In a significant step toward manufacturing large hardware beyond Earth’s atmosphere, a robotic system developed by Caltech with funding from DARPA has autonomously assembled a structural section in low Earth orbit. The experiment, known as ASTRA, is being positioned as the first real demonstration that repeated, piece-by-piece construction can be carried out in space without astronauts physically present.
- Autonomous assembly: ASTRA used a robotic arm to pick bars and joints from motorized dispensers and connect them into repeated sections called bays.
- Compact payload: The entire system, including the arm, dispensers, assembly plate and onboard computer, fits in roughly one cubic meter.
- Partial success: The robot built a third bay on top of two pre-assembled ones, but the planned 1.4-meter antenna was not completed before the mission’s December close.
How the orbital construction actually worked
The core hardware is deceptively simple. A robotic arm takes composite fiber tubes and aluminum joints from motorized dispensers and places them onto an assembly plate, linking them into repeating structural bays. After each bay is finished, a reflective sheet is attached and the assembly is pushed outward so the sequence can begin again.
The reflective surface is a perforated Kapton film tensioned by two opposing cable networks, a design that echoes research into programmable materials capable of changing shape and self-assembling. Caltech developed the system with the University of Pennsylvania, which contributed to the reflective material.
ASTRA started from two pre-built bays and constructed a third in orbit. According to the team, that achievement demonstrates the feasibility of autonomous construction. The full design called for 11 bays forming a 1.4-meter reflector antenna, with far larger structures of hundreds of meters or even kilometers as the long-term ambition.
Launch, communications and mission constraints
The robot reached space on March 30, 2026, lifting off from the Vandenberg Space Force Base aboard Momentus’s Vigoride 7 orbital vehicle. It flew as one of ten payloads on SpaceX’s Transporter-16 mission.
Operating conditions were far from ideal. Communication with the robot was limited to roughly seven or eight windows per day, each lasting about 15 minutes. That constrained schedule shaped how the team planned and verified each assembly step.
The work falls under phase three of DARPA’s NOM4D program, which focuses on efficient manufacturing and design of materials in orbit and on the Moon. The mission was scheduled to close in December, and the antenna was not finished before that deadline.
The human factor behind the autonomy claim
The headline claim of near-zero human intervention deserves nuance. Ground controllers sent commands, released launch restraints and helped resolve unexpected problems. The team had to adjust instructions repeatedly and test fixes on a duplicate robot on the ground before uplinking corrected orders to orbit.
The robot did not operate with zero human involvement; what it performed autonomously was the physical task of assembly.
That distinction matters. The genuinely new element is not a robot working entirely alone, which it did not, but the fact that the most delicate part of the job, manipulating parts and fitting them together in vacuum, can be executed without anyone physically present. That opens the door to building objects in orbit that are too large to launch pre-assembled, from giant antennas to telescopes.
Caution is warranted, as always with early-stage DARPA research. One bay was assembled, not a complete antenna. The mission approaches its December close without meeting its full objective, and substantial ground control was required throughout. Yet as a demonstration that autonomous assembly in space is possible, the experiment delivers. The leap from a single section to kilometers of structure is enormous, and it has now taken its first real step.