Leveraging Large Language Models for Intelligent Power Electronics PCB Routing Optimization

Paper i proceeding, 2025

With the increasing demands for efficiency and integration in power electronics, PCB routing has become a critical factor in ensuring signal integrity. This work introduces a novel automated routing optimization framework that combines large language models (LLMs) with the Astar algorithm. The approach utilizes LLMs to semantically analyze PCB layout characteristics, including network types and trace density, to intelligently determine routing priorities and dynamically adjust pathfinding constraints. Through seamless integration of these analytical capabilities with the Astar algorithm's path optimization, the method effectively reduces trace crossings and minimizes signal interference. Experimental results demonstrate significant improvements over traditional automated routing methods, particularly in optimizing trace paths and reducing congestion. These findings suggest that AI-driven PCB routing strategies can contribute to improved performance in power electronics applications.