Adaptive Levels of Automation for UAV Supervisory Control

Unmanned Aerial Vehicles hold great promise. As engineers provide UAVs with more autonomy, manpower requirements decrease. But increased autonomy replaces judgment with heuristics. It often degrades the remaining human performance. Balance is needed. Phase I demonstrates the feasibility of using game technology to find this balance. “Avant Guard” models a mission with multiple small UAVs. This game simulates look-ahead convoy protection in a hostile urban situation. The game has clear value as a research instrument. Its cognitive challenges are organized into simple tasks. Each task fits cleanly into an overall model of autonomous behavior. Each task is independently automated – at precisely defined levels. Calibrated to established standards, they are easily compared to one another and to the findings of other researchers. Besides this range of static levels, the experimenter can assign an adaptive mechanism to any task. Several distinct natural mechanisms are provided. Alternatively, the researcher may choose to develop an original adaptive mechanism by supplying simple rules to a finite state machine. Avant Guard enables new research in cognitive engineering. The result will be design patterns that promote responsive interplay between man and autonomous entities. The problem of balancing autonomy is increasingly important in both military and civilian enterprises.