RUNNING HEAD: Application of Classic

crumcasteΤεχνίτη Νοημοσύνη και Ρομποτική

17 Νοε 2013 (πριν από 3 χρόνια και 6 μήνες)

95 εμφανίσεις


1







R
UNNING HEAD
:
Application of Classic








Application of
Classic
Human
Information P
rocessin
g Models


to
Unmanned Aircraft Systems

(UAS)








F. Richard Ferraro


University of North Dakota





















Address correspondence to:


F
. R. Ferraro
, Ph.D.

Dept. Psychology

University of North Dakota

Corwin
-
Larimore 215

319 Harvard St., Stop 8380

Grand Forks, ND 58202
-
8380

701
-
777
-
2414

701
-
777
-
3454

f_ferraro@und.nodak.edu

2



Abstract



The cu
rrent paper reviews classic theories
and models
of
information


processing and
pattern recognition in an attempt to apply them to
unmanned


aircraft systems (UAS).
These models deal with pattern recognition and provide


a framework for how UAS pilots a
nd operators recognize and interpret patterns.


UAS pilot and operator training as well as how human factors of UAS piloting are


relevant topics.


Key Words
:
Unmanned Aircraft Systems (UAS), information processing, human


factors


3

Application of Cla
ssic Human Information Processing Models to Unmanned Aircraft


Systems (UAS)


According to McCarley and Wickens (2005), the increased use of Unmanned
Aircraft Systems (UAS) in war
-
time and non
-
war time activities (border patrol
issues, traffic flow analys
is and monitoring, agricultural crop management
issues, etc.) will result in the necessary application of models and theories of
human information processing for adequate performance of the UAS. Much of that
performance relates to human factors. Consider
ing that accident rates for UASs
are approximately 25
-
30 percent higher than for manned aircraft (Williams,
2004), this application is critical.

One overriding theme related to the
Unmanned Aircraft Systems (UAS
s)
c
oncerns how patterns
(broadly defined
;

Sp
oehr & Lehmkuhle, 1982
)
are identified
quickly and accurately

(
Cummings & Guerlain, 2007; Dixon & Wickens, 2006
; Lam,
Mulder & van Paassen, 2007
).
Furthermore, the UAS operator (or pilot) is
physically separated from the vehicle.
Patterns here can be ope
rationally
defined in any number of ways, including identification of faces or components
of faces (Davies, Van Der Willik, & Morrison, 2000), to
targets moving on a
computer screen

or radar monitor
. The relatively simple process of identifying
a pattern
(a face) or items contained within a pattern (objects within a
pattern) is actually a very complex process, demanding much attention and
concentration on the part of the individual attempting to identify the item, or
items, in question (Galotti, 1999). Re
gardless of whether the identification
process is basic or applied, much of what is known about pattern identification
comes from cognitive psychology and the general topic of pattern recognition and
object identification offers several benefits to
UAS hum
an factors research i
n
the form of a) several decades of laboratory
-
based basic research into pattern
recognition (e.g., letter and word identification) and b) an extensive applied
focus in which processes underlying the identification of more real
-
world

4

o
bjects (faces, items in a screening device, etc.) are detailed theoretically
and empirically (Smyth, Collins, Morris, & Lavy, 1994). It is suggested that
the
continued
application of pattern recognition methods and theories associated
with cognitive psych
ology will be positive and relevant for issues related to
UAS human factors research
for the following reasons. First, theoretical models
can be provided which could be used to increase the accuracy of how patterns are
detected,
identified and processed.

It may the case that one specific model
does a better job of predicting pattern recognition accuracy than another model.
Second, application of which model does a better job at identifying objects
c
ould be used to construct a training program for
UAS pil
ots.

In other words,
UAS pilots
could train to become better able at identifying specific
patterns
and
features associated with certain
targets
.

Pattern Recognition: Models and Data


The exact process an individual goes through to identify a pattern (shap
e,
letter, word, face, object
, sound
) is complex and the result of various
physiological processes that are beyond the scope of this review (Hubel & Wisel,
1962). Regardless of the complexity of the underlying physiological, perceptual
and psychological p
rocesses involved in pattern recognition, models put forth
fall into three broad categories and include Template Models, Prototype Models
and Feature Models. These models are important because not only do they offer up
explanations of how patterns are iden
tified, they also permit identification of
patterns as they are presented in different situations and distortions (e.g.,
searching for a
specific target occurring with a field of similar but different
targets,
searching for a
distorted or degraded target
).

This applied nature of
pattern recognition, and the flexibility it affords, would seem to be crucial
for any discussion of
UAS human factors
and related issues.

