Vitiello, Federica Sensing and Data Fusion for Detect and Avoid and Multi-UAV Cooperation in Low Altitude Environments. [Tesi di dottorato]

Anteprima

Testo Vitiello_Federica_36.pdf Download (12MB) | Anteprima

Abstract

UAVs can undoubtedly bring benefits and advantages in many fields by completing tasks which are considered too dull, dirty and dangerous for human operators while offering a fertile technological ground for the development of innovative applications. Still, the safety of their missions, which foresee their integration in both controlled and uncontrolled civil airspace, is not keeping the pace with their increasing popularity, which is also pushing towards their exponential deployment in urban areas. This is all the more crucial if novel concepts for UAV based air transportation methods, such as AAM and UAM, are considered. In these latter cases, high volumes of operations are forecasted to occur mostly in low altitude conditions. In these frameworks, DAA and navigation strategies must be designed to enable safe flights of UAVs by providing detection and collision avoidance solutions, with the former, and reliable navigation, with the latter, thus finally unleashing the full autonomous potential of these novel platforms. These two aspects are the major focuses of this thesis. As far as DAA is concerned, the performance assessment of non-cooperative technologies for the detection and tracking of sUAVs flying in low altitude conditions is tackled. The sensing setup chosen for the application consists of visual cameras and low SWaP RADARs whose performance are analysed both on numerical and experimental bases. The tests clearly show limitations in the standalone implementation of the two sensors, paid in terms of absent distance estimates, for the camera, and coarse angular accuracy, for the RADAR. Innovative fusion solutions are therefore proposed and tested offline on experimental data during ground-to-air campaigns with a low-altitude-flying sUAV. The strategies provide improvements in the angular accuracy of the RADAR-based tracking estimates which likewise positively impacts conflict detection. The experimental approach used for the evaluation sheds a light over the effects of clutter on RADAR data, whose mitigation is either achieved with filtering and centroiding pre-processing steps or by adopting additional RADAR/visual fusion steps acting at detection level. The strategies tested can be also used in the context of AAM/UAM surveillance where performance of distributed sensing exploiting a network of ground-based RADARs and cameras are assessed. In this case, improvements in terms of track coverage can be achieved by adopting fusion solutions which take advantage of the spatial disposition of the sensors. An innovative solution based on a leader-helper approach is used to improve the track coverage of a single RADAR by injecting measurements from another RADAR comprised in the network. Nevertheless, when multiple RADARs operate with overlapping Fields of View, interference can occur. Therefore, a preliminary strategy to filter interference affected RADAR measurements is also proposed. Finally, in the framework of cooperation between multiple UAVs, a “chief-deputy” approach for calibration of onboard magnetometers is designed. The latter foresees a rotation of the chief platform around the deputy to capture relative positioning information by means of visual cameras and GNSS data processing. These can be used to evaluate both onboard and external magnetic disturbances which, if not accounted for, can spoil the quality of heading angle estimates. The strategy proves its efficacy in retrieving sub-degree pointing errors when ground-based and airborne targets are used as control points, thus exceeding the performance of both uncalibrated heading, which does not account for the disturbances, and navigation filter-estimated heading.

Downloads

Actions (login required)

Modifica documento