Gateway is often described as the Moon’s first spaceport, but almost nothing about it resembles an airport on Earth. There will be no runway, no holding pattern drawn against a fixed landscape and no controller watching radar returns from a tower. The station and every spacecraft approaching it will be moving through a three-body gravitational environment in which Earth, the Moon and time all matter.
That creates a deceptively difficult question. If Orion, a lunar lander or a logistics vehicle reaches the Moon before a docking port is available, where can it wait without drifting too far away, wasting propellant or approaching another spacecraft?
A team with members from NASA’s Johnson Space Center, Texas A&M University, a.i. solutions and Purdue University has now published the most detailed integrated treatment yet of that traffic problem. The work does not establish finished flight rules. It uses thousands of simulated trajectories to compare how Gateway and a waiting vehicle could maintain predictable spacing in the Moon’s planned near-rectilinear halo orbit.
Gateway will not circle the Moon like Apollo did
A low lunar orbit looks intuitive on a diagram: a spacecraft repeatedly circles close to the surface. Gateway is planned for something far more elongated, called a near-rectilinear halo orbit, or NRHO. It is associated with the Earth-Moon L2 region and is shaped by the gravity of both bodies rather than being a simple circle centred only on the Moon.
According to NASA’s current Gateway description, the station will pass about 1,500 kilometres from the Moon at its closest and reach roughly 70,000 kilometres at its farthest. One circuit takes about 6.5 days. The route sweeps over the lunar poles and maintains a line of sight to Earth.
The orbit was selected as a compromise. It offers access to the lunar south polar region while demanding less long-term stationkeeping propellant than a low lunar orbit. NASA’s explanation of the choice describes it as combining useful surface access with fuel efficiency and continuous communications.
Calling the path an orbit can make it sound like a permanent rail. It is not. The particular 9:2 lunar-synodic-resonant NRHO modelled by the researchers is slightly unstable. Small navigation errors, imperfect burns, solar radiation pressure and routine spacecraft operations gradually push a vehicle away from its intended path. Gateway and nearby visitors must make periodic corrections.
A port can be busy before anyone docks
Gateway is intended to receive several kinds of visiting spacecraft. Orion would bring astronauts. Human Landing System vehicles would carry crews between the station and the surface. Logistics modules would deliver supplies, while later spacecraft could arrive for science, servicing or other cislunar missions.
Those arrivals will not always align perfectly. A launch can slip. A port may be occupied. Orion may need to deliver a module to one berth, back away and relocate to another. A lander could reach lunar space well before its crew leaves Earth.
The Acta Astronautica study separates the problem into four timescales. Port relocation may take only hours. Short-term loitering can last as long as one 6.5-day revolution. Medium-term loitering covers a vehicle waiting for up to 90 days for an approaching mission. Long-term cases include spacecraft supporting later missions or operating independently near Gateway.
The new analysis concentrates on the medium-term case. That is long enough for tiny errors to accumulate, but close enough to an intended docking that the waiting vehicle should not retreat to a remote storage orbit unless necessary.
The proposed answer is a string of pearls
The most direct strategy places Gateway and the loitering spacecraft on the same broad orbital path but separates them in phase. One crosses a chosen point before the other, like beads spaced along a moving string. Engineers can describe the gap in minutes because a fixed time offset corresponds to very different distances around the highly elongated orbit.
Near perilune, the close approach to the Moon, the spacecraft move rapidly and a short time gap can mean thousands of kilometres. Near apolune, where they are farthest away and moving more slowly, the same time gap shrinks to hundreds or even tens of kilometres.
A vehicle preparing to dock benefits from being close. A smaller phase gap usually shortens the transfer and reduces the propellant required to reach Gateway. Yet bringing it closer increases conjunction risk if navigation or thruster errors cause the two paths to drift together. The traffic problem is therefore a trade between proximity and margin, not a search for one universally ideal parking distance.
This builds on a path already tested in flight. SpaceDaily covered CAPSTONE before its launch as the small pathfinder for Gateway’s orbit. The CubeSat subsequently became the first spacecraft to operate in an NRHO and demonstrated that the planned dynamical neighbourhood could be reached and maintained.
Why following Gateway is better than following a map
The researchers compared two broad forms of stationkeeping. In one, Gateway and the visitor each try to follow separate precomputed reference trajectories. Their errors are independent. Both can remain reasonably close to their own plan while still drifting dangerously close to each other.
The alternative is relative stationkeeping. Gateway continues to target its baseline path, while the loitering craft targets Gateway’s projected motion plus a chosen time offset. If Gateway arrives several minutes early or late at a future perilune, the visitor adjusts to preserve the relative gap rather than blindly following an ideal schedule that Gateway itself has missed.
This is not autonomous formation flight in the everyday sense. The paper models scheduled orbit-maintenance manoeuvres, navigation cut-off times, execution errors and a propagated estimate of where Gateway will be several revolutions later. The important change is the reference: the waiting craft follows the real station’s predicted rhythm.
That relative method added some propellant cost and operational complexity for the loiterer. It also reduced the range of possible phase drift, making close formations more predictable. Around Gateway, predictability is itself a safety resource.
