GAIN-AI: Context-Aware AI Assistant for Lunar EVA Guidance
A novel AI technology named GAIN-AI (Guided Assistant for Intelligent Navigation) has been created to support astronauts during extravehicular activities (EVAs) on the Moon. This system, outlined in a paper available on arXiv (2608.13589), tackles the challenge of delivering prompt procedural guidance amid intricate tasks. It leverages a large language model integrated with structured context, which encompasses EVA procedure documents, real-time telemetry, and error-handling protocols formatted as JSON. The information is presented in three concise augmented reality display units: Goal, Task, and Verification. In assessments involving 111 synthetic EVA scenarios, GAIN-AI achieved a perfect score of 10.0/10 under normal conditions and 8.15/10 in single-fault situations, aiming to reduce distractions by highlighting only the most pertinent information when needed.
Key facts
- GAIN-AI is a context-aware AI assistant for lunar EVA procedural guidance.
- It uses a large language model grounded with structured context.
- The system operates in two layers: grounding and output restructuring.
- Output is formatted into Goal, Task, and Verification units for AR display.
- Evaluated on 111 synthetic EVA scenarios.
- Scored 10.0/10 on nominal conditions.
- Scored 8.15/10 on single-fault scenarios.
- Paper available on arXiv with ID 2608.13589.
Entities
Institutions
- arXiv