In this paper, we propose a vision-based control system which enables a multi-rotor unmanned aerial vehicle (UAV) to track an unmanned ground vehicle (UGV) and land on it by using mainly the input of an integrated vision sensor without external localization systems. Our solution is able to generate and follow agile approaching maneuvers in which the target vehicle may leave the field of view of the UAV's vision sensor. This is particularly relevant in scenarios where external localization systems such as GPS are not available or not reliable. In our approach, the UAV observes the movement of the UGV, predicts its motion and generates a smooth approach trajectory to the predicted position. A 6-DOF controller in cascade form is used to track the trajectory, which can lead to movements were the UGV is lost from the field of view (FOV), and then once the UGV is back on the FOV a normal visual servoing tracking is used for landing. The UAV states required for the control law were obtained from an Extended Kalman Filter in combination with a Mahony complementary filter using only internal sensors. The control law and the landing state machine were implemented in ROS and the simulations were developed on Gazebo based on the Rotors simulator.
Vision-Based UAV Landing on a Moving Platform in GPS Denied Environments Using Motion Prediction
2018-11-01
462143 byte
Conference paper
Electronic Resource
English
Vision Based Optimal Landing On a Moving Platform
British Library Conference Proceedings | 2016
|Monocular Vision based Navigation in GPS-Denied Riverine Environments
British Library Conference Proceedings | 2011
|VISION-BASED AUTONOMOUS UAS LANDING ON A STOCHASTICALLY MOVING PLATFORM
British Library Conference Proceedings | 2021
|