COBRA: An AI-Assisted Circuit-Level Optimizer for Open Source Based RFIC Design

Simone G, Lurz D, Grund M, Schneider F, Loose M, Breun S, Koch M, Weigel R, Franchi N (2026)

Publication Language: English

Publication Type: Conference contribution, Conference Contribution

Publication year: 2026

Publisher: IEEE

Event location: São Paulo BR

URI: https://ieeexplore.ieee.org/document/11706804

DOI: 10.1109/SBCCI69725.2026.11706804

Abstract

This work presents COBRA, a fully open source AI-assisted optimization framework for layout-aware Radio Frequency Integrated Circuit (RFIC) design. This paper also introduces ORCA, an open source pipeline for generating machinelearning surrogates for electromagnetic (EM) simulations. In the proposed workflow, ORCA generates EM surrogate models, which COBRA then integrates into a circuit-level optimization loop using tools such as GDSFactory, Palace, Qucs-S and Xyce. By replacing costly EM simulations with a neural network surrogate and performing only a few EM-based fine-tuning iterations, the computational effort of layout-aware RFIC optimization is significantly reduced. Due to the compatibility with Qucs-S and Xyce, the framework is not limited to passive structures or S-parameter analysis but enables co-simulation and optimization of active and passive components, including nonlinear circuit simulations. Since the proposed framework is fully open source, a neural network for a given geometry needs to be trained only once by the community and can subsequently be reused by other designers. This enables a significant reduction in optimization runtime, reducing typical design cycles from several hours of EM simulations to only a few minutes while maintaining fullwave EM simulation accuracy.

Authors with CRIS profile

Gianluca Simone Lehrstuhl für Intelligente Technische Elektronik und Systeme Martin Grund Lehrstuhl für Intelligente Technische Elektronik und Systeme Michael Loose Lehrstuhl für Intelligente Technische Elektronik und Systeme Sascha Breun Lehrstuhl für Technische Elektronik (LTE) Manuel Koch Lehrstuhl für Intelligente Technische Elektronik und Systeme Robert Weigel Lehrstuhl für Technische Elektronik (LTE) Norman Franchi Lehrstuhl für Intelligente Technische Elektronik und Systeme

Involved external institutions

Otto-von-Guericke-Universität Magdeburg DE Germany (DE)

How to cite

APA:

Simone, G., Lurz, D., Grund, M., Schneider, F., Loose, M., Breun, S.,... Franchi, N. (2026). COBRA: An AI-Assisted Circuit-Level Optimizer for Open Source Based RFIC Design. In Proceedings of the 2026 39th SBC/SBMicro/IEEE Symposium on Integrated Circuits and Systems Design (SBCCI). São Paulo, BR: IEEE.

MLA:

Simone, Gianluca, et al. "COBRA: An AI-Assisted Circuit-Level Optimizer for Open Source Based RFIC Design." Proceedings of the 2026 39th SBC/SBMicro/IEEE Symposium on Integrated Circuits and Systems Design (SBCCI), São Paulo IEEE, 2026.

BibTeX: Download