Advancing beyond conformal conflict resolution in air traffic control: balancing efficiency and conformance

• Main Authors: Guleria, Yash, Pham, Duc-Thinh, Alam, Sameer
• Other Authors: 11th International Conference on Research in Air transportation (ICRAT 2024)
• Format: Conference or Workshop Item
• Language: English
• Published: 2024
• Subjects: Computer and Information Science; Engineering; Air traffic conflict resolution; Machine learning; Supervised learning; Air traffic controllers; Conformal automation; Human-AI teaming
• Online Access: https://hdl.handle.net/10356/179379
• Institution: Nanyang Technological University

While favored by air traffic controllers (ATCOs), ATCO- conformal conflict resolution often lacks operational efficiency due to such potential issues as excessive aircraft deviations and higher separation buffers. This study investigated the potential of advancing beyond conformal conflict resolution advisories in en- route air traffic control by balancing optimality and conformance of the advisories to bridge the gap between ATCO conformance and maneuver efficiency. Firstly, a machine learning (ML) approach is proposed to generate various conflict resolution advisories, includ- ing balanced options. Then, initial human-in-the-loop experiments with 6 experienced ATCOs were conducted, wherein ATCOs were required to rank conflict resolution advisories based on their preferences, for the presented conflict scenarios. Given the trade-off between conformance and efficiency of the conflict resolution ma- neuvers, this research investigated the ATCOs’ preferences among the resolution options provided, for the conflict scenarios. Results indicated a strong preference for conformal conflict resolution in 70 of 114 scenarios collected, with a significant inclination towards balanced resolutions in 62 scenarios, suggesting a move towards more efficient strategies without compromising conformance. Cu- mulatively, conformal and balanced conflict resolutions were the most favorable and second most favorable choices for 77.19% and 75.4% of the scenarios, respectively. ATCO feedback also indicated that such a conflict resolution advisory mechanism could prove useful in improving controller decision-making. These findings highlighted the need for further research into adaptive conflict resolution models that can potentially better balance the tradeoffs between optimal- and conformal resolution models.