Journal of Engineering, Project, and Production Management, 2026, 16(5), 2025-290
Optimizing Cloud Computing Resource Scheduling Strategies Using Reinforcement Learning
Associate Professor, Guilin Tourism University, No.26 Liangfeng Road, Guilin, Guangxi, 541006, China, E-mail: 18607731203@163.com
Project Management
Received November 26, 2025; revised December 25, 2025; accepted January 3, 2026
Available online June 17, 2026
Abstract: The high dynamism and unpredictability of current cloud computing environments often lead to low resource utilization and increased service latency. This paper proposes an improved Soft Actor-Critic (SAC) algorithm. A multi-scale Convolutional Neural Network (CNN) is used to process state information such as task queue length, resource utilization, and virtual machine load, extracting local variance and global trend, respectively. A dual-Q network architecture is designed to suppress overestimation of action values. Entropy coefficient learning adjusts exploration and utilization based on resource fluctuations. Gradient backpropagation adjusts the weights of the composite reward function considering the Service Level Agreement (SLA). A soft update mechanism for the target network ensures training stability. Experimental results demonstrate that the algorithm exhibits high efficiency in highly dynamic environments, achieving an average task latency of 12.2ms and a resource utilization rate of 94.6% within 200 seconds.
Keywords: Cloud computing environment; resource scheduling strategy; SAC algorithm; multi-scale CNN; dual-q network.
Copyright © Journal of Engineering, Project, and Production Management (EPPM-Journal).
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
Requests for reprints and permissions at eppm.journal@gmail.com.
Citation: Peng, Z. (2026). Optimizing Cloud Computing Resource Scheduling Strategies Using Reinforcement Learning. Journal of Engineering, Project, and Production Management, 16(5), 2025-290.
