A Multi-Agent Collaboration-Based Edge Data Processing and Decision-Making Mechanism for the Industrial Internet of Things

Abstract

A large amount of data is produced at the edge of the Industrial Internet of Things, and it needs to be processed in real time and intelligently decided upon; thus, the latency and bandwidth limitations of traditional centralisation have become serious problems. The problems mentioned above will be solved by the new Distributed Processing Framework based on multi-agent cooperation. As an edge node, it is an independent decision-making unit in a multi-agent system, so for the purpose of heterogeneous data synchronization, distributed consensus algorithms are employed, and event-driven protocols are used to improve communication efficiency. Deep reinforcement learning will be employed to obtain the feature representation and dynamically schedule tasks online, and due to resource constraints, it has been optimised to have a small model size. Game theory and policy gradient methods are combined to form a cooperative decision-making scheme in a dynamic environment, and thus the local strategies will be in line with the general goal. The whole stack of the technical system for data interaction and intelligent decision-making in this paper will be introduced to support the application of edge intelligence in the Industrial Internet of Things.