Template Models


The basic assumption behind template models (Uhr, 1963) is that in the
real
world the brain stores away separate "pictures" or images of all objects it

5

encounters. These pictures are akin to what is imaged on the retina and
although they are not exact pictures, they preserve various spatial components
of the object in question.
Each template, then, is an exact replica of whatever
the individual has been exposed to. Many, many such templates are assumed to be
stored in long
-
term memory and they are labeled as such, based on the
individual's experience and exposure to the item in
question. A triangle, thus,
would be represented as a template of three contours each connected at their
respective endpoints which produce an internal representation labeled a
triangle. When we are exposed to a triangle on the real world, we search our
stored templates and do not stop the search until the template is found and
identified. This identification process is called pattern recognition. Despite
its simplicity, the template model is incomplete because it

does not take into
account the almost i
nfinite number of templates that would have to be searched
in order to find or identify a single pattern or object. That is, there are
many triangles out there in various shapes and sizes and in different
orientations. Template models assume that we stor
e each and every possible
template for a speci
fic object. This would soon in
undate the information
processing system, resulting in object recognition failure. As a result,
template models, while interesting, have never been taken too seriously as
adequat
e models of pattern recognition.

In fact, more recent work (i.e.,
Biederman, 1985; Biederman & Copper, 1991) has advanced a model suggesting that
relations among components of object features are critical for individuals
identify progressively more comple
x objects in the environment.
Template models
treat each stimulus as a separate entity, regardless of issues related to shape,
size, orientation and other factors that would render one object different than
another. A separate entity for each template of

an object would, as mentioned,
soon overload the information processing system (Schneider & Shiffrin, 1977)
resulting in poor performance and high error rates.

Prototype Models


6


Prototype models do not treat each stimulus separately and, rather,
examine

similarities among related stimuli
and
are essential for adequate
pattern recognition to occur. In other words, a prototype is like an average
rather than a single event. Thus, over time, we have been exposed to triangles
of varying shape, size, color a
nd orientation. This vast amount of information
is stored and when we are exposed

to a class of items that share

aspects of the
original stimulus, we are able to identify it as such. Thus, we are able to
identify someone we know from a stranger despite t
he fact that both faces have 2
eyes, 1 nose, 1 mouth and 2 ears. These two faces belong to the same class of
stimuli (faces) and we are able to identify any pattern of 2 eyes, 1 nose, 2
ears, and 1 mouth as a face. In prototype models, pattern recogniti
on is the
result of identification as well as abstraction of information from many
sources.

Furthermore, such abstraction of information allows for only the essential
attributes of a stimulus to be required for identification and, in the process,
relies
on attributes that possess the lowest level of variability. In turn,
pattern recognition within a prototype model is both efficient and economical
from the perspective of speed of response to identify a pattern as well as
storage requirements in memory (b
oth short
-
term and long
-
term). There is
considerable evidence that prototypes do exist
psychologically
(e.g., Attneave,
1957; Bransford & Franks, 1971; Posner & Keele, 1968) and Prototype models are
good at accounting for speed and accuracy of pattern rec
ognition under numerous
and varied stimulus presentation formats. This aspect of this model may be
attractive from a
UAS human factors
perspective in

that a prototype
could be
created and then used to test various stimuli against the prototype. Although
appealing, however, prototype models do not add much to the issue of how
information is internally represented or how such prototypes are actually stored
in memory. To remedy this potential problem area, feature models
have been
suggested
.


7

Feature Models


Given the vast physiological evidence suggesting that individual nerve
cells in the brain are maximally responsive to specific stimulus features, it is
no surprise that the feature model of pattern recognition has enjoyed wide
appeal both experimentally a
nd from an applied perspective (Gibson, Shapiro, &
Yonas, 1968; Selfridge, 1959). The basic mechanism behind the feature model is
that once a stimulus is encoded and displayed on the retina, component features
are analyzed and a list of potential features

is assembled. With the pas
sage of
time, this feature list

gradually decreases until enough critical features
remain and a decision is made. In essenc
e, a threshold level is reached

based
on the accumulation of sufficient featural information about the s
timulus in
question. Once the threshold is passed, a pattern recognition decision is made.
One
benefit of
feature
models is that they often combine
relevant aspects of
both the prototype and template models. Feature models adequately addre
ss the
interna
l representation
of the stimulus in question. Feature models are also
quite flexible enough to deal with stimuli that may come to the senses in
different distortions or different orientations. Modern facial recognition
systems currently in use in some ai
rports follow this pattern, being able to
identify individuals based on patterns contai
ned in their eyes, in much the
same
way that patterns on fingerprints are used to identify someone. Feature models
are also adequate because they rely on consistencies
within the human
physiological system, to the level of individual brain cells. While there are
also problems with the feature model (i.e., context effects are often difficult
to reconcile within strict feature models), a pattern recognition system that
re
lies on specific features for identification would appear to hold the most
promise in an applied setting. And, as mentioned, airports are also ready using
such feature model systems in the detection of suspected terrorists.

