报告题目：An Online Interactive Physics-Informed Diffusion-Adversarial Network for Solving Mean Field Games

报 告 人：尹伟石 副教授  长春理工大学

报告时间：2025年7月4日 09：00—10：00

报告地点：数学楼研讨室6

校内联系人： 吕俊良 [email protected]

报告摘要：We propose an online interactive physics-informed diffusion-adversarial network (IPIDAN). This method is designed for solving mean field game (MFG) problems in high-dimensional, complex, and dynamic environments. By employing the variational dyadic structure of mean field games, we transform the dynamic game problem into a static optimization problem. Using a generative adversarial framework, adversarial training approximates the solution to the mean field games. IPIDAN, through the structure of the diffusion generation model, provides the network with greater tunability. It also enhances its ability to model randomness in high-dimensional strategy spaces. By establishing a connection between the diffusion process and the agents’ motion dynamics, we offer greater interpretability for the network. Numerical experiments demonstrate the effectiveness of IPIDAN in solving high-dimensional mean field game models. These results are particularly validated through quadrotor obstacle avoidance experiments in various scenarios.

报告人简介：尹伟石，长春理工大学数学与统计学院副教授，硕士研究生导师。主要研究兴趣是数学物理反问题、机器学习算法的设计与理论分析和微分方程数值解等。在JCP，JCAM，CICP，IPI等期刊发表论文20余篇，主持并参与国家自然科学基金、吉林省科技厅基金和吉林省教育厅基金6项。目前担任中国仿真学会不确定系统分析与仿真专委会委员、Math Review评论员以及IPIA会员。