Large Constellations of Small Satellites: A Survey of Near Future Challenges and Missions

: Constellations of satellites are being proposed in large numbers; most of them are expected to be in orbit within the next decade. They will provide communication to unserved and underserved communities, enable global monitoring of Earth and enhance space observation. Mostly enabled by technology miniaturization, satellite constellations require a coordinated e ﬀ ort to face the technological limits in spacecraft operations and space tra ﬃ c. At the moment in fact, no cost-e ﬀ ective infrastructure is available to withstand coordinated ﬂight of large ﬂeets of satellites. In order for large constellations to be sustainable, there is the need to e ﬃ ciently integrate and use them in the current space framework. This review paper provides an overview of the available experience in constellation operations and statistical trends about upcoming constellations at the moment of writing. It highlights also the tools most often proposed in the analyzed works to overcome constellation management issues, such as applications of machine learning / artiﬁcial intelligence and resource / infrastructure sharing. As such, it is intended to be a useful resource for both identifying emerging trends in satellite constellations, and enabling technologies still requiring substantial development e ﬀ orts.


Introduction
The idea of a constellation of satellites appeared in the market about twenty years ago with Iridium and Globalstar as pioneering examples. They offered worldwide communication links competing with the terrestrial cellular network, having, however, limited success due to service costs mainly [1,2]. We can now say that at that time the business model was not sustainable due to the small market and high initial and maintenance costs.
Access to space is now broadening thanks to technology miniaturization and design experience. With a tremendous forecasted increase in the launch rate for small satellites (from pico-sized to mini-sized) [3], constellations are getting attention again from sustainable businesses [4].
Constellations have their greatest potential in the communication field. The upcoming era of the Internet-Of-Things requires the communication infrastructure to handle huge amounts of data and to guarantee service in any geographical position. Constellations, however, also have great potential in weather science, safety/security and disaster monitoring [5].
In the rapidly emerging business of satellite constellations, it is important to track and update the information about players and trends to guide future developments. While for other emerging satellite related markets such as the one of the nanosatellite revolution, several surveys have already been published, both technical [6] or market oriented [3], the authors are not aware of a similar effort towards satellite constellations. As remarkable exception, the very recent work in [7], is mainly devoted to Space Traffic Management. In [8] constellations are instead considered as part of the wider context of distributed satellite systems (DSS). They are characterized morphologically as those DSS having high Although some constellations still have unpublished years of completion, many are expected to be in orbit by 2022 with a peak in 2020. This reflects the rapid development of the market and the high competitiveness which stresses the shortening of the time-to-market. A clear outlier is the Starlink constellation from SpaceX, with a 4425-element constellation dominating the columns of the years 2019 to 2024-the year in which it is expected to be completed. This delay in the time-to-market is probably due to the large amount of satellites that are expected to be placed.
Looking at the cumulative sum, the number of satellites in orbit is going to increase more than linearly with about 8000 spacecraft in orbit in 2024 due to constellations only. The resulting volume of traffic opens up a question about whether it is possible to sustain this development or not, under several points of view. For example, the current ground segment infrastructures will probably not be able to monitor and control such a large number of satellites. A major satellite ground service provider such as KSAT is already investing for infrastructure enlargement [15]. At the same time, constellation management shall be enhanced to make an efficient use the new infrastructure. This requires new operational architectures towards higher automation, either onboard or on ground, involving, for instance, artificial intelligence and virtual reality [16,17]. A second concern involves the communication, with the RF spectrum becoming possibly overcrowded and the required datathroughput increasingly larger. Lastly, but probably most importantly, the space traffic and debris issues, which may prevent the safe and successful operation of spacecraft [17].  