Doctoral Dissertations

Abstract

The rapid increase in power density and stringent power-integrity requirements in modern System-on-Chip (SoC) platforms have made Power Delivery Network (PDN) design an increasingly complex, multi-stage challenge. Critical decisions must be made both during pre-layout planning, such as stackup configuration, power-plane geometry, and early decoupling capacitor (decap) budgeting, and during post-layout refinements. Traditional heuristic and evolutionary optimization techniques struggle with scalability, require extensive manual iteration, leading to long runtimes and limited adaptability across varying board configurations. To address these challenges, this work proposes a unified reinforcement-learning-driven framework for automated PDN synthesis and decap optimization that spans both pre-layout and post-layout stages. The methodology integrates three complementary components: (1) a hybrid multi-agent reinforcement learning framework for automated pre-layout power-plane and stackup synthesis under domain-specific DC resistance constraints, (2) a coordinated multi-agent decap planning strategy for early-stage multidomain SoCs using a centralized controller to minimize capacitor usage, and (3) a sequential Transformer-decoder-based reinforcement learning model for post-layout decap assignment. Validation on synthetic benchmarks and industry-scale designs demonstrates that the proposed framework consistently produces electrically compliant, space-efficient, and geometrically feasible PDN solutions with substantially reduced runtime compared to traditional optimization methods, establishing a scalable foundation for end-to-end automated PDN design in complex multi-domain electronic systems.

Advisor(s)

Hwang, Chulsoon

Committee Member(s)

Beetner, Daryl G.
Kim, DongHyun (Bill)
Wang, Hanfeng
Jiang, Lijun

Department(s)

Electrical and Computer Engineering

Degree Name

Ph. D. in Electrical Engineering

Publisher

Missouri University of Science and Technology

Publication Date

Spring 2026

Journal article titles appearing in thesis/dissertation

Paper I, found on pages 6–44, has been submitted to IEEE Transactions on Signal and Power Integrity.

Paper II, found on pages 45–78, has been submitted to IEEE Transactions on Signal and Power Integrity.

Paper III, found on pages 79–115, is intended for submission to IEEE Transactions on Signal and Power Integrity.

Pagination

xiii, 118 pages

Note about bibliography

Includes_bibliographical_references_(pages 42, 77 & 114)

Rights

© 2026 Haran Manoharan , All Rights Reserved

Document Type

Dissertation - Open Access

File Type

text

Language

English

Thesis Number

T 12599

Download

Share

 
COinS