In order to study the task unloading algorithm for mobile edge computation, this paper proposes an artificial intelligence-based approach. First, a load-unloading model is developed for multi-dependent multi-service nodes within large-scale non-homogeneous mobile edge computing, and then an advanced in-depth training algorithm is used to optimize the task in combination with real-world application options for mobile edge computing. Unloading strategy. Finally, the unloading strategy comprehensively compares energy consumption, cost, load balancing, delays, network operation, and average execution time and analyzes the advantages and disadvantages of each unloading strategy. The simulation results show that the edge algorithm distributes all sub-tasks evenly to all peripheral servers, so the decision to lower central processing unit (CPU) usage to peripheral servers is kept between 20% and 80%, which ensures load balancing performance. As for the Deep Q Network (DQN) algorithm, these two algorithms are better than DQN because the DRQN algorithm and the HERDRQN algorithm are less commonly used on the edge server with the lowest average performance and power ratio, while the CPU utilization is [80%, 100%]. power consumption.algorithm. This proves that the strategy created by the HERDRQN algorithm is scientific and effective in solving the task of unloading the task.
Cite this article as: Y. Yuan, M. Asif Ikbal and A. Alam, "Task unloading algorithm for mobile edge computing based on artificial intelligence," Electrica., 22(3), 387-394, 2022.