Prioritized replay dueling double DQN for intelligent task scheduling in IoT Fog Cloud computing

Task scheduling remains a challenge in IoT-Fog-Cloud systems due to dynamic workloads and heterogeneous resources. Ineffective scheduling strategies increase latency, energy consumption, and Service Level Agreement (SLA) violations. To address these limitations, this paper proposes a Prioritized Experience Replay Dueling Double Deep Q-Network (PER-Dueling Double-DQN) framework for task scheduling in IoT-Fog-Cloud architectures. The proposed framework enables the agent to learn task-resource mappings by combining a Dueling network architecture, Prioritized Experience Replay, and Double-DQN learning. A multi-objective reward function jointly considers latency, energy consumption, deadline compliance, and QoS requirements, while Double-DQN and PER improve value estimation and learning efficiency. The experimental findings indicate that the proposed framework achieves lower mean latency, energy consumption, and SLA violation rates, together with higher cumulative reward, than the evaluated Double-DQN, DQN, DDPG, LSTM-RNN, FCFS, and EDF baselines under the considered simulation settings.