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

Informacje na temat projektu

SynthAIR

Identyfikator umowy o grant: 101114847

DOI

10.3030/101114847

Projekt został zamknięty

Data podpisania przez KE 15 Czerwca 2023

Data rozpoczęcia 1 Września 2023

Data zakończenia 28 Lutego 2026

Finansowanie w ramach

• Climate, Energy and Mobility

Koszt całkowity

€ 1 215 003,75

Wkład UE

€ 993 550,00

Koordynowany przez

SINTEF AS
Norway

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

Okres sprawozdawczy: 2023-09-01 do 2024-08-31

Podsumowanie kontekstu i ogólnych celów projektu

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.

Prace wykonane od początku projektu do końca okresu sprawozdawczego oraz najważniejsze dotychczasowe rezultaty

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.

Innowacyjność oraz oczekiwany potencjalny wpływ (w tym dotychczasowe znaczenie społeczno-gospodarcze i szersze implikacje społeczne projektu)

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.