A ROS Based Automatic Control Implementation for Precision Landing on Slow Moving Platforms Using a Cooperative Fleet of Rotary-Wing UAVs

Abstract

In this paper we present an industrial implementation of an efficient method to solve the problem of the automatic precision landing for rotary-wing UAVs, ready to be used inside a cooperative fleet of drones. The realized software module and tests are part of a large industrial R&D Vitrociset project called SWARM: an AI-Enabled Command and Control (C&C) system, able to execute and review ISR missions for mini/micro cooperative fleets of heterogeneous UAVs. Preparatory to the presented results, it was the identification of a non-linear mathematical model as well as the realization of a robust PID-based control system capable of controlling a single drone of the fleet. A discrete-time Kalman filter was integrated and tested to estimate the possible displacement of the landing points, in order to improve the control law through predictive connotations in case of slow moving tags. The presented approach is featured by the balance between computational efficiency and versatility, in particular in the discovering stage of multiple and different AprilTag during the landing phase. The still under test software module uses the Open Source Robotic Operating System (ROS) libraries for both the acquisition of the data necessary to the control laws, and for the execution of the computer vision algorithms implemented for the precision landing. After analyses and simulations campaigns in a synthetic environment and multiple hardware in the loop (HIL) stress tests, the final prototype algorithm was deployed on a commercial-off-the-shelf mini-class UAV, demonstrating landing capacity on a fixed target with an error of less than ten centimeters; moreover, with slow-moving tags, appreciable tracking abilities emerged on sufficiently smooth trajectories. A special interface with the HIL flight controller was then integrated, with the capability of using its telemetry data for distributing them to all the members of the cooperative fleet, making it possible to access the real-time estimate of the states of each single drone, and making each one of them aware of the selected landing areas of the others, by navigation sensors data fusion with a five meters GPS precision.