An agent-based approach for air traffic conflict resolution in a flow-centric airspace

• Main Authors: Guleria, Yash, Pham, Duc-Thinh, Alam, Sameer
• Other Authors: School of Mechanical and Aerospace Engineering
• Format: Conference or Workshop Item
• Language: English
• Published: 2023
• Subjects: Engineering; Air Traffic Control; Flow-Centric Operations
• Online Access: https://hdl.handle.net/10356/170973; https://2023.ieee-itsc.org/
• Institution: Nanyang Technological University

The air traffic control paradigm is shifting from sector-based operations to flow-centric approaches to address the scalability limitations of geographically-bound air traffic sectors and to meet the growing demands of air traffic. These future concepts of operations differ from traditional air traffic operations, especially in maintaining safe separation between flights. Flow-centric operations are characterized by maintaining safe separation between traffic flows (both interflow as well as intra-flow), in contrast to current standards of maintaining separation between pairs of aircraft. This paper proposes a novel approach for resolving air traffic conflicts in flow-centric en-route airspace by employing a combination of a model-free Deep Reinforcement Learning policy and a self-stabilizing graph structure. The problem is formulated as a sequential decision-making task in a large action space, requiring a series of decisions to be made over time to resolve potential conflicts at both the inter-flow and intra-flow levels, while adhering to the flow plans and subsequently reaching the destination. Model performance is evaluated by measuring the frequency of safe separations achieved and the efficiency of the maneuvers (deviation from the flow plans). Despite the intra-flow and inter-flow speed uncertainties and the dynamic behavior of the shape of the flows due to variations in the number of aircraft in each flow in every scenario, the proposed approach achieves safe separations for 100% of the scenarios evaluated. The results also demonstrate that despite the induced delay due to conflict resolution maneuvers, the flows closely adhere to their original flow plan. This approach can be used to develop intelligent conflict resolution advisory tools in a flowcentric airspace paradigm.