Intermediate layer: Alternative Scenarios

The layers of patterns presented so far apply to the current situation (what actually took place). We now introduce a higher layer of patterns, related to "what could have taken place?". Keep in mind that our layers are somehow similar to levels of language; they will allow us to express things with greater accuracy. So here, we want to express notions related to the intent, such as:

Let's clarify now where the intent intervenes. For an air traffic controller, an important part of the work lies in the identification of future conflicts between aircraft, so the intent of the aircraft refers to how the controller anticipates future conflicts. In other words, we can reproduce the conflict detection process of the air traffic controller by extrapolating the intended future of the aircraft, and by checking where these extrapolated trajectories will be conflicting.
Similarly, the pilot (aided by the onboard avionics) also has intents: The aircraft electronic equipment informs him about the optimal profile, and he intends to "stick" to this profile.
So far we have identified in our methodology two kinds of alternative scenarios:

These scenarios are actually patterns in the sense that we will use them as building bricks for higher level patterns (such as workload, as we will see in the top level layer). These two scenarios are detailed below.

The horizontal "intended route" scenario

An introductory example of a horizontal scenario

We start with an introductory example, illustrated on the left Figure. We see an aircraft trajectory, together with its flight plan route. We consider the initial position of the aircraft, shown on the bottom of the Figure (the black disk). In reality, the aircraft did not follow its intended route; we see that it was instructed a Direct towards a remote Waypoint of the flight plan route.
The intended route of the aircraft, from the initial position, is in the Figure. We make use of the lower level layer of pattern (the identification of controller's instructions) where we identified aircraft maneuvers with regards to the flight plan route, and from the initial position, we "know" that we are in Direct towards a Waypoint. Therefore, the intended route is to capture this Waypoint, and from this Waypoint to follow the flight plan route, as illustrated by the red profile in the Figure.
In this introductory example, the extrapolation of the profile is made easier by the two following particularities:

  • The straight line extrapolated from the initial position intersects the flight plan route;
  • this intersection corresponds to a Waypoint of the flight plan route.

A complex example of a horizontal scenario

The previous example corresponds to the easiest configuration due to the two particularities mentioned above. In practice, many configurations are likely to occur, and it is sometimes necessary to define rules to construct the intended route. The Figure to the left illustrates such a case, where initially the aircraft has diverged from its intended route (the flight plan including the STAR, and the APCH).
The notion of intended route, here, has no meaning since the deviation from the intended route resulted obviously from a controller's instruction, and similarly, it is a controller's instruction that will bring back the aircraft on its route. So we need to guess under which logic, for this individual aircraft, the controller might operate. Here we have designed the following rule: we considered how far the controller might let the aircraft deviate, and this "how far" is defined by pairs of consecutive waypoints of the flight plan and STAR route. We choose the "pair of consecutive waypoints" which intersects the deviated route by the longest distance D from the initial position, as . Then, we construct the horizontal scenario as "going forward during distance D," and then resuming the flight plan route followed by the STAR and the APCH, as .

In summary, the notion of intended route requires specific rules depending on the initial configuration. This initial configuration has been captured by lower level patterns, so we are here in a process similar to deep learning, the only difference being that we have to define our own rules, whereas Deep Learning creates rules through a training process with a predefined data sample.
In other words, we stick to the paradigm of multilayer pattern identification, but we implement this paradigm by defining our own rules, instead of letting the system discover these rules through training. This is possible, of course, due to the fact that the rules that we have to determine are pretty straightforward and applicable to specific operational situations, and that these operational situations have been captured in lower level patterns.

The vertical "optimal profile" scenario

This optimal scenario implements an optimized descent profile in terms of fuel consumption. The rationale for this optimization is to reduce the aircraft Thrust, and here again the rules to apply depend on the operational configuration. We will limit ourselves to the case of landing aircraft, which suffices to illustrate the principle. However, the same process could be applied to landing aircraft.

General Guidelines for Optimizing the Landing Aircraft Profile

As depicted in the left figure, we examine a cruising aircraft (at altitude $H$) preparing to land (at altitude 0). A key principle is that higher cruising altitudes lead to more economical flights. This is derived from Equation (2), where in cruise (constant speed), Thrust equals Drag (since $\frac{\mathrm{d}V}{\mathrm{d}t}=0$ in this equation). Our performance model confirms that Drag decreases with altitude.
Another observation from the Performance model is that the aircraft must lose energy during the transition from cruise to landing, decreasing both in altitude and speed. Therefore, the optimal descent profile is as follows:

  • The descent (to lose energy);
  • the aircraft continues to cruise until it can descend with this lowest permissible thrust.

As seen, the vertical scenario's implementation heavily relies on the Performance model. The lowest permissible thrust, often termed Idle Thrust, can be estimated from air traffic data files or as a ratio of maximum thrust (this ratio can also be estimated from real data files). Without delving into overly technical details, we now present an illustrative implementation of an optimal descent vertical scenario, utilizing the performance model.

Additional Guidelines (Including Operational Context)

When determining the optimal descent profile, the operational context with specific constraints must be considered. Firstly, the speed profile remains unchanged, captured by a lower-level pattern, the speed schedule (expressed in altitude and speed, as illustrated to the right). Secondly, the optimal descent profile must adhere to vertical constraints defined for the flight. These constraints apply to specific Waypoints in terminal procedures (STAR and APCH), expressed as FL (below/above/equal to) XXX on the Waypoint.
The left figure illustrates the slope diagram of a flight alongside vertical constraints. Two tooltips are shown, one for FL $\leq$ 270, and one for FL=100. The aircraft (at FL=120), below the vertical constraints.
The optimal descent profile . It complies with the vertical constraints and is above the current profile.