SyntAIr - IMPROVED ATM AUTOMATION AND SIMULATION THROUGH AI-BASED UNIVERSAL MODELS FOR SYNTHETIC DATA GENERATION

Informazioni relative al progetto

SynthAIR

ID dell’accordo di sovvenzione: 101114847

DOI

10.3030/101114847

Progetto chiuso

Data della firma CE 15 Giugno 2023

Data di avvio 1 Settembre 2023

Data di completamento 28 Febbraio 2026

Finanziato da

• Climate, Energy and Mobility

Costo totale

€ 1 215 003,75

Contributo UE

€ 993 550,00

Coordinato da

SINTEF AS
Norway

Periodic Reporting for period 1 - SynthAIR (SyntAIr - IMPROVED ATM AUTOMATION AND SIMULATION THROUGH AI-BASED UNIVERSAL MODELS FOR SYNTHETIC DATA GENERATION)

Periodo di rendicontazione: 2023-09-01 al 2024-08-31

Sintesi del contesto e degli obiettivi generali del progetto

The SynthAIR project is designed to address critical challenges in Air Traffic Management (ATM), including data scarcity, and the need for enhanced decision-making tools. In the context of increasingly complex airspace management and rising air traffic, traditional data sources often fall short in providing the comprehensive, high-quality datasets required for effective ATM operations. SynthAIR aims to fill this gap by developing advanced AI-based models that generate realistic synthetic datasets, mirroring real-world air traffic scenarios. These datasets will support improved simulations, predictive analytics, and decision-making processes within ATM, ultimately leading to safer and more efficient air traffic operations.

The project is strategically aligned with European Union policies on digital transformation and innovation in transportation, contributing to the broader goals of enhancing aviation safety, reducing delays, and improving environmental sustainability. By providing validated synthetic data and AI tools, SynthAIR is expected to significantly impact the capacity and resilience of ATM systems, enabling more informed and timely decisions. The project's pathway to impact includes rigorous validation exercises, stakeholder engagement, and recommendations for integrating synthetic data into existing ATM frameworks.

Lavoro eseguito dall’inizio del progetto fino alla fine del periodo coperto dalla relazione e principali risultati finora ottenuti

SynthAIR focused on developing and validating AI models to generate high-quality synthetic data for Air Traffic Management (ATM). Key activities included:

- Model Development: Advanced AI models like GANs and VAEs were created to generate synthetic datasets that accurately simulate air traffic scenarios, such as flight trajectories and delays.
- Data Generation: These models produced synthetic datasets tailored to specific ATM use cases, ensuring the data’s diversity, fidelity, and operational relevance.
- Validation: The synthetic data was rigorously tested through simulations and statistical assessments, ensuring it meets the required operational standards.

Main Achievements
- Validated Synthetic Datasets: High-fidelity synthetic datasets were successfully developed, closely mirroring real-world ATM data.
- Robust AI Models: The project delivered AI models capable of generating realistic and operationally useful synthetic data.
- Evaluation Metrics: A comprehensive set of metrics for evaluating the synthetic data generated.

Progressi oltre lo stato dell’arte e potenziale impatto previsto (incluso l’impatto socioeconomico e le implicazioni sociali più ampie del progetto fino ad ora)

Results
- Synthetic ATM Datasets: SynthAIR generated synthetic datasets, including airborne flight trajectories, landing trajectories, and go-arounds, designed to closely replicate real-world ATM scenarios. These datasets were created using advanced AI models such as TimeGAN and TCVAE, ensuring high fidelity and operational relevance​​.
- AI Model Development: Four AI models, including TCVAE with VampPrior and TimeVQVAE, were developed and validated. These models demonstrated the capability to generate realistic, multivariate time series data critical for ATM applications​.
- Validation and Evaluation Framework: The project established a comprehensive evaluation framework, using metrics like Fréchet Inception Distance (FID) and t-SNE visualizations, to assess the quality and applicability of the synthetic datasets​.

Potential Impacts
- Enhanced ATM Simulations: The synthetic datasets provide a foundation for improving ATM simulations, aiding in more accurate predictions and better decision-making processes.
- Further Research and Adaptation: Continued refinement of the models and datasets, particularly in integrating additional factors like weather data, will be essential for broader applicability and adoption in real-world ATM environments.
- Regulatory and Standardization Needs: Collaboration with industry stakeholders to ensure that the synthetic data aligns with ATM operational standards will be crucial for future integration and widespread use.