Data-Driven Modeling and Analysis Framework for HMI

Introduction

  • Objective: to design a method to intelligently classify intent conflict events and enable detailed analysis by subject matter experts (SMEs)

Consensus-based Multi-step Clustering and Model Creation

  • Intent conflicts (or mode confusions) detected by the intent-based mode-confusion detection algorithm are taken as inputs.
  • Similar types of HMI issues (or mode confusions) are automatically grouped into several consensus clusters
  • Each consensus cluster can be succinctly represented by ‘representative’ event

HMI Issue Detection Results

  • Data reduction capability of consensus-based multi-step clustering technique
  • Example: Cluster 32
    • Feature consistency: 83%
    • No. primary conflicts: 11
    • No. intent conflict events: 8,593
    • Intent conflicts only in the vertical dimension
    • Long-duration delays in autopilot beginning descent after a new MCP target is selected
  • Intent: -1: Descend, 0: Constant Altitude, 1: Climb
  • VMODE: 1: Vertical Auto, 2: Altitude Hold
  • TMODE: 9: Mach off
  • Flight Phase: 5: Cruise, 6: Descent

Related Publication

  • A. Vaidya, S. Lee, and I. Hwang, “Data-Driven Modeling and Analysis Framework for Cockpit Human–Machine Interaction Issues ,” AIAA Journal of Aerospace Information Systems,October 2016, (online) http://dx.doi.org/10.2514/1.I01046