AstroForge is putting AI in command of its next spacecraft
Original reporting by TechCrunch
AstroForge is a space startup leveraging artificial intelligence to achieve unprecedented autonomy in spacecraft operations, specifically for asteroid mining. Traditionally, complex space missions, such as NASA's OSIRIS-REx rendezvous with an asteroid, demand vast teams of human flight controllers—with a hundred operators per eight-hour shift communicating constantly from Earth. This labor-intensive model, while proven, presents a significant barrier for agile ventures like AstroForge, which lack NASA-scale resources. Moreover, reliance on Earth-based communications introduces vulnerabilities, as AstroForge experienced when it struggled to gain control of its Odin prototype due to limited ground antenna access and narrow communication windows.
A New Approach
These challenges spurred AstroForge to fundamentally rethink spacecraft command. The company has developed "Solo," an in-house transformer-based AI model designed to manage spacecraft operations autonomously. This represents a radical departure from the industry norm, where spacecraft autonomy primarily relies on traditional control algorithms, with neural networks only recently seeing limited deployment for satellite positioning. Solo aims to enable a future where spacecraft can diagnose and resolve their own anomalies, potentially operating without constant human oversight or even direct radio communication with Earth. AstroForge plans to test Solo in "shadow mode" on its DeepSpace-2 mission by late 2026, paving the way for its fully autonomous "Autonomy-1" mission in 2027. This bold move seeks to redefine the economics and capabilities of deep-space exploration.
AstroForge’s audacious reliance on AI for full spacecraft autonomy, culminating in a potential mission devoid of Earth-receiving radios, marks a decisive pivot in space operations. Driven by the practical limitations faced by startups, this approach directly challenges the long-standing model of human-intensive ground control and robust communication links, as exemplified by NASA’s complex missions. The upcoming "shadow mode" flight of Solo on DeepSpace-2 and its intended full deployment on Autonomy-1 represent a high-stakes validation of whether transformer-based intelligence can indeed independently navigate and resolve anomalies in the unforgiving vacuum of space.
A New Frontier
Should AstroForge succeed, the ripple effects across the space industry would be profound. This shift towards onboard, self-governing intelligence could democratize access to space, dramatically reducing operational costs and making ambitious missions viable for a far wider array of entities beyond traditional state agencies. Crucially, it unlocks deeper, more distant exploration by mitigating the challenges of light-speed communication delays and limited ground station availability. The promise of spacecraft capable of diagnosing and resolving their own issues, as envisioned by AstroForge, suggests a future of more resilient, self-sufficient missions. Ultimately, this pioneering work, even if constrained to specific tasks, paves the way for a paradigm where AI is not just an assistant, but the central architect of mission survival and success, redefining humanity's reach into the cosmos.
Frequently asked questions
- Why is AI becoming important for space mission control and spacecraft operations?
- AI is becoming crucial for space missions to overcome challenges like limited human resources and vast communication delays. Startups, in particular, lack the extensive ground teams that large agencies employ. Implementing AI enables spacecraft to operate more independently, resolve unexpected issues onboard, and function reliably far from Earth, reducing reliance on constant human intervention and costly ground infrastructure.
- How does AstroForge's "Solo" AI system control autonomous spacecraft operations?
- AstroForge's "Solo" is an in-house developed, transformer-based AI control stack. It integrates traditional control algorithms with specialized models trained on specific spacecraft subsystems, such as power or navigation. An overarching intelligence layer, trained on thousands of sensor inputs, enables the spacecraft to monitor its status, identify anomalies, and potentially resolve problems autonomously without immediate human intervention from Earth.
- When will the first fully autonomous AI-controlled spacecraft be launched into deep space?
- AstroForge plans to launch its "Solo" AI control system in "shadow mode" onboard the DeepSpace-2 mission by late 2026 for testing. The company's first fully autonomous mission, "Autonomy-1," is slated for 2027. This mission aims for significant independence, potentially operating without radios to receive commands from Earth, demonstrating the spacecraft's ability to manage its operations and resolve issues independently in deep space.