Looking at the cumulative sum, the number of satellites in orbit is going to increase more than linearly with about 8000 spacecraft in orbit in 2024 due to constellations only. The resulting volume of traffic opens up a question about whether it is possible to sustain this development or not, under several points of view. For example, the current ground segment infrastructures will probably not be able to monitor and control such a large number of satellites. A major satellite ground service provider such as KSAT is already investing for infrastructure enlargement [15]. At the same time, constellation management shall be enhanced to make an efficient use the new infrastructure. This requires new operational architectures towards higher automation, either onboard or on ground, involving, for instance, artificial intelligence and virtual reality [16,17]. A second concern involves the communication, with the RF spectrum becoming possibly overcrowded and the required data-throughput increasingly larger. Lastly, but probably most importantly, the space traffic and debris issues, which may prevent the safe and successful operation of spacecraft [17]. Although some constellations still have unpublished years of completion, many are expected to be in orbit by 2022 with a peak in 2020. This reflects the rapid development of the market and the high competitiveness which stresses the shortening of the time-to-market. A clear outlier is the Starlink constellation from SpaceX, with a 4425-element constellation dominating the columns of the years 2019 to 2024-the year in which it is expected to be completed. This delay in the time-to-market is probably due to the large amount of satellites that are expected to be placed.
Looking at the cumulative sum, the number of satellites in orbit is going to increase more than linearly with about 8000 spacecraft in orbit in 2024 due to constellations only. The resulting volume of traffic opens up a question about whether it is possible to sustain this development or not, under several points of view. For example, the current ground segment infrastructures will probably not be able to monitor and control such a large number of satellites. A major satellite ground service provider such as KSAT is already investing for infrastructure enlargement [15]. At the same time, constellation management shall be enhanced to make an efficient use the new infrastructure. This requires new operational architectures towards higher automation, either onboard or on ground, involving, for instance, artificial intelligence and virtual reality [16,17]. A second concern involves the communication, with the RF spectrum becoming possibly overcrowded and the required datathroughput increasingly larger. Lastly, but probably most importantly, the space traffic and debris issues, which may prevent the safe and successful operation of spacecraft [17].  Table A1 in Appendix A for details). Black, dotted line represents cumulative sum. Not-classified (NC) stands for companies that have not published expected time-to-service. As far as the size is concerned, the proposed constellations are quite widespread. According to Figure 3, roughly half of the constellations are relatively small, containing less than 50 elements. Constellations with numbers of elements between 50 and 150 elements appear to be quite appealing as well, with a 22% share. Larger size slots are instead of particularly low interest. Finally, a considerable 17% features non-declared sizes, as anticipated.
Satellite sizes show a clear bias towards micro and nano classes, reflecting the trend towards the miniaturization of satellite platforms in general ( Figure 4). Although 35% of the constellation projects do not declare size, 32% belong to nano-class and 18% to micro-class. Pico, mini, medium and large classes reflect the minority, with a total share of 15%. This means that constellations are going to be composed of satellites weighting mainly from 1 to 100 kg.  Table A1 in Appendix for details). Black, dotted line represents cumulative sum. Not-classified (NC) stands for companies that have not published expected time-to-service.
As far as the size is concerned, the proposed constellations are quite widespread. According to Figure 3, roughly half of the constellations are relatively small, containing less than 50 elements. Constellations with numbers of elements between 50 and 150 elements appear to be quite appealing as well, with a 22% share. Larger size slots are instead of particularly low interest. Finally, a considerable 17% features non-declared sizes, as anticipated.
Satellite sizes show a clear bias towards micro and nano classes, reflecting the trend towards the miniaturization of satellite platforms in general ( Figure 4). Although 35% of the constellation projects do not declare size, 32% belong to nano-class and 18% to micro-class. Pico, mini, medium and large classes reflect the minority, with a total share of 15%. This means that constellations are going to be composed of satellites weighting mainly from 1 to 100 kg.   . Constellations by satellite sizes: mass < 1 kg = pico; 1 kg < mass < 10 kg = nano; 10 kg < mass < 100 kg = micro; 100 kg < mass < 500 kg = mini; 500 kg < mass <1000 kg = medium, mass > 1000 kg = large.

Future Challenges
Large constellations will require a paradigm shift with respect to the way space missions are currently handled, with major challenges involving technical, management and regulatory aspects. The main ones are discussed hereafter by analyzing the work that has been done so far and outlining . Constellations by satellite sizes: mass < 1 kg = pico; 1 kg < mass < 10 kg = nano; 10 kg < mass < 100 kg = micro; 100 kg < mass < 500 kg = mini; 500 kg < mass <1000 kg = medium, mass > 1000 kg = large.

Future Challenges
Large constellations will require a paradigm shift with respect to the way space missions are currently handled, with major challenges involving technical, management and regulatory aspects. The main ones are discussed hereafter by analyzing the work that has been done so far and outlining possible future developments.

