GPS Aided Beam Prediction and Tracking for UAV mmWave Communication

This is a research code repository for GPS-aided beam prediction and tracking for mmWave communication.

In this work, there are 3 folders inside folder notebooks:

1. Folder minmaxgeo_uebsvector contains the proposed model for beam prediction and tracking that uses min-max normalized UAV's geodetic position (latitude and longitude) and UAV-BS unit vector as input. This folder contains the code implementation of this work: GPS-Aided Deep Learning for Beam Prediction and Tracking in UAV mmWave Communication.
2. Folder uebsvector_logscaledheight contains the proposed model for beam prediction and tracking that uses UAV-BS unit vector and log scaled UAV's height as input. Read the tl;dr article about it here.
3. Folder baseline contains the baseline models for beam prediction and tracking as described in [3] (NOTE: these implementations are not the official code from the original authors).

Notebook Content:

1. 00_test_drone_cnn_ed_rnn_experiment and 01_drone_cnn_ed_rnn_experiment.ipynb: a notebook to train the proposed model.
2. 02_visualization_combination.ipynb and 03_visualization.ipynb: a notebook to visualize the evaluation metrics.
3. 00_test_onnx.ipynb: a notebook to measure the inference time using PyTorch model and ONNX model. Read the article about it here
4. 00_test_dataset_label.ipynb: visualize label distribution using various data set splitting methods.
5. 00_test_drone_base_prediction.ipynb: train a baseline beam prediction model[3].
6. 00_test_drone_base_tracking.ipynb: train a baseline beam tracking model[3].

How to run the code

1. Clone this repo.
2. Enter the repo directory through the terminal.
3. Run poetry install to install the dependencies (run pip install poetry if you haven't installed poetry yet).
4. Run poetry update to update the libraries version.
5. Run poetry shell to activate the virtual environment.
6. Download end extract the zip file data set Scenario 23 from DeepSense6G website [1].
7. Create folder data/raw/ inside the repo.
8. Put the data set into folder data/raw/. The dataset folder should be named in the format of Scenario{scenario_number}. This folder should contains scenario{scenario_number}.csv file.
9. To run the jupyter notebook, open the jupyter notebook and make sure to set the kernel to the poetry environment.
10. Inside the notebook, make sure to change the repository path to the correct path (e.g. sys.path.append('F:/repo/gpsbeam')).
11. Run the notebook.

How to Reproduce the Result

1. To reproduce the proposed model result, run 01_drone_cnn_ed_rnn_experiment.ipynb.
2. Visualize the result by running 02_visualization_combination.ipynb and 02_visualization.ipynb.

Additional Information

• The preprocessed dataset will be saved in data/processed/.
• Experiment result will be saved in data/experiment_result/.

Thanks to

This repository is inspired by the following repositories:

• Position-Beam-Prediction
• lidar_beam_tracking
• Vision-Position-Beam-Prediction

Thanks to the tutorials from the following sources:

• Deepsense Workshop Spring 2024
• DeepSense6G Tutorials

Reference:

[1] A. Alkhateeb, G. Charan, T. Osman, A. Hredzak, and N. Srinivas, “DeepSense 6G: large-scale real-world multi-modal sensing and communication datasets,” to be available on arXiv, 2022. [Online]. Available: https://www.DeepSense6G.net

[2] Charan, G., Hredzak, A., Stoddard, C., Berrey, B., Seth, M., Nunez, H., & Alkhateeb, A. (2022, December). Towards real-world 6G drone communication: Position and camera aided beam prediction. In GLOBECOM 2022-2022 IEEE Global Communications Conference (pp. 2951-2956). IEEE.

[3] Charan, G., & Alkhateeb, A. (2024). Sensing-Aided 6G Drone Communications: Real-World Datasets and Demonstration. arXiv preprint arXiv:2412.04734. [Online].Available: https://arxiv.org/abs/2412.04734