Research on Industrial Cyber Range Based on Multi-agent Cooperative Optimization

Combining the characteristics of multi-agents and the prototype structure of the industrial cyber range (ICR), a multi-agent cooperative optimization ICR routing algorithm is proposed. It is used to relieve the pressure of the upper computer computing interaction between the information layer of the industrial control network and the field layer. Use the dependency relationship between environment perception and interactive decision-making of multi-agent reinforcement learning to establish an ICR multi-agent cooperative model, decompose ICR into the central brain of the industrial control network and distributed intelligent routing modules, abstract each module into an agent, and apply it to the industrial control network. The learning of historical data realizes the optimal feedback of the agent to the Part I environment and computing resource requirements, and improves the congestion environment of the partial industrial control network. The experimental results show that, compared with the Q-Learning algorithm, multi-agent cooperative optimization of the ICR routing algorithm improves the utility of the industrial control network in the Part I area to a certain extent.