Page 25 - Innovations in Intelligent Machines

Predicting Operator Capacity for Supervisory Control of Multiple UAVs 13
found that if system reliability decreased in the control of multiple UAVs, trust
declined with increasing numbers of vehicles but improved when the human
was actively involved in planning and executing decisions. These results are
similar to those experimentally found by Dixon et al. in that systems that
cause distrust reduce operator capacity [6]. Moreover, cultural components of
trust cannot be ignored. Tactical pilots have expressed inherent distrust of
UAVs as wingmen, and in general do not want UAVs operating near friendly
forces [7].
Reliability of the automation is only one of many variables that will deter-
mine operator capacity in MUAV control. The level of control and the context
of the operator’s tasks are also critical factors in determining operator capac-
ity. Control of multiple UAVs as wingmen assigned to a single seat fighter has
been found to be “unfeasible” when the operator’s task was primarily naviga-
ting the UAVs and identifying targets [8]. In this experimental study, the level
of autonomy of the vehicles was judged insufficient to allow the operator to
handle the team of UAVs. When UAVs were given more automatic functions
such as target recognition and path planning, overall workload was reduced.
In contrast to the previous UAVs-as-wingmen experimental study [6]
that determined that high levels of autonomy promotes overall performance,
Ruff et al. [5] experimentally determined that higher levels of automation
can actually degrade performance when operators attempted to control up
to four UAVs. Results showed that management-by-consent (in which a
human must approve an automated solution before execution) was superior to
management-by-exception (where the automation gives the operator a period
of time to reject the solution). In their scenarios, their implementation of
management-by-consent provided the best situation awareness ratings and
the best performance scores for controlling up to four UAVs.
These previous studies experimentally examined a small subset of UAVs
and beyond showing how an increasing number of vehicles impacted operator
performance, they were not attempting to predict any maximum capacity. In
terms of actually predicting how many UAVs a single operator control, there is
very little research. Cummings and Guerlain [9] showed that operators could
experimentally control up to 12 Tactical Tomahawk missiles given significant
missile autonomy. However, these predictions are experimentally-based which
limits their generalizability. Given the rapid acquisition of UAVs in the mili-
tary, which will soon follow in the commercial section, predictive modeling
for operator capacity will be critical for determining an overall system archi-
tecture. Moreover, given the range of vehicles with an even larger subset of
functionalities, it is critical to develop a more generalizable predictive mod-
eling methodology that is not solely based on expensive human-in-the-loop
experiments, which are particularly limited for application to revolutionary
systems.
In an attempt to address this gap, in the next section of this paper, we will
extend a predictive model for operator capacity in the control of unmanned
ground vehicles to a UAV domain [10], such that it could be used to predict