Context

To ensure capacity in control centers, the airspace has been designed to streamline and separate traffic, thereby guaranteeing the required safety levels. Traffic is thus structured vertically (flight levels), laterally (airways), and temporally (speed regulation when approaching TMAs, or “Miles in Trail” in the United States). If we consider, for example, the optimal vertical profile of a commercial aircraft (see Figure 1), it does not feature flight levels but consists of a continuous climb and descent (shown in green in the figure 1). However, the operational profile is constrained by flight levels to assist air traffic controllers in separating aircraft vertically. In the horizontal plane, aircraft are also required to follow a network of routes that structures the traffic for controllers, helping to ensure sector capacity. This network, together with sectorization, reduces the number of crossing points and localizes them in the central part of the sectors to improve controller efficiency. At night, when traffic is light, it is common for controllers to give direct routings to aircraft to improve flight performance. When traffic increases, controllers tend to return aircraft to predefined routes and sometimes impose speed constraints for regulation in subsequent sectors (en-route or TMA). If all traffic were operated to optimize kerosene consumption (wind-optimized routes, optimal vertical profile, optimal speed profile), it would be impossible for controllers, using current tools, to manage this type of traffic efficiently, and the airspace would quickly become under-capacitated. Control centers are essentially dimensioned for peak traffic periods, and at certain times, some sectors within the center experience overcapacity. In these sectors, it would therefore be possible, using a decision support tool, to propose environmentally efficient trajectories at the pre-tactical level. Using a catalog of trajectories proposed by airlines, ranked from least to most constrained (with or without flight levels, wind-optimized, etc.), and taking into account the capacities of the sectors crossed by these trajectories, it is then possible to optimize all trajectories in order to minimize the environmental impact of traffic while respecting the capacity of the airspace. These aircraft would then have an optimized 4D profile but would also be more complex to manage and should be weighted with a higher complexity coefficient. In addition to trajectory optimization, we propose a joint optimization of airspace to favor CCO (Continuous Climb Operations) and CDO (Continuous Descent Operations). The idea is to move towards DAC level 5 and DAC 6 (DAC : Dynamic Airspace Configuration), also incorporating green ops aspects—i.e., optimizing dynamism to achieve a good balance between environmental (ENV) and capacity (CAPA) indicators. This approach could also include ASM (Airspace Management) aspects such as military zones. At the scale of a few aircraft, a human could select those likely to benefit from ”green” trajectories. However, at the scale of a control center, the induced complexity would be too great. This thesis proposes to develop artificial intelligence-based algorithms to select the aircraft that could benefit from more environmentally efficient trajectories while ensuring compliance with the capacities of the impacted sectors. In addition the algorithm will jointly optimize the airspace configuration. This approach could then be extended to the European level in collaboration with the “Network Manager.”

State of the Art

In the literature, such a problem is partially addressed. Some research initiatives address the problem of airspace congestion by adapting the demand to the current available capacity. This can be done at the strategic level or at the pretactical level [4, 5, 3, 6, 1, 7, 8]. This problem induces high combinatorics. If we consider that for each flight one has 10 departure time slots, 5 flight level option and 5 horizontal route options, the overall options for one flight is equal to 10x5x5=250. Having 36000 flights every day over the European airspace the induce combinatorics is given 250(36000). The associated non linear airspace congestion creates a strong dependency between flights which prevent the use of separability approaches. Meta-heuristics are then good candidates to address this problem. Some other approaches try to adapt the capacity to the demand by optimally designing airspace sectors [2, 10] or by optimizing the airspace configuration which consists in creating optimal groups of elementary sectors based on the underlying demand [9]. Today’s airspace is structured as a multitude of 3D blocks known as elementary sectors. These indivisible spaces can accept a limited quantity of traffic per period of time, called capacity. In order to manage air traffic, these volumes can be grouped together to form control sectors, assigned to ATCOs to perform their control duty. Thus, the airspace configuration problem consists of finding the optimal layout of the space to accommodate the air traffic, the demand in this case. The airspace configuration process is performed locally within each Area Control Center (ACC). It consists of preparing the plan of configuration for the day, called opening scheme, to accommodate the traffic expected with regards to the operational constraints. The main ones are the available number of ATCOs, which directly impact the ability to open sectors, and the ease with which the solution can be implemented in an real world environment. In particular, transitions from one configuration of space to another can be more or less complex to put in place, or even impossible.This task is usually the responsibility of ATFCM experts, using their experiences and operational rules to propose the best airspace configuration. However, the division of space into a multitude of elementary bricks makes this a highly combinatorial problem. The average number of possible configurations for a given ACC is close to 50000. One of them has to be chosen every 20 minutes by in order adapt the capacity to the demand. Such configurations are interdependent inducing non separability of the underlying optimization problem. European Airspace has 62 ACC the overall combinatoric for such a configuration problem is given by : 62x72x50000=223.2x106 .

Objective

Usually, the two problems are addressed separately by means of a bi-level optimization scheme [11]. The first objective of this thesis is to develop a global optimization methodology able to address the whole problem (trajectory + configuration optimization) at the same time for the case study of the European airspace. The second objective of this thesis is to use the methodology developed in the first step to maximize the number of CCO and CDO operations in the airspace.

Agenda

The PhD will be organized as follows : • Literature review (6 months) • Mathematical modeling of the first step (3 months) • Developement of the global optimization of the first step (6 months)