Pose Estimation for Multicopters Based on Monocular Vision and AprilTag

This paper deals with the problem of pose estimation for multicopters by using a downward-looking monocular camera and AprilTag. First, successive AprilTag images are captured by the downward-looking camera and the relative pose between them is estimated using a vision algorithm directly. Then, a Kalman filter combined with a constant-velocity process model is adopted to improve the accuracy and robustness of pose estimation. Experimental results show that the proposed pose estimation method can satisfy closed-loop control requirements of multicopters.