Human-Agent Teaming for Multi-Robot Control and Effects of Individual Differences

Abstract

A military multitasking environment was simulated and three experiments were conducted to examine the effects of an intelligent agent, RoboLeader, on the performance of robotics operators. In the first experiment, RoboLeader was perfectly reliable; in the second experiment, RoboLeader’s recommendations were manipulated to be either false-alarm prone or miss prone, with a reliability level of either 60% or 90%; in the third experiment, the capabilities of RoboLeader were expanded to deal more specifically with dynamic retasking requirements based on various battlefield developments in pursuit of moving targets in urban environments. Results of Experiment 1 showed that RoboLeader, when perfectly reliable, was effective in reducing the operators’ mission times in target search tasks, although significant benefits of RoboLeader on the operators’ concurrent task performance and workload were not observed. Results of Experiment 2 show that the type of RoboLeader imperfection affected operator’s performance of tasks involving visual scanning (target detection, route editing, and situation awareness). Results of Experiment 3 showed that participants’ primary task benefited from RoboLeader on all level-of-autonomy conditions compared with the manual performance. Across the experiments, participants with higher spatial ability consistently outperformed those with lower spatial ability in tasks that required the most visual scanning, regardless of the experimental conditions. Participants’ self-assessed attentional control and video gaming experience were found to impact their overall multitasking performance. Frequent video gamers also demonstrated significantly better situation awareness of the mission environments.