Graph-based decision support for Air Traffic Management

Sparksee, our high-performance graph database, sits at the core of the solution: modelling flights, airspace, and delay causes as a single connected network rather than isolated data points.

When bad weather, capacity constraints, or unexpected events ripple through European airspace, a delay in one sector rarely stays contained, it propagates through a tightly coupled network of flights, airports, and regulations. Today’s tools are good at predicting that a delay will happen, but they offer little help comparing what actions a Network Manager, airline, or airport could take, and what each option would mean for the wider network.

GRAPH-DECIDE addresses this by representing the entire operational picture (flights, sectors, airports, ATFM regulations, delay causes, and capacity constraints) as a dynamic, attributed graph built on Sparksee. This graph-native representation is what makes it possible to apply Graph Convolutional Neural Networks (GCNNs), which learn from the relational structure of the network, how a delay in one place couples to capacity elsewhere in ways traditional prediction models simply can’t capture. On top of this, a multi-criteria decision layer translates those predictions into concrete, comparable options, weighing trade-offs like fuel cost, on-time performance, and passenger disruption.

Over 12 months, the project will deliver a proof-of-concept platform validated together with airline and airport experts, laying the groundwork for graph-based decision optimization to be matured within future SESAR initiatives.

This project reflects the kind of work Sparsity does best: putting our own graph technology, Sparksee, at the service of complex, safety-critical decisions that keep networks whether cars, cities, or continents of air traffic running more safely and reliably.