Autonomous Landing of a Low-Cost Quadrotor on a Moving Target

This paper addresses the problem of a low-cost quadrotor on its autonomous landing on a moving target. Since a variant of methods are proposed, there still remains troublesome in practice. In this paper, a simple yet effective autonomous landing system is designed which tries to tackle three practical problems: robust target tracking, re-discovery when guidance information unavailable and effective scheduling on low-cost platform. In detail, unlike most methods mainly focus on either vision or expensive Real-time kinematic (RTK) based techniques, the designed system firstly fuses visual and low-cost position data in Kalman filter manner to robustly and economically improve the tracking performance. Secondly, a heuristic local random searching strategy is applied to cope with the situation where no guidance information is available. Thirdly, a behavior tree based method is utilized to help the target detection, tracking and landing tasks run more efficiently on a computationally limited quadrotor. For demonstration, experimental results are conducted to show the performance and feasibility of the proposed method.