Model predictive flight control using adaptive support vector regression

This paper explores an application of support vector regression (SVR) to model predictive control (MPC). SVR is employed to identify a dynamic system from input–output data, and the identified model is used for predicting the future states in the MPC framework. In order to deal with constant and dynamic uncertainties, an online adaptation algorithm is designed using the gradient descent (GD) method and the adjusted SVR model is fed to the MPC optimizer. In addition, the convergence property of the adaptation rule and the condition for the convergence of the MPC optimization are discussed using discrete-time Lyapunov stability analysis. Finally, the proposed approach is applied to identification and flight control of an unmanned aerial vehicle (UAV) lateral dynamics.