An Adaptive Resource-Aware Node Selection for Federated Learning in IoT Using Genetic Optimization

초록

The performance of Federated Learning (FL) in Internet of Things (IoT) environments strongly depends on efficient resource management under heterogeneous device and network constraints. Although FL enables decentralized model training without sharing raw data, balancing operational cost and Quality of Service (QoS) remains a major challenge for scalable and stable learning in large-scale IoT systems. This paper proposes a Genetic Algorithm (GA)-based optimization framework for federated IoT environments that jointly minimizes communication cost and maximizes QoS through adaptive client selection. The proposed approach formulates node participation as a normalized multi-objective optimization problem and dynamically selects suitable devices at each training round according to their resource availability and network conditions. Simulation results in heterogeneous IoT scenarios demonstrate that the proposed method significantly reduces communication cost while maintaining high QoS and stable global model convergence, highlighting its effectiveness for resource-efficient federated learning in real-world IoT deployments.

키워드