Title: Edge computing and dynamic scheduling control of internet of things based on DQN reinforcement learning scheduling algorithm

Authors: Pengkang Xing; Changjie Wu

Addresses: School of Electronic Information Engineering, Henan Polytechnic Institute, Nanyang, 473000, China ' School of Electronic Information Engineering, Henan Polytechnic Institute, Nanyang, 473000, China

Abstract: This study proposes a dynamic scheduling control method for internet of things (IoT) systems based on a deep Q-network (DQN) algorithm, combining convolutional neural networks and Q-learning. Feature information is extracted via a convolutional neural network, and its parameters are optimised through Q-learning to develop a deep Q-learning algorithm for IoT edge computing and dynamic orchestration. The goal is to improve resource utilisation and reduce energy consumption. Experiments show the proposed method achieves a low error rate of 0.31%, a computation speed of 6.7 bps, and a space occupancy rate of 27.8%, outperforming weighted fair queuing (1.22%, 37.9%) and highest response ratio next (1.73%, 53.2%). IoT resource utilisation reached 92.1% and system stability 93.8%. These results demonstrate the algorithm's superior accuracy, efficiency, and reliability, offering a cost-effective solution for dynamic scheduling and optimisation in IoT edge computing environments.

Keywords: internet of things; convolutional neural networks; Q-learning algorithms; QLAs; deep Q-network algorithms; edge computing; dynamic scheduling control.

DOI: 10.1504/IJSCC.2026.152969

International Journal of Systems, Control and Communications, 2026 Vol.17 No.2, pp.216 - 232

Received: 25 Apr 2025
Accepted: 10 Jun 2025
Published online: 17 Apr 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article