Intelligent Resource Adaptation for Diversified Service Requirements in Industrial IoT
Weiting Zhang, Yi-Qian He, Tao Zhang, Chenhao Ying, Jiawen Kang
Abstract
In this paper, we investigate the problem of resource adaptation for diversified service requirements in the industrial Internet of things (IIoT). An intelligent resource adaptation scheme is designed to integrate scheduling of computing and networking resources, and make task offloading decisions, thereby satisfying the diversified requirements within IIoT. To provide the necessary integrating resource support for different requirements, we first introduce a cloud-edge-end collaborative IIoT architecture that leverages rich resources and global coverage of cloud computing, combined with the real-time processing capabilities and low network resource consumption of edge and local computing. To obtain an optimal resource adaptation strategy, we formulate an optimization problem with the objective of maximizing the system’s long-term key quality indicator (KQI). Due to the coupled constraints among resource adaptation decisions, it is difficult to solve this problem directly. Additionally, diversified service requirements and dynamic network conditions of IIoT increase the complexity of finding solutions. Thus, we reformulate the formulated problem as a joint optimization problem involving task offloading and resource allocation. Then, we transform the joint optimization problem into a Markov decision process (MDP) and propose a KQI-oriented deep reinforcement learning (DRL)-based algorithm for task offloading and integrated resource scheduling (KODDQN). This algorithm intelligently handles the diversified service requirements by sensing key information, such as the state of computing and networking resources within the IIoT. Extensive simulation results show that the proposed scheme can improve the system’s long-term KQI by 2% compared to classical DDQN-based schemes.