AstroForge Builds Solo AI to Troubleshoot Spacecraft Without Earth Commands

Solo is meant to diagnose faults onboard after AstroForge’s past communication troubles, but it has not yet operated in space.

By 2 min read
AstroForge Builds Solo AI to Troubleshoot Spacecraft Without Earth Commands
AstroForge Builds Solo AI to Troubleshoot Spacecraft Without Earth Commands

Listen to this story

The audio brief

About 1:36
0:001:36
Read transcript
AstroForge is preparing Solo, an onboard system meant to diagnose spacecraft problems without waiting for commands from Earth. The company plans to test it first in shadow mode on DeepSpace-2, before giving it a larger role on a planned autonomous mission in 2027. The motivation is clear: AstroForge’s Odin spacecraft launched in 2025, but communications problems meant the company could not regain control. Both of its prototype spacecraft encountered anomalies that prevented them from achieving most of their mission objectives. Solo is designed for narrower situations, not as a general-purpose spacecraft pilot. It combines conventional flight controls with models built for individual spacecraft systems, plus a transformer-based intelligence layer trained on data from roughly 2,500 sensors. AstroForge says it could, for example, connect a power problem with star-tracker trouble after a spacecraft loses its position, then switch off the affected system. That remains a proposed capability, not one demonstrated in space. DeepSpace-2 is expected to launch by the end of 2026 alongside Intuitive Machines’ third lunar mission. Solo will first observe operations without controlling them. The planned 2027 Autonomy-1 mission is tied to Stock Space’s first rocket and has NASA backing. CEO Matthew Gialich currently expects the spacecraft to carry no radio able to receive Earth commands, although that choice could change. The key constraint is simple: Solo’s hardest test has not happened yet.

Story brief

3 key points

AstroForge is preparing Solo, a constrained onboard autonomy system designed to diagnose and respond to spacecraft anomalies without immediate Earth commands. The software will first run in shadow mode on DeepSpace-2, targeted for launch by the end of 2026, before a planned 2027 Autonomy-1 mission. Solo combines traditional flight controls with system-specific models and data from roughly 2,500 sensors. Its value is...

  1. 01

    Solo combines conventional controls, system-specific models, and an intelligence layer trained on data from about 2,500 spacecraft sensors.

  2. 02

    DeepSpace-2 will test Solo in shadow mode before AstroForge considers handing it operational responsibility.

  3. 03

    AstroForge says both prototype spacecraft encountered anomalies; Odins 2025 communications failure prevented regained control.

AstroForge is pursuing a spacecraft that can respond to problems without waiting for Earth. It has built Solo, an onboard control system it plans to test on DeepSpace-2 before a planned autonomous flight in 2027.

The failure Solo is meant to address

AstroForge’s Odin launched into deep space in 2025, but communication difficulties meant the company could not regain control. Both of AstroForge’s prototype spacecraft have suffered anomalies that prevented them from achieving most mission objectives.

A narrow system for specific failures

AstroForge describes Solo as constrained autonomy, not a general-purpose spacecraft pilot. The stack combines conventional control software, models trained on test data for individual systems, and an intelligence layer trained on data from about 2,500 spacecraft sensors.

The intended job is anomaly response. AstroForge says Solo could link a power problem to star-tracker trouble after a craft loses its position, then toggle the affected system. That proposed capability remains unproven in orbit.

Traditional control algorithms remain common in spacecraft autonomy because neural-network reliability is a concern. AstroForge is adding its model layer to conventional controls rather than claiming it has solved general spacecraft autonomy.

The software around the experiment

DeepSpace-2’s baseline software is designed to handle operations onboard rather than through continuous direct commanding after launch.

Two planned operating modes

  • Safe Mode is among the modes AstroForge describes for DeepSpace-2.
  • Minimum Power Mode is another named mode in the planned architecture.

A staged path to a harder handoff

DeepSpace-2 is expected to launch alongside Intuitive Machines’ third lunar mission by the end of 2026. AstroForge says Solo will run in shadow mode so its engineers can put the system through its paces before the Autonomy-1 mission.

The 2027 mission is planned for Stock Space’s first rocket, with NASA backing and an aim of gathering scientific data about the Sun. Chief executive Matthew Gialich said he does not currently plan to include radios able to receive commands from Earth, though that decision could change before launch.

Sources

  1. astroforge.comFlight Software at AstroForge - AstroForge
  2. techcrunch.comAstroForge is putting AI in command of its next spacecraft | TechCrunch

Loading discussion...

YOUR READING SPACE

Notifications