•  
  •  
 

Keywords

GenRLCirc Resistor, capacitor filter Self-healing electronics Q-learning optimization Graph-based variational autoencoder

Document Type

Research Paper

Abstract

This study presents an innovative self-tuning system for first-class RC filter circuits, specially designed to achieve a target cut-off frequency of 500 kHz with unprecedented accuracy and high energy efficiency. The proposed model is based on a multivariate adaptive tuning algorithm that synchronously adjusts both the resistance (R) and the capacitance (C). The effectiveness of the model was verified through a three-level methodology that included theoretical modeling using Maxwell's equations, digital simulation using the MATLAB/Simulink environment, and practical testing with an accurate spectrometer. The results demonstrated a standard frequency accuracy of 99.9969% and a relative error of less than 0.0031%, surpassing the accuracy and energy consumption of previous studies. It also recorded a low power consumption of 785.42 microwatts, with an improvement of 15-40% compared to conventional designs. The system achieved rapid convergence in less than five iterations, three times faster than traditional algorithms, as well as superior thermal stability of ±0.001%/°C in the range of -20 to +70 °C. This algorithm represents a revolutionary advancement in the field of automatic tuning of analog systems, opening up new horizons for applications in low-power wireless communication (5G/6G), implantable medical electronics, and precision terminal computing. This design also enhances the principle of complete autonomy on the chip and improves efficiency in complex, realistic environments.

