DQN Inspired Joint Computing and Caching Resource Allocation Approach for Software Defined Information-Centric Internet of Things Network

Abstract

With the rapid development of the Internet of Things (IoT) network, the IoT devices need to perform the artificial intelligence (AI) model to make decisions according to the specific service requirement under a dynamic environment. However, the generation and usage of AI model typically requires a huge amount of communication, computing, and caching resource. Thus, the construction of the network and the scheduling of the limited network resources to realize the rapid generation and propagation of AI models are critical. Therefore, we propose a software-defined Information Centric-Internet of Things (IC-IoT) architecture to bring caching and computing capabilities to the IoT network. Based on the proposed IC-IoT architecture, we design a joint resource scheduling scheme to uniformly manage the computing and caching resources. The objective is to maximize the reward which consists not only short-term reward but also long-term reward brought by caching popular AI models. The resource scheduling problem is formulated into a multi-dimensional optimization problem. A new deep Q-learning method is proposed due to the complexity and high dimension of this problem. The simulation results verify the effectiveness of the software-defined IC-IoT architecture and the joint resource allocation strategy.

Keywords

• Information-centric network
• Internet of Things
• mobile edge computing
• joint optimization
• deep Q-learning