HEART: Achieving Timely Multi-Model Training for Vehicle-Edge-Cloud-Integrated Hierarchical Federated Learning

Researchers introduce HEART, a novel framework for efficient multi-model federated learning across vehicle-edge-cloud architectures that addresses training latency and resource allocation challenges in IoV systems. The solution combines hybrid synchronous-asynchronous aggregation with optimized task scheduling using particle swarm optimization and genetic algorithms.

The emergence of Internet of Vehicles (IoV) systems demands ML solutions capable of handling distributed, mobile data sources without centralizing sensitive vehicular information. HEART addresses a genuine technical gap in hierarchical federated learning by tackling the underexplored problem of simultaneous multi-model training in dynamic edge environments. Traditional federated learning frameworks assume single-task scenarios; HEART's contribution lies in managing competing resource demands when vehicles execute parallel ML workloads while maintaining mobility across network edges.

The technical innovation combines established optimization techniques—particle swarm optimization and genetic algorithms—with greedy scheduling to balance computational efficiency against training time. This hybrid approach acknowledges the NP-hard nature of the underlying optimization problem, providing practical heuristics rather than theoretical guarantees. The framework's two-stage design separates task allocation from execution priority, enabling scalability across heterogeneous vehicular networks.

For the autonomous vehicle and smart transportation ecosystem, efficient federated learning infrastructure directly impacts deployment timelines for collaborative AI systems. Reducing training latency while maintaining model quality accelerates development cycles for safety-critical applications like collision avoidance and traffic prediction. The approach also reduces bandwidth consumption by implementing smart aggregation rules, addressing practical constraints in vehicle-to-edge connectivity.

The research represents incremental but meaningful progress in federated learning optimization rather than breakthrough innovation. Real-world validation on actual datasets strengthens the work's credibility, though the paper's focus on algorithmic efficiency may understate deployment challenges in highly dynamic vehicular environments. Future work should address fault tolerance and partial vehicle dropout scenarios common in practice.

• →HEART framework enables simultaneous multi-model training in vehicle-edge-cloud systems while minimizing global training latency.
• →Hybrid synchronous-asynchronous aggregation rules prevent model obsolescence caused by vehicular mobility in federated networks.
• →Two-stage optimization combining PSO and genetic algorithms with greedy scheduling addresses NP-hard resource allocation problems.
• →Framework demonstrates superior performance over existing methods on real-world datasets, validating practical applicability.
• →Efficient federated learning infrastructure accelerates autonomous vehicle and smart transportation AI development cycles.