Fast convex model predictive trajectory planning for robot-assisted landings

Abstract Robot-assisted landings for aircrafts which are able to take off and land vertically offer the capability to speed up and automate landing and turnaround processes and are particularly advantageous for challenging landing conditions such as wind gusts or a moving landing platform. To perform an assisted landing, a precise and robust control approach is required, covering movement synchronization, catching and decelerating. In this paper a novel approach on model predictive trajectory planning for robot-assisted landings of unmanned aerial vehicles is designed, implemented and analyzed. The approach enables an unmanned aerial vehicle to be caught in hovering flight state using a 6-axis robotic arm with a positive-locking gripper. The trajectory planning utilizes a simplified robot model assuming an approximately linear mapping between Cartesian velocities and joint velocities, justified by a limited robot workspace. The aircraft’s flight trajectory is predicted using a Kalman filter to overcome the robot’s delay behavior. Catching tolerances between aircraft and gripper result in slackness for the synchronization process and are exploited during the trajectory planning. The suitability of the approach presented is analyzed and discussed in simulations, laboratory tests and several flight tests.