Distributed satellite systems

Replacing one large satellite with many small ones changes what is possible and what is hard. Some measurements require simultaneous observation from separated points and cannot be made any other way. Some functions become harder because the system must now coordinate itself.

Why distribute

The first motivation is the one that is genuinely enabling rather than merely economic. A single spacecraft measuring a changing field cannot tell whether a variation is spatial or temporal, because it moved and time passed simultaneously. Four spacecraft in a tetrahedron can, which is why Cluster and MMS fly that formation.

Formation flight

Formation flight means actively maintaining relative geometry, which is different from a constellation where satellites merely occupy assigned slots.

From Rendezvous and docking, relative motion in a circular orbit follows the Clohessy-Wiltshire equations, and the key results carry directly.

Free-drift relative orbits exist. A properly established relative orbit is a closed ellipse elongated two to one along-track, traversed once per orbit with no propellant. Formations are designed around these, because anything else costs continuous thrust.

Differential drag and $J_2$ break formations. Satellites at slightly different altitudes experience different drag and different nodal precession, so a formation shears apart unless controlled. In low orbit, differential drag can also be used as a control input by changing attitude to vary each vehicle’s ballistic coefficient, which is propellantless formation control.

GRACE is the clearest illustration of measurement by formation. Its science product is the change in separation between two satellites, measured to microns, as they pass over mass anomalies. The satellites are the instrument, and neither alone measures anything.

Inter-satellite links

Connecting satellites to each other rather than only to the ground changes the architecture fundamentally.

Optical links are now operational at constellation scale, and their advantages are decisive where they work: no spectrum licensing, no interference, very high capacity, and a beam narrow enough to be difficult to intercept.

The difficulty is pointing. Beamwidths are microradians, and the terminals are on platforms moving at kilometres per second relative to each other. Acquisition requires a search, tracking requires continuous closed-loop control, and point-ahead compensation is needed because the receiving satellite moves measurably during the light travel time.

The architectural payoff is that a constellation with inter-satellite links no longer needs a gateway in view of every serving satellite. Traffic hops through the mesh to a satellite that can see one, which is what makes service over oceans and polar regions possible without landing rights everywhere. See Communications satellites.

Routing in a moving topology

The distinguishing feature of a satellite network is that its topology is deterministic. Terrestrial routing protocols discover an unknown topology and converge; a satellite network’s future connectivity is computable from orbital mechanics arbitrarily far ahead.

That makes reactive protocols the wrong tool. Contact graph routing instead plans over a schedule of known future contacts, which is both more efficient and more stable.

Where connectivity is intermittent, as in deep space, delay tolerant networking applies. The bundle protocol stores data at each hop and forwards it when a link becomes available, with custody transfer so responsibility for delivery passes explicitly. There is never an end-to-end path, which breaks every assumption TCP makes, and DTN exists precisely because of that.

Fractionated spacecraft

The more radical proposition is to split a spacecraft’s functions across separate free-flying modules: one carrying the payload, another power, another communications, connected wirelessly.

The claimed advantages are that modules can be launched, upgraded and replaced independently, that a failure removes one function rather than the vehicle, and that the system can be reconfigured in orbit.

The concept has been studied extensively, notably in DARPA’s System F6, and has not been deployed. The obstacles are that wireless power transfer between modules is inefficient, that maintaining formation costs propellant continuously, that the interfaces are more complex than the internal ones they replace, and that the reliability argument is weakened by adding formation control as a new failure mode.

The honest assessment is that the idea is intellectually attractive and has not survived contact with the mass and power budget. Constellations of complete small satellites captured most of the resilience benefit without the coordination cost.

Coordination

A distributed system must agree on things, and it must do so over links that come and go.

Autonomy is required rather than preferred, because the ground cannot arbitrate in real time. Ground-in-the-loop coordination reintroduces the contact window constraint described in Autonomous spacecraft, and a formation that must wait for a pass to resolve a conflict is not a formation.

The practical pattern for large constellations is hierarchical: each satellite handles its own safety autonomously, the constellation operator plans at fleet level on the ground, and a small set of decisions is delegated to the space segment where latency demands it.

References

• Alfriend, K. T. et al. Spacecraft Formation Flying: Dynamics, Control and Navigation, Butterworth-Heinemann, 2010.
• Tapley, B. D. et al. “GRACE Measurements of Mass Variability in the Earth System”, Science, 305:503, 2004.
• Burch, J. L. et al. “Magnetospheric Multiscale Overview and Science Objectives”, Space Science Reviews, 199:5, 2016.
• Burleigh, S. et al. “Delay-tolerant networking: an approach to interplanetary Internet”, IEEE Communications Magazine, 41:128, 2003.
• Brown, O. and Eremenko, P. “Fractionated Space Architectures: A Vision for Responsive Space”, AIAA RS4-2006.
• Toyoshima, M. “Recent trends in space laser communications for small satellites”, Optical Engineering, 60:034106, 2021.