The design and implementation of a system for the automatic generation of narrative debriefs for AUV Missions
Johnson, Nicholas Adam R.
MetadataShow full item record
Increased autonomy allows autonomous underwater vehicles to act without direct support or supervision. This requires increased complexity, however, and a deficit of trust may form between operators and these complex machines, though previous research has shown this can be reduced through repeated experience with the system in question. Regardless of whether a mission is performed with real vehicles or their simulated counterparts, effective debrief represents the most efficient method for performing an analysis of the mission. A novel system is presented to maximise the effectiveness of a debrief by ordering the mission events using a narrative structure, which has been shown to be the quickest and most effective way of communicating information and building a situation model inside a person’s mind. Mission logs are de-constructed and analysed, then optimisation algorithms used to generate a coherent discourse based on the events of the missions with any required exposition. This is then combined with a timed mission playback and additional visual information to form an automated mission debrief. This approach was contrasted with two alternative techniques: a simpler chronological ordering; and a facsimile of the current state of the art. Results show that participant recall accuracy was higher and the need for redundant delivery of information was lower when compared to either of the baselines. Also apparent is a need for debriefs to be adapted to individual users and scenarios. Results are discussed in full, along with suggestions for future avenues of research.