Dissemination
Radišić T., Andraši .P., Novak D., Rogošić T.: The Proposal of a Concept of Artificial Situational Awareness in ATC in Engineering Power : Bulletin of the Croatian Academy of Engineering, Vol. 15 (2), 2020
This concept of operations is developed as a base for human-machine distributed situational awareness that will be used as support in en-route ATC monitoring tasks. It proposes the monitoring tasks which could be assigned to AI in this project. The concept of operations describes the expected changes between the current concept of operations, future concepts which do not consider humanmachine distributed situational awareness, and the proposed concept which includes the AI into the team situational awareness.
The analysis shows which monitoring tasks exist, which of them can be automated in different scenarios (medium/high automation), and most importantly what are requirements for their automation in terms of needed data, changes in operations, changes in the user interface, and the possible effect on human operators.
Three individual ML modules are developed in the AISA project to study a shared situational awareness of AI and ATCOs. This deliverable describes the design, development and validation of the ML trajectory prediction module. It aims at predicting the true aircraft track and future positions of flights with the initially filed flight plan and the current aircraft state as input. Therefore, a two-step process is established. First a neural network is trained to predict the static aircraft track without any prediction in the time domain. Afterwards the current aircraft state is combined with the predicted track to determine a concrete future position prediction. ADS-B data from The OpenSky Network and flight plan data from the DDR2 database from EUROCONTROL is going to be used as database.
This work deals with monitoring tasks focusing on situational awareness. This approach applies ML techniques to perform predictions about separation infringements and safety metrics associated with the intrinsic characteristics of the separation between an aircraft pair. Herein, this work focuses on the concept of Situation of Interest (SI). One SI is when an aircraft pair is expected to intersect with a horizontal separation lower than a pre-defined separation and infringe the vertical separation minima. The safety metrics are the Minimum Distance, the distance and the time to reach the Minimum Distance for each aircraft pair. Moreover, it has been developed two approaches with similar roles of the Air Traffic Controller’s (ATCO) team. The Static mode focuses on planner ATCO. This mode predicts SI and their safety metrics when an aircraft pierces into the airspace with the aircraft located within the airspace. The Dynamic mode focuses on tactical ATCO. This mode predicts SI and their safety metrics throughout the aircraft's evolution within the airspace, and it receives the 4DT prediction of the aircraft within the airspace.
Air traffic complexity models use different bases for determining complexity. The goal is to build an air traffic complexity model which will be able to determine complexity regardless of the traffic situation, observed airspace, or air traffic controller. This is done by implementing a novel solution that uses air traffic controller tasks, defined depending on the traffic situation, to determine air traffic complexity. This approach offers a solution to certain problems recognized in previous complexity models.
In this deliverable (D 4.1) we describe an architecture for the data and metadata in such a KG and for the software components for incrementally processing and querying the data and metadata in the KG. The Proof-of-Concept KG System exemplifies this architecture and has the purpose of guiding further developments in WP 4 and WP 5. The proposed architecture of a KG system facilitates SPARQL Queries Capturing Monitoring Tasks based on traffic/airspace data converted to RDF. It further accomodates the integration with other components such as the Reasoning Engine in Prolog developed in Task 4.2 and Machine Learning Modules developed in WP3. The main goal is to develop and assess the concept of the artificial SA based on a KG from a functional perspective rather than to consider requirements of a real-time life system. In this deliverable we also describe the UML-to-RDFS/SHACL mapper which facilitates the transformation from information exchange models such as AIXM and FIXM modeled in UML to KG schemas in RDFS and SHACL.
