Cloud-Edge Collaborative Computing Offloading Method for IoT Terminals

With the booming IoT industry in recent years, smart devices are carrying more computationally intensive tasks. In the scenario of using drones for smart device inspection in the energy internet, it is often necessary to inspect towers, wires and their surrounding artificial and natural environments along transmission lines over long distances and take a lot of images and videos for inspection. However, the Unmanned Aerial Vehicle (UAV) has weak computing power and relies on a small capacity battery for operation, which cannot take on too many computing tasks; and the geographical location and task requirements of the UAV change from time to time, which easily leads to uneven load on the edge server nodes, increasing equipment energy consumption and reducing network performance. To solve the above problems, this paper investigates the cloud-edge collaborative task offloading mechanism based on deep reinforcement learning. This paper first constructs a three-layer system model of UAV-base station-cloud server for the intelligent inspection scenario and establishes a mathematical model with system power consumption as the optimization target and service processing delay and load balance among base stations as the constraints based on different demands in the actual scenario. Then, the deep reinforcement learning algorithm optimized by proximal policy optimization (PPO) is used for simulation to solve the offloading decision matrix in a given time and network environment, and the performance of the model is verified. The simulation results show that the algorithm in this paper can reduce the energy consumption of the system while ensuring the delay and load balance.