Page 28 - Innovations in Intelligent Machines

16 M.L. Cummings et al.
WTI is a subset of IT, it is not explicitly included (although the measurement
technique of IT will determine whether or not WTI should be included in the
denominator.)

X Y Z
WT = WTI i + WTQ j + WTSA k (2)
i=1 j=1 k=1
NT
FO = Y Z + 1 (3)
IT + WTQ +
j=1 k=1 WTSA k
While the revised fan-out (3) includes more variables than the original
version, the issue could be raised that the additional elements may not pro-
vide any meaningful or measurable improvement over the original equation
which is simpler and easier to model. Thus to determine how this modification
affects the fan-out estimate, we conducted an experiment with a UAV simu-
lation test bed, holding constant the number of vehicles a person controlled.
We then measured all times associated with equations 1 and 3 to demonstrate
the predictions made by each equation. The next section will describe the
experiment and results from this effort.

3.2 Experimental Analysis of the Fan-out Equations

In order to study operator control of multiple UAVs, a dual screen simulation
test bed named the Multi-Aerial Unmanned Vehicle Experiment (MAUVE)
interface was developed (Fig. 3). This interface allows an operator to effec-
tively supervise four independent homogeneous UAVs simultaneously, and
intervene as the situation requires. In this simulation, users take on the role
of an operator responsible for supervising four UAVs tasked with destroying
a set of time-sensitive targets in a suppression of enemy air defenses (SEAD)
mission. The left side of the display provides geo-spatial information as well
as a command panel to redirect individual UAVs. The right side of the display
provides temporal scheduling decision support in addition to data link “chat
windows” commonly in use in the military today [12]. Details of the display
design such as color mappings and icon design are discussed elsewhere [13].
The four UAVs launched with a pre-determined mission plan, so initial
target assignments and routes were already completed. The operator’s pri-
mary job in the MAUVE simulation was to monitor each UAV’s progress,
replan aspects of the mission in reaction to unexpected events and in some
cases manually execute mission critical actions such as arming and firing of
payloads. The UAVs supervised by participants in MAUVE were capable of 6
high-level actions: traveling en route to targets, loitering at specific locations,
arming payloads, firing payloads, performing battle damage assessment, and
returning to base, generally in this order.
In the MAUVE simulations, flight control was fully automated as was the
basic navigation control loop in terms of heading control. Operators were occa-
sionally required to replan route segments due to pop-up threat areas so the