EUROCONTROL calibration of optimised approach spacing tool (COAST) with use of machine learning models

White paper

COAST allows air navigation service providers (ANSPs) to optimise the calibration of their final approach separation delivery tool for safe and efficient time-based separation (TBS) operations, using machine learning (ML) models trained on historical local airport traffic and MET data.

Arrival runway capacity/throughput is directly linked to the applicable minimum longitudinal separation between traffic on final approach, as well as the separation delivery performance by the Air Traffic Controllers.

The separation criteria are based on surveillance and on wake turbulence, and typically based on longitudinal distance minima under ATS surveillance service.

In order to safely deliver the separation at the separation delivery point (typically the runway threshold), the Controllers usually takes some buffers when spacing the traffic on final. These buffers are taken by experience of the expected traffic speed behaviour (compliance to ATC speed instruction and typical landing speed) and can lead to an average over-separation between 0.5 NM and 1 NM (above the applicable minima). These buffers also result from the uncertainty related to the compression between traffic on final due to the natural catch-up and the difference in final approach speeds.

In addition, headwind conditions on final approach cause a reduction of the aircraft ground speed which results in increased time separation for each aircraft pair, a reduction of the landing rate, and a lack of stability of the runway throughput during arrival operations. This has a negative impact not only on the achieved capacity, but also on the predictability of operations, time and fuel efficiency, and the environment (emissions). The impact on predictability for core hubs is particularly important at the network level. The service disruption caused by the reduction in achieved runway throughput compared to declared capacity in medium and strong headwinds has a significant impact on the overall network performance and is particularly exacerbated if this occurs on the first rotation of the day because of the impact on all the other rotations throughout the day.

Time-Based Separation (TBS) in final approach is an operational solution, which uses time instead of distance to separate aircraft on their final approach to a runway. In order to apply this concept, approach and tower air traffic controllers need to be supported by a separation delivery tool which

  • provides a distance indicator (final target distance – FTD), enabling to visualise surveillance display the distance corresponding to the applicable TBS minima, and taking in account the prevailing wind conditions;
  • integrates all applicable separation minima and spacing needs.

This separation delivery tool, providing separation indicators between arrival pairs on final approach, also enables an increase in separation performance when providing a second indicator (Intermediate Target Distance – ITD): a spacing indicator to optimise the compression buffers and ensuring optimum runway delivery (ORD).

For being operationally deployed, a separation delivery tool allowing the Time-Based Separation application but also improving ATCO performance and management of complex business rules (separation/spacing), has to be demonstrated as fully reliable.

The move from distance to time-based rules allowing efficient and safe separation management requests to properly model/predict aircraft speed and behaviour in short final and the associated uncertainty. A too conservative definition of buffer can lead to a reduction of efficiency whereas making use of advanced Machine Learning techniques for aircraft behaviour prediction allows improvements of separation delivery compared to today while maintaining or even reducing the associated ATCO workload.

COAST (Calibration of Optimised Arrival Spacing Tool) developed by EUROCONTROL allows ANSPs to train Machine Learning Models on historical traffic and MET data. Those models, once integrated into a separation delivery tool, allow the computation of the most efficient separation to apply and the associated buffer ensuring a level of safety in line with the TBS Safety Case and the EUROCONTROL TBS-ORD Tool Specification.

With the use of a separation delivery tool and COAST, runway throughput can be increase (up to 6% at major airports) even under low wind conditions thanks to the optimisation of the spacing and separation performance. In addition, TBS operations will deliver mitigation of the headwind effect, and provide operational resilience and stabilised arrival throughput across headwind conditions.

COAST requires regular updates of the models in order to cover progressively the complete traffic fleet and also to cope with new aircraft and operational evolutions. Learn more in the full document below.


EUROCONTROL COAST with use of Machine Learning models - White paper