The design problem tackled by Task 4.2 is to improve accessing the KG from Prolog by designing a KG-Prolog mapper that takes care of data interchange and mapping between Prolog engine and KG, so that Prolog programmers can easily develop Prolog programs, which read from and write to the KG. We investigate schema-oblivious and schema-aware KG-Prolog mapping. The schema-oblivious approach can be realized easily but is unwieldy for Prolog programmers when it comes to reading complex KG data. Schema-aware KG-Prolog mapping provides the contents of the KG in a form amenable to Prolog programmers according to the KG schema. We implement the schema-aware approach in three different variants and conduct preliminary performance studies for comparison. We provide a full integration of Prolog engine and AISA KG system for the schema-oblivious approach together with one variant of the schema-aware approach.
D4.1
The UML to RDFS/SHACL mapper
https://github.com/jku-win-dke/AISA-XMI-Mapper/
The Proof-of-concept KG system prototyp
https://github.com/bneumayr/aisa-kg-system/
D4.2
The KG-Prolog mapper
https://github.com/jku-win-dke/AISA-KG-Prolog-Mapper/
A knowledge graph is a collection of descriptions of various data which is put into context and then provides a framework for future work. This populated knowledge graph provides insight into traffic situations and all the data needed to describe it with enough detail so it would be functional both in real life and as a simulation. This deliverable will serve as a valuable input into further development of the conceptual model and will be a basis for the development of the deliverable D4.4 Facts, rules and queries capturing en-route ATC operations.
This deliverable gives an insight into the knowledge engineering needed for ensuring all relevant rules and facts about en-route ATC operations have been captured. The relevant facts and rules have mostly been gathered via interviews with licenced ATCOs and by closely examining FIXM and AIXM. The KG-Prolog Mapper is used for converting RDF facts from the KG into Prolog facts. Lastly, SPARQL queries that will be used to monitor the traffic situation and ATCOs have been developed and explained in this deliverable that can be considered a continuation of D4.3 Populated knowledge graph.
This deliverable presents the risk assessment of the AISA project that focuses on performing a safety analysis by identifying hazards, analysing them, and their risks (based on probability and severity) and providing mitigation measures. This work analyses the whole system that covers novel technologies based on artificial intelligence. This risk assessment provides valuable information for the further development of the AISA system that could be applied as potential safety requirements.
This report presents the results of two simulation experiments performed with an AI-based situation awareness system (AI SA system) developed in the AISA project to check the accuracy of the AI SA system’s estimations and predictions and its capability to contribute to human-machine team situation awareness.
The AISA website is obviously the main tool for communication, a central reference for the project and a data repository of all the main results. It is also a dissemination tool, e.g. the technical deliverables, papers will be accessible there, but the main purpose is to support communication, either indirectly or directly. Indirectly, it means that there are and will be information there, other communication tools can refer to and directly, when there is an events page for example where participants can register for a workshop.
Project promotion in AISA was based on an integrated approach combining Dissemination (making results available), Communication (making sure that potential stakeholders are aware of the project and its results and establishing dialogue with some of them), and Exploitation Planning (planning of measures to encourage use of results after project completion).
This document describes the Data Management Plan (DMP) of the AISA project according to the guidelines described in the Guidelines on Data Management in Horizon 2020 document [1]. As such, this DMP describes the data management life cycle for all datasets to be collected, processed, or generated by AISA project during its research activities. It details all data the project will collect and generate, how it will be exploited or made accessible for verification and re-use, and how it will be curated and preserved.
Authors: Dr. Javier Pérez-Castán (Universidad Politécnica de Madrid), Tomislav Radisic, Thomas Feuerle, J. Bowen Varela, L. Pérez-Sanz, L. Serrano-Mira
Abstract:
This article evaluates the application of Machine Learning (ML) classification techniques applied to air traffic conflict detection. The methodology develops a static approach in which the conflict detection prediction is performed when an aircraft pierces into the airspace. Conflict detection does not evaluate separation infringements but a Situation of Interest (SI). An aircraft pair constitutes an SI when is expected to cross with a longitudinal separation lower than 10 Nautical Miles (NM) and a vertical separation lower than 1000 feet (ft). Therefore, the ML predictor classifies aircraft pairs between SI or No SI pairs. Air traffic information is extracted from OpenSky that provides ADS-B trajectories. ADS-B trajectories do not offer enough situations to be evaluated. Hence, the authors performed simulations varying the entry time of the trajectories to the airspace within the same time period. The methodology was applied to a portion of Switzerland airspace, and simulations provided a 5% rate of SI samples. Cost-sensitive techniques were used to handle the strong imbalance of the database. Two experiments were performed: the Pure model considered the whole database, and the Hybrid model considered aircraft pairs that intersect longitudinally lower than 20 NM and vertically lower than 1000 ft. The Hybrid model provided the best results achieving 72% recall, representing the success rate of Missed alerts and 82% accuracy, which means the whole predictions' success rate.
