Scalability of Robotic Controllers: Effects of Progressive Autonomy on Intelligence, Surveillance, and Reconnaissance Robotic Tasks

Abstract

Abstract : Multiple task demands of robotic operators can result in cognitive overload. We investigated the utility of offloading some of the operator tasks by examining three levels of automation: (1) teleoperation, (2) semi-autonomy (teleoperation with obstacle avoidance) and, (3) autonomy (reflexive waypoint navigation with obstacle avoidance) for intelligence, surveillance, and reconnaissance of an area. Using a within-subjects design, twenty-seven Soldiers navigated a robot along a route littered with obstacles while looking for unexploded ordinance using the three levels of automation. They completed each condition twice, once while concurrently engaged in secondary cognitive tasks, and once without the secondary tasks. The secondary tasks interfered with course navigation in the teleoperation and semi-autonomous control conditions, but not in the autonomous control condition. Mental workload ratings were also higher in the teleoperation and semi-autonomous conditions than in the autonomous condition. There were more driving errors in the teleoperation condition than in the semi-autonomous and autonomous conditions. Although the automatic obstacle avoidance feature of the semi-autonomous condition reduced errors, it adversely impacted course completion times and made the operators feel as though they were competing with the feature for control of the robot. The Soldiers preferred full autonomy for all the navigation/maneuver tasks.