Sensing and control of a quadrotor using a visual inertial fusion method

Abstract

A visual inertial fusion method is proposed in this paper for the state estimation and control of a low-cost Unmanned Aerial Vehicle. A binary template matching algorithm is combined with a gradient based algorithm to compute optic flow (OF). The proposed OF method is capable of handling large displacement, illumination variation and gives subpixel accuracy. With a ground plane assumption, the Jacobian motion model is employed to solve for the unscaled linear velocity, which is fused with inertial measurements in an Extended Kalman Filter (EKF) framework to estimate metric speed and altitude. A number of flight tests have been conducted both indoors and outdoors to evaluate the performance of the proposed approach.