AI-Driven Digital Twins for Mobile Networks: Simulation & Optimization

With the rapid development of mobile networks, the types of services and devices carried by the mobile network show diversified development trends. The vast scale of the network and the complex service requirements present significant challenges to the design, optimization, and operation of mobile networks. In conventional network optimization methods, optimization algorithms cannot be directly applied to real-world networks and are more challenging to scale to large-scale networks. The existing network optimization model based on expert experience and mathematical modelling will not be able to cope with the complex control and operation of the new generation of mobile networks, and new methods are urgently needed to break through the existing technical difficulties and traditional optimization framework.

This talk will introduce the mobile network digital twin, which is used for intelligent network optimization and features AI-driven simulation and decision-making as its core components. The goal of the AI-driven digital twin of the mobile network is to simulate the mobile network on the digital twin platform, allowing different network configurations to be tested for their effectiveness. Then, the decision-making part will update the configuration solutions based on the feedback iteratively and finally output the optimized network configuration adapted to the real-world network environment. This talk will discuss modeling technology jointly driven by data and knowledge, and how to create accurate digital twin models of users' mobile behavior, network service requirements, the environment for wireless propagation, and network performance. Also, this talk will introduce intelligent decision-making algorithms for large-scale network optimization. Based on these, this talk will discuss how to implement the mobile network digital twin system and present the performance of the proposed new network optimization framework on a prototype system.