivers and onboard sensors to provide accurate relative navigation solutions in real time.
To enable the proposed architectures for the satellite cluster, a high data rate is required at the ISL, as discussed in Table 1. Especially, the intra-PHY split requires the data rate in the ISL up to approximately 86 Gb/s. As a solution to achieve such a high data rate, free-space optical (FSO) communications can be used for the ISLs instead of the RF interface. However, several challenges must be addressed to apply the FSO for ISLs between the leader and follower satellites [13]. As slightly different orbits between the leader and follower make the difference in relative angular motion of two satellites, a point-ahead angle prediction is required to compensate for the difference. In addition, due to the narrow beamwidth of the FSO and large ISL distance, the algorithms for acquisition, tracking, and pointing must be accurate. Since the capacity of the current FSO technology may not satisfy the fronthaul data rate requirement of the intra-PHY split, the compression of fronthaul data can be a promising approach for the proposed architecture [14].
Time and frequency synchronization is a critical issue for the satellite cluster. The signals from the satellites may experience different propagation delays caused by the following factors:
In addition, the relative velocity of the satellites in the cluster may be dissimilar due to unequal elevation angles, which causes different Doppler shifts in the signals from the satellites. This misalignment in time and frequency would degrade the performance of the cooperative transmissions. The precompensation for the timing difference and Doppler shift can be done to resolve this time and frequency uncertainty. For example, the timing and frequency offsets can be precompensated by obtaining the position of the UE through global navigation satellite systems and the speed and position of the satellite through the ephemeris information [1].
As the leader plays an important role in the operation of the cluster, the followers are highly dependent on the leader. In the worst case, where the leader breaks down, the loss of the leader would render the whole cluster unusable, which is the inherent problem due to the leader–follower relationship. To mitigate this problem, multiple leaders may be placed in the cluster. The additional leader may work as a follower in the normal operation or remain inactive as a spare leader. If the original leader fails, the additional leader would start to serve as a real leader.
In this section, we discuss possible application scenarios where the merits of the satellite clusters for LEO megaconstellations can be exploited.
To expand application scenarios of satellite communications, direct access to smartphones is an essential requirement. Compared to the conventional UE in satellite communications, smartphones may have a lower antenna gain, for example, up to 5 dB less gain, and lower transmit power due to less battery capacity. Therefore, enhancement of the link budget is required for the direct access. Recent trials for direct access to smartphones include iPhone14’s satellite communication feature and the cooperation between SpaceX and T-Mobile, but both provide low data rate services in some limited applications. The cooperative transmissions among clustered satellites can be one of the promising approaches to this end.
As various services demanding extremely high computational loads emerge, distributed computing is a key requirement in future networks. The satellites in a cluster may work as edge nodes performing distributed computing. For example, the followers may offload the computation tasks of the leader in a distributed manner. With the preconfigured topology of the cluster, stable and expectable computing performance can be achieved. With unclustered satellites, however, dynamic group formation and ISL management are required to incorporate multiple LEOs moving independently.
The satellite clusters can be utilized for localization of UEs instead of the existing GPS. It means that a UE can estimate its position without an additional GPS receiver. Because LEO satellites are located at much lower altitudes than GPS satellites, they have the potential to deliver several benefits in terms of navigation, precise point positioning/timing, and location-enabled communications. With the clustered LEO satellites whose relative positions are maintained, the aforementioned benefits can be more easily obtained than unclustered LEO satellites. This is because the synchronized transmissions of the reference signals among multiple satellites can be simply implemented. However, the clock accuracy of LEO satellites is very critical for positioning through LEO constellations because errors in clock estimates degrade the accuracy of the global navigation satellite systems measurements. LEO satellites can be equipped with low-cost chip-scale atomic clocks to enhance the accuracy.
When GEO satellite networks coordinate with the LEO satellite clusters, the GEO satellites can serve the UEs with low requirements on the latency and throughput, while shorter latency and higher throughput are provided by the LEO clusters. This could balance the load between the two satellite networks, and make a better usage of the resources of the whole network. The DC operation as in 5G NR [15] could be considered for LEO and GEO satellite networks to improve throughput and reduce service interruption. For example, the GEOs are configured as the leaders, while the LEOs are the followers.
In this article, we introduced the concept of a satellite cluster for satellite communications achieving high throughput by cooperative transmissions. The characteristics, formation, and cooperative transmission schemes of the satellite cluster were discussed and stochastic geometry-based simulations were performed to evaluate the coverage and capacity performance. We also proposed the practical architectures for clustered satellite networks based on the 3GPP standard and discussed the related challenges. The future applications of the clustered LEO satellite networks were also discussed.
This work was supported by an Institute of Information and communications Technology Planning and Evaluation grant funded by the Korea government (MSIT) (2021-0-00847, Development of 3D Spatial Satellite Communications Technology). Heejung Yu and Junil Choi are the corresponding authors of this article.
Dong-Hyun Jung (dhjung@etri.re.kr) is a senior researcher at the Radio & Satellite Research Division, Electronics and Telecommunications Research Institute, 34129, South Korea. He received his M.S. degree in electrical engineering from Seoul National University in 2017 and is currently pursuing his Ph.D. degree in electrical engineering from the Korea Advanced Institute of Science and Technology.
