This repository contains an autonomous framework designed to bridge the gap between unstructured Natural Language (NL) mission requests and automated planning environments.
The pipeline ingests high-level operational intent, extracts relevant entity requirements, grounds the parameters using deterministic orbital visibility constraints, and structuralizes the output into a verifiable Intermediate Representation (IR). This IR simplifies the synthesis of planner-ready files, such as PDDL (Planning Domain Definition Language) domains and problems.
The project is structured into two independent, specialized environments to decouple the data processing infrastructure from the resulting research assets:
ConstellationContextDatasetGeneration/
│
├── README.md <- This main project roadmap (You are here).
├── .gitignore <- Strict environment and runtime cache exclusions.
│
├── dataset_generator/ <- Core Pipeline Infrastructure & LLM Orfchestration.
│ ├── README.md <- Local environment setup, dependencies, and execution guides.
│ └── ... (Scripts, LLM agent boundaries, and Docker deployment assets)
│
└── dataset/ <- Research Dataset Assets & Visualizations.
├── README.md <- Dataset characterization, schemas, and metrics.
└── ... (JSON index summaries, raw assets, and generated charts)