Data-Driven Conflict Resolution Generator

  • The current-day separation assurance is provided by air traffic controllers, but most of conflict resolution approaches are model-based (using flight dynamics)

  • Objective: to build a model that generates resolution methods (called Data-Driven Resolution Generator, D2RG) based on knowledge extracted and characterized from flight data which provides separation assurance

Conflict Dataset Construction

  • Resolution Method (Label)

  • Conflict Situation (Features)

Resolution Type: Hierarchical Classification

  • Conflict type matching: use flight phases

  • Upper-level classifier: maneuvering dimensions

  • Lower-level classifier: resolution types within specific dimension

  • Output: Ranked list of resolution types with corresponding likelihoods

  • Example: true label = Direct-to

Resolution Parameters: Optimal Parameter Design

  • Use analytic or data-driven approach to find resolution parameters, which guarantee separation assurance

<Horizontal/Speed>

Analytically find optimal heading/speed adjustment which can resolve conflict

<Vertical>

Use classification algorithm with the same feature set

D2RG Test Results with Simulated Flight Data

  • Among 54,300 flights over CONUS, there are 14,599 conflicts

  • Resolution Types: Accuracy 84.12%

  • Resolution Parameters: D2RG guarantees separation assurance, i.e. all the conflicts are safely resolved

Related Publication

  • K. Kim, R. Deshmukh and I. Hwang, “Development of Data-Driven Conflict Resolution Generator for En-Route Airspace,” AIAA Journal of Aerospace Information Systems, (submitted on December 10, 2017; under review)