Welcome to the official code repository for our paper, "Aggregation Mechanisms in Federated Learning for Enhancing Robotic Visual Obstacle Avoidance." In this repository, you will find the necessary code to replicate our Federated Learning (FL) experiments aimed at improving visual obstacle avoidance in robotics.
mixed_fusion.ipynb: The primary notebook that implements Federated Learning with various aggregation mechanisms for visual obstacle avoidance.evaluation.ipynb: This notebook is used for performance evaluation of the saved models.draw.ipynb: A notebook responsible for generating figures for the paper.traditional_centralized_learning/: This directory contains code for conducting traditional centralized learning for comparison with our FL approach.
Ensure that your system meets the following requirements:
- Correct Cuda version compatible with your hardware
- Appropriate Python version
- Compatible version of PyTorch
-
Create a new conda environment:
conda create -n fl python=x.x
-
Activate the new environment:
conda activate fl
-
Dependencies Installation Install the required dependencies by running the following commands:
conda install pytorch torchvision torchaudio cudatoolkit=xx.x -c pytorch -c nvidia conda install jupyter notebook conda install tqdm conda install scikit-learn conda install matplotlib conda install tensorboard
Please replace python=x.x and cudatoolkit=xx.x with the versions that match your system.
Make sure to activate the conda environment before launching Jupyter Notebook. Then just simply run our codes.
We welcome contributions to this project. Please fork the repository and submit a pull request for review.
Distributed under the MIT License. See LICENSE for more information.
For any queries regarding the code or the paper, please open an issue in this repository.
The authors wish to thank doctoral student Xianjia Yu for his support and assistance.