Energy-Efficient Solution Based on Reinforcement Learning Approach in Fog Networks

With the recent development of delay-sensitive Internet of Things (IoT) applications, the energy consumption has drawn a significant attention. With the growing popularity of the Fog computing, it is expected to be an effective solution to meet not only low latency, but also decreasing the energy consumed by the system. This paper studies the energy efficiency and quality of service (QoS) issues in IoT-Fog-Cloud systems by proposing a joint optimization of resource allocation and workload dispatching over a fog-cloud system. The joint communication and computing optimization problem is formulated by a Nash Equilibrium problem (NEP), which allows the trade-off between consumed energy by the system and QoS. To break the curse of large-scale systems, we propose a Reinforcement Learning-based algorithm that allows users to learn the optimal policy without having a priori knowledge of the dynamic statistics of the system. Finally, we conduct simulation experiments based and comparison two benchmarks. Evaluations and comparisons demonstrate the efficiency of our proposal.