Operator Fatigue Detection Tech Review Finds ASTiDTM & OptalertTM Most Effective

Operator Fatigue
Detection Technology Review
© 2008 Caterpillar All Rights Reserved. CAT, CATERPILLAR, their respective
logos,“Caterpillar Yellow,” and the POWER EDGE trade dress as well as corporate and
product identity used herein, are trademarks of Caterpillar and may not be used without
permission. Cat and Caterpillar are registered trademarks of Caterpillar Inc., 100 N.E.
Adams, Peoria IL 61629.
Executive Summary » Introduction » Technology Review » Driving Simulation Study
» Project Recommendations » Acknowledgements » Appendix 1: Team Member Bios »
Appendix 2: Expert Reviewer Bios » Appendix 3: Product Summaries

© 2008 Caterpillar All Rights Reserved. CAT, CATERPILLAR, their respective logos,“Caterpillar Yellow,” and the POWER EDGE trade dress as well as corporate and product identity used herein, are
trademarks of Caterpillar and may not be used without permission. Cat and Caterpillar are registered trademarks of Caterpillar Inc., 100 N.E. Adams, Peoria IL 61629.
1 / Operator Fatigue Detection Technology Review
The technology review identified 22 technologies
that were commercially available or considered as
emerging technologies with potential for near term
commercialization. The team then proceeded to
gather data from the technology supplier, the scientific
literature and customers (if available) to gather a good
understanding of the background of the technology.
The team then created a list of technology features and
capabilities that would be used for evaluating the 22
technologies. The set of features was developed into a
rating matrix consisting of 16 feature categories with 93
total features. The team provided a 1 to 10 weighting
to each category and feature based on the importance
of that category or feature to the mining customer.
In addition to weighting the matrix for the mining
customer, a set of weights was also established through
discussions with fatigue industry experts in order to
represent the population needs in general. To provide an
unbiased, objective assessment of these technologies
the team invited 5 international experts from a variety
of related industries (transportation research, mining
research, biosignal analysis, human factors and
ergonomics research) to provide their input to the matrix
for all of the technologies. The input from all experts was
consolidated and the technologies were ordered from
best to worst in terms of the experts’ scoring. The top
5 technologies based on the mining industry weightings
were: ASTiD™ (Pernix), FaceLab (Seeing Machines),
HaulCheck (Accumine), Optalert™ (Sleep Diagnostics)
and the Driver State Monitor (Delphi).
At the present time ASTiD™, HaulCheck, and Optalert
have been trialed in a mining application. Results from
the ASTiD™ and Optalert™ field trial are encouraging.
FaceLab and the Driver State Monitor have only been
trialed in on-highways applications. Plans are in place
for a mining trial of the Seeing Machines product
in the coming months. To compliment the field
trial results from the three technologies mentioned
above, the team conducted trials of the Driver State
Monitor (Delphi) and FaceLab (Seeing Machines)
using an interactive driving simulator. Simulator trials
of the Delphi and Seeing Machines devices showed
very good correlation with driving errors and fatigue;
however the data suggested that more work could
be done to increase the robustness of both systems
before they would be ready for commercial release.
In response, both Delphi and Seeing Machines are
now examining what additional development would
be required to adapt their technologies for the mining
industry customer.
An additional outcome of the technology review and
simulator research was an in-depth look at the use of
head-nod sensors for fatigue detection. Two head-
nod sensor products were included in the technology
review, both of which had been used previously in the
mining industry. The two products received extremely
low scores from the expert reviewers. Use of one of
these devices was included in the interactive driving
simulator study to provide an objective measure of the
head-nod sensors’ effectiveness and to determine if the
low ratings are warranted. The device was plagued with
numerous false alarms due to typical driving-related head
movements and true alarms only accounted for 1% of
the fatigue related driving errors.
Caterpillar launched the fatigue technology review project in January 2006. Since
that time Caterpillar has conducted an in depth review of available and emerging
fatigue detection technologies. In addition to the original scope of simply reviewing
fatigue detection technologies, the team included follow-up evaluations of 3 of the
top 6 technologies using an interactive driving simulator. The results provided
in this report summarize the activities from both the technology review and the
follow-up evaluation.
I. Executive Summary

