This research project investigates the application of vision-based methods for controlling the motion of quadrotor UAVs. The desired task is to adjust the position and orientation of the UAV using visual measurements from an observed static object. The overall control strategy is based on a combination of classical image-based visual servoing technique (IBVS) and PID control algorithm. The PID controllers are designed to control the longitudinal velocity, lateral velocity, rate of change of altitude, and yaw rate while the IBVS approach is adopted to generate the corresponding reference signals that will guide the UAV to complete the required task. In contrast to most conventional vision-based control methods that rely on the reconstruction of the 3D Cartesian coordinates of the target from image features in the camera frame, the IBVS method uses visual measurements directly in the control law design. The elimination of complex target pose estimation schemes makes the system more efficient and less sensitive to camera calibration errors. Even though IBVS theory is well established in the literature, the method has been applied to a few real plants with particular emphasis on six degree-of-freedom robotic systems. The successful implementation of IBVS in UAV applications is less common and rarely reported in underactuated quadrotor systems for which only four of the six degrees-of-freedom can be controlled independently. The goal of this paper is to apply the IBVS method to control a quadrotor vehicle and test the model through numerical simulations.
The main objective in designing an IBVS controller is minimizing the error between the current image features and the desired image features in the image plane. The image features are the projected pixel coordinates of points of interest from the observed target. The IBVS control design provides reference signals for the translational and rotational velocity components of the camera (vx, vy, vz, ωx, ωy, ωz) that ensure exponential decrease of the error. Assuming the relationship between the camera frame and the quadrotor body frame is known, the camera velocity signals can be transformed to desired velocity signals for the quadrotor. Due to the underactuated properties of the quadrotor system only four of the six reference signals are used directly for guidance (vx, vy, vz , ωz). The angular rates (ωx, ωy) are coupled with the x and y velocities. The tracking of the generated reference velocities is achieved through several inner PID control loops. The combined IBVS-PID controlled quadrotor system is shown in Figure 1. The simulation results showing the time history of the quadrotor states are also presented in Figure 2, where the desired task is to guide the quadrotor from the initial position and orientation (x,y,z) = (0.5,0.5,2) m and (φ,θ,ψ)= (0,0,5) deg to the final state (x,y,z) = (0,0,2) m, (φ,θ,ψ)= (0,0,0) deg.
