QL-PPTS:A Machine Learning Scheduling Scheme on Edge Computing Environments

Abstract Multiple Access Edge computing (MEC) is a computing model that can extend cloud computing and storage capabilities to the edge of cellular networks, close to IoT devices. MEC enables computation-constrained IoT devices to execute computation-intensive or low-latency applications at the edge of the network. However, in order to execute these applications efficiently, a reasonable scheduling strategy is required. Because of the particularity of edge environment, the traditional scheduling algorithm is not suitable. In this paper, we propose a new scheduling algorithm called QL-PPTS that combines reinforcement learning with a look-ahead list scheduling algorithm PPTS to reduce the makespan. we use QL to regard the rank of predictive cost matrix (rankP CM) of PPTS as the immediate reward. After the convergence of the Q-table, the priority queue of tasks is obtained according to the greedy principle. Experiments show that the QL-PPTS algorithm outperform the QL-HEFT.Specially, QL-PPTS can outperform QL-HEFT by 9% in terms of efficiency.