Аli Sultan Mayya1

1 St. Petersburg State Electrotechnical University "LETI" (St. Petersburg, Russia)
1 alimayya1357@gmail.com

The rapid development of telemedicine and wearable ophthalmological devices imposes increased requirements on automated optical positioning systems capable of ensuring accurate and stable pupil tracking without the use of external cameras or complex hardware components. Of particular relevance is the development of compact direct-drive control systems that can operate under limited computational and energy resources while remaining robust to parametric uncertainty and external disturbances.

The purpose of this study is to develop a state-space mathematical model of a direct-drive pupil tracking system and to perform a comparative analysis of control strategies based on a linear–quadratic regulator (LQR), a neural network controller using a multilayer perceptron (MLP), and a reinforcement learning controller implemented with the TD3 algorithm.

An electromechanical model of a lens positioning system driven by two DC motors was developed, taking into account both mechanical and electrical dynamics. Three control approaches for tracking predefined discrete pupil positions were implemented and analyzed. The results demonstrate that the LQR controller provides stable and predictable performance under nominal operating conditions. The MLP-based neural controller achieves comparable, and in some cases higher, tracking accuracy than the LQR (RMSE = 0.1089 versus RMSE = 0.1557). The reinforcement learning controller exhibits comparable accuracy along the X-axis (RMSE = 0.1140) but shows a significantly larger error along the Y-axis (RMSE = 1.5928), indicating incomplete convergence of the learned policy under the selected training configuration.

The results demonstrate the applicability of classical and neural network–based control methods in automated pupil tracking systems for wearable telemedical devices. The proposed model and comparative analysis are particularly relevant for compact direct-drive ophthalmological systems operating under constraints on the integration of additional sensors or external observation hardware, enabling reliable pupil positioning with a minimal measurement set.

Mayya Ali Sultan. State-space modeling and comparative analysis of automated control methods for a direct-drive pupil tracking system in teleophthalmology. Biomedicine Radioengineering. 2026. V. 29. № 3. P. 37–42. DOI: https:// doi.org/10.18127/ j15604136-202603-06 (In Russian)

1. Oftal'mologicheskaya pomoshch', narusheniya zreniya i slepota [Elektronnyj resurs]. Vsemirnaya organizaciya zdravoohraneniya. URL: https://www.who.int/health-topics/blindness-and-vision-loss#tab=tab_1.
2. Gadzharavala S.N., Pelkovski Dzh.N. Preimushchestva i ogranicheniya telemediciny. The Journal for Nurse Practitioners. 2021. № 2 (17). P. 218–221.
3. Pratiba V., Rema M. Teleoftal'mologiya kak model' okazaniya oftal'mologicheskoj pomoshchi v sel'skih i malodostupnyh regionah Indii. International Journal of Family Medicine. 2011. V. 2011. Article 683267.
4. CHzhou Yu. Bazovaya model' dlya obobshchaemogo vyyavleniya zabolevanij po izobrazheniyam setchatki. Nature. 2023. V. 622. № 7981. P. 156–163.
5. Shanmugam M.P. i dr. Fundus-s"emka s ispol'zovaniem mobil'nogo telefona: obzor metodov. Indian Journal of Ophthalmology. 2014. T. 62. № 9. S. 960 (In Russian).
6. Li S. i dr. Gibridnyj LQR-regulyator s pryamoj i obratnoj svyaz'yu na osnove prognoznoj modeli dlya upravleniya urovnem vody v otkrytyh kanalah. Journal of Hydroinformatics. 2024. T. 27. № 1. S. 33–50 (In Russian).
7. Snell R.S., Lemp M.A. Klinicheskaya anatomiya glaza. Oxford: Blackwell Science Ltd., 1998. XV, 344 s. ISBN 978-0-632-04344-6.
8. Chzhu CH., Fudzimura K., Czi C. Obnaruzhenie i otslezhivanie glaz v real'nom vremeni pri razlichnyh usloviyah osveshcheniya. Proceedings of the 2002 Symposium on Eye Tracking Research & Applications. New York: ACM. 2002. P. 139–144.
9. Bagua H. Sravnitel'nyj analiz PID- i LQR-regulyatorov dlya upravleniya skorost'yu dvigatelya postoyannogo toka. The Journal of Engineering and Exact Sciences. 2023. V. 9. № 12. Article 19429.
10. Fudzimoto S., van Hoof H., Meger D. Ustranenie oshibok approksimacii funkcij v actor-critic metodah obucheniya s podkrepleniem. arXiv preprint. 2018. arXiv:1802.09477.
11. Dargazani A. Glubinnoe obuchenie s podkrepleniem dlya intellektual'nogo upravleniya robotami: koncepcii, obzor literatury i perspektivy. arXiv. 2021. arXiv:2105.13806.
12. Precizionnyj mini-dvigatel' M10 s planetarnym reduktorom (DC 2,5–5 V, 92 ob/min) dlya medicinskogo oborudovaniya. URL: https://aliexpress.ru/item/1005005879748285.html?sku_id=12000034684854420