Automated Power Plane And Stackup Synthesis For Package PDNs Using Reinforcement Learning

Author

• Haran Manoharan
• Chulsoon Hwang, Missouri University of Science and TechnologyFollow

Abstract

Modern packages have more power domains, making manual Power Delivery Network (PDN) design slow and heavily trial-and-error. This work presents an automated framework for power plane and stackup synthesis using reinforcement learning. Given chip pin maps and Ball Grid Array (BGA) ball assignments, the proposed algorithm selects power layers and generates plane shapes that satisfy target DC Resistance (DCR) constraints for all power domains. The algorithm combines graph-based methods with iterative shape expansion, while a physics-based engine evaluates resistance at each step. The approach performs on-the-fly optimization and generalizes across diverse package configurations. The proposed work has been tested across multiple test cases and provides a solution within minutes.

Recommended Citation

H. Manoharan and C. Hwang, "Automated Power Plane And Stackup Synthesis For Package PDNs Using Reinforcement Learning," Final Program 2026 Asia Pacific International Symposium and Exhibition on Electromagnetic Compatibility Apemc 2026, Institute of Electrical and Electronics Engineers, Jan 2026.

The definitive version is available at https://doi.org/10.1109/APEMC65388.2026.11593654

Department(s)

Electrical and Computer Engineering

Comments

National Science Foundation, Grant IIP-1916535

Keywords and Phrases

Package PDN; power plane routing; reinforcement learning; stackup synthesis

Document Type

Article - Conference proceedings

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2026 Institute of Electrical and Electronics Engineers, All rights reserved.

Publication Date

01 Jan 2026