AstroForge is putting AI in command of its next spacecraft
AstroForge is training a transformer AI to run its next asteroidâmining spacecraft without Earth control.
Why Now
The company plans its first fully autonomous launch in 2027, backed by NASA, after earlier prototype failures.
What Happened
AstroForge built a transformerâbased stack called Solo, combining traditional control, subsystem models, and a 2,500âsensor intelligence layer. Two prototypes, Odin (2025) and DeepSpaceâ2 (2026), suffered anomalies and communication issues, prompting the shift to onboard AI.
Why It Matters
If successful, it could reduce reliance on costly ground networks and enable smaller firms to operate deepâspace missions. It also tests neuralânetwork reliability in critical aerospace contexts.
The Limitation
The approach is unproven; early missions faced failures, and the AIâs performance in realâworld anomalies remains unvalidated.
What You Can Do
Follow AstroForgeâs launch updates and study their Solo architecture for insights into constrained autonomy.
Source
Read original sourceWhy we picked this
New transformer-based AI model on a space probe is core AI content.