A Caltech robot built part of a structure in orbit, on its own

The robot is called ASTRA, short for Autonomous In-Space Tensioned Reflector Assembly. Caltech built it, and DARPA, the research arm of the US Department of Defense, paid for it.

It was meant to build a reflector antenna, a dish-shaped surface that collects or sends radio signals, 1.4 meters wide. The frame of that dish is a truss. A truss is a light frame made of straight tubes joined at points, like a crane arm or a bridge.

Each repeating piece of the truss is called a bay. ASTRA flew with two bays already built. Then it assembled a third bay by itself, in orbit.

DARPA says this is the first time a structure has been built in space by an automated system, with no astronauts connecting the pieces.

It is simpler than you might think. DARPA describes a relatively simple gantry robot, a machine that slides along three straight lines, like a crane over a workbench.

It picks up composite tubes (strong, light tubes made of fibers) and aluminum connectors from storage dispensers. It joins them into a new bay. Then it pushes the finished bay outward and starts the next one.

DARPA's program manager, Andrew Detor, says the plan is to build things the way you build with toys. "You can think of the Caltech approach as similar to Tinkertoys or K'nex construction toys," he said.

In real life Think of a flat-pack bookshelf. The box is small and easy to ship. You build the shelf where you want it. ASTRA does the same, except the shelf is a space antenna and the living room is orbit.

Today there are two ways to make a big space structure. Both have limits, and DARPA says so plainly.

• Build it on Earth: the whole thing must fold up to fit inside the rocket's nose cone, called the fairing. That caps its size.
• Build it by hand in space: astronauts join big pieces that were sent up over several launches. That is slow and costly.

A third way is to ship a box of parts and let a robot put them together. Caltech's Sergio Pellegrino, who leads the ASTRA demonstration, puts it simply: "Once you get to the size limit where structures cannot be prebuilt on the ground and unfolded, you have to think about how to build them in space."

Why it matters: DARPA wants structures hundreds of meters or even kilometers across. "Nothing approaching that scale can be achieved today," Detor said.

ASTRA launched on March 30, 2026, on SpaceX's Transporter-16 mission from Vandenberg Space Force Base in California. Transporter missions are shared rides, where many small customers split one rocket.

It rode on Momentus's Vigoride 7 vehicle, which carried 10 payloads, meaning 10 separate experiments, into low Earth orbit. That is the zone a few hundred miles above the ground where most satellites fly.

ASTRA filled about one cubic meter. In that small box were the two pre-built bays, a robotic arm and the dispensers loaded with parts for the rest.

Momentus hosted the payload and provided communications, power and imaging during the test. Another SpaceX story from this week shows how busy that company's schedule has become.

Very fast for space hardware. DARPA says the program was not first planned around a flight at all.

When good results came in during the second phase of the program, Detor decided to pivot the last phase to a real space test. "We went from the idea of an on-orbit demo to launching a working prototype into space in 15 months," he said.

The Caltech team had even less time for the hardware itself. They had under 12 months to build, test and deliver it to Momentus. Pellegrino called it "an enormous undertaking for our small team."

• Phases 1 and 2: the team worked out the idea and the design, and tested it in labs on Earth.
• Phase 3: the space test, launched in March 2026.

Space is not forgiving. Some things went wrong, and the team had to solve them from the ground.

The engineers could contact ASTRA only about seven or eight times a day, and each window lasted about 15 minutes. Three people sent the commands: Charles Sommer, Alan Truong and graduate student Tara Venkatadri.

The robot did the assembly automatically. But the team watched photos sent from orbit. When a step failed, they studied the pictures and tried again.

To find fixes, they used a twin. A spare copy of ASTRA sits in Pellegrino's lab at Caltech. "Every time we confirm there is an issue in space, we come to the laboratory and try to mimic the problem on the ground," Sommer said.

The catch: because of these problems and limited time, ASTRA built only one bay, not the whole antenna.

Jaden Suh, a former Caltech postdoctoral scholar who is now a professor at the University of Texas at Austin, said the setbacks are the point of a test like this. "There are a lot of differences in the space environment compared to down on Earth," he said. "There are always lots of little surprises."

The full design is more than a frame. After building a bay, ASTRA is meant to attach a thin film full of tiny holes. It forms two nets that are pulled against each other by springs, which keeps the surface in shape.

Caltech and the University of Pennsylvania invented the film together. Professor Igor Bargatin led the Penn team. The film is lighter than the knitted metal mesh used on many folding antennas. It needs less tension, so the frame holding it can be lighter too.

Light matters because every extra kilogram makes a rocket's job harder. The full ASTRA design has 11 connected bays, each attached to the nets and pushed outward.

The mission ends in December. The team does not expect ASTRA to finish all 11 bays by then.

Even so, Caltech says the test taught the team ways to simplify the robot and the building steps. The team believes it has enough evidence that the method works.

DARPA stresses that this is not the end. "The NOM4D Caltech demonstration is not an endpoint; it's a beginning," Detor said.

There is one more NOM4D milestone ahead. Later this year, a team from the University of Illinois Urbana-Champaign will send a project to the International Space Station. It will not assemble anything. It will test a special resin, a liquid plastic that hardens, to make a strong composite part in space with little energy.

Detor lists some ideas: large space depots or habitats, big solar arrays and huge radio antennas. Each needs size, which a rocket's fairing does not allow.

Caltech adds big telescopes, which gather more light and can see farther. Pellegrino also co-directs Caltech's Space Solar Power Project, which wants to collect sunlight in space and send the power to Earth without wires.

Building in orbit has a history. The robotic arm Canadarm2 helped assemble the International Space Station, but it moved large parts that were already made on Earth. In 2014, a 3D printer on the station printed a part for itself. NASA's Archinaut One planned to make beams and a solar array in orbit, but the project ended in 2023.

ASTRA is different because it builds a bigger object out of a handful of small pieces.

• No new gadget yet: this is a proof of concept. You will not use anything from it this year.
• Bigger tools later: larger telescopes and antennas could one day see farther and carry more data.
• Rockets matter less: if robots assemble parts in orbit, the size of the rocket stops being the main limit.
• Watch the next tests: the space station composite experiment is the next step in the program.