Título: Neural Network -Based DF -MRAC of Nonlinear Aerospace Systems: A Comparative Study of RBF, WNN, and ELM Models
Autores: Mayron Pantoja Cardoso, Aline Silva Lima & Lucia Valeria Ramos de Arruda
Resumo: This paper investigates the application of Derivative-Free Model Reference Adaptive Control (DF-MRAC) to the attitude control dynamics of finless rockets utilizing Thrust Vector Control (TVC). The proposed approach integrates an adaptive model that applies different neural network architectures for system uncertainty approximation to ensure precise trajectory tracking. Specifically, this work evaluates and compares the performance of Radial Basis Function (RBF) networks, Wavelet Neural Networks (WNN), and Extreme Learning Machines (ELM) in estimating uncertainties within the DF-MRAC framework. Simulation results demonstrate the effectiveness of the DF-MRAC approach in managing rapid dynamic changes and disturbances while maintaining stability and alignment with reference trajectories, and provide a comparative analysis of the RBF, WNN, and ELM networks in this context. The results highlight the potential of DF-MRAC for control in aerospace systems under uncertain conditions and offer insights into the suitability of different neural network architectures for uncertainty estimation.
Palavras-chave: Derivative-Free Adaptive Control (DF-MRAC); Thrust Vector Control (TVC); Rocket Attitude Control; Neural Networks; Radial Basis Functions (RBF); Wavelet Neural Networks (WNN); Extreme Learning Machines (ELM).
Páginas: 8
Código DOI: 10.21528/CBIC2025-1166563
Artigo em PDF: CBIC_2025_paper1166563.pdf
Arquivo BibTeX:
CBIC_2025_1166563.bib