Balanced Truncation for UAV Longitudinal Dynamic Models Preserving Fidelity and Enabling Efficient Embedded Control
Year: 2025
Type: article
Abstract: This study investigates the effectiveness of Balanced Gramian Truncation for reducing the order of a fourth-order single-input single-output longitudinal UAV model, with the objective of retaining essential dynamics while alleviating computational burden for real-time control. Reduced models of orders three, two, and one are derived and quantitatively assessed under diverse excitations including step, impulse, ramp, sinusoidal, and mixed signals. The third-order reduction achieves a maximum step error of 0.01396, RMS error of 0.00927, and H∞ norm of 0.01409, while the second-order model yields corresponding values of 0.01245, 0.00323, and 0.01829; in contrast, the first-order model suffers severe degradation with a step error of 1.4083, RMS of 0.67589, and H∞ norm of 1.516. For impulse input, the third- and second-order models limit RMS errors to 0.00868 and 0.01906, whereas the first-order model reaches 0.17814. Ramp tests highlight the accuracy of the second- and third-order models with maximum errors below 0.08, compared to 4.60 for the first-order case. Frequency-domain analysis shows maximum magnitude deviations of 0.01406 and 0.00882 and phase deviations of 20.13° and 89.57° for the third- and second-order models, whereas the first-order model exceeds 1.5 in magnitude error and nearly 90° in phase shift. These results consistently indicate that two- and three-state surrogates preserve the fidelity of the original dynamics while reducing state dimension by up to 75%, thereby offering a practical foundation for efficient controller synthesis and embedded UAV implementation.
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7160949649
