Trust, Attention, Situation Awareness, and Workload in Multi-AUV Control: Part 1 – Anomaly Detection and AUV Monitoring

Heterogenous autonomous underwater vehicles are useful in rapid environmental assessment (REA) where operating environments are dynamic and uncertain, and human physiological barriers exist. However, monitoring vehicles with poor communication bandwidth and lengthy missions can reduce operator attention, situation awareness (SA) and trust in autonomous systems. In this first of two studies, 95 participants operated a highly parameterised multi-drone dual-task simulator of a real-world prototype system, using data from field trials. The primary task involved monitoring a 30-minute underwater REA, fixing drone navigational (“drift”) errors to induce attention, and the secondary task involved anomaly detection (image classification) of underwater images to simulate post-mission analysis. Participants periodically reported trust and SA. We assessed trust, attention, and SA in different auditory environments. Bayesian multilevel modelling found that trust remained stable despite intermittent errors, induced attention did not improve situational awareness, and tone-based alerts in noisy environments led to faster responses to drone errors.