Velocity Tuning for Air Traffic Control
Summary
In this thesis we study air traffic conflict resolution via speed control. Each aircraft is assumed to have a fixed route and no speed change during its route.
We use a fast and intuitive approach to the conflict resolution which finds an exact solution using velocity tuning, a concept from robot motion planning.
The choice in priority for a pair of conflicting aircrafts introduces a binary branch in the solution space. We solve the problem using linear programming with branching constraints and reduce the search space by using a branch cutting solution method.
In this study we will reveal that the order in which the branching constraints are entered in a linear solver influences the calculation time.
When using a random initialization of our branching constraints, the branch cutting solution method has outliers of a factor 100 above its average runtime.
We introduce a novel method for the dynamic reordering of the decision tree, which removes these outliers.
Furthermore, we show that changing the initial order of the branching constraints can have an effect similar to the dynamic reordering, and suggest and compare multiple initial sortings.
Our computational study reveals that using an initial sorting based on geometrical properties of the branching constraints can result in a substantial reduction in computation time, enabling the use for real time applications.
We will run our tests using synthetic scenarios in a single-layered air sector based on CTA Amsterdam South 1.
With our proposed solution method a realistically sized problem can be solved within seconds, and a extremely complicated scenario with more than 100 conflict points can be solved within minutes.