A Multi-Agent LLM System for Natural Language Querying of Operational Knowledge Graphs in Satellite Ground Stations

Operational data discoverability: Satellite ground-station maintenance produces large volumes of operational data that are difficult to discover through traditional query interfaces.

Time-critical maintenance decision latency: Traditional query interfaces slow decision making in time-critical ground-station maintenance contexts.

Operator troubleshooting support: Operators need better support for troubleshooting using historical maintenance information.

Operational data discoverability: Satellite ground-station maintenance produces large volumes of operational data that are difficult to discover through traditional query interfaces.

Time-critical maintenance decision latency: Traditional query interfaces slow decision making in time-critical ground-station maintenance contexts.

Multi-agent LLM KG RAG system: The paper proposes a deployed multi-agent system combining Large Language Models, knowledge graphs, and Retrieval-Augmented Generation for conversational operational querying.

Specialized agent workflow: Specialized agents collaborate to perform intent mapping, multi-hop reasoning, and explainable synthesis.

Domain operational knowledge graph: The system includes a domain knowledge graph that operationalizes antenna-system context for satellite ground-station maintenance.

Operator-in-the-loop integration: The work provides integration lessons for incorporating operators into the system loop.

Multi-agent LLM KG RAG system: The paper proposes a deployed multi-agent system combining Large Language Models, knowledge graphs, and Retrieval-Augmented Generation for conversational operational querying.

Deployment at Fucino Space Centre:

Transparency and reliability enhancement:

Conversational retrieval architecture contribution: