Distributed Intelligent Scheduling for Multi-Level eMBB Services in IoT

The integration and coexistence of multiple tasks is one of the significant characteristics of the Internet of Things (IoT) scenario. In response to the demand for differentiated QoS guarantee of multiple tasks in scenarios, this article studies the distributed multiple task scheduling problem with user active decision assistance. Based on the QoS requirements of different tasks, reward functions related to task profits, latency cost, resource and power consumption are defined. A user local Markov decision process model for distributed task scheduling is modeled, and an intelligent solution is designed based on a policy gradient algorithm with a baseline. The paper implements scenario simulation of the distributed scheduling process for multiple tasks, adopting data distribution to train the user local multiple task scheduling request decision agent. The paper also compared the performance of the proposed intelligent algorithm with classical scheduling algorithms, verifying the convergence and effectiveness of the proposed algorithm.