Precision Landing Tests of Tethered Multicopter and VTOL UAV on Moving Landing Pad on a Lake.

Autonomous take-off and landing on a moving landing pad are extraordinarily complex and challenging functionalities of modern UAVs, especially if they must be performed in windy environments. The article presents research focused on achieving such functionalities for two kinds of UAVs, i.e., a tethered multicopter and VTOL. Both vehicles are supported by a landing pad navigation station, which communicates with their ROS-based onboard computer. The computer integrates navigational data from the UAV and the landing pad navigational station through the utilization of an extended Kalman filter, which is a typical approach in such applications. The novelty of the presented system is extending navigational data with data from the ultra wide band (UWB) system, and this makes it possible to achieve a landing accuracy of about 1 m. In the research, landing tests were carried out in real conditions on a lake for both UAVs. In the tests, a special mobile landing pad was built and based on a barge. The results show that the expected accuracy of 1 m is indeed achieved, and both UAVs are ready to be tested in real conditions on a ferry.

Influence of Insufficient Dataset Augmentation on IoU and Detection Threshold in CNN Training for Object Detection on Aerial Images.

The objects and events detection tasks are being performed progressively often by robotic systems like unmanned aerial vehicles (UAV) or unmanned surface vehicles (USV). Autonomous operations and intelligent sensing are becoming standard in numerous scenarios such as supervision or even search and rescue (SAR) missions. The low cost of autonomous vehicles, vision sensors and portable computers allows the incorporation of the deep learning, mainly convolutional neural networks (CNN) in these solutions. Many systems meant for custom purposes rely on insufficient training datasets, what may cause a decrease of effectiveness. Moreover, the system's accuracy is usually dependent on the returned bounding boxes highlighting the supposed targets. In desktop applications, precise localisation might not be particularly relevant; however, in real situations, with low visibility and non-optimal camera orientation, it becomes crucial. One of the solutions for dataset enhancement is its augmentation. The presented work is an attempt to evaluate the influence of the training images augmentation on the detection parameters important for the effectiveness of neural networks in the context of object detection. In this research, network appraisal relies on the detection confidence and bounding box prediction accuracy (IoU). All the applied image modifications were simple pattern and colour alterations. The obtained results imply that there is a measurable impact of the augmentation process on the localisation accuracy. It was concluded that a positive or negative influence is related to the complexity and variability of the objects classes.