In summary, of the 22 technologies only three of the
top rated technologies are immediately available:
ASTiD™ (Pernix), HaulCheck (Accumine) and
Optalert™ (Sleep Diagnostics). Of these technologies
only ASTiD™ and Optalert™ can be considered
as fatigue detection technologies. The HaulCheck
product only measures lane position and vehicle
proximity; notifying the operator only after they have
deviated dangerously out of their lane regardless
of their level of fatigue. Alternatively, ASTiD™ and
Optalert™ are both designed to detect the early
effects of drowsiness, both have gone through
field trials with promising results and they are both
established on sound scientific research from well
respected fatigue laboratories. Therefore, for the
purposes of identifying effective and predictive
technologies for drowsiness detection, only ASTiD™
and Optalert™ are recommended. Both technologies
are viable options for use as supporting technologies
to an asset’s overall fatigue management program.
When implementing new technologies, it is easy
to forget or ignore the most important aspects of a
successful fatigue management program; ensuring
that people recognize and take responsibility for
their own fitness for work, taking into consideration
the frontline supervisors and their understanding
and management of their workgroups, and the
development of a culture within our businesses that
encourages reporting of and action on drowsiness and
fatigue risks. To improve the likelihood for success
of both the new technology and the accompanying
fatigue management programs, it is important to
utilize the appropriate change management process.
This process ensures that end user contribution is
sourced, intervention strategies are agreed on and
timely, and that appropriate communications and
support are established prior to, during and following
the implementation.
When discussing the potential for decreasing the
operational risk of 24/7 operations through the
implementation of safety-enhancement technologies,
it is easy to allow technology to supersede and
overshadow the importance of good people-centric
policies. These devices should not be relied on as
the panacea for managing fatigue. In fact, these
technologies only serve as a last line of protection
when all other fatigue management policies and
procedures have been put into place.
Executive Summary

II. INTRODUCTION
Within the public sector, driver drowsiness and
inattention are significant factors contributing to
commercial truck crashes accounting for 1,200 deaths
and 76,000 injuries annually at an estimated cost of
$12.4 billion to the commercial trucking industry. In the
surface mining industry, 93% of haulage truck accidents
are due to human error. 60-70% of human error
accidents were found to be fatigue-related. Fatigued
drivers are often not aware of their condition, frequently
driving for 3-30 seconds with their eyes totally closed.
While operator fatigue is predominantly a people
management issue, there is available technology that
can be adapted to assist in the detection of the onset
of fatigue and interface with the operator to prevent
an incident, and subsequently, allow remedial actions
to be taken. The on-highway trucking, automotive, and
mining industry have looked to technology to provide
supplementary solutions to the driver/operator fatigue
issue. Numerous technologies have surfaced, but none
have been clearly identified as the ideal solution in terms
of accuracy or wide spread operator acceptance.
This project contained two major components: 1) An
in-depth technology review and 2) A driving simulation
study of leading fatigue technologies.
Operator fatigue is one of the most prevalent root causes of earth moving equipment
accidents within the mining industry. Sleep deprivation, fatigue and drowsiness
decrease awareness, attention, and increase reaction time.

III. TECHNOLOGY REVIEW
A. Objectives
The purpose of this project was to conduct a
comprehensive review of all existing and emerging
fatigue detection technologies. This is the first time
that a wide range of experts from various backgrounds
has used the same methodology to quantitatively and
objectively assess fatigue management technologies.
The end result of this project was a comprehensive
objective assessment of the available fatigue/alertness
technologies as well as identifying the merits of
emerging technologies that may become available
in the future. The goals of the technology review
project are to:
• Identify and conduct a detailed review of all available
fatigue and alertness technologies currently being
utilized for detecting driver fatigue worldwide.
• Identify what research is being considered or is in fact
underway in regards to detection technology.
• Identify the gaps within available detection technology
compared to current research and recommend
methods to resolve this variance.
• Provide a detailed report and summary to mining
industry companies on the application and
effectiveness of identified detection technology
systems that may support the mining industry. (May
include technology currently being utilized in the
aviation, military, commercial trucks and/or motor
vehicles industries)
B. Team Members
The technology review project included individuals from a
customer, Caterpillar and from the shift work consulting
firm, CIRCADIAN™. Detailed team member biographies
are included in the appendices.
Team Members:
David Edwards, Caterpillar Inc.
Acacia Aguirre, CIRCADIAN™
Bill Davis, CIRCADIAN™
Todd Dawson, CIRCADIAN™
Udo Trutschel, CIRCADIAN™
C. Methods
The methodology was very structured to maintain a high
degree of objectivity. Below is the outline of the main
tasks for completing the technology review.
1. Identify all available technologies. This was done
through web searches, interviews with fatigue
industry experts and mining customers, prior team
experience and an extensive patent review on
alertness/fatigue technologies.
2. Identify most promising technologies. This shorter
list of technologies was determined by availability
of system, previous or current experience of the
technology in the mining industry, potential for use in
mining, and current stage of development
3. Gather information on most promising technologies.
Whenever possible, users in the mining industry were
contacted for input regarding the technology and any
outcomes or data that was available. The information
was gathered from, but not limited to public domain
reports, interviews with users and interviews with
technology suppliers.
4. Develop diagnostic Objective Matrix Tool. Experts
at CIRCADIAN™ with input and advice from both
customer and CAT developed the matrix.
5. Score each fatigue/alertness technology. Using
the matrix, each technology was scored by internal
CIRCADIAN™ experts.
6. External expert scoring. External experts were used
to broaden the arena of expertise to leaders in optics
(which are often utilized in fatigue technologies),
psychology, ergonomics, medicine, mining and
transportation. These external experts also used the
matrix for technology scoring.
7. Mining weights. To close the circle of experience,
representatives from a customer were asked to
identify the most important aspects of a fatigue
detection device with regard to the mining industry.
8. Develop composite scores. All of the data gathered
was incorporated into a final composite score for each
technology. This provides an overall view of the most
promising fatigue and alertness technologies.

