Optimizing data interaction strategies for unreliable agents in multiplex networked industrial environments

Optimizing data interaction strategies for unreliable agents in multiplex networked industrial environments

연구 분야: Infrastructure

학회: CCF Transactions on High Performance Computing

Multiplex networked industrial environments are complex scenarios that include multiple heterogeneous industrial subsystems (such as production, logistics, energy, etc.). These subsystems operate independently, but are highly interconnected and collaborate to complete production tasks. In such environments, numerous agents (such as sensors, controllers, actuators, etc.) are responsible for information gathering, decision-making, and control. However, these agents are often unreliable due to cost, technology and environmental constraints. In order to optimize existing collaborative strategies, appropriate methods must be used to select reliable agents for data interaction. Previous methods, in the pursuit of high data integrity, tend to overlook network layer and agent load balancing issues, resulting in most requests being concentrated on a single network layer or agent, significantly increasing request latency. To address these issues, this paper proposes a data fusion method based on the unreliability of agent data interactions to optimize execution target screening. First, agent credibility is introduced to describe reliability and combined with entropy theory to mathematically model the overall optimization goal, proposing the concept of group entropy. Then, based on group entropy, network layer collaboration strategies are heuristically optimized by partially migrating requests from problematic network layers to others for execution. Within network layers, more suitable agents to execute requests are also selected based on group entropy using entropy change rates. Finally, comparative experiments show that the proposed method effectively mitigates the impact of unreliable agent data interaction on the fusion process, improving data integrity while reducing request latency.