This paper introduces CL-DiffPhyCon, a closed-loop diffusion-based control framework for PDE-governed physical systems. The method decouples synchronous denoising, allowing for asynchronous control signal generation at different physical time steps, significantly improving computational efficiency. Evaluations on Burgers' equation and incompressible fluid control demonstrate superior performance and faster sampling efficiency compared to existing generative control methods.

“CL-DiffPhyCon introduces an asynchronous diffusion-based closed-loop control method for PDE systems, achieving efficient and adaptive real-time control.”

“CL-DiffPhyCon 提出了一种基于异步扩散的闭环控制方法，实现高效且自适应的 PDE 系统控制。”