PhD defense - Md Siddiqur Rahman

5 May 2022

On May 3, 2022, Md Siddiqur Rahman successfully defended his dissertation in Toulouse at ENAC. Prof. Schultz was invited as a member of the PhD committee to evaluate the results obtained. The thesis is entitled "Deep Learning for En-Route Aircraft Conflict Resolution - two complementary approaches" and elaborates on two different approaches for conflict resolution in air traffic.

A situation is called a conflict when aircraft along their route cannot maintain the required minimum separation from each other. Earlier models for assisting controllers in conflict resolution were based on mathematical and statistical approaches. Recent successes of deep neural network models in various fields have reignited research interest in automated aircraft conflict resolution. Conflicts are resolved by air traffic controllers giving instructions to pilots to change routes based on different aircraft positions and flight paths. 

The first model uses a neural network that provides a classification with multiple labels as output (conflict resolutions by course changes). The trajectories (time, latitude, longitude, altitude, and heading) of all aircraft involved in the scenario are used as input parameters. Compared to other machine learning models (Support Vector Machines, K-Nearest Neighbor Classifier, Logistic Regression) that use multiple single-label classifiers, the developed model achieved the best results.

However, since this model is not suitable to handle a variable number of aircraft efficiently, a convolutional neural network was designed in the next step (also multi-label classification). For this purpose, the input data (conflict scene) were graphically processed and displayed.