Robust post-stall perching with a fixed-wing UAV

Abstract

Consider a bird perching on a branch. In the presence of environmental disturbances and complicated fluid flow, the animal exploits post-stall pressure drag to rapidly decelerate and land on a perch with a precision far beyond the capabilities of our best aircraft control systems. In this thesis, I present a controller synthesis technique for achieving robust, post-stall perching with a fixed-wing Unmanned Aerial Vehicle (UAV). Using sum-of-squares (SOS) programming and building on a novel control synthesis approach known as LQR-Trees, I demonstrate the ability to improve the robustness of the post-stall perching manuever to variable initial conditions, modeling error, and external disturbances. I achieve this by developing methods for carrying out rigorous robust verification of the nonlinear aircraft model along a time-varying nominal perching trajectory in the presence of both dynamic and parametric uncertainty. I also present methods for carrying out stochastic verification in the presence of Gaussian acceleration uncertainty and adaptive control techniques for improving controller performance in the presence of parametric uncertainty. Using the robust verification techniques for dynamic uncertainty, I proceed to generate a robust LQR-Tree controller and test that controller on real hardware using a small 24 inch wing span glider and a Vicon motion capture studio. The experiments show successful perching for 94 percent of the 147 flights launched between 6 and 8 m/s. I further demonstrate robustness to initial pitch variations by launching the glider by hand and showing repeatable successful perching results. Following these experiments, I then build a magnetic field sensing system capable of estimating the position of the perching UAV using the magnetic field generated by a powerline. I demonstrate on hardware that the aircraft is still capable of successfully executing a closed-loop perching maneuver indoors using the lower fidelity state estimates. I then move the entire experiment outdoors and begin testing the UAV's perching performance in the presence of wind gusts. Very quickly, it becomes apparent that the glider can not reliably perch in wind. To address this, I describe a control approach for mitigating the effects of wind on aircraft performance and propose an experiment for testing this approach.