Experimental Evaluation of a Scalable Autonomy Concept for Cognitive Agents Aboard Reconnaissance UAVs

The concept of scalable autonomy with a guided access to lower automation levelsallows human UAV pilots to select the level of autonomy and enables the onboard automation to understand the pilot’s intent and offer support. To evaluate this concept, we conducted flightand simulationexperiments with German military personnel performing reconnaissance missions with a small UAV. We compared threeconfigurations. The high autonomy configuration completely prevents access to low-level functions. The naive approach configuration allows unrestricted access to all lower-level functions. The guided access configuration restricts access by forcing the pilot to communicate his intent first by entering mission tasks. As dependent variables, we measured mission performance and workload by the achievement of objectives, questionnaires (e.g. NASA-TLX) and secondary task performance evaluation.The low-level access configurations improved mission performance significantly, while keeping theworkload on a normal level. The subjective workload wasevenslightly reduced