Periodic Reporting for period 1 - HAIKU (Human AI teaming Knowledge and Understanding for aviation safety)
Période du rapport: 2022-09-01 au 2024-02-29
HAIKU envisions developing Human-Centred AI-Based Intelligent Assistants for safe, secure, trustworthy, and effective Human-AI partnerships in aviation systems. Anchored in a truly human-centric approach, the HAIKU goal is to pave the way for AI integration in aviation, crafting Intelligent Assistant prototypes that embody human values and dynamically evolve based on user interactions. Starting from users’ needs, HAIKU prioritize integrating technology to enhance human activities, ultimately improving safety within aviation operations.
The HAIKU consortium is composed of 15 partners, bringing together Human Factors expertise, domain’s key end-users and technology suppliers of excellence.
Human-AI partnership, Explainability, Future aviation workforce and skills, Safety culture, Societal acceptance of AI, Acceptable means of compliance for AI, and Safety, HP, Security and Liability assessment are the main transversal areas of work in HAIKU.
The core streams of work are 6 aviation use cases:
- Use Case 1: Intelligent Assistant in the cockpit to assist in ‘startle response’ adverse events
- Use Case 2: Intelligent Assistant in the cockpit to assist in route planning/replanning
- Use Case 3: Digital Intelligent Assistant for Urban Air Mobility coordinator to assist in traffic management
- Use Case 4: Intelligent Assistant for tower (and remote tower) controllers to assist in routine and repetitive tasks for aircraft on approach
- Use Case 5: Intelligent Assistant to improve airport safety through data analysis
- Use Case 6: Airport Intelligent Assistant to monitor risk factor conditions associated with indoor spread of infectious diseases
The HAIKU consortium is composed of 15 partners, bringing together Human Factors expertise, domain’s key end-users and technology suppliers of excellence.
Human-AI partnership, Explainability, Future aviation workforce and skills, Safety culture, Societal acceptance of AI, Acceptable means of compliance for AI, and Safety, HP, Security and Liability assessment are the main transversal areas of work in HAIKU.
The core streams of work are 6 aviation use cases:
- Use Case 1: Intelligent Assistant in the cockpit to assist in ‘startle response’ adverse events
- Use Case 2: Intelligent Assistant in the cockpit to assist in route planning/replanning
- Use Case 3: Digital Intelligent Assistant for Urban Air Mobility coordinator to assist in traffic management
- Use Case 4: Intelligent Assistant for tower (and remote tower) controllers to assist in routine and repetitive tasks for aircraft on approach
- Use Case 5: Intelligent Assistant to improve airport safety through data analysis
- Use Case 6: Airport Intelligent Assistant to monitor risk factor conditions associated with indoor spread of infectious diseases
WP2 Main activities performed in M1-M18:
- Review of the state-of-the-art of human-centred AI guiding principles and definition of a set of high-level principles for HAIKU;
- Definition of the HAIKU vision;
- Exploration and analysis of the technological, operational and societal future trends, ending up with the development of the 2030 and 2050 future landscapes for each HAIKU target segment (Airport, Air Traffic Management, Cockpit, Urban Air Mobility);
- Review of the existing societal acceptance models and preliminary analysis of societal acceptance concerning the six HAIKU Intelligent Assistants prototypes;
- Coordination of the engagement of external targeted end-users and stakeholders
Main achievements: Development of the 2030 and 2050 landscapes as a key reference to design Intelligent Assistants for future operations; The results of the preliminary analysis of societal acceptance as a key input to refine the Intelligent Assistant design; The establishment of a solid network of external end-users and stakeholders interested and willing to contribute to HAIKU activities, counting around 100 experts from all over the world, contributing to the creation of an AI community in aviation.
WP3 Main activities performed in M1-M18:
- Review of applications of AI and Digital Assistant Technology in different industrial sectors (e.g aviation, automotive)
- Review of typical HF constructs and discussion about applicability in AI systems,
- Review of sensors and technologies for human monitoring, analysis of applicability in HAIKU UCs,
- Development of a framework for H-AI teaming and automation, named LACC-LOA,
- Development of 17 different concepts of Intelligent Assistants, for the ATM, UAM, Cockpit and Airport segments,
- Development of a provisional human-centred framework for validating the project’s Use Case (UC) Intelligent Assistants, including their mapping on EASA’s AI levels.
