The core of the tactical layer is a routine which, given some flight trajectories planned by the air companies, simulates the real course of the flights, detects the conflicts between aircraft and solve them. Beneath is shown a replay of the conflict solving procedure by the routine is an Italian ACC.
And here is another zoomed view. Colours code altitudes (also shown in Flight Levels besides the aircraft here).
The tactical layer is able to simulate different scenarios, in particular free-routing scenarios. “free-routing” refers to the ability of pilots to fly “straight” (along the great circle), instead of following pre-defined navigation points. Free-routing procedures are already implemented in different areas of Europe, but only for airspaces with a low degree of complexity. It will be a challenge for main dense traffic areas to switch to free-routing unless the conflict-solving procedures are changed. The tactical layer is here to help.
Indeed, the tactical layer features some kind of “super-controller” able to solve conflicts efficiently in a given area. It has a limited horizon and have noisy information on the real positions of the aircraft, hence modelling a real controller in this regard. This way, we were able to catch some interesting effect, like the fact that with free-routing, the number of conflicts will drop a lot. More importantly, the relationships between number of potential conflicts and actual ones change when one goes to free-routing, as shown below.
(the figure is on its way)
The colour codes the efficiency of the trajectories, i.e. their best potential length over their actual one. An efficiency of 1 corresponds to the pure free-routing scenario. In abscissa, we plotted the number of potential conflicts (conflicts predicted based on the strategic trajectories) and in ordinate, we plotted the number of real conflicts (during the tactical phase).
In fact, given some pre-deconflicted trajectories, the free-routing seems more fragile to imperfect information, since the number of flights increases more rapidly for high efficiency values.
We also compared the behaviour of the controller to what would be expected from a human controller, based on complexity metrics. The results show that our controller reacts in much the same way than the human, in the sense that what is complex for the human is complex for the algorithm and the other way around. This way, the complexity computed in the free-routing scenario could be used as a basis for the possible reaction of humans to the corresponding situations. Moreover, the algorithm could help the controllers to make decisions in these situations.