Calibrating Wireless Ray Tracing using Local Phase Error Estimates

Code repository for paper "Calibrating Wireless Ray Tracing for Digital Twinning using Local Phase Error Estimates"

If you use this software, please cite it as

@article{ruah2023calibrating,
  title={Calibrating Wireless Ray Tracing for Digital Twinning using Local Phase Error Estimates},
  author={Ruah, Clement and Simeone, Osvaldo and Hoydis, Jakob and Al-Hashimi, Bashir},
  journal={arXiv preprint arXiv:2312.12625},
  year={2023},
  online={https://arxiv.org/abs/2312.12625}
}

Project structure

• src/: main codebase. Implements synthetic channel observations and material parameters calibration using any of the presented schemes, which comprise:
  • the proposed Phase Error-Aware Calibration (PEAC) scheme;
  • the Phase Error-Oblivious Calibration (PEOC) baseline;
  • and the Uniform Phase Error Calibration (UPEC) baseline.
• study/: codebase containing the experiments presented in the paper.
• blender/: source files of the 3D models used in the experiments.
• assets/ : Mitsuba exports of the Blender source files.
• logs/: experimental data is stored in this folder by default. Can be set to a custom folder by setting the environment variable LOGS_FOLDER in .env.
• scripts/: bash utils and scripts to launch the experiments.

Setup environment

Pre-requisites

• Download and install miniconda
• CPU-based: download and install LLVM
• GPU-based: follow TensorFlow GPU support tutorial for the required drivers.
  Note: for Linux-based systems, an installation script is provided in ./scripts/create-gpu-env-linux.sh.

Create conda environment

• Install conda environment by running the command:
  conda env create -f ./environment.yaml
• Source the environment by running:
  conda activate phase-aware-rt-calibration

Update python environment

conda env update -f ./environment.yaml

Install notebook kernel

Add the installed conda environment to Jupyter by setting a notebook kernel.
Inside the phase-aware-rt-calibration environment, run:
python -m ipykernel install --user --name phase-aware-rt-calibration

Run Calibration and Coverage Map Experiments

By default, experimental data is stored in ./logs and plots are stored in ./logs/saved_plots/.

Urban scenario "The Strand Campus"

Run calibration

bash scripts/the_strand/run_calibration_phase_noise_concentration.sh
bash scripts/the_strand/run_calibration_phase_noise_snr.sh
bash scripts/the_strand/run_calibration_pos_noise.sh

Calibration Plots

bash scripts/the_strand/run_plots.sh

Compute Coverage Maps

bash scripts/the_strand/run_coverage_maps.sh

Coverage Map Plots

bash scripts/the_strand/run_coverage_maps_plots.sh

Toy Example

Run calibration

bash scripts/toy_example/run_calibration.sh

Calibration Plots

bash scripts/toy_example/run_plots.sh

Experiments with FDTD-generated data

Run calibration

bash scripts/toy_example_maxwell/run_calibration.sh

Calibration Plots

bash scripts/toy_example_maxwell/run_plots.sh

Run FDTD Experiments

Instructions to run the FDTD simulations of Maxwell's equations can be found in a separate README.md.

Miscellaneous

Extract Mitsuba scene

Prerequisites:

• Download and install Blender
• Install Mitsuba add-on for Blender

Generate Mitsuba export:

• Open the .blend scene file in ./blender/
• In Blender, go to File > Export > Mitsuba (.xml)
• Save file with Forward: Y Forward and Up: Z Up

Extract Bezier curves from Blender files

Drawing curves inside Blender can be useful to easily define a set of coordinates (e.g., Rx positions) to be used during simulation. These can be extracted from the .blend source file as follows:

• Prerequisite: Add Blender installation folder to environment PATH
• Run:
  blender --background <BLENDER_SCENE_PATH> --python ./blender/extract_curves.py --output-path=<JSON_OUTPUT_PATH>"

Example: blender --background ./blender/the_strand/the_strand.blend --python ./blender/extract_curves.py --output-path="./assets/the_strand/curves.json"