Vision Based Robust Autonomous Landing of a Quadrotor on a Moving Target

Autonomous landing is the most difficult maneuver in a quadrotor mission because of challenges such as external disturbances and localization error, both of which cause deviation from the desired trajectory and may result in a crash. In this paper, we present a guidance framework that allows a multirotor to land on a moving target accurately. The framework utilizes onboard vision to detect and estimate the landing target parameters. We then analyze the effects of environmental disturbances, and abrupt changes in the motion of the landing target to allow us to investigate them from a computer vision perspective. This will aid in development of a robust autonomous landing strategy for a moving target. We present robustness results through outdoor hardware experiments.