Selection of critical nodes in drone airways graphs via graph neural networks

Abstract

This Master Thesis has two distinct parts. The first one mod- els an application of Graph Neural Networks (GNN) for the identifica- tion of critical nodes in graphs that correspond to traffic networks. We call critical nodes those that can compromise the traffic flow in some subgraphs of the network. Specifically, the example data for the demon- stration corresponds to the Vienna subway network, hence the linear subgraphs correspond to the subway lines with intersections at some key subway stations. Those critical nodes relative to a subway line compro- mise the traffic flow at this line, therefore, we propose three GNN based approaches for the identification of such critical nodes, reporting encour- aging results. The second part of the Master Thesis illustrates the back- ground research work on drone airspace management and a discussion of how the reported results may have some relevance for this emerging dif- ficult problem. The main idea is that the urban airspace for drones, that may be carrying out delivery of either persons (aerotaxis) or goods, can be structured along airways that mimic the existing network of streets. The computational example explored in part one of the Master Thesis, thus, becomes relevant for the development of intelligent drone airspace management.