Page 37 - Innovations in Intelligent Machines

Predicting Operator Capacity for Supervisory Control of Multiple UAVs 25
the revised model is both more conservative and closer to the actual num-
ber of vehicles under successful control. While under low workload, both the
experiment and prediction indication that operators could have controlled
more vehicles than four, the only high workload scenario in which operators
demonstrated any spare capacity was with the super-active (management-
by-exception) decision support. Moreover, wait time caused by the lack of
situation awareness dominated overall wait time. In addition, this research
demonstrates that both workload and automated decision support can dra-
matically affect wait times and thus, operator capacity.
While more pessimistic than the original fan-out equation (1), the revised
fan-out equation can really only be helpful for broad “ballpark” predictions
of operator capacity. This methodology could provide system engineers with
a system feasibility metric for early manning estimations, but what primar-
ily limits either version of the fan-out equation is the inability to represent
any kind of cost trade space. Theoretically fan-out, revised or otherwise, will
predict the maximum number of vehicles an operator can effectively control,
but what is effective is often a dynamic constraint. Moreover, the current
equations for calculating fan-out do not take into account explicit perfor-
mance constraints. In light of the need to link fan-out to some measure of
performance, as well as the inevitability of wait times introduced by human
interaction, we propose that instead of a simple maximum limit prediction,
we should instead find the optimal number of UAVs such that the mission
performance is maximized.

3.4 The Overall Cost Function
Maximizing UAV mission performance is achieved when the overall per-
formance of all of the vehicles, or the team performance, is maximized.
Consider multiple UAVs that need to visit multiple targets, either for destruc-
tion (SEAD missions as discussed previously) or imaging (typical of Intelli-
gence, Search, and Reconnaissance (ISR) missions). A possible cost function
is expressed in (4):

C = Total Fuel Cost + Total Cost of Missed Targets
+Total Operational Cost (4)

Total Fuel Cost is the amount of fuel spent by all the vehicles for the
duration of the mission multiplied by the cost of consuming that fuel. The
Total Cost of Missed Targets is the number of targets not eliminated by
any of the UAVs multiplied by the cost of missing a single target. The
Total Operational Cost is the total operation time for the mission multiplied
by some operational cost per time unit, which would include costs such as
maintenance and ground station operation costs. This more detailed cost func-
tion is given in (5).