Model Summary


8


The three models

outlined above collectively have much to offer from an
applied perspective to
UAS i
ssues in aviation. While no single model can
account for all the data, each model consists of several attractive components
t
hat when combined offer
an adequate theoretica
l model that will be sufficiently
useful in matters related to
UAS performance and operation
.

Data: Task Conditions and Subject Characteristics


Much of the literature (Johnson, Paivio, & Clark, 1996)

on object and
pattern recognition has tested several ex
perimental manipulations and their
effects of performance, including picture naming, face processing, and object
and pattern recognition. For the present review, the specific task conditions
that seem relevant are practice
,
context
, and ability
.

Practic
e


As a general observation, practice tends to speed up processing
requirements of objects, including responses to those objects (Bartram, 1973;
Snodgrass & Hirschman, 1994). The term repetition priming is often invoked to
address this issue: repetition o
f a stimulus (even if one
r
epetition) has been
shown to dramatically speed up responses on subsequent stimuli that are
identical. There is also evidence that this practice effect can last up to 6
weeks, even though explicit memory of the items presented i
s no longer valid
(Brown, et al, 1991). Thus, a subj
ect cannot overtly recognize an

object, but
their response time performance indicates that they have indeed identified it by
responding faster on each repetition. Such repetition is not affected by obje
ct
size, retinal position, or left
-
right orientation. This is important, as such
variables are hardly ever controlled sufficiently in the applied world or in
UAS
situations.
This characteristic may be especially relevant as the same type of
objects are r
epeatedly viewed by
UAS operators.

Context


Context can be both facil
itating or inhibiting
. A shoe
, in a background
of other shoes, may be difficult to identify. Conversely, a shoe presented

9

within a background of other
related but different
objects (soc
ks, razor, comb)
is
relatively easy to pick out. However, the context of the background
information is critical. Some items may "pop out", depending on the background
(Treisman & Gelade, 1980). This is especially the case if the object that pops
out is
somehow dissimilar to
other objects. Thus, an item in

a
UAS display
that
obviously does not belong there should be identified much easier than an o
bject
that is likely to be in
the display
. In other words, object identification can
be primed (or facilita
ted) by prior presentation of a verbal or non
-
verbal
stimulus semantically related to the object in question (Boyce & Pollatsek,
1992; Lupker & Williams, 1989). In an applied setting, it may be helpful to
auditorily present the names of items that are res
tricted in
the UAS display
, as
well as including pictures of such items. In this way, the
UAS operator
will
have
at least
two modalities (sight, sound) with which to use
for i
dentifying
potentially illegal or harmful objects
, T
he more codes available, the

better
performance typically will be (Paivio, 1971, 1986).

Ability


There are vast individual differences in many aspects of information
processing, especially within the domains of pattern recognition and object
identification (Paivio, 1971, 1986). Obje
ct naming and identification is also
related to both verbal and non
-
verbal ability. Many of these relationships have
come from performance based on standardized tests of object naming and
identification, such as the Boston Naming Test (Kaplan, Goodglass,
& Weintraub,
1983). One recommendation might be to have annual or semi
-
annual tests of
naming
ability
given to
the
UAS operator
s
,

and use tests that are standardized
for various ages and education levels (such as the BNT).

Summary


This
selective
review

was initiated to determine relevant human
information processing models that would have ready application to areas and
issues related to how
UAS operators
perform their job
.
Continued a
pplication of

10

various aspects of the three
classic

models
of pattern
recognition
outlined
would be a
recommendation in
current
and future UAS operator training.
It is
hoped that the integration of human information processing models and applied
human factors
performance activities will result in better overall
UAS operator

effici
ency
.


11

References


Attneave, F. (1957). Transfer of experience with class schemata. Journal
of Experimental Psychology, 54, 81
-
88.


Bartram, D. J. (1973). The effects of familiarity and practice on naming
pictures of objects. Memory & Cognition,

1, 101
-
105.


Biederman, I. (1985). Human image understanding: recent research and a
theory. Computer Vision, Graphics and Image Processing, 32, 29
-
73.


Biederman, I., & Copper, E. (1991). Priming contour
-
deleted images:
Evidence from intermediate repr
esentations in visual object recognition.
Cognitive Psychology, 23, 393
-
419.


Boyce, S. J., & Polletsek, A. (1992). Identification of objects in
scenes: The role of scene background in object naming. Journal of Experimental
Psychology: Learning, Memor
y & Cognition, 18, 531
-
543.


Bransford, J. D., & Franks, J. J. (1971). The abstraction of linguistic
ideas. Cognitive Psychology, 2, 331
-
350.


Brown, A. S., Neblett, D. R., Jones, T. C., & Mitchell, D. B. (1991).
Transfer of processing in repetition pri
ming: Some inappropriate findings.
Journal of Experimental Psychology: Learning, Memory, & Cognition, 17, 514
-
525.


