Adaptive nonlinear model predictive path-following control for a fixed-wing unmanned aerial vehicle

Abstract

This paper presents an adaptive Nonlinear Model Predictive Control (NMPC) for the path tracking control of a fixed-wing unmanned aircraft. The objective is to minimize the mean and maximum error between the reference trajectory and the UAV. Navigating in a cluttered environment requires accurate tracking. However linear controllers cannot provide good tracking performance due to nonlinearities that arise in the system dynamics and physical limitations such as actuator saturation and state constraints. NMPC provides an alternative since it can combine multiple objectives and constraints which minimize the objective function. However, computational complexity is a major barrier to the real time implementation of the NMPC. An indirect approach which uses gradient descent methods can speed up the optimization but it is dicult to specify a proper termination condition of the optimization. If a decreasing cost metric is used, it can cause control input oscillations. We propose a new optimization termination metric which can remove the control input oscillations. This can be achieved by adding the actuator slew limit to the optimization termination requirement in addition to the cost monotonocity. In addition, we propose an adaptive NMPC which varies the control horizon according to the path curvature profile for tight tracking. Simulation results show that the proposed optimization algorithm can remove control input oscillations and track the trajectory more accurately than the conventional fixed horizon NMPC.