Constellation Management
With such a number of active elements in orbit, their management is a fundamental point of interest. Since constellations have not been widely used in the past, not too much work has been done in this sense. However, some key experiences can certainly be located in projects such as Galileo (European GNSS) [18], Globalstar constellations [19] and other GNSS constellations (GPS, BeiDou, etc.).
In [18,19] different strategies have been explored, all driven by the common objective of enhancing the level of automation, so that more satellites will not translate into a proportional increase of managing effort. Proposed solutions fall into two main categories: (a) Optimization of automatic satellite tracking (such as telemetry download). (b) Automatic failure detection, so that the operator does not need to manually check the satellite's status of health.
Still, there are a great number of non-automatic operations left to do; [19,20] pointed out some desired operational improvements, along with lessons learnt from operating constellations. Some points have been found to be very real and worth being analyzed here:

1.
Splitting between payload operations and spacecraft operations, possibly with dedicating ground segments to each of the two.

2.
Increasing automation onboard the spacecraft, which is not always possible due to satellite size constraints. In the latter case, expert systems (intended as an ensemble of algorithms, and machines aiding the operator's decisions, usually associated with artificial intelligence, are applied for high level tasks-prediction, planning, diagnosis, repair, etc.) shall be deployed on ground.

3.
Taking not only the ground segment into account but also the operations from the initial phases of the constellation design.

4.
Expert systems shall be designed to assist operators and keep the workload constant during constellation operations, e.g., mitigating the heavier workload during launch and early orbit phase.
As per point 2, the need for automation, a relevant example is the collision avoidance assessment and maneuver planning, which is now largely manual. This problem is addressed in [7], where the need for an increased accuracy in orbit determination is pushed forward as a means to reduce false alarms and implement automatic orbital corrections.
In favor of point 2, the authors of [21] propose an onboard automated management system, based on artificial intelligence. In their implementation, failures are not only detected, but also handled automatically towards a resolution along with a re-scheduling of the original plan. The main drawback of the proposed approach is its need for an intersatellite-link, which is usually not affordable for low-cost strategies. Moreover, it implies intensive intersatellite communication contributing to RF spectrum crowding.
Another approach is proposed in [22] for an Earth Observation constellation. It consists of an autonomous (re-)planning strategy for maximizing the total science return of observations over time.
Despite not reaching the degree of automation as in [21], the automatic re-planning decreases the ground workload, allowing the operator to concentrate more on the goal, rather than the path to it. Thus, this is also in favor of point 4.
Regarding point 3, a few aspects that shall be included at the beginning of the constellation design operation-wise will be now discussed. The automation logic given above is one example. In fact, it impacts greatly on the development complicating the space segment design, at least software-wise. The replacing strategy is another operational aspect that must be taken into account from the beginning: a replacing or spare strategy is the policy adopted by the operators to substitute failed or terminated satellites of the constellation. Most of the constellations need such a policy to guarantee a 24/7 service, affecting launching phases and satellite design (reliability). The authors in [23] give an example of work in this sense through the application of inventory management approach to the space field. They propose a set of parking orbits and in-plane spares which are refurnished from ground following an optimal policy (i.e., minimize the total Expected Spare Strategy Cost). These spare satellites are then moved into the constellation when needed. Notice that the reliability of the constellation can be regarded also from a building process perspective as stated in [24] (rather than satellite design only) which suggests, for instance, adopting a Failure Mode and Effect Analysis for a robust mass production.
The deployment strategy is also fundamental to be included in point 3, given the constellation size trend and the competitiveness of the field. Spacecraft deployment must be accounted for since the beginning because it has a significant impact on the lifecycle cost. In fact, it affects both the number of launches and the complexity of the satellite to be launched. In principle one launch for every orbital plane is needed, also the complexity of the onboard propulsion system (if any) changes based on the post-launch operations to be performed.
Few interesting works in this respect are [25][26][27][28]. The approach studied in [25] consists of deploying the spacecraft gradually as they are needed by the market, which is shown to reduce the life cycle cost of a constellation significantly, of about 20% when applied to the Globalstar case study. Similarly, [28] consider a staged deployment which is optimized using genetic algorithms. This approach seems to be particularly appealing as it makes the constellation size adaptable to the market reaction, which is very difficult to predict as Iridium and Globalstar experiences have shown.
Other efficient deployment strategies are compared in [26] such as J2 driven deployment and carrier-vehicle deployment. These methods allow one to configure the orbits in space avoiding multiple launches. The use of Earth-Moon Lagrangian point L1 is envisaged in [27], however, this solution seems to be convenient only in combination with carrier vehicles.

Communication Issues
Communication is a key point when operating a constellation, in fact:

•
On-Board Automation is unlikely to grow to the point of allowing fully autonomous fleet management: a large amount of satellites will thus need to communicate frequently with ground.

