|Black-box Optimization of PID Controllers for Aircraft Maneuvering Control
Dohyung Kim* and Hyun-Shik Oh
International Journal of Control, Automation, and Systems, vol. 20, no. 3, pp.703-714, 2022
Abstract : In this paper, we propose a new method for auto-tuning an aircraft maneuvering controller using blackbox optimization. Assuming that we do not have a deep understanding of the complex nature and behavior of the controlled aircraft model, we propose a data-efficient Proportional Integral Derivatives (PID) tuning method with explorations on the aircraft responses from the sampled control inputs. More specifically, we utilize Bayesian optimization (BO) with Gaussian process (GP) to create a black-box model of the aircraft response corresponding to the sampled control parameters. We tested the feasibility and performance of the proposed data-driven tuning method with a six degrees of freedom (6DoF) nonlinear aircraft model. We also experimented with various GP kernel structures and hyperparameters to find the most suitable kernel function. Compared to the conventional tuning method, our proposed method shows shorter flight time and smaller deviations from the waypoints. The proposed data-driven tuning method can be an alternative to traditional model-based or model-free tuning methods, especially when the objective functions are expensive to evaluate and only a small number of experiments are available.
Aircraft maneuvering control, Bayesian optimization, black-box modeling, Gaussian process, PID controller.