A Flexible Delegation-Type Interface Enhances System Performance in Human Supervision of Multiple Robots: Empirical Studies With RoboFlag

Three experiments and a computational analysis were conducted to investigate the effects of a delegation-type interface on human supervision of simulated multiple unmanned vehicles. Participants supervised up to eight robots using automated behaviors ("plays"), manual (waypoint) control, or both to capture the flag of an opponent with an equal number of robots, using a simple form of a delegation-type interface, Playbook. Experiment 1 showed that the delegation interface increased mission success rate and reduced mission completion time when the opponent "posture" was unpredictably offensive or defensive. Experiment 2 showed that performance was superior when operators could flexibly use both automated behaviors and manual control, although there was a small increase in subjective workload. Experiment 3 investigated additional dimensions of flexibility by comparing delegation interfaces to restricted interfaces. Eight interfaces were tested, varying in the level of abstraction at which robot behavior could be tasked and the level of aggregation (single or multiple robots) to which plays could be assigned. Performance was superior with flexible interfaces for four robots, but this benefit was eliminated when eight robots had to be supervised. Finally, a computational analysis using task-network modeling and Monte Carlo simulation gave results that closely paralleled the empirical data on changes in workload across interface type. The results provide initial empirical evidence for the efficacy of delegation-type interfaces in human supervision of a team of multiple autonomous robots.