•
Constellations are designed usually for real time-24/7-purposes, requiring data down/up-load at any time.
The above two points rise concerns about Radio Frequency (RF) spectrum partitioning. An overcrowded RF spectrum may indeed cause physical interference of adjacent RF signals. At the same time, the traffic capacity of the communication infrastructure shall grow in parallel with the data volume travelling in the RF channels.
The research community is actively working towards these two aspects. One of the most promising approach towards infrastructure optimization is the sharing and integration between space and ground communication networks, especially in view of the upcoming 5G service [29]. Work in this sense can be found since the turn of the century, see e.g., [30] where the possibility of integrating a satellite network with a terrestrial network is envisaged using an IP-based communication. Few years later, authors of [31] provided a survey of mobile satellite systems endorsing IP-based communication, discouraging however communication satellites other than GEO, mainly for cost efficiency. Despite the miniaturization trend in LEO spacecraft was already established by then, the recommendation towards GEO is not surprising, given the negative experiences of the firsts LEO constellations like Iridium and Globalstar. More recently, due to the rising interest in satellite constellations authors of [32,33] brought the attention again onto the potential of LEO satellites as communication infrastructure. The first studies the integration of the 5G ground network with a space-based network emphasizing its potential for global connectivity. The second studies the efficient implementation of inter-satellite link through a routing algorithm. The algorithm takes into account maximum available link time and remnant bandwidth to increase the total traffic capacity of the network in the presence of handover.
A more practical solution to the frequent communication needed by low-autonomy spacecraft is a conventional network of existing ground stations. This solution could be already feasible upon standards definition for communication (e.g., CCSDS) and hardware interfaces, in fact ground station providers like KSAT or KRATOS are going towards this direction. KRATOS for example designed a device called quantumCMD [34], a small computer able to operate up to 4 satellites when integrated in a ground station. The power of the device is its scalability with number of satellites and ground stations.
Solutions against the increasingly crowded spectrum are devoted mainly towards (a) spectrum sharing and (b) enhancement of regulation. Examples can be found in [35], where the possibility of a Database-Assisted spectrum sharing is pushed forward. Using this approach, the temporarily unused spectrum could be reallocated for a more efficient use. Note that also [35] uses the keyword "sharing" between satellite and terrestrial networks.
A completely different approach to avoid RF spectrum overcrowding consists of moving to the optical part of the spectrum. Optical communication promises higher data rates using smaller and lighter terminals, even though due its high sensitivity to atmospheric conditions it is more suited for free-space inter-satellite links rather than satellite-to-ground [36] (enabling communication between constellation elements is an asset by its own, though). An optical communication system conceived for LEO constellations is described in [37] and is currently at an advanced development stage.

