Weighted explainable federated learning for privacy-preserving and scalable energy optimization in autonomous vehicular networks

This is the authors' abstract. We don't add key points for this paper.

The rise of electric and autonomous vehicles in smart cities poses significant challenges in vehicular energy management, including unoptimized consumption, inefficient grid utilization, and unpredictable traffic dynamics. Traditional centralized machine learning models and cloud-based Energy Management Systems (EMSs) often struggle with real-time adaptability, high-dimensional data processing, and privacy concerns. While Federated Learning (FL) offers a decentralized solution by enabling edge devices to collaboratively train models without sharing raw data, conventional FL typically treats all client contributions equally—regardless of their data volume, quality, or contextual relevance. This limits model generalization in heterogeneous vehicular environments. To address this, we propose a Weighted Explainable Federated Learning (WEFL) framework that enhances conventional FL by assigning dynamic importance to client updates based on factors such as data relevance and local model performance. The framework also integrates Explainable AI (XAI) methods to improve interpretability, transparency, and regulatory compliance. The proposed WEFL-XAI model ensures privacy-preserving, real-time, and adaptive vehicular energy optimization, leveraging traffic patterns, vehicle energy states, and grid load conditions. Experimental evaluations demonstrate that our Multi-Layer Perceptron (MLP)-based Weighted Federated Learning (WFL) model achieves an R 2 of 86.84% for energy consumption and 74.16 % for traffic density, reflecting a strong trade-off between performance and privacy. While these values are lower than centralized MLP benchmarks, the WFL model outperforms standard FL baselines by offering enhanced privacy preservation, interpretability, and decentralized scalability, making it a more viable choice for real-world smart mobility deployments.