AstroForge is developing the Solo transformer model for autonomous spacecraft control
AstroForge is developing the autonomous Solo system, a transformer model for controlling a spacecraft without the need for ground control. It grew out of communication problems with Odin in 2025, and the company plans the first flight for 2027 with support from NASA.
AstroForge, a company developing technology for asteroid mining, is building an autonomous spacecraft control system called Solo. It is an internally developed transformer model that combines traditional control algorithms with models trained on data from approximately 2500 spacecraft sensors for specific subsystems such as power or navigation. According to Armand Awad, head of flight software at AstroForge, the system should be able to handle tasks such as identifying and resolving anomalies—for example, linking a power outage to a problem with a star tracker and fixing the fault without intervention from Earth.
The project grew out of the experience with Odin, a spacecraft that AstroForge launched into space in 2025 but could not establish sufficient communication with because of the limited number of ground antennas capable of long-distance transmission and the short communication windows. According to CEO Matthew Gialich, the company faced a choice between building its own ground antenna network (estimated at around 200 million dollars for five antennas around the world) and trying to replace that need with a model onboard the spacecraft.
AstroForge plans its first autonomous flight, a mission named Autonomy-1, for 2027, on the first rocket from Stoke Space. According to the source, NASA will support the mission, which is intended to collect scientific data about the Sun. Gialich said he does not plan to have radios capable of receiving signals from Earth onboard for this mission. Before that, however, the Solo system will fly in “shadow mode” on the third spacecraft from AstroForge, DeepSpace-2, which is set to be part of the third mission to the Moon by Intuitive Machines, planned for the end of 2026—where engineers from the company will be able to test it without direct control over the flight.
AstroForge was founded in 2022 and has raised 56 million dollars in venture capital. Both of the test spacecraft flown by the company so far experienced anomalies during flight that prevented them from meeting most of their mission objectives. The vast majority of spacecraft autonomy still relies on traditional control algorithms because of concerns about the unreliability of neural networks; the first use of a neural network to control the attitude of a satellite in orbit took place only last year.
Why it matters
It provides a concrete example of deploying transformer models beyond text and agent applications—in an environment where a communication failure means the loss of an entire mission and where traditional control algorithms have so far predominated because of concerns about the unreliability of neural networks. For the spaceflight and robotics sectors, it is a test of whether AI can replace some of the costly ground infrastructure and human oversight involved in resolving onboard anomalies.
Relevant practical impact
What this means
For a business
This case shows that transformer models are beginning to make their way into the control of safety-critical systems beyond typical chat and agent deployments, providing relevant context for companies considering deploying AI where failures are costly and human oversight is limited.
Development More business impacts →Check the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.