Skip to main content
European Commission logo
polski polski
CORDIS - Wyniki badań wspieranych przez UE
CORDIS
CORDIS Web 30th anniversary CORDIS Web 30th anniversary

Programme Category

Article available in the following languages:

EN

Digitalisation and Automation principles for ATM

 

Proposals should select a specific ATM operational environment, present a vision of a higher level of automation in this operational environment (which may include the delegation of control to the automation) and address one or more of the specific challenges above that hinder the application of machine learning methods for the further automation of ATM (e.g. Transparency, Generalization). Proposals should aim at providing a better understanding of this challenge(s) and investigate innovative methods to address the(se) challenge(s) in ATM. Proposals may make assumptions about the availability of technology and/or operations enablers (e.g. data link), but need to state them clearly. This topic covers ground and airborne automation that impacts ATM.

Projects addressing the Transparency of automated systems incorporating machine learning methods required for cooperative human machine systems should identify which information needs to be provided to enable the human operator to cooperatively work together with the automation. Based on the identified information requirements, the project should select or develop and assess suitable machine learning methods for ATM automation that are able to provide these kind of information and assurances. The project may investigate the applicability of methods from the domain of Explainable Artificial Intelligence (XAI).

Projects addressing the Generalisation and the adaptation of the algorithms to changes in the operational environment, should investigate methods to estimate and increase the ability of an automated systems to handle a situations that were not foreseen during the development and training. These methods should enable automation to adapt to changes of the environment, like the change of behaviour of some actors (e.g. modification of operational procedures), the entrance of new actors or unforeseen traffic or weather situations. Projects may explore the possibility to apply algorithms able to learn during operation in order to adapt to optimise operations based on changes in the environment. Projects may also investigate the effects of uncertainty added to operations by these new methods.

Research activities may take aspects related to certification into account as required. However, this shall not be the primary focus of research in the proposal as this will be addressed in topic SESAR-ER4- 09-2019.

Proposals can also suggest to address other challenges of applying AI machine learning for ATM Automation other than these mentioned above if justification is provided.

Increasing the Automation in ATM is considered as a key to significantly improve ATM performance. However, ATM is a continuous 24-7 set of services where the complexity of the ATM system, its fallback modes and necessary recovery steps has proved to be a major challenge for the introduction of further automation, and this has consequently slowed down the advancement of automation in ATM, especially in the most congested areas of Europe. The latest progress in the domain of Artificial intelligence and in particular Machine Learning may open new possibilities for further automation in ATM in high-density operations and some new applications have already been developed in SESAR.The application of Machine Learning for Automation in ATM also comes with new challenges, including sound assurance arguments that need to be solved to avoid a negative impact. In particular safety, business continuity and cyber security issues need to be proposed at an early stage of development of the automation concept.

One challenge of increasing automation is related to transparency of the automated system. Any automated assistance system needs to be able to provide the human operator with all information necessary to enable an understanding of the reasons for its behaviour and/or decisions. Otherwise the system may not be accepted or trusted by the operator, thus negating the theoretical benefits of the automation.

Another challenge is the Generalization of results from Machine Learning methods. Differences between the data used for training and the data feed to a trained algorithm can lead to unexpected results, including that not all situations can be anticipated during the training. Additionally, due to this behaviour, the system might not be able to adapt to changes of behaviour of other actors that could
not be anticipated during the training of the system. In order to certify systems based on Machine- Learning, methods are needed to demonstrate that in delegating the control to these system sufficient assurance can be provided that it does not raise safety risks beyond what can be mitigated by other measures. Moreover, concerns have been raised that this behaviour may add uncertainty or unexpected behaviour to the ATM system.

Projects are expected to provide principles that could enable higher levels of automation that are predicted to lead to an improvement of ATM performance, in particular cost efficiency, capacity and safety.