Abstract
Problems such as urban transportation networks require dynamic routing optimization that adapts to rapidly changing traffic conditions caused by congestion, accidents, and demand fluctuations. Classical routing algorithms such as Dijkstra and A* are computationally efficient but require complete path re-computation with each change, making them unsuitable for real-time dynamic environments. Classical Graph Neural Networks (GNNs) capture spatial dependencies effectively but encounter the over-smoothing problem, where increasing network depth to model long-range dependencies causes node representations to become increasingly similar, resulting in loss of discriminative information necessary for effective routing decisions.
This PhD research addresses these fundamental limitations through the development of Hybrid Quantum Graph Neural Networks (HQGNNs) that leverage quantum entanglement to capture non-local correlations across the network without requiring deep message-passing architectures. By embedding graph signals into quantum state space, HQGNNs enable the representation of complex long-range congestion propagation patterns while preserving local discriminability—a combination that is fundamentally difficult to achieve in purely classical architectures.
The resulting framework in the scope of this thesis will be evaluated with City of Antwerp data. The city of Antwerp, with its dense urban core and complex multi-modal transportation infrastructure, serves as an ideal testbed for evaluating HQGNN performance in real-world scenarios. Antwerp's network exhibits characteristic challenges including recurring bottlenecks at key intersections, variable traffic patterns influenced by port activity, and the need for coordination between private vehicles, public transit, and commercial freight. Applying HQGNNs to Antwerp's transportation system enables validation of the proposed approach against traffic data while demonstrating practical scalability and adaptability to actual urban routing constraints.
Researcher(s)
Research team(s)
Project type(s)