Gyeongrae Im (imgrae@etri.re.kr) is a senior researcher at the Radio & Satellite Research Division, Electronics and Telecommunications Research Institute, Daejeon 34129, South Korea. He received his Ph.D. in electrical engineering from Seoul National University in 2019.
Joon-Gyu Ryu (jgryurt@etri.re.kr) is a principal researcher at the Radio & Satellite Research Division, Electronics and Telecommunications Research Institute, Daejeon 34129, South Korea. He received his Ph.D. in radio and information communications engineering from Chungnam National University in 2014.
Seungkeun Park (seungkp@etri.re.kr) is assistant vice president for the Radio & Satellite Research Division, Electronics and Telecommunications Research Institute, Daejeon 34129, South Korea. He received his Ph.D. in information communication engineering from the University of Chungbuk in 2004.
Heejung Yu (heejungyu@korea.ac.kr) is a professor in the Department of Electronics and Information Engineering at Korea University, Sejong 30019, South Korea. He received his Ph.D. in electrical engineering from the Korea Advanced Institute of Science and Technology, Daejeon, South Korea, in 2011.
Junil Choi (junil@kaist.ac.kr) is an associate professor at the School of Electrical Engineering at the Korea Advanced Institute of Science and Technology, Daejeon 34141, South Korea. He received his Ph.D. degree in electrical and computer engineering from Purdue University in 2015.
[1] “Solutions for NR to support non-terrestrial networks (NTN),” 3rd Generation Partnership Project, Sophia Antipolis, France, 3GPP Tech. Rep. 38.821 v16.0.0, Dec. 2019.
[2] R. Radhakrishnan et al., “Survey of inter-satellite communication for small satellite systems: Physical layer to network layer view,” IEEE Commun. Surveys Tuts., vol. 18, no. 4, pp. 2442–2473, 4th Quart., 2016, doi: 10.1109/COMST.2016.2564990.
[3] G.-P. Liu and S. Zhang, “A survey on formation control of small satellites,” Proc. IEEE, vol. 106, no. 3, pp. 440–457, Mar. 2018, doi: 10.1109/JPROC.2018.2794879.
[4] “Study on NR to support non-terrestrial networks,” 3rd Generation Partnership Project, Sophia Antipolis, France, 3GPP Tech. Rep. 38.811 v15.4.0, Sep. 2020.
[5] Q.-Y. Yu et al., “Virtual multi-beamforming for distributed satellite clusters in space information networks,” IEEE Wireless Commun., vol. 23, no. 1, pp. 95–101, Feb. 2016, doi: 10.1109/MWC.2016.7422411.
[6] J. K. Eyer et al., “A formation flying control algorithm for the CanX-4&5 low Earth orbit nanosatellite mission,” Space Technol., vol. 27, no. 4, pp. 147–158, Sep. 2007.
[7] A. M. Popov et al., “Development and simulation of motion control system for small satellites formation,” Electronics, vol. 10, no. 24, Dec. 2021, Art. no. 3111, doi: 10.3390/electronics10243111.
[8] R. Wang, M. A. Kishk, and M.-S. Alouini, “Ultra-dense LEO satellite-based communication systems: A novel modeling technique,” IEEE Commun. Mag., vol. 60, no. 4, pp. 25–31, Apr. 2022, doi: 10.1109/MCOM.001.2100800.
[9] J. Chu et al., “Robust design for NOMA-based multibeam LEO satellite internet of things,” IEEE Internet Things J., vol. 8, no. 3, pp. 1959–1970, Feb. 2021, doi: 10.1109/JIOT.2020.3015995.
[10] “Study on new radio access technology: Radio access architecture and interfaces,” 3rd Generation Partnership Project, Sophia Antipolis, France, 3GPP Tech. Rep. 38.801 v14.0.0, Mar. 2017.
[11] “5G wireless fronthaul requirements in a passive optical network context,” International Telecommunication Union, Geneva, Switzerland, ITU-T Rec. G. Sup. 66, Sep. 2020.
[12] S. Bandyopadhyay et al., “Review of formation flying and constellation missions using nanosatellites,” J. Spacecraft Rockets, vol. 53, no. 3, pp. 567–578, Mar. 2016, doi: 10.2514/1.A33291.
[13] A. U. Chaudhry and H. Yanikomeroglu, “Free space optics for next-generation satellite networks,” IEEE Consum. Electron. Mag., vol. 10, no. 6, pp. 21–31, Nov. 2021, doi: 10.1109/MCE.2020.3029772.
[14] H. Yu and J. Joung, “Optimization of frame structure and fronthaul compression for uplink C-RAN under time-varying channels,” IEEE Trans. Wireless Commun., vol. 20, no. 2, pp. 1278–1292, Feb. 2021, doi: 10.1109/TWC.2020.3032432.
[15] J. F. Monserrat et al., “Multi-radio dual connectivity for 5G small cells interworking,” IEEE Commun. Standards Mag., vol. 4, no. 3, pp. 30–36, Sep. 2020, doi: 10.1109/MCOMSTD.001.1800030.
Digital Object Identifier 10.1109/MVT.2023.3262360