Ten minutes was too close in the simulations
The team tested nominal separations of 60, 30 and 10 minutes. Each case ran through 100 Monte Carlo trials over 56 NRHO revolutions, about one year, with randomised navigation, manoeuvre-execution and solar-pressure errors. The one-year span exceeded the 90-day medium-term definition so that the method could settle and its longer behaviour could be seen.
With a 60-minute offset and the noisier spacecraft assumptions, the vehicles remained about 5,400 kilometres apart near perilune and roughly 400 kilometres apart near apolune. At 30 minutes, typical separation was about 2,900 kilometres near perilune and 100 to 300 kilometres near apolune.
The 10-minute case was different. Mean separation near perilune remained just above 1,000 kilometres, but the visiting vehicle came within 220 kilometres there in the trials. Near apolune, the minimum fell to six kilometres with the noisier assumptions and three kilometres for the quieter spacecraft case. The intended ten-minute phase difference sometimes narrowed to roughly two minutes or less.
The authors did not label six kilometres as a universal collision boundary. They explicitly noted that an exact safe range is difficult to define because navigation uncertainty, vehicle modelling and burn errors interact with the sensitive dynamics. They did conclude that approaches on the order of tens of kilometres were likely unsafe for medium-term loitering under the model, and they did not recommend the ten-minute option they tested.
Independent flight plans allowed startling close approaches
A second comparison shows why relative control matters. Two vehicles began 30 minutes apart, but each maintained its own separate reference orbit. In some trials the visitor changed from trailing Gateway to leading it. The pair came within 1.5 kilometres near perilune and four kilometres near apolune, despite both trying to follow planned trajectories.
The relative method held phase variations to about plus or minus 20 minutes in the model, compared with roughly plus or minus 35 minutes when the spacecraft followed independent baselines with different targeting settings. Tighter control on both independent paths also maintained separation, but required Gateway itself to operate more precisely.
The lesson is not that a 30-minute nominal gap is automatically safe. It is that traffic rules must be built around relative motion and uncertainty. A clearance based only on two perfect paper orbits could miss the way real vehicles wander.
Safety has a small but measurable fuel price
Keeping a tighter schedule required more frequent or larger corrections. In one seven-year set of single-spacecraft simulations, the loosest targeting case used an average 1.4 metres per second of velocity change each year, but its largest timing deviation averaged more than 50 minutes. A tighter case raised annual cost to 1.9 metres per second and cut that average maximum deviation to 7.5 minutes.
The difference, about half a metre per second per year, is modest in propulsion terms. Operational effort also matters, however. Burns below a minimum threshold can be waived, simplifying work on the ground. Tighter control meant fewer opportunities to skip a manoeuvre.
Targeting settings need not remain fixed for the station’s life. The simulations alternated loose and tight periods, suggesting that Gateway could conserve propellant and operational attention between missions, then tighten its adherence before visiting vehicles arrive. The response was not immediate: after changing settings, the trajectory took several revolutions to settle into its new pattern.
Real spacecraft are noisier than mathematical points
Spacecraft disturb their own trajectories. Reaction wheels must be unloaded with thrusters. Docking contact and separation springs impart small forces. Plumes, vented gas, attitude manoeuvres and the pressure of sunlight all matter when the desired path is slightly unstable.
The researchers first used a comparatively simple model for the multi-vehicle cases, then tested tighter Gateway targeting in a higher-fidelity simulation containing assumed Artemis docking, undocking, plume, contact, venting and attitude-control disturbances. The improved adherence survived that more realistic test.
That validation should not be stretched too far. The higher-fidelity section modelled one Gateway stack across mission events; it was not a complete repetition of every two-spacecraft loitering case. The study proposes and stress-tests strategies, but flight controllers will still need vehicle-specific navigation performance, propulsion behaviour, failure cases and agreed safety margins before turning them into operational rules.
The physical station is also still being assembled on Earth. Earlier SpaceDaily coverage followed HALO’s preparation and its future docking interfaces, while a more recent report described the control centre Germany is developing for European Gateway elements. The traffic analysis belongs to that same long preparation: the procedures must exist before several expensive vehicles converge.
A traffic system, not lunar air traffic control
The language of airports is useful because Gateway is meant to be a place where crews, cargo and vehicles meet before continuing elsewhere. It becomes misleading if it implies a fixed building or an earthly controller issuing simple headings. No spacecraft can stop in space. Waiting means following another trajectory, measuring it continuously and spending propellant to remain where mission designers intended.
Nor does the paper create a complete cislunar traffic-management authority. It addresses a defined orbital configuration, a set of assumed errors and a particular medium-term loitering problem. Debris, failed manoeuvres, communications outages, emergency departures and traffic outside Gateway’s immediate neighbourhood require additional work.
What the research supplies is a mathematical beginning: identify the kinds of waiting that missions will require, control vehicles relative to one another, simulate uncertainty rather than perfect flight and accept that closer parking costs safety margin. At the Moon’s first planned spaceport, the essential infrastructure may be invisible. It will consist of predicted trajectories, navigation updates and small burns that keep every moving arrival in its proper place.