Main achievements: Delivery of the state-of-the-art review, documented in D3.1; Analysis of Human Monitoring applicability and added value in HAIKU UCs, documented in D3.1; Delivery of the first draft HAT design framework, documented in D3.1; Delivery of 17 concepts, documented in D3.2; First provisional validation framework, documented in D3.3.
WP4 Main activities performed in M1-M18:
- Definition of a common, harmonised approach to design and development for all the UCs,
- Preliminary identification of High Level requirements (HLRs) and Human-AI Teaming (HAT) requirements,
- Mapping of HAT requirements onto the eight HAT constructs identified in D3.3
- Development of low fidelity prototypes for Use Cases 1, 4, 5, 6.
Main achievements: Definition of requirements, documented in D4.1; Definition of the Intelligent Assistants architecture, first iteration, documented in D4.4 for UC 1, 4, 5, 6; Definition of the first mock-up of the HMI for each use case, documented in D4.4 for UC 1, 4, 5, 6 and in D6.2 for all the UCs. D4.4 will be updated at M24, when UC2 and 3 will be documented in D4.5 (M24 and M36); Data acquisition, analysis and modelling for UC1, 4, 5, 6. Results reported in D6.2; Development of the AI components for all the UCs, reported in D4.4 for UC 1, 4, 5, 6 and in D6.2 for all the UCs. D4.4 will be updated at M24, when UC2 and 3 will be documented in D4.5 (M24 and M36).
WP5 main activities performed in M1-M18:
- identification of XAI framework (Construal Level Theory),
- application of CLT in all the HAIKU UC to define UC-specific XAI strategy.
Main achievements: Definition of the HAIKU XAI framework and application in all the UCs, documented in D5.1; XAI validation carried out for UC4 and documented in D6.2.
WP6 main activities performed in M1-M18:
- harmonisation of reference scenarios and ConOps for each Intelligent Assistant,
- coordination and delivery of validation strategies and plans across UCs,
- execution of the first Validation session for all the UCs, data collection and analysis.
Main achievements: first iteration of the Intelligent Assistant design; D6.1 containing the validation strategy and plan; D6.2 containing the validation results and insights for the next validation round.
WP7 main activities performed in M1-M18:
- delivery of the regulatory landscape for AI in aviation, covering ethical and legal frameworks,
- identification of an integrated framework for safety, security, Human Factors and liability assurance, including applicable methods and techniques,
- iterative application of the framework for the assessment of the UCs.
Main achievements: D7.1 delivered, being updated now; D7.2 delivered, to be updated at M25; D7.3 delivered, UC1 and 6 assessed at M18.
WP8 main activities performed in M1-M18:
- Determine how human roles will evolve and look like by 2030 and beyond (up to 2050);
- Identify the skill-sets the future workforce would need to effectively and safely work and team-up with AI in the future;
- Detail how best to educate, train and organise operators to work in an AI-based environment;
- Explore how safety culture could be affected (positively as well as negatively) by a transition to AI integration into aviation operations and make recommendations on safeguards to monitor and maintain safety culture
Main achievements: An overview on current way of performing selection, training and testing in aviation; Human-AI Teaming roadmap for Commercial Air Transport; The evolution of human role in ATM & UTM and Flight Operations; Airports’ impact areas map (D8.1); Concept of “AI CRM” for pilots (SPIC, SPO) (D8.2); The Impact of AI on Future Aviation Safety Culture and initial safeguards
- Review of the state-of-the-art of human-centred AI guiding principles and definition of a set of high-level principles for HAIKU;
- Definition of the HAIKU vision;
- Exploration and analysis of the technological, operational and societal future trends, ending up with the development of the 2030 and 2050 future landscapes for each HAIKU target segment (Airport, Air Traffic Management, Cockpit, Urban Air Mobility);
- Review of the existing societal acceptance models and preliminary analysis of societal acceptance concerning the six HAIKU Intelligent Assistants prototypes;
- Coordination of the engagement of external targeted end-users and stakeholders
Main achievements: Development of the 2030 and 2050 landscapes as a key reference to design Intelligent Assistants for future operations; The results of the preliminary analysis of societal acceptance as a key input to refine the Intelligent Assistant design; The establishment of a solid network of external end-users and stakeholders interested and willing to contribute to HAIKU activities, counting around 100 experts from all over the world, contributing to the creation of an AI community in aviation.