Cummings, M. L., & Guerlain, S. (2007). Developing operator capacity
estimates of supervisory control of autonomous vehicles. Human Facto
rs, 49, 1
-
15.


Davies, G., Van Der Willik, P., & Morrison, L. J. (2000). Facial
composite production: A comparison of mechanical and computer
-
driven systems.
Journal of Applied Psychology, 85, 119
-
124.


Dixon, S. R., & Wickens, C. D. (2006). Automation
reliability in unmanned
aerial vehicle control: A reliance
-
compliance model of automation dependence in
high workload. Human Factors, 48, 474
-
486.


Galotti, K. M. (1999). Cognitive psychology in and out of the laboratory
(2nd Ed.). Belmont, CA: Brooks
-
Cole.


12


Gibson, E. J., Shapiro, F., & Yonas, A. (1968). Confusion matrices of
graphic patterns obtained with a latency measure. The analysis of reading
skill: A program of basic and applied research. Final Report project No. 5
-
1213. Cornell University,
78
-
96.


Hubel, D. H., & Wiesel, T. N. (1968). Receptive fields, binocular
interaction, and functional architecture in monkey striate cortex. Journal of
Physiology, 168, 215
-
243.


Johnson, C. J., Paivio, A., & Clark, J. M. (1996). Cognitive components
of

picture naming. Psychological Bulletin, 120, 113
-
139.


Kaplan, E., Goodgla
ss, H., & Weintraub, S. (
1983). Boston Naming Test:
Revised edition. Philadelphia: Lea & Febiger.


Lam, T. M., Mulder, M., & van Paassen, M. M. (2007). Haptic interface for
UAV
collision avoidance. International Journal of Aviation Psychology, 17, 167
-
195.


Lupker, S., & Williams, B. A. (1989). Rhyme priming

of pictures and
words: A lexical activation account. Journal of Experimental Pscyhology:
Learning, Memory, & Cognition,

15, 1033
-
1046.


McCarley, J. S., & Wickens, C. D. (2005). Human factors implications of
UAVs in the national airspace. Technical Report, Aviation Human Factors
Division, Institute of Aviation, University of Illinois at Urbana
-
Champaign.


Paivio, A. (197
1). Imagery and verbal processes. NY: Holt, Reinhart &
Winston.


Posner, M., & Keele, S (1968). On the genesis of abstract ideas. Journal
of Experimental Psychology, 77, 353
-
363.


Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic
proce
ssing. I. Detection, search, and attention. Psychological Review, 84, 1
-
66.


Selfridge, O. G. (1959). Pandemonium: A paradigm of learning. In: The
mechanization of thought processes. London: Her Majesty's Stationary Office.


13


Smith, M. M., Collins, A. F
., Morris, P. E., & Levy, P. (1994). Cognition
in action (2nd Ed.). Hillsdale, NJ: LEA.


Snodgrass, J. G., & Hirschman, E. (1994). Dissociations among implicit
and explicit memory tests: The role of stimulus similarity. Journal of
Experimental Psycholo
gy: Learning, Memory, & Cogntion, 20, 150
-
160.


Spoehr, K. T., & Lehmkule, S. W. (1982). Visual information processing.
San Francisco, CA: W. H. Freeman.


Treisman A., & Gelade, G. (1980). Feature
-
integration theory of
attention. Cognitive Psychology,
12, 97
-
136.


Uhr, L. (1963). "Pattern recognition" computers as models of form
perception. Psychological Bulletin, 60, 40
-
73.


Williams, K. W. (2004). A summary of unmanned aircraft accident/incident
data: Human factors implications. Technical Report D
OT/FAA/AM
-
04/24).
Washington, DC: Office of Aerospace Medicine, FAA.


14

10/30/
08


Kathryn Kelly, Ph.D.

Editor

Psychology Journal

P. O. Box 176

Natchitoches, LA 71458


Dear Dr.
Kelly
:

Thanks for your email, dated 10/29/08, regarding
a
manuscript (
Application

of Classic Human Information Processing M
odels
to Unmanned Aircraft Systems

(UAS)
) which I
submitted for possible
publication in
Psychology Journal.
I have
revised the manuscript based on your helpful feedback and now attached a revised
copy. As before,

t
his manuscript has not been published elsewhere, nor is it
concurrently under review at any other professional journal. I am the
corresponding author and if you have further comments or questions, please do
not hesitate to contact me.

Thank you in advan
ce.




Sincerely,




F. R. Ferraro
, Ph.D.

Dept. Psychology

University of North Dakota

Corwin
-
Larimore 215

319 Harvard St., Stop 8380

Grand Forks, ND 58202
-
8380

701
-
777
-
2414

701
-
777
-
3454

f_ferraro@und.nodak.edu