Hierarchical Federated Learning Framework for Driving Range Estimation of Battery Electric Vehicle

Overview

Federated Learning (FL) enables devices to cooperatively train a machine learning (ML) model without sharing their private data. It has two key obstacles, however: a communication bottleneck and data heterogeneity. To overcome the communication challenge, I present a probabilistic device selection strategy that allows fewer devices to participate in training. To address the issue of data heterogeneity, we created a hierarchical FL to customize edge models for the devices.

The hierarchical federated learning framework

Experiment

Hardware

• NVIDIA GTX 1080Ti

Software

• IDE: PyCharm
• Programing Language: Python3
• Machine Learning Framework: PyTorch
• Cloud plateform: AWS

Prediction Results

Pulbications

Based on this project, I complete

• 2 patents (filed by USPTO)
• 1 conference paper (IEEE Intelligent Vehicle 2023)