Szczegóły publikacji
Opis bibliograficzny
RL-based direct control vs adaptive feedback controller for nonlinear MIMO systems / Patryk BAŁAZY, Krzysztof LALIK // W: MMAR 2025 [Dokument elektroniczny] : 29th international conference on Methods and Models in Automation and Robotics : 26–29 August 2025, Międzyzdroje, Poland : technical papers : on line proceedings. — Wersja do Windows. — Dane tekstowe. — Piscataway : IEEE, cop. 2025. — ( International Conference on Methods and Models in Automation and Robotics ; ISSN 2835-2815 ). — USB ISBN: 979-8-3315-2648-1. — Print on Demand(PoD) ISBN: 979-8-3315-2650-4. — e-ISBN: 979-8-3315-2649-8. — S. 312–316. — Wymagania systemowe: Adobe Reader. — Bibliogr. s. 315–316, Abstr.
Autorzy (2)
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 162293 |
|---|---|
| Data dodania do BaDAP | 2025-09-12 |
| Tekst źródłowy | URL |
| DOI | 10.1109/MMAR65820.2025.11150783 |
| Rok publikacji | 2025 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Wydawca | Institute of Electrical and Electronics Engineers (IEEE) |
| Konferencja | International Conference on Methods and Models in Automation and Robotics 2025 |
| Czasopismo/seria | International Conference on Methods and Models in Automation and Robotics |
Abstract
This paper presents a comparative analysis of two reinforcement learning (RL) approaches for controlling nonlinear multiple-input multiple-output (MIMO) systems: direct control, where the RL agent outputs control signals applied directly to the system actuators, and adaptive feedback control, where the RL agent dynamically adjusts the parameters of a feedback controller based on real-time measurements of the system’s output. We discuss the theoretical foundations underlying each approach, emphasizing the potential advantages in handling complex nonlinear interactions inherent in MIMO systems, as well as the practical implementation challenges that may arise. Specifically, the direct control approach offers flexibility and a significant capability to manage intricate nonlinear dynamics but encounters critical challenges concerning stability guarantees, interpretability of learned policies, and computational complexity in real-time scenarios. In contrast, the adaptive feedback control method enhances interpretability and stability by integrating RL into established classical control frameworks, yet its overall performance heavily depends on the effectiveness of the selected feedback controller structure. By analyzing and highlighting the strengths and limitations of both approaches, this study aims to deliver valuable insights into the practical applicability of reinforcement learning techniques in the control of nonlinear MIMO systems, and to propose potential avenues and strategies for future research and development efforts within this increasingly significant area.
