Energy-Efficient Task Orchestration in the Edge-Cloud Continuum Using Deep Reinforcement and Federated Learning for Sustainable IOT

Efficient orchestration in the edge–cloud continuum is essential for reducing energy consumption and meeting latency requirements in large-scale IoT systems. This article presents a hybrid deep reinforcement learning (DRL) and federated learning (FL) framework that dynamically allocates computation across IoT, edge, fog, and cloud layers. The DRL agent learns energy-efficient scheduling strategies through a latency-aware reward design, while FL enables decentralized model training without exposing raw data. Experimental evaluation demonstrates up to 31.6% lower energy consumption and 28.4% latency reduction compared to existing heuristics. Results also show rapid learning convergence within 200 episodes, indicating strong adaptability under changing network and workload conditions. These findings confirm the effectiveness of the proposed framework in improving energy efficiency, latency performance, and scalability for next-generation IoT deployments.