Boeing trained satellites to translate telemetry into a language understandable by people so that anyone can comprehend it.

New LLM Technology for Autonomous Space Telemetry Analysis

During the early launches of Boeing’s Starliner spacecraft, problems arose due to software code errors and incorrect interpretation of telemetry data. If that data had been communicated to the launch team in a human‑readable language, the outcome might have been different. Therefore the company developed a small language model (LLM) capable of running on low‑power satellite computers and converting raw signals into accessible reports.

Challenges with Commercial Platforms
At the start of the project engineers were unsure whether standard commercial satellite platforms could handle large language models. In space, processor power is less important than radiation tolerance, stable operation without overheating, and minimal power consumption.

Testing and Adaptation
During ground laboratory tests, Boeing Space Mission Systems successfully adapted the LLM to equipment that meets space requirements. The model now analyzes satellite telemetry and generates system status reports in natural language instead of traditional raw data processing.

> “The equipment manufacturer originally claimed it was impossible due to strict constraints, but we found a way to implement the idea,” said AI Lab director Arvel Chapell III.

Benefits
1. Reduced latency – data is processed onboard immediately rather than sent to Earth for subsequent analysis.

2. Increased autonomy – operators can ask the satellite questions and receive understandable answers almost in a dialog mode.

3. Safety – the model is tied to physical system parameters, reducing the risk of “hallucinations” and increasing reliability for critical operations.

Potential for Updating Existing Constellations
Because the technology runs on standard hardware, it can be deployed software‑wise into already launched satellites. This means upgrading most space systems could be completed in months rather than years, as would be required with a full platform replacement.

Thus, the new LLM opens the way to more efficient and autonomous satellite management, improving reliability and reducing maintenance costs.