[home] [Personal Program] [Help]

---

Session: Data-driven testing
Session starts: Thursday 20 June, 16:00
Presentation starts: 16:30
Room: Room 1.1

---

Leandro Lustosa (ISAE-SUPAERO)

Abstract:
Previous work [1] develops a real-time wing deformation estimation technique for Very Flexible Aircraft (VFA) stabilization and control. In particular, this capability supports the design of novel Stabilization Augmentation System (SAS) architectures for VFA that exhibit structural dynamic modes (e.g., first bending, twisting) coupling with traditional flight dynamics modes (e.g., short period, phugoid) and therefore require real-time measurements of structural deformation, on top of the traditional rigid-body flight mechanics estimates, for simultaneous active stabilization of rigid and flexible states. The method [1] uses an array of rate-gyros linearly installed along the wingspan and a single sighting device on the wing root that tracks a small number of lumped visual markers on fixed stations on the wing surface. While the estimation errors in standalone rate-gyro angular rates integration would diverge with time and the measurement availability bandwidth in standalone embedded computer vision tracking systems is low (when compared to typical attitude controller bandwidths), the previously proposed Extended Kalman Filter (EKF)-based fusion estimation algorithm yields in simulation high-bandwidth estimates with bounded error covariances in time. Additionally, the technique does not rely on structural dynamic models and thus can be more easily adapted from one aircraft to another. The current paper contributes to this effort by examining its hardware implementation challenges. In particular, the EKF requires accurate knowledge of the camera configuration (i.e., its position, tilt angle, and focal length). We show that ruler-based and camera resectioning measurement values yield poor results, are challenging to fine-tune by hand, and ultimately call for more sophisticated techniques. The present work proposes an optimization-based method for calibrating the critical parameters (i.e., intrinsic and extrinsic camera parameters) through the help of a motion capture system. It also studies its observability characteristics under different sets of free parameters to tune. The resulting fine-tuned EKF shows significantly reduced estimation errors over manual calibration and sets the groundwork for its use in uncertain real-world conditions. REFERENCES [1] Lustosa, L. R., Kolmanovsky, I., Cesnik, C. E. S., and Vetrano, F. (2021). Aided inertial estimation of wing shape. Journal of Guidance, Control, and Dynamics, 44(2), 210-219.