Space Mining Startup AstroForge Bets on AI for Its Next Deep-Space Mission

AstroForge has announced that its upcoming Autonomy-1 spacecraft will use a compact transformer-based AI model to help command the mission, giving the space mining startup a more autonomous approach to operating far from Earth.

The company is turning to AI because deep-space missions cannot rely on immediate instructions from ground teams. At large distances, communication delays make real-time control impractical, so spacecraft need to handle more decisions on their own.

Autonomy-1 is designed to move beyond the fixed command sequences and rule-based systems commonly used in spacecraft operations. AstroForge plans to use its onboard model to interpret sensor information and assist with navigation and other mission decisions without constantly waiting for direction from Earth.

The AI architecture is based on transformers, the same general type of technology used in systems such as ChatGPT. AstroForge, however, is working with a much smaller model because spacecraft face tight limits on computing power and energy.

That creates a difficult engineering problem. The onboard system must perform reliably with constrained hardware while also operating in the radiation-heavy environment of space. Any AI responsible for spacecraft decisions also needs to remain dependable when unexpected conditions arise.

The company’s approach could give Autonomy-1 more flexibility than a spacecraft that follows only predetermined instructions. Rather than depending entirely on commands uploaded before or during the mission, the spacecraft would have greater ability to respond to what its sensors detect.

AstroForge is developing the technology around its broader asteroid-mining ambitions. Autonomous decision-making could become particularly important for missions that operate too far from Earth for continuous human involvement.

NASA and other space organizations have previously experimented with AI in space, but AstroForge is placing the technology closer to the center of spacecraft operations. The company’s plan puts an AI model in a more direct command role rather than limiting it to supporting tasks.

The mission will also test whether transformer models can be compressed enough to work within the practical limits of spacecraft hardware. Models of this type typically require substantial computing resources, which creates a very different challenge when they must run onboard a spacecraft with strict power and processing constraints.

If the system performs as intended, Autonomy-1 could demonstrate a model for more independent deep-space missions. Spacecraft could handle a larger share of navigation and operational decisions locally instead of depending on constant instructions from Earth.

For AstroForge, the immediate test is more specific: proving that a compact transformer model can operate reliably enough to take on meaningful spacecraft responsibilities during a deep-space mission.

Autonomy-1 will therefore serve as both a spacecraft mission and a test of how much operational control can realistically be delegated to onboard AI when direct human oversight is limited.

This analysis is based on reporting from the tech buzz.

Image courtesy of AstroForge.

This article was generated with AI assistance and reviewed for accuracy and quality.