A robot has built part of a structure in orbit without an astronaut holding the wrench. Caltech’s Autonomous In-Space Tensioned Reflector Assembly, or ASTRA, used a compact robotic builder to assemble a new section of a lightweight truss hundreds of miles above Earth. DARPA and Caltech describe it as the first automated construction of a structure in space.
The important detail is what ASTRA actually finished. The payload launched with two truss sections already assembled. In orbit, its robot picked up separate tubes and joints and constructed a third section, while also working with a thin reflective material. The full 11-section, 1.4-meter reflector antenna was not completed. That limitation does not erase the first, but it keeps the achievement in the realm of a hard-won proof of concept rather than a finished orbital factory.
What the robot built
ASTRA launched on March 30, 2026, aboard SpaceX’s Transporter-16 mission as a payload hosted by Momentus on its Vigoride 7 orbital vehicle. The entire experiment occupied roughly one cubic meter. Inside were two prebuilt truss bays, an arm-like manipulator, dispensers holding composite-fiber tubes and aluminum joints, and the parts intended for the remaining bays.
After launch restraints were released, the two completed bays were pushed outward. The robot then assembled the third bay on a work plate. It retrieved individual components, positioned them, and joined them into the repeating triangular frame. The planned system also attaches a perforated reflective film to form two cable-net surfaces held against each other by springs.
That is a much more specific accomplishment than saying a robot “built an antenna.” It did not finish the antenna, install a working radio-frequency system, or construct something kilometers wide. It demonstrated that a box of parts can become a larger structural section after reaching orbit.
Autonomous did not mean unsupervised
The word “autonomous” needs some daylight around it. Engineers on Earth sent commands to ASTRA and examined photographs of the results. The machine then carried out its assembly operations without astronauts physically manipulating the components. When something behaved unexpectedly, the team used a flight-spare copy in Sergio Pellegrino’s Caltech laboratory to reproduce the problem and test a workaround.
That troubleshooting happened through a narrow communications keyhole. Caltech says the team had only seven or eight contact windows per day, each lasting about 15 minutes. The engineers had to diagnose a problem from downloaded images, recreate it on the laboratory twin, choose a fix, and wait for another opportunity to send instructions.
Space was doing what space does best: turning a straightforward mechanical sequence into a boss battle.
The unfinished antenna matters
The complete ASTRA design called for 11 bays around a 1.4-meter reflector. Technical problems and limited operating time prevented the payload from completing that structure, and Caltech said it appeared unlikely to do so before the mission ends in December.
That shortfall is part of the result, not an embarrassing footnote. A machine that works in a laboratory can face different temperatures, lighting, friction, material behavior, communications delays, and launch-related changes in orbit. The test exposed places where the hardware and procedures need to be simpler. It also showed that the basic pick-position-connect-deploy sequence could work beyond Earth.
DARPA compares the concept to building with K’Nex or Tinkertoys. The analogy is useful because the proposed advantage comes from repetition. A rocket would carry compact stocks of standard parts instead of a complete structure folded to fit inside a payload fairing. A robot could join those pieces after launch, one bay at a time.
Why build anything after launch?
Rocket fairings impose a stubborn ceiling on spacecraft design. Engineers can fold an antenna or telescope mirror and deploy it later, or astronauts can assemble large pieces, but both options become increasingly complicated as structures grow. In-space construction offers a third path.
DARPA’s long-range examples include giant radio antennas, solar arrays, depots, habitats, and optical systems. Caltech points to astronomy, communications, and space-based solar power. Those possibilities remain future applications, not capabilities proved by this experiment. ASTRA built one additional bay, not a kilometer-scale observatory.
Still, the visual is unusually concrete: a small machine launched with parts, reached orbit, and made itself larger. That is the payoff. The experiment moved automated space construction from laboratory hardware and animation to an imperfect physical build in the environment where the idea is meant to work.
Watch the official demonstration
The following Caltech video includes on-orbit imagery and explanatory material from the ASTRA demonstration.
Sources
Primary reporting and technical details: Caltech’s ASTRA mission report and DARPA’s NOM4D announcement. The official Caltech video source is available on YouTube.
Featured image: The ASTRA truss builder after constructing its third bay in orbit. Credit: Caltech/Momentus, via Caltech News. Used as an official Caltech news asset for journalistic reporting under Caltech’s media-use terms.
