Adaptive Resource Allocation Method Based on Deep Q Network for Industrial Internet of Things
Xiaolong Lai, Qing Hu, Weixin Wang, Li Fei, Ying Huang
Abstract
The transformation and upgrading of the industrial manufacturing field brings important development opportunities for the promotion and deployment of the Industrial Internet of Things (IIoT). How to connect huge number of IIoT nodes to the current IP-based Internet for promoting the integration of industrialization and informationization is a key issue towards IIoT. Targeting at the insufficiency that existing algorithms only support a single channel and utilize limited number of time slots, which results in the deteriorated performance for solving the problem of resource allocation for nodes with different priorities. In this paper the self-similarity of the observation data of the IIoT nodes is evaluated to dynamically prioritize the nodes, and then Deep Q Network (DQN) algorithm is adopted to design an adaptive resource allocation method that takes into account the characteristics of diverse node priorities. Simulation results demonstrate that the proposed method can effectively improve network throughput and reduce data transmission delay in comparison with existing methods.