Periodic Reporting for period 1 - TRUSTY (TRUSTWORTHY INTELLIGENT SYSTEM FOR REMOTE DIGITAL TOWER)
Berichtszeitraum: 2023-09-01 bis 2024-08-31
Overall, the goal of TRUSTY is to provide adaptation in the level of transparency to enhance the trustworthiness of AI-powered decisions in the context of RDTs. While in an actual tower, operators have direct visual access to the taxiway and runway monitoring, the RDTs concept only provides such information through video transmission with
a warning and the corresponding explanation. To deliver trustworthiness in an AI-powered intelligent system TRUSTY will consider several approaches, and they are listed:
• ‘Self-explainable and Self-learning’ system for critical decision-making
• ‘Transparent ML’ models incorporating interpretability, fairness, and accountability
• ‘Interactive data visualization and multimodal human-machine interface/interactions (HMI), i.e. Graphical User Interface (GUI)’ for smart and efficient decision support
• ‘Adaptive level of explanation’ regarding the user's cognitive state.
• “HCAI” to enhance the trustworthiness of AI-powered systems.
• “Human-machine collaboration (HMC) or Human-AI teaming (HAIT)” to consider user feedback to insure some computation flexibility and the users’ acceptability.
The measurable objectives are categories: i) Scientific, innovation, and research objectives (SIO) focusing on the research to deliver a rigorous and self-standing methodology to drive the implementation and define its operational principles; 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, evaluation of (cost) effectiveness and the scalability potential of the ecosystem,
the spread of excellence gained and applicability for the ATM ecosystem.
a warning and the corresponding explanation. To deliver trustworthiness in an AI-powered intelligent system TRUSTY will consider several approaches, and they are listed:
• ‘Self-explainable and Self-learning’ system for critical decision-making
• ‘Transparent ML’ models incorporating interpretability, fairness, and accountability
• ‘Interactive data visualization and multimodal human-machine interface/interactions (HMI), i.e. Graphical User Interface (GUI)’ for smart and efficient decision support
• ‘Adaptive level of explanation’ regarding the user's cognitive state.
• “HCAI” to enhance the trustworthiness of AI-powered systems.
• “Human-machine collaboration (HMC) or Human-AI teaming (HAIT)” to consider user feedback to insure some computation flexibility and the users’ acceptability.
The measurable objectives are categories: i) Scientific, innovation, and research objectives (SIO) focusing on the research to deliver a rigorous and self-standing methodology to drive the implementation and define its operational principles; 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, evaluation of (cost) effectiveness and the scalability potential of the ecosystem,
the spread of excellence gained and applicability for the ATM ecosystem.
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 full field as both D3.1 and D3.2 are submitted. D3.1 is accepted and D3.2 is under review. M3 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 is under review. And D6.2 is due in PR2
ISO3: Communication, dissemination, and exploitation. D6.2 is due in PR2. And D7.1 is approved.
ISO1: Active engagement of main stakeholders in the definition of requirements and system conceptualization during and the rest of the project. D3.2 is under review. And D6.2 is due in PR2
ISO3: Communication, dissemination, and exploitation. D6.2 is due in PR2. And D7.1 is approved.
According to the D3.1 study defines the possible definition, specifications for the taxiway and runway monitoring and decision support in RDTs application domain and consequently the functional requirements of the XAI, human-centred XAI, HMI and 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 will also identify the SotA techniques and approach for interactive data visualisation, human-centric AI model development and hAIi 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 is 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.
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 is 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.