Predictive Control with Constraints: Application to Lateral Flight Control of a Fixed-Wing Unmanned Aerial Vehicle

Abstract

This paper explores the feasibility of using the predictive control strategy to design the lateral flight controller for a fixed-wing unmanned aerial vehicle (UAV). One of the primary features of predictive control is the control theory’s ability to consider the future implication of current control actions by explicitly computing the predicted system response over a finite horizon, i.e. the receding horizon control strategy. The optimal control input is then determined by optimizing some measure of predicted performance. Another important advantage that comes with the receding horizon strategy is the ability to respect constraints imposed upon the control inputs, state and output variables. The systematic handling of constraint violations is achieved by solving a constrained optimization problem, e.g. the quadratic programming (QP) problem at each sampling instant. The capability makes the predictive control highly appealing in the design of flight controllers for aircraft. The control surfaces of a fixed-wing aircraft are often physically limited in terms of deflection angle and angular speed. Furthermore, the operation of an aircraft often needs to be confined within a safe flight envelop which means it is vital to ensure some flight parameters, e.g. the aircraft’s roll rate, do not exceed the corresponding safety limits during flight. In this work, the predictive control strategy, formulated as a constrained optimization problem is implemented on the lateral flight control of the Spoonbill UAV – an existing UAV research platform of the Remotely-Piloted Vehicle and Microsatellite Research Laboratory, National Cheng Kung University, Taiwan. The synthesis of the lateral flight controller for the Spoonbill UAV within the framework of constrained predictive control is presented and discussed. Finally, successful flight test results demonstrate the viability of the predictive control strategy and show that despite the heavy computational requirement due to the online constrained optimization algorithm, it is both feasible and practical to implement the predictive control strategy on a high-bandwidth control application such as the inner-loop control of a fixed-wing UAV.