Public Deliverables

Official project deliverables made publicly available through the European Commission’s CORDIS platform and other repositories.

Authors: SINTEF
Date: January 23, 2024
DOI: 10.5281/zenodo.13935203
Abstract: This deliverable describes the concept outline of the SynthAIr project. The main objective of SynthAIr is to explore and define AI-based methods for synthetic data generation in the domain of ATM systems due to the limitation of AI-based tools development caused by the lack of sufficient data (e.g., safety-related data) and the challenge of generalizing those AI-based models. The project investigates data-driven methods for synthetic data generation because they require 1) less user knowledge expertise (i.e., no need to derive the explicit model of the distribution) and 2) better generalization capabilities. More specifically, inspired by recent advances in computer vision and language technology, SynthAIr proposes the concept of a Universal Time Series Generator (UTG). A UTG is a model trained on several different time series and capable of generating a synthetic dataset representing a new dataset, simply conditioned by a compressed representation of it. In the aviation domain, this generator can be trained on data related to a few airports and then used to generate synthetic data for a new airport. The same principle can be applied to define a Universal Time Series Forecaster (UTF), which is capable of making predictions in a new environment (i.e., data from a new airport) without any additional training.

Authors: TUD, SINTEF, DEEPBLUE, EUROCONTROL
Date: May 14, 2024
DOI: 10.5281/zenodo.13935162
Abstract: This deliverable discusses the state-of-the-art related to the use cases considered in the project, as well as relevant synthetic data modelling techniques to be used for elaboration of use cases. Based on multiple data-, modelling-, and stakeholder-related criteria, two promising use cases were selected for further elaboration in the project. The literature review serves as a starting point for the activities in WP3 (Synthetic Data Generation for Multivariate Time Series for ATM-automation) and WP4 (Universal Time Series Model for Prediction and Data Generation for ATM-automation), based on the selected use cases.

Authors: TUD, SINTEF, DEEPBLUE, EUROCONTROL
Date: 22 January 2025
CORDIS Link: Download
Abstract: This deliverable discusses the definition of the different use cases proposed in the SynthAIr project. Starting from the definition of the downstream problems, the existing models in the literature and the relevance of each use case to aviation stakeholders to the evolution of the use cases within the life of the project. Based on feasibility and relevance, the use cases were then sorted into a priority list to formulate a starting point for WP3 (Synthetic Data Generation for Multivariate Time Series for ATM-automation).

Authors: SINTEF, TUD, DEEPBLUE, EUROCONTROL
Date: 08 October 2024
CORDIS Link: Download
Abstract: This deliverable presents the initial machine learning models developed for synthetic data generation as part of the SynthAIr project. It introduces four advanced AI methods designed to generate synthetic multivariate time series data in the context of Air Traffic Management (ATM). The document outlines the methodology, model architectures, and a comprehensive evaluation framework. Preliminary results showcase the models’ potential to generate realistic ATM data, particularly aircraft trajectories, while also identifying key challenges and limitations. A guide to the open-source code is included to promote reproducibility and support further development within the ATM research community.

Authors: SINTEF, TUD, DEEPBLUE, EUROCONTROL
Date: 08 October 2024
CORDIS Link: Download
Abstract: This deliverable presents the synthetic Air Traffic Management (ATM) datasets generated as part of the SynthAIr project. It focuses on datasets created using the initial machine learning models outlined in deliverable D3.1. These datasets offer a high-fidelity representation of aircraft trajectory characteristics, realistic variability, and class-conditional generation. The document provides detailed information on the datasets’ characteristics, including key variables, distributions, and spatial-temporal coverage. It also includes a user guide for accessing and utilizing the open-source data, along with a summary of known limitations. By providing diverse, high-fidelity synthetic data that is easily accessible, this deliverable supports ongoing ATM research and development, in alignment with SynthAIr’s objectives to advance ATM automation and simulation capabilities.

Authors: SINTEF, TUD, DEEPBLUE, EUROCONTROL
Date: 22 November 2024
CORDIS Link: Download
Abstract: This deliverable presents the initial development of general time series embeddings for ATM data. It demonstrates how the rich latent representations learned by variational autoencoder models can serve as general-purpose time series embeddings without requiring additional model development. The document outlines an evaluation framework for assessing embedding quality and presents initial results focusing on trajectory clustering and visualization. While these results are promising, we also identify areas for future enhancement. This work lays the groundwork for developing more refined embedding models in subsequent deliverables, ultimately contributing to enhanced ATM automation and simulation capabilities through improved data representations.

Authors: DEEPBLUE, TUD, SINTEF, EUROCONTROL
Date: 30 October 2024
CORDIS Link: Download
Abstract: This document is the initial version of the SynthAIr Exploratory Research Plan (ERP). It describes the Exploratory research plan which will guide the preparation and execution of the validation exercises of the SynthAIr solution for TRL1 (Basic Technology Research / Research to Prove Feasibility).

Authors: DEEPBLUE
Date: July 04, 2024
DOI: 10.5281/zenodo.13935230
Abstract: This document is the Communication, Dissemination and Exploitation Initial Plan for SynthAIr. It contains detailed information about the Communication and Dissemination strategy, and the preliminary Exploitation strategy. Targets, key messages, information about branding, channels, social media, publications, events and overall KPIs both for communication and dissemination actions are detailed in this document.