Machine Learning classification techniques applied to static air traffic conflict detection Presentation at the 11th EASN International Conference, September 2021 article is available at: https://iopscience.iop.org/article/10.1088/1757-899X/1226/1/012019
Authors: Javier Alberto Pérez-Castán, Luis Pérez-Sanz,Lidia Serrano-Mira, Francisco Javier Saéz-Hernando,Irene Rodríguez Gauxachs and Víctor Fernando Gómez-Comendador
Abstract:
Given the ongoing interest in the application of Machine Learning (ML) techniques, the development of new Air Traffic Control (ATC) tools is paramount for the improvement of the management of the air transport system. This article develops an ATC tool based on ML techniques for conflict detection. The methodology develops a data-driven approach that predicts separation infringements between aircraft within airspace. The methodology exploits two different ML algorithms: classification and regression. Classification algorithms denote aircraft pairs as a Situation of Interest (SI), i.e., when two aircraft are predicted to cross with a separation lower than 10 Nautical Miles (NM) and 1000 feet. Regression algorithms predict the minimum separation expected between an aircraft pair. This data-driven approach extracts ADS-B trajectories from the OpenSky Network. In addition, the historical ADS-B trajectories work as 4D trajectory predictions to be used as inputs for the database. Conflict and SI are simulated by performing temporary modifications to ensure that the aircraft pierces into the airspace in the same time period. The methodology is applied to Switzerland’s airspace. The results show that the ML algorithms could perform conflict prediction with high-accuracy metrics: 99% for SI classification and 1.5 NM for RMSE.
Article is available at: https://www.mdpi.com/2226-4310/9/2/67
Author: Roland Guraly (Slot Consulting)
Abstract:
The usage of artificial intelligence (AI) is spreading nowadays and by now it has reached the safety critical industries as well, but it is not clear when and how it is going to be used in related primary systems. Air transport is a good example for such an industry. Although the introduction of new technologies in this sector should be maintained with an obvious care, scientific projects have already started to explore the possible benefits AI can bring for air traffic controllers. One of those projects is the AISA (AI Situational Awareness Foundation for Advancing Automation) which found that for safe implementation of advanced automation concepts in air traffic control, the human and the machine should share the same situational awareness.
Article is available at: Introducing artificial intelligence in air traffic control
Authors: Javier A. Pérez-Castán, Luis Pérez Sanz, Marta Fernández-Castellano, Tomislav Radišic, Kristina Samardžic and Ivan Tukaric
Abstract:
…This work deals with the learning assurance process for ML-based systems in the field of air traffic control. A conflict detection tool has been developed to identify separation infringements among aircraft pairs, and the ML algorithm used for classification and regression was extreme gradient boosting. This paper analyses the validity and adaptability of EASA W-shaped methodology for ML-based systems…
Article is available at: https://www.mdpi.com/1424-8220/22/19/7680/htm
Previous related work can be found at the website of the BEST project: www.project-best.eu
Funded by
Contact information
Prof. Tomislav Radišić
University of Zagreb
tradisic@fpz.unizg.hr
This project has received funding from the SESAR Joint Undertaking under Grant Agreement No. 892618, under European Union’s Horizon 2020 Research and Innovation programme.
Copyright © 2020 AISA Consortium