Flying to an asteroid and landing on it without mistakes presents significant challenges for space missions.
While NASA missions incorporate redundancy and automation, much of the operational work still relies on large teams of flight controllers communicating with spacecraft from Earth. For instance, the OSIRIS-REx mission, which encountered an asteroid in 2018, required 100 operators per eight-hour shift.
AstroForge's Autonomous Vision
Such extensive resources are often unavailable to startups. AstroForge, a company focused on developing technology for asteroid mining, is addressing this by integrating artificial intelligence. The company has created an autonomous control stack for its spacecraft, dubbed “Solo,” which is an in-house developed transformer-based model.
AstroForge intends to launch its first autonomous spacecraft in 2027, utilizing the inaugural rocket from Stoke Space. This mission, supported by NASA, is designed to collect scientific data about the sun.
This represents a notable advancement, as most spacecraft autonomy currently relies on traditional control algorithms due to concerns regarding the reliability of neural networks. The first instance of a neural network controlling a satellite’s orbital positioning occurred only last year.
Learning from Past Missions
Founded in 2022 and backed by $56 million in venture funding, AstroForge has previously launched two prototype spacecraft. Both experienced anomalies that hindered them from achieving most mission objectives. In 2025, their Odin spacecraft was sent into deep space, but the company encountered significant communication difficulties. The limited number of large antennas on Earth capable of transmitting to spacecraft hundreds of thousands of miles away, combined with narrow communication windows, posed substantial challenges.
Ultimately, AstroForge was unable to regain control of Odin. This experience prompted the company to explore alternative solutions: could a spacecraft be equipped with sufficient onboard intelligence to resolve its own problems?
Matthew Gialich, AstroForge’s co-founder and CEO, reflected on the incident, stating, “Would that have been recoverable with all the data on the spacecraft? I don’t know, but I can tell you nothing onboard tried it, and I would love something onboard to try if the spacecraft is unrecoverable at launch.”
Gialich further elaborated on the strategic decision, asking, “The trade for me is: Do I go build my own ground network, which is going to cost [around] $200 million to put up five dishes around the world and then do operations on it, or do I try to remove it with a model?”
The Solo AI Architecture
Armand Awad, AstroForge’s head of flight software, explained that the company chose to capitalize on recent advancements in transformer models. Their control stack integrates traditional control algorithms with models trained on test data for specific subsystems, such as power generation or navigation. An overarching intelligence layer, trained on data from approximately 2,500 spacecraft sensors, completes the system.
Gialich clarified the scope of their ambition: “I’m not saying I’m going to make general spacecraft autonomy or general autonomy for the world. I’m making a constrained autonomy at a very low sensor input, following the basic training of a transformer model.”
Theoretically, the AI agent controlling the spacecraft will handle tasks like anomaly resolution. Awad envisioned scenarios where the AI could detect a loss of spatial position, correlate it with a power anomaly affecting its star tracker, and then autonomously rectify the issue, potentially by restarting the component.
Testing and Future Deployment
AstroForge’s third vehicle, DeepSpace-2, is slated for launch by the end of 2026, alongside Intuitive Machines’ third moon mission. Solo will be onboard this vehicle, operating in a “shadow mode” to allow AstroForge engineers to rigorously test its capabilities before the full Autonomy-1 mission.
Regarding the commitment to a truly independent AI agent, Gialich stated, “I don’t plan on flying radios that can receive from Earth on Autonomy-1. We have to go all in right now. The team’s probably going to talk me into it by the time we fly it. But right now, I’m telling them no radios.”