References

N. Mchirgui, N. Quadar, H. Kraiem, A. Lakhssassi, The Applications and Challenges of Digital Twin Technology in Smart Grids: A Comprehensive Review, Appl. Sci., 14 (2024) 10933. https://doi.org/10.3390/app142310933 O. Peckham, J. Raines, E. Bulsink, M. Goudswaard, J. Gopsill, D. Barton, A. Nassehi, B. Hicks, Artificial Intelligence in Generative Design: A Structured Review of Trends and Opportunities in Techniques and Applications, Designs, 9 (2025) 79. https://doi.org/10.3390/designs9040079 C. Yu, J. Gao, W. Cao, and X. Zhang, “AnalogGenie: A graph representation and generation model for analog IC topology,” in Proceedings of the International Conference on Learning Representations (ICLR), 2025, 1–12. C. Yu, J. Gao, W. Cao, and X. Zhang, “AnalogGenie-Lite: A lightweight graph-based generative model for large-scale analog IC topology generation,” in Proceedings of the International Conference on Machine Learning, 2025, 1–10. W. Li, C. Yu, J. Gao, X. Zhang, “LaMAGIC: Language Model-based Analog IC Generation with Circuit-Level Graph Representation,” arXiv preprint, arXiv:2407,18269 (2024) 1–14. https://arxiv.org/abs/2407.18269 Z. Huang, X. Zhang, C. Yu, “CktGen: Specification-Conditioned Variational Autoencoder for Analog Circuit Topology Generation,” arXiv preprint, arXiv:2410, 00995 (2024) 1–12. https://arxiv.org/abs/2410.00995 X. Liu, Y. Wang, L. Zhang, “GraphVAE for Circuit Topology Generation and Optimization,” IEEE Trans. Comput.-Aided Design Integr. Circuits Syst., 42 (2023)2801–2814. https://doi.org/10.1109/TCAD.2023.3245678 S. Sun, F. Wang, S. Yaldiz, X. Li, L. Pileggi, A. Natarajan, M. Ferriss, J.-O. Plouchart, B. Sadhu, B. Parker, A. Valdes-Garcia, M. A. T. Sanduleanu, J. Tierno, and D. Friedman, “Indirect performance sensing for on-chip self-healing of analog and RF circuits,” IEEE Transactions on Circuits and Systems I: Regular Papers, 61, 2014, 2243–2252. https://doi.org/10.1109/TCSI.2014.2333311. S. B. N. Premakumari, G. Sundaram, M. Rivera, P. Wheeler, R. E. P. Guzmán‏, Reinforcement Q-Learning-Based Adaptive Encryption Model for Cyberthreat Mitigation in Wireless Sensor Networks, Sensors, 25 (2025) 2056. https://doi.org/10.3390/s25072056 A. Aghanim, H. Chekenbah, O. Oulhaj, and R. Lasri, Q-Learning Empowered Cavity Filter Tuning with Epsilon Decay Strategy, Progress In Electromagnetics Research C, 140 (2024) 31–40. DOI: 10.2528/PIERC23111903. N. S. K. Somayaji and P. Li, Pareto optimization of analog circuits using reinforcement learning, ACM Trans. Des. Autom. Electron. Syst., 29 (2024) 1–14. http://dx.doi.org/10.1145/3640463 T. Braun, T. Korzyzkowske, L. Putzar, J. Mietzner, and P. A. Hoeher, Realtime spectrum monitoring via reinforcement learning—A comparison between Q-learning and heuristic methods, Sensors, 24, 2024, 573. https://doi.org/10.3390/s24020573 A. Aghanim, Q-learning empowered cavity filter tuning with epsilon-greedy strategy, Prog. Electromagn. Res. C, 140 (2024) 31–40. https://doi.org/10.2528/PIERC23111903 M. Asad, H. Arslan, and H. M. Furqan, “Adaptive Q-Learning Based Spectrum Tuning for Cognitive Radio Networks,” IEEE Access, 12, 2024, 45123–45135. https://doi.org/10.1109/ACCESS.2024.3387542 Y. Wang, X. Li, and Z. Zhang, Fast Convergence Q-Learning Algorithm for Frequency Control in Analog Systems, Electron. Lett., 60 (2024) 25–27. https://doi.org/10.1049/el.2023.0245 S. Kumar, R. Singh, and M. Sharma, Energy-Constrained Q-Learning for Real-Time Analog Control, IEEE Transactions on Circuits and Systems I: Regular Papers, 71, 2024, 1902–1914. https://doi.org/10.1109/TCSI.2024.3378214 K. Settaluri, A. Haj-Ali, Q. Huang, K. Hakhamaneshi, and B. Nikolić, AutoCkt: Deep reinforcement learning of analog circuit designs, in Proc. Design, Automation & Test in Europe (DATE), 2020, 490–495. https://doi.org/10.23919/DATE48585.2020.9116200 P. Gao, T. Yu, F. Wang, and R.-Y. Yuan, Automated Design and Optimization of Distributed Filter Circuits Using Reinforcement Learning, J. Comput. Des. Eng., 11 (2024) 60–76. https://academic.oup.com/jcde/article/11/5/60/7715024. S.-W. Hong, Y. Tae, D. Lee, G. Park, J. Lim, K. Cho, C. Jeong, M.-J. Park, and J. Han, Analog Circuit Design Automation via Sequential RL Agents and gm/ID Methodology, IEEE Access, 12 (2024) 104473–104489. https://dblp.org/pid/16/3415-1.html#j46 Z. Wang and Y. Ou, Learning Human Strategies for Tuning Cavity Filters with Continuous Reinforcement Learning, Appl. Sci., 12 (2022) 2409. https://doi.org/10.3390/app12052409 A. Aghanim, O. Otman, A. Oukaira, and R. Lasri, Optimizing Q-Learning for Automated Cavity/Combline Filter Tuning at 941 MHz, EPJ Web of Conferences, 326 (2025) 01006. https://doi.org/10.1051/epjconf/202532601006. C. J. C. H. Watkins and P. Dayan, Q-Learning, Machine Learning, 8 (1992) 279–292. https://doi.org/10.1007/BF00992698. M. Tokic, Adaptive ε-Greedy Exploration in Reinforcement Learning Based on Value Differences, in: KI 2010: Advances in Artificial Intelligence (LNCS 6359), (2010) 203–210. https://doi.org/10.1007/978-3-642-16111-7_23 Z. Zhao and L. Zhang, Analog Integrated Circuit Topology Synthesis with Deep Reinforcement Learning, IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 41 (2022) 5138–5151. https://doi.org/10.1109/TCAD.2022.3153437.

Highlights

Graph-VAE, Q-learning, and self-repair were integrated on-chip without external control On-chip learning operated at ≈15 nJ per 100 episodes, enabling ultra-low-power IoT use The Soft-Reset mechanism recovered from faults in under 30 ms, preventing Q-table drift Tuning errors ranged from 45–55 Hz, consistently within ±5 Hz and better than all baselines Dense on-chip ReRAM Q-tables enabled 1–2 Hz frequency bins for high-precision applications

DOI

10.30684/etj.2025.159870.1951

First Page

705

Last Page

715

Download

Share

COinS