Periodic Reporting for period 1 - JARVIS (JUST A RATHER VERY INTELLIGENT SYSTEM)
Période du rapport: 2023-06-01 au 2024-12-31
The fast penetration of automation and Artificial Intelligence (AI) is boosting the adoption of automated and autonomous systems across industries. Known as digitalisation, this trend is also rapidly changing aviation. Driven by the increasing complexity of the aviation ecosystem (aircraft, air traffic control – ATC, airports), digitalisation provides solutions in the form of Digital Assistants (DAs) that, by teaming with their human counterparts (pilots, ATC operators, airport operators), support the execution of tasks to ensure safe and profitable operations in complex scenarios. The JARVIS project, led by Collins Aerospace, aims at developing and validating three ATM solutions: an Airborne DA (AIR-DA, TRL4), an ATC-DA (TRL4), and an Airport DA (AP-DA, TRL6).
The AIR-DA will increase the level of automation in the flight deck and thanks to AI-based machine intelligence will act as an enabler for reduced crew operations and single pilot operations. The adoption of the AIR-DA will allow pilots to manage complex scenarios without compromising safety or security, while reducing pilot workload. The ATC-DA will increase the level of automation in air traffic control, where environmental Key Performance Indicators (KPIs) and airspace capacity management will benefit from the adoption of AI-based technologies. ATC-DA targets support for both tactical conflict resolution , tactical recommendations, flight plans syntactical errors correction and short term traffic forecasting. Finally, the AP-DA will increase the level of automation in airports, enhancing safety and for scenarios such as intrusion detection to enhance runways operations as of aircraft turnaround and terminal passengers' flows
The adoption of AI-driven technologies in the aviation ecosystem represents a promising breakthrough but also entails challenges. JARVIS will address key challenges common to the three different DAs:
I. Assured AI design, to deliver trustworthy, explainable, safe, and ethical decision-making algorithms;
II. Human AI Teaming (HAT), to deliver human-centric designs to maximise the teamwork between humans and autonomous systems;
III. Big data and cloud infrastructures for the proper management of data moving from a centralised architecture to a more edge-to-cloud architecture.
These challenges are addressed in collaborations with regulatory agencies (including EASA) and the international AI community.
The AIR-DA will increase the level of automation in the flight deck and thanks to AI-based machine intelligence will act as an enabler for reduced crew operations and single pilot operations. The adoption of the AIR-DA will allow pilots to manage complex scenarios without compromising safety or security, while reducing pilot workload. The ATC-DA will increase the level of automation in air traffic control, where environmental Key Performance Indicators (KPIs) and airspace capacity management will benefit from the adoption of AI-based technologies. ATC-DA targets support for both tactical conflict resolution , tactical recommendations, flight plans syntactical errors correction and short term traffic forecasting. Finally, the AP-DA will increase the level of automation in airports, enhancing safety and for scenarios such as intrusion detection to enhance runways operations as of aircraft turnaround and terminal passengers' flows
The adoption of AI-driven technologies in the aviation ecosystem represents a promising breakthrough but also entails challenges. JARVIS will address key challenges common to the three different DAs:
I. Assured AI design, to deliver trustworthy, explainable, safe, and ethical decision-making algorithms;
II. Human AI Teaming (HAT), to deliver human-centric designs to maximise the teamwork between humans and autonomous systems;
III. Big data and cloud infrastructures for the proper management of data moving from a centralised architecture to a more edge-to-cloud architecture.
These challenges are addressed in collaborations with regulatory agencies (including EASA) and the international AI community.
In the first half of the project, most efforts were concentrated on the investigation, the high-level and detailed design, and the preliminary development of the DAs’ features. Each DA developed in JARVIS is defined in terms of features by its scope and its capacities of teaming in delivering human tasks. Such features define the functionality (behaviour) of a product or a system that is required to meet its user needs.
