TRUSTWORTHY INTELLIGENT SYSTEM FOR REMOTE DIGITAL TOWER

Projektinformationen

TRUSTY

ID Finanzhilfevereinbarung: 101114838

DOI

10.3030/101114838

Projekt abgeschlossen

EK-Unterschriftsdatum 14 Juni 2023

Startdatum 1 September 2023

Enddatum 28 Februar 2026

Finanziert unter

• Climate, Energy and Mobility

Gesamtkosten

€ 999 967,50

EU-Beitrag

€ 999 967,50

Koordiniert durch

MALARDALENS UNIVERSITET
Sweden

Periodic Reporting for period 2 - TRUSTY (TRUSTWORTHY INTELLIGENT SYSTEM FOR REMOTE DIGITAL TOWER)

Berichtszeitraum: 2024-09-01 bis 2026-02-28

Zusammenfassung vom Kontext und den Gesamtzielen des Projekts

The measurable objectives are categories: i) Scientific, innovation, and research objectives (SIO) focusing on the research to deliver a rigorous and self-standing methodology; ii) Technological objectives (TO) focusing on a prototypical system and the delivery
and deployment in close to the real environment; iii) User-centric design (UCD) focusing on user acceptability in the domain of RDTs and iv) Impact and societal objectives (IO) with a specific focus on providing relevant impact and considering socio-economic aspects,
the spread of excellence gained and applicability for the ATM ecosystem.

The TRUSTY project introduces APSARA (AI-Powered Situational Awareness for Remote Airfields), an innovative AI-driven solution designed to enhance safety and efficiency in Remote Digital Tower operations. APSARA integrates multimodal data streams to provide Air Traffic Controllers (ATCOs) with a unified, actionable output for improved decision-making.
APSARA leverages advanced computer vision and AI algorithms to analyse airfield video feeds. This enables real-time detection of objects and events, such as aircraft, vehicles, and potential hazards, on runways and taxiways. This capability allows for early identification of risks like runway incursions or wind shear.
The system also processes ATC and cockpit audio communications, converting speech into text and interpreting it using semantic tools (e.g. Word2Vec). This approach allows APSARA to identify critical patterns and keywords associated with high-risk situations, triggering timely alerts for ATCOs.

A key focus of ASPARA is human-AI trust and adaptive interaction. APSARA incorporates neurophysiological monitoring (EEG, EDA) to assess operator states such as workload, stress, vigilance, and acceptance. These indices inform trust modelling and adaptive system behaviour, enabling dynamic adjustments in HMI transparency and AI explainability. This fosters confidence, resilience, and effective collaboration between humans and AI.
While ASPARA’s current Technology Readiness Level (TRL) is low, the solution anticipates significant improvements in safety, operational capacity, and situational awareness. By consolidating multimodal data into a single, informative output, the solution empowers ATCOs to manage aircraft movements with greater precision and timeliness, even under challenging conditions. Future R&I phases will enable real-time trust-aware adaptation, reinforcing Europe’s leadership in safe, transparent, and human-centric AI for air traffic management.

Arbeit, die ab Beginn des Projekts bis zum Ende des durch den Bericht erfassten Berichtszeitraums geleistet wurde, und die wichtigsten bis dahin erzielten Ergebnisse

SIO1: Provide a clear definition and a technical work plan of a trustworthy intelligent system with identified Key Performance Indicators (KPIs)/Key Value Indicators (KVIs) based on SotA study and gap analysis. The goal is fullfield as both D3.1 and D3.2. as accepted and M2 is completed.
SIO2: Provide a “Self-explainable” and “Self-learning system” for critical decision-making based on MML considering robust and resilient ML models to the tasks taxiway and runway inspection and misalignment warning.D4.2 is accepted and M3 is completed.
SIO3: Provide ‘Transparent ML models’ incorporating interpretability, fairness, and accountability based on human-centred XAI and active learning.D4.4 is accepted and M3 is completed.
SIO4: Provide an ‘Adaptive level of explanation’ regarding the user's cognitive state based on human factors. D5.4 is accepted and M4 is completed.
SIO5: Provide a conceptual framework for building a trustworthy intelligent system.D6.2 is accepted and M5, M6 are completed.

TO1: Robust and resilient multi-modal machine learning (MML) models,D4.2 is accepted and M3 is completed.
TO2: Transparent ML models with a human-centred explanation, D4.4 is accepted and M3 is completed.
TO3: Framework for HAIT D5.4 is accepted and M4 is completed.
TO4: Smart HMI and GUI for intelligent decision support.D5.2 is accepted and M4 is completed.

ISO1: Active engagement of main stakeholders in the definition of requirements and system conceptualization during and the rest of the project.D3.2 and D6.2 are accepted.
ISO3: Communication, dissemination, and exploitation. D6.2 is accepted.

Fortschritte, die über den aktuellen Stand der Technik hinausgehen und voraussichtliche potenzielle Auswirkungen (einschließlich der bis dato erzielten sozioökonomischen Auswirkungen und weiter gefassten gesellschaftlichen Auswirkungen des Projekts)

According to the D3.1 study defines the possible definition, specifications for the taxiway and runway monitoring and decision support in remote digital towers (RDTs) application domain and consequently the functional requirements of the XAI, human-centred XAI, human-machine interface/interactions (HMI), and Human-AI teaming (HAIT) solutions in a customized way. Here, it identifies the transparency in AI based on a systematic literature review on AI explainability, HMI and GUI with human-centred XAI in the domain of RDTs, i.e. taxiway and runway monitoring. The task also identified the SotA techniques and approach for interactive data visualisation, human-centric AI model development and HMI interfaces, and HAIT in RDT domain.

Again,considering D3.2 the key stakeholders engagement through the Workshop #1 on measuring trust and the Workshop #2 on tower ATCO tasks and gap analysis, identified operational needs for tower controllers and established Key Performance Indicators (KPIs) essential for trust in Explainable AI (XAI) systems. The primary objective was to identify key operational tasks within RDT that can benefit from enhanced XAI support and to present a comprehensive development roadmap for integrating transparency in an AI-driven Decision Support System (DSS) based on the feedback from end-users.

Research Outcomes Across the publications advance the development of trusted, AI‑enabled Remote Digital Towers (RDTs) by addressing technical, human‑factors, and operational challenges. They show that AI can significantly enhance situational awareness, decision support, and operational resilience, but only when designed with transparency, explainability, and calibrated human–AI collaboration at the forefront.

Across the work, researchers demonstrate that multimodal machine learning—integrating video, audio, sensor data, and contextual information—improves event detection, anomaly recognition, and controller support in complex air‑traffic scenarios.