WP3 Main activities performed in M1-M18:
- Review of applications of AI and Digital Assistant Technology in different industrial sectors (e.g aviation, automotive)
- Review of typical HF constructs and discussion about applicability in AI systems,
- Review of sensors and technologies for human monitoring, analysis of applicability in HAIKU UCs,
- Development of a framework for H-AI teaming and automation, named LACC-LOA,
- Development of 17 different concepts of Intelligent Assistants, for the ATM, UAM, Cockpit and Airport segments,
- Development of a provisional human-centred framework for validating the project’s Use Case (UC) Intelligent Assistants, including their mapping on EASA’s AI levels.
Main achievements: Delivery of the state-of-the-art review, documented in D3.1; Analysis of Human Monitoring applicability and added value in HAIKU UCs, documented in D3.1; Delivery of the first draft HAT design framework, documented in D3.1; Delivery of 17 concepts, documented in D3.2; First provisional validation framework, documented in D3.3.
WP4 Main activities performed in M1-M18:
- Definition of a common, harmonised approach to design and development for all the UCs,
- Preliminary identification of High Level requirements (HLRs) and Human-AI Teaming (HAT) requirements,
- Mapping of HAT requirements onto the eight HAT constructs identified in D3.3
- Development of low fidelity prototypes for Use Cases 1, 4, 5, 6.
Main achievements: Definition of requirements, documented in D4.1; Definition of the Intelligent Assistants architecture, first iteration, documented in D4.4 for UC 1, 4, 5, 6; Definition of the first mock-up of the HMI for each use case, documented in D4.4 for UC 1, 4, 5, 6 and in D6.2 for all the UCs. D4.4 will be updated at M24, when UC2 and 3 will be documented in D4.5 (M24 and M36); Data acquisition, analysis and modelling for UC1, 4, 5, 6. Results reported in D6.2; Development of the AI components for all the UCs, reported in D4.4 for UC 1, 4, 5, 6 and in D6.2 for all the UCs. D4.4 will be updated at M24, when UC2 and 3 will be documented in D4.5 (M24 and M36).
WP5 main activities performed in M1-M18:
- identification of XAI framework (Construal Level Theory),
- application of CLT in all the HAIKU UC to define UC-specific XAI strategy.
Main achievements: Definition of the HAIKU XAI framework and application in all the UCs, documented in D5.1; XAI validation carried out for UC4 and documented in D6.2.
WP6 main activities performed in M1-M18:
- harmonisation of reference scenarios and ConOps for each Intelligent Assistant,
- coordination and delivery of validation strategies and plans across UCs,
- execution of the first Validation session for all the UCs, data collection and analysis.
Main achievements: first iteration of the Intelligent Assistant design; D6.1 containing the validation strategy and plan; D6.2 containing the validation results and insights for the next validation round.
WP7 main activities performed in M1-M18:
- delivery of the regulatory landscape for AI in aviation, covering ethical and legal frameworks,
- identification of an integrated framework for safety, security, Human Factors and liability assurance, including applicable methods and techniques,
- iterative application of the framework for the assessment of the UCs.
Main achievements: D7.1 delivered, being updated now; D7.2 delivered, to be updated at M25; D7.3 delivered, UC1 and 6 assessed at M18.
WP8 main activities performed in M1-M18:
- Determine how human roles will evolve and look like by 2030 and beyond (up to 2050);
- Identify the skill-sets the future workforce would need to effectively and safely work and team-up with AI in the future;
- Detail how best to educate, train and organise operators to work in an AI-based environment;
- Explore how safety culture could be affected (positively as well as negatively) by a transition to AI integration into aviation operations and make recommendations on safeguards to monitor and maintain safety culture
Main achievements: An overview on current way of performing selection, training and testing in aviation; Human-AI Teaming roadmap for Commercial Air Transport; The evolution of human role in ATM & UTM and Flight Operations; Airports’ impact areas map (D8.1); Concept of “AI CRM” for pilots (SPIC, SPO) (D8.2); The Impact of AI on Future Aviation Safety Culture and initial safeguards
Still on-going and not mature enough at this stage