Reinforcement Deep Learning Approach for Multi-User Task Offloading in Edge-Cloud Joint Computing Systems

To enhance system utility in multi-user task offloading, a reinforcement deep learning-based task offloading scheme within an edge-cloud joint computing framework is proposed. This scheme leverages deep reinforcement learning to optimize the collaborative allocation of resources between edge and cloud, improving decision-making for task offloading modes. A reinforcement learning algorithm based on submodular theory is developed to fully utilize both computing and communication resources in edge and cloud environments. Simulation results show that the proposed scheme significantly reduces execution delays and energy consumption. Even under resource-constrained conditions with multiple users, the system maintains stable performance and high efficiency.