UAV Landing on General Moving Platforms Without Markers

Multirotor Unmanned Aerial Vehicles (UAVs) and their ability to hover and maneuver in the air make them the best vehicle for applications requiring quick package deliveries. Our assumed example system is a collaboration of a UAV and a land vehicle capable of storing multiple packages and charging the UAV’s batteries. Another example is quickly exploring some unknown region where the UAV scans the unexplored region and returns to the moving vehicle when required knowledge is gathered. An active research problem in such systems is about takeoff and landing the UAV on the moving vehicle. The landing problem, however, is quite challenging without unique markers like visual tags, hard-coded trajectory targets, etc. We present a markerless controller that uses only computer vision and calculates the optimum trajectory to land the UAV on the moving platform. This paper presents the work in a two-stage implementation process. First, the detection of a moving vehicle and finding the centroid by applying optical flow and contour technique on image feed from an RGB camera, and secondly, following the ground vehicle by matching the velocity by utilizing a visual servoing process followed by simple PID controller to descend in the smooth trajectory towards the landing point.