Vision Assited 5G Beam Selection with Federated Learning

This repository includes code for the project that studies the using vehicle sensor data (e.g., LiDAR) to reduce the 5G beam selection overheads. The project is informed and inspired by the following sources and repositories:

• https://github.com/galidor/ITU_Beam_Selection_TF
• https://arxiv.org/pdf/2102.02802

The new contributions of these project include:

• Evaluate the impact of Low-Rank Adaptation (LoRA) on model complexity and beam selection performance
• Use Flower Federated AI Framework to implement the federated LIDAR aided beam selection