AstroForge is putting AI in command of its next spacecraft

AstroForge, a startup developing asteroid mining technology, is turning to artificial intelligence (AI) to command its future spacecraft, "Solo." The company plans to launch its first autonomous spacecraft in 2027, backed by NASA, which will gather scientific data about the sun. While most spacecraft autonomy relies on traditional control algorithms due to concerns about neural network unreliability, AstroForge's AI-powered control stack, named "Solo," aims to overcome these limitations.

The company has developed Solo, a transformer-based model, for its spacecraft, which incorporates traditional control algorithms, models trained on specific subsystems, and an overall intelligence layer trained on about 2,500 sensors within the spacecraft. This approach was inspired by the recent use of neural networks to control a satellite's positioning in orbit.

Despite previous anomalies in its prototype spacecraft, AstroForge has persisted in its efforts. In 2025, its Odin spacecraft was launched into deep space but proved difficult to communicate with, leading AstroForge to consider the possibility of autonomous problem-solving onboard. Matthew Gialich, AstroForge's co-founder and CEO, expressed interest in a model onboard to attempt self-recovery if the spacecraft were to fail during launch.

Armand Awad, AstroForge's head of flight software, believes that this constrained autonomy will allow the spacecraft to handle tasks like anomaly resolution, such as realizing a power anomaly, correlating it with a star tracker issue, and fixing the problem. AstroForge's third vehicle, DeepSpace-2, is currently scheduled to launch alongside Intuitive Machines' third moon mission, expected to head for space by the end of 2026.

Solo will fly in "shadow mode" during this mission, allowing AstroForge's engineers to test the AI agent before the Autonomy-1 mission.