FLAD: Federated Learning for LLM-based Autonomous Driving in Vehicle-Edge-Cloud Networks

The Federated LLM-based Autonomous Driving (FLAD) framework has been introduced to tackle the challenges of training large language models (LLMs) for autonomous driving, particularly high computation transmission costs and privacy concerns. FLAD employs federated learning, allowing autonomous vehicles to collaboratively train models without sharing raw data, thus preserving privacy. Key innovations include a cloud-edge-vehicle collaborative architecture that minimizes communication delays, an intelligent parallelized training mechanism that optimizes efficiency, and a knowledge distillation method for personalizing LLMs based on diverse edge data. Prototyped using NVIDIA Jetsons, FLAD demonstrates superior end-to-end performance in autonomous driving while effectively utilizing distributed vehicular resources, marking a significant advancement in the field of AI and autonomous systems.