EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Mathematical Foundations of AI-Assisted VLSI: A Comparative Review of Physical Design, Optimization, Verification and Testing

Radhika Kavra, Lakshika Chutani, Vaibhav Maheshwari

Boletim da Sociedade Paranaense de Matemática · 28.10.2026 · 44 (14) · 1-19

DOI: 10.5269/bspm.85935 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Boletim da Sociedade Paranaense de Matemática, online erschienen am 28.10.2026. Verfasst von Radhika Kavra und 2 weiteren Autor:innen. Die Arbeit nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

As Moore's Law slows, advanced-node VLSI design faces increasing computational complexity. Meeting the requirements of 7 nm and smaller technologies involves solving highly complex, often NP-hard, optimization problems. Traditional EDA methods based on deterministic algorithms, such as routing, SAT solving, and sparse matrix computations ($Ax=b$), become increasingly computationally expensive for billion-transistor designs. Consequently, learning-based approaches are gaining attention as complementary solutions. This review examines the use of Graph Neural Networks (GNNs), Reinforcement Learning (RL), and meta-heuristic optimization techniques for improving placement, optimization, verification, and testing in modern VLSI design. This review examines four key application domains: ($S1$) learning-based methods for physical design optimization, ($S2$) neural-network surrogate models for circuit tuning and performance prediction, ($S3$) AI-assisted logic synthesis and verification on And-Inverter Graphs (AIGs), and ($S4$) machine-learning-guided Automatic Test Pattern Generation (ATPG) to reduce backtracking during testing. Emerging applications in hardware security, including the detection of layout-level Trojans, are also discussed. Additionally, the review highlights the trade-offs associated with AI-based approaches, particularly the balance between computational efficiency and prediction accuracy. Overall, the surveyed literature demonstrates how AI-driven models can accelerate complex optimization tasks and support modern VLSI design workflows.

Zitieren

APA

Kavra, R., Chutani, L., & Maheshwari, V. (2026). Mathematical Foundations of AI-Assisted VLSI: A Comparative Review of Physical Design, Optimization, Verification and Testing. Boletim da Sociedade Paranaense de Matemática, 44(14), 1-19. https://doi.org/10.5269/bspm.85935

Vancouver

Kavra R, Chutani L, Maheshwari V. Mathematical Foundations of AI-Assisted VLSI: A Comparative Review of Physical Design, Optimization, Verification and Testing. Boletim da Sociedade Paranaense de Matemática. 2026;44(14):1-19. doi:10.5269/bspm.85935

BibTeX
@article{Kavra2026,
  author = {Kavra, Radhika and Chutani, Lakshika and Maheshwari, Vaibhav},
  title = {Mathematical Foundations of AI-Assisted VLSI: A Comparative Review of Physical Design, Optimization, Verification and Testing},
  journal = {Boletim da Sociedade Paranaense de Matemática},
  year = {2026},
  volume = {44},
  number = {14},
  pages = {1--19},
  doi = {10.5269/bspm.85935}
}
Kostenlos starten →