Subsystem 08 of 12 · Autonomy
Lunar Habitat ACT, the Autonomous Construction Testbed·Autonomy·Concept definition, revision A·August 2026·See it in the tour

Two point six seconds is too long to drive anything

Earth is 1.28 light seconds away. Every command arrives late and every consequence is seen later still, which makes remote operation of machinery unsafe and autonomy a requirement rather than a feature.

2.6 sround trip
1.3 mblind travel per command
94%machine hours unsupervised, target
Contents
  1. The number that decides everything
  2. Where the decisions are made
  3. What runs unsupervised
  4. Edge compute
  5. Failure modes, and the ones that matter
  6. What is not yet known
  7. Increment plan
  8. Sources and further reading
Section 01

The number that decides everything

Light takes 1.28 seconds to cross the 384,400 km between Earth and the Moon, so a command and its acknowledgement take 2.56 seconds at absolute best. Add ground processing, relay hops and the operator's own reaction time and three to four seconds is realistic.

For a rover moving at a walking pace of 0.5 m/s, that is 1.3 metres travelled between the operator seeing something and the machine responding to it. A boulder that appears at the edge of a camera frame is already under the wheel by the time the stop command lands.

Apollo drove the lunar roving vehicle with a human aboard, which sidesteps this entirely. An uncrewed construction site does not have that option, and the whole point of this base is that machines do the dangerous, repetitive work while the crew do the parts that need judgement.

1.28 sone way
2.56 stheoretical round trip
3.4 srealistic, with processing
1.7 mblind travel at 0.5 m/s
Section 02

Where the decisions are made

Earth goals and review Base supervisor site scheduling Crew exceptions Machine autonomy task execution Safety layer hard stops Actuators daily plan tasks override
Earth sets goals on a daily cadence and reviews what happened. It never commands an actuator. The base supervisor turns goals into scheduled tasks. The machines execute them and handle everything that happens inside a task. The safety layer sits below all of it and can only stop things, never start them.
The rule that makes this safe

The safety layer is the lowest and simplest thing in the stack, it is separately verified, and it has exactly one authority: to stop. It cannot be talked out of a stop by a higher layer, and nothing above it can command motion the safety layer has inhibited. Complexity goes upward, authority to move goes downward, and the two never cross.

Section 03

What runs unsupervised

TaskAutonomy levelHuman roleWhy
Regolith excavationrepetitive, boundedFullSet the face and the quotaWell defined, low consequence, endlessly repetitive
Haulage between pointsknown routeFullNoneGraded roads and a surveyed route. This is the easy case.
Laser sinteringthe pad and roadsFullApprove the planBounded work area, continuous quality feedback from the ultrasonic head
Paver placementbermsSupervisedWatch and approvePlacement errors compound, so a human checks each course
ISRU plant operationin situ resource utilisation, continuousFullException onlySteady state process with clear limits
Regolith placement over modulesbulk shielding, a later incrementTeleoperatedDirect controlA dropped load on a pressure vessel is unrecoverable
Any work near the habitatwithin 15 mSupervisedCrew presentConsequence, not difficulty
Anything near a personextravehicular activity (EVA) in progressInhibitedMachine stopsThe safety layer, not a policy
Overall94% of machine hours unsupervisedDesign targetNo flight precedent yet, see section 06
Section 04

Edge compute

All of this runs on the base rather than in orbit or on Earth, because a decision loop that crosses a radio link is a decision loop that can be interrupted. The compute is modest by terrestrial standards and enormous by spacecraft standards, which is the interesting part. The figures in the table are allocations for the concept, not measured hardware.

ElementSpecPowerNote
Perceptionstereo and lidar24 TOPS78 WTrillion operations per second (TOPS). Radiation tolerant accelerator, derated from the commercial part
Planning and schedulingsite supervisor8 core, 32 GB42 WRuns the daily plan and the task queue
Safety layerseparate hardwareLockstep microcontroller (MCU)6 WFormally verified, does one thing
Storagelogs and models8 TB12 WEverything is logged, because Earth reviews after the fact
Total138 W0.5 percent of the habitat load

138 watts to remove most of the need for a human to be outside. Against an EVA hour that costs 0.043 mSv of crew dose, the open surface rate the radiation document carries, plus the hours of preparation a conventional airlock adds before every EVA, that is the best trade in the entire architecture.

