Design and validation of specific payloads for unmanned aerial systems to perform contact inspection tasks in large infrastructures

Resumo

Nowadays UAV (Unmanned Aerial Vehicles), as known as drones, are widely used for different kinds of infrastructure inspection. The main advantages of the use of these vehicles are their facility to reach and get information of hard accessible areas. In this way, inspections that used to suppose the installations of scaffolds or the use of cranes, can now be performed by these vehicles, making these tasks more efficient and cost-effective. Most of the inspection tasks that are already carried out by UAVs use remote sensing payloads, such as different kinds of cameras or LiDAR (Light Detection and Ranging) sensors. Some other inspections use sensors that have to be in contact with the structure in other to make reliable measurement, such as resistivimeters or ultrasonic sensors. Usually, these sensors are used to detect internal issues. The next step in the use of UAV in infrastructure inspection is the development of UAVs to perform contact-based inspections. The main objective of this thesis is to optimize contact inspection tasks in large structures, such as bridges or dams, using UAV technology. In order to perform these inspections, a specific UAV payload was developed. This payload is completely independent of the flight controller, in this way the payload is adaptable to different UAV configurations and flight controllers. Also, it was developed to be adaptable to different contact-based sensors. The neighbourhood of a large structure can be considered a GPS-denied zone, so an alternative positioning system is needed. To solve this problem, an on-board positioning system has been designed, which consist of two measurement units in the front of the UAV. In total, four payloads were developed, testing each one and using the experimental data to improve the payload performance in each iteration. Also, this thesis goes in depth with the autonomous navigation in indoor environments, developing two different path planning algorithms focused in two different scenarios: navigation on an unknown environment and in a known environment. The first scenario refers to navigation in an environment about which the system has no information previous the navigation. The navigation in this case is focused on the optimization of the scan positions to get the 3D information of the environment. It is known as the NBV (Next Best View) problem. The second scenario is the navigation in a known environment, where the system has the complete information about the environment and its 3D geometry. In this case, the system uses this information to calculate the paths to navigate from one position to another. In this field a path planning algorithm was developed that uses a point cloud of the environment to calculate the navigation graph, that contains the information of the rooms’ connection, using it to calculate the rooms that the UAV have to cross to navigate from one room to another. All the developments presented in this thesis have been tested in real case environments, showing the obtained results and the conclusions of each development. This Doctoral Dissertation is structured as a compendium of articles and contains eight peer reviewed scientific articles, six of them published in different scientific journals, indexed on the Journal Citation Report (JCR), and two are double-blind peer reviewed conference proceedings.