Vision-Based Control for Autonomous Quadrotor UAVs

Abstract

Recently, the interest in three-dimensional (3D) modeling of large targets such as ancient artifacts,buildings, or even cities has been increasing in many areas of research. While many methods exist for creating such models out of lower-dimensional data, the process of collecting this data still remains relatively unexplored. In this work, we propose using a team of Unmanned Aerial Vehicles (UAVs) for autonomous collection of photographic data for large-scale 3D modeling. We use this application as the motivation behind designing a vision-based control method for an autonomous team of quadrotor helicopters, or quadcopters. Specifically, we develop three types of control laws that give rise to two behaviors: a Waypoint-Navigation Controller that allows for Trajectory-Following Behavior and two Leader-Follower Controllers that allow for Formation Flight Behavior. We present Matlab simulations to demonstrate the capability of the proposed control design to drive a single UAV to reach desired poses and follow 3D trajectories as well as drive UAV teams to traverse desired paths while maintaining 2D formations. Furthermore, we implement the proposed controllers on a Parrot AR.Drone quadcopter. Experimental results demonstrate the UAV successfully navigating waypoints, following trajectories, and moving into formations with respect to simulated team members.