Space Traffic Management
Among the issues to be faced in the "constellation race" the space traffic management is probably the most critical, yet not directly faced. [7,38] are the few works authors are aware of that discuss this topic, the first specifically within the large constellations framework and the second in general.
Part of this topic is closely related to space debris. Debris are already a problem that is being faced actively with surveillance networks (e.g., the JSpOC, Joint Space Operation Center [39]) and avoidance maneuvers from the spacecraft operators. Will this network be able to withstand, i.e., generate alerts and provide further assistance, also for the future traffic? What happens if a constellation is very valuable but cannot embark a propulsion system for collision avoidance? First steps towards answering these questions are found in very recent studies: [7] discusses the changes in LEO population environment due to large constellations, while authors of [40] advocate the need for updating state-of-the art space debris modelling as a result of the evolving debris environment.
A preventive approach is already taking place thanks to debris mitigation policies i.e., making the spacecraft reenter at the end of its life. [40] suggests that the constancy of the rate at which spacecraft fragmentations (historically the main cause of space debris) occurs, despite the drastic increase in the number of orbiting spacecraft [41], is an indication that mitigation efforts put in place are being successful. Whether or not such efforts will remain effective when hundreds of dismissed or failed constellation satellites will be deorbiting almost simultaneously, is still an open point.
Alternative solutions to the debris problem include active removal [42] and space-based surveillance networks [43,44]. On-Orbit servicing is a further option to decrease the amount of failed or dead satellites that become a debris. Its implementation, however, is still costly and technically challenging [45].
Regulatory aspects of space traffic management are instead poorly covered. Currently, once a free orbital slot has been identified, the common practice consists of seeking for a technically and economically viable solution to reach such a slot. Not much attention, instead, is payed to interferences affecting other operators, that might be caused while the spacecraft are reaching the target orbit or during de-orbiting at the end of life. The authors in [7] envisage an architecture similar to the air traffic management with traffic zones (orbital slots) and "flight plans." Though up to now it has been safe to Aerospace 2020, 7, 133 8 of 18 assume that space is so large that satellite operations do not interfere with each other, this may not be true in the near future. For instance, with thousands more spacecraft in orbit, an Earth observation satellite may find unexpectedly another one in its field of view, or a region of space may become so overcrowded as to impact the quality of space observations from ground. One such event was indeed experienced after the launch of the first Starlink satellites ( Figure 5).
Besides the regulatory part, there are also technical challenges to be overcome. In [46,47] for instance, machine learning through support vector machines is used both to monitor satellite health and address management issues, highlighting a great potential of neural networks for enhancing space traffic management. Other approaches, such as space transponders and enhanced tracking with corner reflectors or onboard GPS receivers, are pushed forward in [7].
Novel concepts of conjunction assessment services are also on their way. In [38] a prototype of a ground-based service that can interface with all subscribed satellite operators (scalable solution) was presented. Besides integrating different object databases and giving alerts similarly to JSpOC, it can compute the most suitable avoidance maneuvers. Thanks to the global situational awareness of the service, such a maneuver can ensure minimum fuel consumption while avoiding "cascade maneuvers." Moreover, after suggesting the maneuver directly to the involved spacecraft operators, it can update its database and inform the other operators when the maneuver is accomplished.
Another interesting attempt is discussed in [48], wherein the Australian Government is financing a conjunction assessment service featuring a ground-based laser "deviator." The aim is to maneuver small uncooperative objects remotely from ground using a laser beam, which is theoretically feasible but with great technical challenges due to the laser power needed.

Discussion
The increasing trend of launching very small satellites into space has been clear and well established for the last two decades. Smaller satellites are in turn paving the way to constellations, which are gaining widespread interest. Spacecraft constellations are appealing, especially in three fields, namely, (i) communications, for global coverage, (ii) Earth observation, for near real time measurements and (iii) space observation for continuous monitoring/surveillance.
During this survey, about one hundred companies have been found proposing constellations with varying numbers of satellites. Satellite sizes range mainly from nano to micro-sizes, i.e., from 1 kg up to 100 kg. Most of them are expected to be active in orbit before 2025. The most common number of elements for the constellations is below 150 units.
A review was carried out on the challenges that satellite constellations will have to face to become more sustainable, with a focus on three categories: constellation management, communication and space traffic. The level of maturity reached by these three areas is, however, not homogeneous: communication is probably the most matured, where relevant work is being done

Discussion
The increasing trend of launching very small satellites into space has been clear and well established for the last two decades. Smaller satellites are in turn paving the way to constellations, which are gaining widespread interest. Spacecraft constellations are appealing, especially in three fields, namely, (i) communications, for global coverage, (ii) Earth observation, for near real time measurements and (iii) space observation for continuous monitoring/surveillance.
During this survey, about one hundred companies have been found proposing constellations with varying numbers of satellites. Satellite sizes range mainly from nano to micro-sizes, i.e., from 1 kg up to 100 kg. Most of them are expected to be active in orbit before 2025. The most common number of elements for the constellations is below 150 units.
A review was carried out on the challenges that satellite constellations will have to face to become more sustainable, with a focus on three categories: constellation management, communication and space traffic. The level of maturity reached by these three areas is, however, not homogeneous: communication is probably the most matured, where relevant work is being done concerning infrastructure integration and protocol efficiency. Some significant past experience in constellation management was reviewed during this survey; still, the need for an improved level of automation is clear. In this respect, artificial intelligence seems to be a valuable option, together with infrastructure sharing for rapid development and commercial viability. On the other hand, space traffic management is mostly unprepared, with significant developments only in terms of debris countermeasures. Some extensions of the air traffic regulation are expected in the years to come to mitigate the current free-space policy covered by the Outer Space Treaty.
Although many technical challenges are still being addressed, the amount of work that has been analyzed during this review suggests good chances of success for large constellation missions.

Conflicts of Interest:
The authors declare no conflict of interest.
Appendix A Table A1. List of constellations surveyed in this work. "?" stands for unknown; "*" stands for "dormant or cancelled project".

Company
No