Energy-Aware Dynamic Offloading over IoT-based Mobile Edge Networks

In this research, we look at the challenge of scheduling jobs in a software-defined fog-IoT network while keeping latency and energy consumption in mind. We suggest the potent Deep Q-Learning process as a first step toward resolving the online job allocation and scheduler problem. The other guarantees economical energy use by preserving battery life within the limitations imposed by app reliance to reduce network latency. Next, we propose a deep reinforcement learning (DRL) method for adaptive scheduling tasks as well as allocation in SDN-enabled edge networks, and we factor in the task arrival process as part of this approach. Through thorough testing, we compared the proposed approach to three cutting-edge deep learning techniques (namely, deterministic, random, and A3C agents). Our method was shown to be better to other methods via several simulations. We also emphasise our design’s energy efficiency, which may result in savings of up to 87% compared to more traditional methods. We have proved that the offloading system can do up to 50 percent more work in the same amount of time. The results demonstrate that the proposed method may be utilised to proactively optimise scheduling as well as allocation of complicated activities with dependencies in decentralised Fog IoT networks.