Section 05

Failure modes, and the ones that matter

FailureEffectDetectionResponse
Perception false negativemisses an obstacleCollisionLidar and bumper disagree with the cameraSafety layer stops on any disagreement. Slow, and correct.
Planner deadlocktwo machines want the same spaceWork stopsWatchdog on task progressBoth stop, the supervisor re sequences. Stopping is always safe.
Link loss to Earthexpected dailyNo effectNot neededThe base was never waiting for Earth. Logs queue and forward.
Link loss to the basea machine goes out of rangeMachine is aloneHeartbeat timeout, 5 sSafe stop, then autonomous return along the recorded track.
Model driftconditions changeDegraded decisionsEarth review of the logsModels are updated on a weekly cadence, never in real time.
Why nothing updates itself in real time

An autonomous system that changes its own behaviour between reviews is a system nobody can certify. Models are frozen for a week at a time, every decision is logged, and Earth reviews the logs before a new model is uplinked. It is slower than continuous learning and it is the only version of this that is defensible.

Section 06

What is not yet known

The 94 percent unsupervised figure is a target with no flight precedent. The longest autonomous drives on another world are Perseverance's: 347.7 m in a single Martian day and 699.9 m without human review, on Mars, with a ground team planning each sol. NASA's ISRU Pilot Excavator is designed to dig and deliver 10 tonnes of regolith over 11 days with its own autonomous digging software, and it has not flown. Teleoperation studies show task performance falling sharply between half a second and three seconds of delay, which supports section 01, but nobody has yet run construction machinery on the Moon at a 2.6 second loop to measure how much a human in the cupola can safely take back. The 24 TOPS radiation tolerant accelerator, the lockstep safety controller and the 138 W total are allocations, not parts that have flown. The safety layer's stop authority has to be verified against every actuator on the site before the first autonomous task, and how that is done for a machine whose models change between weekly uploads has not been settled.

Section 07

Increment plan

Increment 1, teleoperation with local safety

Everything is driven from the cupola console inside the horizon, with the safety layer already in place. No autonomy yet, but the architecture it will sit in is built first.

Increment 2, autonomous haulage

The easy case: known route, graded road, bounded speed. This is where the operational confidence comes from.

Increment 3, autonomous excavation and sintering

The two tasks that consume the most machine hours. Getting these unsupervised is what actually reduces crew EVA time.

Increment 4, multi machine coordination

Excavator, hauler and gantry working the same site simultaneously, sequenced by the base supervisor. This is the step that makes the construction rate anything other than linear.

Sources

Sources and further reading

Every figure in this document traces to one of these. Agency documents and peer reviewed work first, reporting only where it is the primary record.

NASA 2020 Technology Taxonomy, TX10 Autonomous SystemsThe autonomy definitions used in section 03
NASA ISRU Pilot Excavator, technology readiness level 5 design overview, 2024The 10 tonne, 11 day autonomous excavation target
Jet Propulsion Laboratory, autonomous systems help Perseverance do more science on MarsThe single drive and no review distance records
Autonomous robotics is driving Perseverance rover's progress on Mars, Science Robotics 2023Peer reviewed account of AutoNav
Low latency lunar surface telerobotics from Earth Moon libration points, NASA 2011The 0.3 to 0.4 s real time threshold and the 2.6 s farside case
Effect of time delay on lunar sampling teleoperations, Applied Ergonomics 2023Performance at 0, 0.5 and 3 s delay
Apollo Lunar Roving Vehicle, NASA Space Science Data Coordinated ArchiveThe crewed baseline this architecture is measured against
RTCA DO 178C, software considerations in airborne systems and equipment certificationThe separation of authority described in section 02
First measurements of the radiation dose on the lunar surface, Science Advances 2020The Chang'e 4 dosimeter behind the EVA dose in section 04