A Novel Deep Reinforcement Learning Methodology for EMI Mitigation in Multiscenario Pinmap Design

Intelligent pin assignment (pinmap) of ball grid array (BGA) packages for electromagnetic interference (EMI) mitigation is becoming a new trend. A high-performance design methodology tackling different pinmap design scenarios is urgently needed. This article proposes an efficient pinmap design methodology based on flexible deep reinforcement learning (DRL) definitions and addresses three typical pinmap design scenarios. An improved DRL model is designed to mitigate the radiated emissions of pinmaps, which has an excellent search capability. The results show that the proposed methodology has better search performance to find the final solutions with minimized radiated emissions than conventional DRL-based pinmap design approaches across various pinmap design scenarios.