How efontesp90/llsim5g works
Project Type and Purpose
• LLSim5G is an open-source, Python-based link-level simulator for heterogeneous 5G networks (TNs, NTNs, UAVs).
• It is not a full-stack web app but a modular simulation tool that reads configuration, generates scenarios, applies channel models, computes link metrics, and writes outputs.
Main Architectural Components
A. Configuration & Initialization
• SimulationConfiguration (PDF + Excel under simulation_settings/)
• main.py (entry point)
B. Scenario Generator
• scenario/ (defines BSs, EDs, mobility models, topology)
C. Channel Modeling Layer
• channel_models/geometry/ (3D geometry)
• channel_models/path_loss_models_* (TR 38.901, 38.811)
• channel_models/ff_models_* (fast-fading)
• channel_models/spectroscopic_data, atmospheric absorption
D. Link Computation Engine
• link_computation/link_computation.py
• Uses scenario definitions + channel model outputs to compute per-link SINR, etc.
E. Link-to-System Adaptation
• link_to_system_adaptation/ (bler_curves.py, link_to_system.py)
• Converts SINR → BLER, CQI metrics
F. Output & Persistence
• output/ (writes Excel, pickle, videos under output/metrics and output/scenario)
G. Validation & Testing
• validation/ (scripts to validate atmospheric absorption, antenna models, HDF5 realizations)
H. Environment & Dependencies
• requirements.txt, myenv/ (Python venv)
• External libs: numpy, pandas, h5py, openpyxl, etc.
Data Flows & Interactions
• main.py loads SimulationConfiguration → passes parameters to the Scenario Generator and Channel Modeling Layer.
• Scenario Generator defines nodes (BS, ED) positions + mobility profiles → feeds Link Computation.
• Channel Modeling Layer processes geometry & path-loss → provides channel coefficients to Link Computation.
• Link Computation Engine iterates over simulation time, invoking channel models per link, producing SINR results.
• Link‐to‐System Adaptation consumes SINR values → yields BLER & CQI.
• Output Layer aggregates and writes metrics & scenario traces.
• Validation modules can be invoked independently to verify specific model fidelity.
Architectural Patterns & Principles
• Layered pipeline: initialization → scenario → channel models → computation → adaptation → output.
• Modular plug-in structure: each channel model (path-loss, fast-fading) exists as a separate module.
• Separation of concerns: scenario generation, physics modeling, computation, and output are in distinct packages.