کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

Robust Output-Feedback Control using a Fusion of Model-Based and Data-Informed Models

نویسندگان :

Reza Norouzi¹ Khalil Alipour² Bahram Tarvirdizadeh³ Majid Sorouri⁴ Mohammad Ghamari⁵

1- Advanced Service Robots (ASR) Laboratory, Department of Mechatronics Engineering, School of Intelligent Systems Engineering, College of Interdisciplinary Science and Technology, University of Tehran 2- Advanced Service Robots (ASR) Laboratory, Department of Mechatronics Engineering, School of Intelligent Systems Engineering, College of Interdisciplinary Science and Technology, University of Tehran 3- Advanced Service Robots (ASR) Laboratory, Department of Mechatronics Engineering, School of Intelligent Systems Engineering, College of Interdisciplinary Science and Technology, University of Tehran 4- Dept. of Electronicl Eng. Maynooth University 5- Electrical Engineering Department California Polytechnic State University

کلمات کلیدی :

Robust Control،Output-Feedback Control،Data-Driven Control،State Estimation،LMI

چکیده :

The design of robust controllers for systems with uncertain dynamics is frequently impeded by the practical constraint of incomplete state measurement. Such robust output-feedback capability is especially critical in robotics and mechatronics, where only limited sensing is available and uncertain interactions with the environment can significantly affect performance and safety. This paper introduces an innovative framework for robust output-feedback control of discrete-time linear time-invariant (LTI) systems, reconciling theoretical full-state feedback designs with practical implementations. We start by making a strong state observer, and we make sure that the design of the observer keeps the estimation error stable across all systems in the model-based uncertainty set. Using this observer, we then create a new Hybrid-Uncertainty-Set (𝑺_𝑯𝑼𝑺). This set is the intersection of the a priori physical constraints and a newly created data-informed set. The data-informed set is based on available input-output data and carefully takes into account the state estimation error. In this enhanced uncertainty characterization, we present manageable Linear Matrix Inequality (LMI) formulations for the design of both robustly safe and optimal output-feedback controllers. The proposed methodology's effectiveness is confirmed via simulations of an autonomous vehicle lane keeping and a boost converter voltage control, illustrating its practical utility and elucidating the inherent performance trade-offs associated with output-feedback.