Petri net-based modelling of human–automation conflicts in aviation

Analyses of aviation safety reports reveal that human–machine conflicts induced by poor automation design are remarkable precursors of accidents. A review of different crew–automation conflicting scenarios shows that they have a common denominator: the autopilot behaviour interferes with the pilot's goal regarding the flight guidance via ‘hidden’ mode transitions. Considering both the human operator and the machine (i.e. the autopilot or the decision functions) as agents, the authors propose a Petri net model of those conflicting interactions, which allows them to be detected as deadlocks in the Petri net. In order to test the authors' Petri net model, they designed an autoflight system that was formally analysed to detect conflicting situations. The authors identified three conflicting situations that were integrated in an experimental scenario in a flight simulator with 10 general aviation pilots. The results showed that the conflicts that the authors had a-priori identified as critical had impacted the pilots' performance. Indeed, the first conflict remained unnoticed by eight participants and led to a potential collision with another aircraft. The second conflict was detected by all the participants but three of them did not manage the situation correctly. The last conflict was also detected by all the participants but provoked typical automation surprise situation as only one declared that he had understood the autopilot behaviour. These behavioural results are discussed in terms of workload and number of fired ‘hidden’ transitions. Eventually, this study reveals that both formal and experimental approaches are complementary to identify and assess the criticality of human–automation conflicts.

Language

  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01523117
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Apr 24 2014 11:58AM