摘要: |
Forty two pilots flew a systems integration lab (SIL) UAV simulator, while controlling either two or four UAVs with imperfect automation in a dual task experiment. One task required supervising the flight path trajectories of the UAVs, the other involved monitoring a 3D image window for camouflaged tanks, a perceptually demanding vigilance task. The latter was sometimes supported by an imperfectly reliable automated target recognition system. Its reliability was either 0.9, or, in two versions, around 0.6. The latter two differed according to the threshold setting, biasing automation errors toward miss-prone or false-alarm prone performance. In addition, some pilots used a baseline (no automation) condition. The results indicated a pronounced effect of increasing workload from two to four UAVs. This effect was much not buffered by automation, and effected performance on both the UAV task and the tank task. Performance on the tank task was slightly improved by high reliability automation, and degraded by lower reliability automation, particularly when the latter was false alarm prone. The difference in automation threshold setting is interpreted in terms of the model of reliance and compliance proposed by Meyer. / NOTE: Technical rept. / Supplementary Notes: Prepared in cooperation with Micro Analysis and Design, Boulder, CO. Sponsored by Department of the Army, Washington, DC. / Availability Note: Order this product from NTIS by: phone at 1-800-553-NTIS (U.S. customers); (703)605-6000 (other countries); fax at (703)605-6900; and email at orders@ntis.gov. NTIS is located at 5285 Port Royal Road, Springfield, VA, 22161, USA. |