AIR-DA investigates three features as illustrated in the attached JARVIS concept figures:
• ATC instruction and clearances management
• Flight strategy advisement
• Arrival and runway predictions
ATC-DA investigates four features:
• Conflict Resolution Advisor
• Tactical Opportunities Recommender
• Short Term Forecaster
• Flight Plan Corrector
AP-DA investigates five features:
• Detection of vehicle taxiway incursion
• Detection of foreign object debris in the runway environment
• Detection of wildlife activities in the runway environment
• Prediction of aircraft turnaround process
• Prediction of terminal passenger flows
The activities performed so far have contributed to the design of the three Digital Assistant solutions, including the operational concept through their respective features that reflect the different functionalities that will be provided, description of the new operating method, the necessary requirements, the expected impact, the technical development, and preparation of prototypes. In addition, the solutions have been focused on preparing the validation exercises identifying the use cases to be addressed. The safety, security and human performance assessment plans have been part of the validation preparation, ensuring all these aspects will be measured and reported to demonstrate the feasibility of the JARVIS solutions.
JARVIS solutions have also been interacting with a transversal activity (Foundational AI) whose objectives are to support the development of three key topics (Human AI Teaming -HAT, Assured AI Design, and Big Data & Cloud Infrastructures) within the solutions and organise the collaboration with EASA through regular meetings with the JARVIS teams. In the first half of the project, the Foundational AI activity has defined a common methodology for the three whitepapers to be released at the end of the project, set up a digital collaboration space and reinforced engagement of the solutions. At mid-project, a paper on HAT challenges has been submitted, and small-scale studies on HAT and two abstracts are being prepared for submission.
AIR-DA investigates three features as illustrated in the attached JARVIS concept figures:
• ATC instruction and clearances management
• Flight strategy advisement
• Arrival and runway predictions
ATC-DA investigates four features:
• Conflict Resolution Advisor
• Tactical Opportunities Recommender
• Short Term Forecaster
• Flight Plan Corrector
AP-DA investigates five features:
• Detection of vehicle taxiway incursion
• Detection of foreign object debris in the runway environment
• Detection of wildlife activities in the runway environment
• Prediction of aircraft turnaround process
• Prediction of terminal passenger flows
The activities performed so far have contributed to the design of the three Digital Assistant solutions, including the operational concept through their respective features that reflect the different functionalities that will be provided, description of the new operating method, the necessary requirements, the expected impact, the technical development, and preparation of prototypes. In addition, the solutions have been focused on preparing the validation exercises identifying the use cases to be addressed. The safety, security and human performance assessment plans have been part of the validation preparation, ensuring all these aspects will be measured and reported to demonstrate the feasibility of the JARVIS solutions.
JARVIS solutions have also been interacting with a transversal activity (Foundational AI) whose objectives are to support the development of three key topics (Human AI Teaming -HAT, Assured AI Design, and Big Data & Cloud Infrastructures) within the solutions and organise the collaboration with EASA through regular meetings with the JARVIS teams. In the first half of the project, the Foundational AI activity has defined a common methodology for the three whitepapers to be released at the end of the project, set up a digital collaboration space and reinforced engagement of the solutions. At mid-project, a paper on HAT challenges has been submitted, and small-scale studies on HAT and two abstracts are being prepared for submission.
The results and outcomes of the JARVIS project are expected to contribute to the following impacts:
• Environment: Optimization of aircraft trajectories, reduced fuel consumption and potentially reducing the environmental footprint.
• Capacity: With AI, airspace capacity shortages could be addressed, enabling dynamic configurations of the airspace and allowing dynamic space separation between aircraft.
• Cost efficiency: Enriching aviation datasets, AI will help air traffic controllers, pilot, stakeholders and decision-makers to bring benefits in cost-efficiency.
• Safety: The research will be done to ensure at least the same level of safety now existent.
• Operational efficiency: Digital Assistants will increase predictability in airspace traffic and boost punctuality.
The strategy for the exploitation of the JARVIS results and the corresponding actions to ensure their successful transfer into tangible benefits for the aviation community will be further developed in the second phase of the project.
• Environment: Optimization of aircraft trajectories, reduced fuel consumption and potentially reducing the environmental footprint.
• Capacity: With AI, airspace capacity shortages could be addressed, enabling dynamic configurations of the airspace and allowing dynamic space separation between aircraft.
• Cost efficiency: Enriching aviation datasets, AI will help air traffic controllers, pilot, stakeholders and decision-makers to bring benefits in cost-efficiency.
• Safety: The research will be done to ensure at least the same level of safety now existent.
• Operational efficiency: Digital Assistants will increase predictability in airspace traffic and boost punctuality.
The strategy for the exploitation of the JARVIS results and the corresponding actions to ensure their successful transfer into tangible benefits for the aviation community will be further developed in the second phase of the project.