|Black-box Modeling for Aircraft Maneuver Control with Bayesian Optimization
Dohyung Kim, Hyun-Shik Oh, and Il-Chul Moon*
International Journal of Control, Automation, and Systems, vol. 17, no. 6, pp.1558-1568, 2019
Abstract : This paper proposes a new method of designing a data-driven controller for aircraft maneuver. Assuming that we do not have knowledge of the controller and the controlled aircraft, we propose a controller design with explorations of the control inputs and their responses from the aircraft. Speciﬁcally, we utilize Bayesian optimization (BO) with Gaussian process (GP) regression for black-box modeling of the aircraft responses from the explored controls, which are selected as samples to experiment with BO. We tested the proposed controller with a rigid six degrees of freedom(6DoF) nonlinear aircraft model by varying the kernel structures of the GPregressions. Our proposed method shows shorter ﬂight times and smaller deviations navigating ﬁxed waypoints compared to the tuned Proportional Integral Derivatives (PID) controller. The proposed controller can be an alternative to PID control, particularly when both controller structure and controlled plant model information are unknown.
Aircraft maneuver, Bayesian optimization, black-box modeling, full 6DoF dynamics, Gaussian process.