Management of Multiple Uavs with Imperfect Automation

Forty-two participants manipulated two or four unmanned aerial vehicles (UAVs) while monitoring an adjacent display for camouflaged tanks. They were sometimes supported in the tank-detection task by an automated target recognition system that operated at either a reliability of .9 with a threshold designed to provide an equal number of false-alarms and misses, .6 with a low threshold producing more automation false alarms, or .6 with a high threshold producing more automation misses. As taskload doubled, performance on the UAV task was significantly reduced. The effect was not mediated by the presence of automated aids, though the aids did influence performance in the tank-detection task. Tank detection was improved by both the highly-reliable aid and (to a lesser extent) by the miss-prone aid, but was not improved (and sometimes hurt) by the false-alarm prone condition. The results support the independence model of reliance and compliance proposed by Meyer (2001).