Optimization-Based Design of a Model Reference Adaptive Attitude Controller for a Micro-Quadcopter
IEEE ACCESS, cilt.1, sa.1, ss.1, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 1 Sayı: 1
- Basım Tarihi: 2026
- Doi Numarası: 10.1109/access.2026.3752138
- Dergi Adı: IEEE ACCESS
- Derginin Tarandığı İndeksler: Scopus, Science Citation Index Expanded (SCI-EXPANDED), Compendex, INSPEC, Directory of Open Access Journals
- Sayfa Sayıları: ss.1
- Erciyes Üniversitesi Adresli: Evet
Özet
This paper presents the design, optimization, and comparative evaluation of a model reference adaptive control (MRAC) architecture for attitude control of a micro-quadcopter. The linearized hover dynamics are modeled as three decoupled double integrators, and a Lyapunov-based adaptive law on each axis guarantees bounded tracking error and stable parameter evolution. Rather than trial-and-error tuning, the reference model parameters and adaptation gains are determined simultaneously by sequential quadratic programming with a multi-start strategy, minimizing a composite cost combining integral and worst-case error criteria with control effort. A proportional-integral-derivative (PID) benchmark is tuned within the identical framework, isolating the effect of architecture from tuning quality. Simulation-based validation uses the full nonlinear coupled attitude dynamics, a 30 ms actuator lag, gust and turbulence torques, an offset payload, and sensor noise at the uncertainty of the onboard inertial unit. The PID benchmark attains superior nominal accuracy and, through its integral action, superior disturbance rejection. The adaptive controller is instead distinguished by its reduced sensitivity to inertia variation over the tested range of −50 % to +150 % and under in-flight payload change; by its stability at attitude excursions where the fixed-gain design loses it; and by the far smaller actuator demand measurement noise induces in it, where the unfiltered fixed-gain benchmark saturates for 60 % of the simulation. This reduced sensitivity is confined to parametric change. A constant trim moment, which the adaptive law has no integral action to null, degrades pitch tracking by 732 %, while costing the PID benchmark 1 %.