Page 26 - Innovations in Intelligent Machines

14 M.L. Cummings et al.
operator capacity, regardless of vehicle dynamics, communication latency,
decision support, and display designs.

3 Predicting Operator Capacity through Temporal
Constraints
While little research has been published concerning the development of a
predictive operator capacity model for UAVs, there has been some previous
work in the unmanned ground vehicle (robot) domain. Coining the term “fan-
out” to mean the number of robots a human can effectively control, Olsen et al.
[10, 11] propose that the number of homogeneous robots or vehicles a single
individual can control is given by:
NT + IT NT
FO = = + 1 (1)
IT IT
In this equation, FO (fan-out) is dependent on NT (Neglect Time), the
expected amount of time that a robot can be ignored before its performance
drops below some acceptable threshold, and IT (Interaction Time) which is
the average time it takes for a human to interact with the robot to ensure it
is still working towards mission accomplishment. Figure 1 demonstrates the
relationship of IT and NT.
While originally intended for ground-based robots, this work has direct
relevance to more general human supervisory control (HSC) tasks where oper-
ators are attempting to simultaneously manage multiple entities, such as in
the case of UAVs. Because the fan-out adheres to Occam’s Razor, it provides
a generalizable methodology that could be used regardless of the domain, the
human-computer interface, and even communication latency problems. How-
ever, as appealing as it is due to its simplicity, in terms of human-automation
interaction, the fan-out approach lacks two critical considerations: 1) The
important of including wait times caused by human-vehicle interaction, and
2) How to link fan-out to measurable “effective” performance. These issues
will be discussed in the subsequent section.

IT
Can insert ITs
for additional
robots here

NT

Segment
IT+NT
Fig. 1. The relationship of NT and IT for a Single Vehicle