A box of loose parts rode to orbit in March. Once in orbit, a robot used some of those parts to build a new section of a lightweight frame. Astronauts never handled the pieces. The robot retrieved them from dispensers and joined them aboard a spacecraft circling Earth.

The team at Caltech calls the experiment Autonomous In-Space Tensioned Reflector Assembly, or ASTRA. The Defense Advanced Research Projects Agency, which funded the work, describes it as the first automated assembly of a structure from individual components in space. ASTRA carried two sections made on Earth and built a third in orbit. A complete antenna would require 11 sections.

The distinction matters for designs that need to grow beyond the room available inside a rocket. The Webb telescope folded its mirror for launch and unfolded it in space. Orbital assembly gives engineers a different option. They can pack individual pieces together for launch and build a frame where it will operate.

Sergio Pellegrino, an aerospace and civil engineering professor, led the team. Caltech designed the demonstration around a reflector 1.4 meters across, a modest size for testing a method that might eventually support much larger space telescopes.

ASTRA flew from Vandenberg Space Force Base on March 30, 2026, aboard SpaceX’s Transporter-16 mission. Momentus carried the roughly cubic-meter payload aboard its Vigoride 7 spacecraft. The host supplied power, communications and imaging. Those services let engineers focus the experiment on building the frame, much as other orbital experiments rely on an existing platform.

Caltech developed and tested the design on Earth before moving to flight hardware. The team completed the payload, tested it and delivered it to Momentus in less than a year. That schedule covered the box itself and its ability to work with the host spacecraft.

The payload held composite fiber tubes, aluminum joints and two prebuilt bays. A bay is one repeatable section of the truss, the framework that would support the reflector around its edge. Diagonal braces help each section hold its shape. By repeating the same geometry, the builder can use the same parts and movements as it works around the perimeter.

After the launch restraints released, the machine pushed the two prebuilt bays outward. Its manipulator could move along three straight axes, reaching into motorized dispensers and positioning pieces on an assembly plate. Precise alignment was the essential task. As with other robotic grippers, success depended on how the hardware handled physical contact. In orbit, ASTRA joined the parts for the third bay.

The robot carried out the assembly movements on board. Charles Sommer, Alan Truong and Tara Venkatadri sent commands through Vigoride and checked the results in downloaded photographs. Photos gave the team a way to see how far a sequence had progressed before deciding what to try next. The approach illustrates how spacecraft operations depend on coordination between flight hardware and people on the ground.

The team typically had seven or eight opportunities a day to communicate with the spacecraft, each lasting about 15 minutes. A problem confirmed in flight could be hard to diagnose within a single window. Caltech kept a flight spare in the lab so engineers could reproduce the behavior and test a workaround before sending instructions back up.

Those short contacts shaped the pace of the build. An engineer could identify a problem in a photo, try to reproduce it on the ground and then wait for the next opportunity to send a fix. The team repeated operations as it worked through surprises.

Former Caltech postdoctoral scholar Jaden Suh, now at the University of Texas at Austin, helped develop the system. The team had already tested assembly in the lab, but the flight spare remained valuable throughout the mission. Ground replicas give engineers a place to check how parts interact without using scarce time in orbit. That testing has a parallel in other experiments that prepare hardware for orbiting spacecraft.

The planned reflector pairs the truss with a perforated film made of Kapton. Caltech and collaborators led by Igor Bargatin at the University of Pennsylvania co-invented the reflective material. The design uses springs to hold two netlike surfaces under tension. The film needs less initial tension than conventional knitted metal mesh, which can reduce the load on the frame and allow a lighter structure.

That matters because a reflector must maintain a useful shape after it deploys. A larger aperture can collect more signal, but its performance also depends on the accuracy of its surface. That requirement varies with wavelength. Future optical mirrors would need their own precision designs, while the ASTRA experiment was conceived around a lightweight antenna reflector. The performance of radio telescopes depends on both the collecting structure and how its signals are measured.

The plan called for the machine to attach the film as it built and deployed the bays. Caltech expects time to run short before the December mission end, leaving the full 11-bay reflector unfinished. The immediate engineering result is that the robot could build a new bay in orbit. Larger reflector designs will require more work on repeated assembly, surface shape and how the finished structure behaves.

A rocket’s fairing, the protective shell around its payload, sets a practical limit on how much preassembled hardware can fit. Folding can stretch that limit, as Webb’s mirror demonstrated. Crews can also join large modules in orbit. ASTRA explores a third approach in which a machine builds a structure from smaller components packed for transport.

DARPA’s Novel Orbital and Moon Manufacturing, Materials, and Mass-efficient Design program supported the work. Its longer-term goals include structures hundreds of meters or more across. Potential applications include radio antennas, frames for space habitats and arrays that collect solar power.

Each application places different demands on stiffness, shape and assembly speed. The same program also plans a separate experiment led by the University of Illinois Urbana-Champaign to test curing composite material aboard the International Space Station.

Building from smaller parts could also change how a spacecraft is packed. Long members and joints can be stored close together, leaving the robot to create the empty space between them after launch. An antenna needs that span to collect signals, but a rocket gains little from carrying the span already assembled.

For ASTRA’s engineers, the next job is to simplify the builder and make each step repeatable. A machine that adds one bay has shown that the parts can be handled and joined in orbit. A machine that can reliably add dozens would let designers begin with the structure a mission needs and work backward to the parts a rocket can carry.