Kinematic Visual Servo Control of a Quadrotor aerial vehicle

Abstract

Visual systems are key sensors for control of small scale unmanned aerial vehicles. In this paper we investigate a range of image based visual servo control algorithms for positioning of flying vehicles capable of hover. The image based outer control loop for translation kinematics is coupled to a high-gain inner control loop that regulates translational velocities and full attitude dynamics. Zero and first order image moments are used as visual features for the control design. Perspective projection moments with suitable scaling along with a classical image based visual servo control design lead to satisfactory transients and asymptotic stability of the closed-loop system when the image plane remains parallel to the target. However, the system response may lack robustness for aggressive manoeuvres. In order to overcome this problem, several control schemes, based on spherical image moments, are designed and their performance is analysed. All designed control laws have been tested on a kinematic robotic manipulator to demonstrate the relative strengths and weaknesses of thedifferent image based visual servo control designs. The three most promising control algorithms have been successfully implemented on an autonomous aerial vehicle showing excellent performances in all three cases.