AI-Driven Mobile Network Optimization for Connected Vehicle Energy Management

Abstract

AI-driven mobile network optimization is becoming increasingly vital for enhancing the performance of connected vehicles, particularly in managing energy consumption during real-time mobility. With the rise of electric and autonomous vehicles, ensuring continuous, efficient, and adaptive connectivity has become critical for energy-efficient operations. However, traditional centralized network optimization techniques often suffer from high latency, lack of scalability, privacy concerns, and inadequate adaptability to dynamic driving environments. These limitations hinder the ability to provide real-time, energy-efficient decisions for connected vehicles. To address these issues, this study proposes a novel Federated Neuroevolutionary Learning (FNL) framework for energy-efficient optimization of mobile networks. The framework combines federated learning for decentralized privacy-preserving model training with neuroevolutionary algorithms that evolve neural network architectures and parameters based on vehicle energy profiles and network metrics. The proposed method enables collaborative, edge-driven optimization of bandwidth allocation, handover management, and communication routes without requiring the sharing of raw vehicle data. It adapts to changing vehicular contexts and supports real-time, energy-aware decision-making across distributed edge nodes. The proposed FNL method achieved 5.72 kWh energy use, 92.5% handover success, 0.098 loss, 78% bandwidth efficiency, 60.3 ms latency, and 91.9% packet delivery demonstrating superior performance across all metrics.