About

Context and Motivation

Autonomous agents are introduced in many tasks to discharge ATCOs in their activity. Being able to coordinate with them is crucial, especially when facing unanticipated events (such as alarms, pilot requests, emergency situations, etc.). To prevent this coordination from inducing additional workload and complexity it is essential to rely on a human centric approach based on human-autonomy teaming and explainability models, as well as physiological state monitoring.​

DIALOG will design an assistant that helps ATCOs respond to pilots’ requests in an efficient collaboration-promoting manner.

Methodology

PHASE 1
———————————––
OPERATIONAL ANALYSIS & BASELINE DEFINITION:
This phase focuses on understanding challenges in AI-ATM collaboration. Activities include reviewing ATCO decision-making models, defining operational gaps, drafting OSED v1, and outlining validation needs based on SESAR’s AI Roadmap.

PHASE 2
———————————––
REQUIREMENTS DEFINITION & VALIDATION PLANNING:
This phase formalizes requirements to guide testing and simulations. Key activities include refining OSED (v2), defining functional and technical requirements for AI-assistance, developing the Exploratory Research Plan (ERP), and identifying performance indicators and risk factors for AI-enabled tools.

PHASE 3
———————————––
PROOF-OF CONCEPT & SIMULATION-BASED VALIDATION
This phase tests AI-assisted human-AI teaming through HITL simulations. Key activities include developing DIALOG prototypes, conducting validation with relevant scenarios, testing workload monitoring and task allocation, and refining models based on feedback.

PHASE 4
———————————––
REQUIREMENTS DEFINITION & VALIDATION PLANNING:
Consolidates validation results and assesses DIALOG’s operational, economic, and usability impact, including finalizing OSED v3, documenting validation in the ERR, developing AI-human guidelines, and conducting an Economic Evaluation.

DIALOG Framework

KEY CHALLENGES FOR MAKING AUTOMATION A TEAM PLAYER

‘If not well managed, the introduction of AI in ATM will add complexity to an already complex system’

DIALOG’S SOLUTIONS

  • Automatic Speech Recognition to transcribe pilot-ATCO communications
  • Hierarchical Task Analysis (HTA) to identify ATCO’s goals and tasks
  • Attention tracking to identify current task
  • Attention tracking and HTA to identify the task to come
  • Physiological monitoring (cardiac, cerebral, occular, electrodermal activty) to characterize ATCOs’ workload and attention0
  • Machine learning and data fusion to classify state
  • Theories from the human-human joint action to identify essential information sharing
  • Human-Autonomy teaming and task sharing as a framework for efficient collaboration
  • CWP prototype designed with high level of explainability