D. Product list and reviews
Identifying commercially available
and emerging technologies
Commercially available products and emerging
technologies were identified primarily through web
searches, literature reviews, interviews with subject
matter experts and a patent search on alertness/fatigue
technologies. This process resulted in identifying 35
technologies. Each product or technology was then
evaluated based on its current availability, history of
experience in the mining industry (past, present or future
trials), feasibility of implementation within a heavy mining
equipment operator station and the technology’s current
stage of development if it was not commercially available.
This process narrowed the original 35 technologies down
to a much smaller list of 22 products and/or technologies
(Table 1). Table 1 shows a complete list of the 22
reviewed technologies as well as their developer and their
general technology category. A quick breakdown of the
reviewed technologies shows eye feature analysis as the
predominant technology type followed by lane deviation
systems (Figure 1).
Gather information on most
promising technologies
Once the final 22 technologies were identified, a more
thorough investigation was conducted to ensure the
Company
AcuMine
Advanced Safety Concepts
ARRB Transport Research
AssistWare Technologies
Atlas Research Ltd
Attention Technologies
Delphi Corporation
International Mining Technologies
Iteris Inc
MCJ
Mobileye NV
Neurocom
Ospat Pty
Pernix
Precision Control Design Inc
Muirhead/Remote Control Tech.
Security Electronic Systems
Seeing Machines
Sleep Diagnostics
Smart Eye
SMI
Welkin
HaulCheck
PASS
Fatigue Management System
SafeTrac
NOV Alert
Driver Fatigue Monitor
Driver State Monitor
Voice Commander System
Lane Departure
EyeCheck
Vision/Radar Sensor
EDVTCS
OSPAT
ASTID™
SleepWatch
Fatigue Warning System
Sleep Control Helmet System
Facelab
Optalert™
AntiSleep
InSight
Nap Zapper
Lane Deviation
Head Nodding Detection
Mental Reaction Time
Lane Deviation
Muscle Tone Analysis
Eye Blink Detection
Eye Blink Detection
Lane Deviation
Fitness for Duty System
Lane Deviation
Skin Conductance
Fitness for Duty System
Steering/Machine Movement
Activity Monitor
Eye Feature Monitoring
Eye Feature Monitoring
Eye and Head Monitoring
Eye and Head Monitoring
Product Technology
Table 1: Final Technology List

evaluators had as much technical information as was
publicly available for each of the technologies. Users of
these technologies in the mining industry were contacted
for input regarding the technology and any outcomes from
technology trials if data was available. Additional information
was gathered from, but not limited to the following sources:
public domain reports and interviews with users and vendors.
Product briefs were developed by the team as a means of
quickly educating stakeholders and our external experts
on the 22 technologies can be found in the appendices.
E. Assessment Matrix
Development of matrix/feature criteria
Through interviews with mining customers and fatigue
experts a list of technology features was created
including descriptive, technical and functional criteria. A
total of 93 features were selected across 16 categories.
Features and feature categories were finalized using
input from both the fatigue and mining industry.
1.0 Focus of technology—This category describes
what the technology is monitoring. Recent studies
suggest that users strongly prefer systems that
require as little personal monitoring and contact with
the technology as possible. The preference is for
systems that monitor vehicles instead of people.
1.1 Vehicle Monitoring
1.2 Operator Monitoring
2.0 System capabilities—System capabilities are
important to identify the spectrum of metrics that can
be measured or are included as part of the technology.
2.1 Accident Mitigation (lane deviation, passing instructor)
2.2 Collision Warning
2.3 Operator Performance Evaluation
2.4 Operator Fatigue Prediction
2.5 Microsleep Detection
2.6 Operator/Dispatch Assistant
3.0 Primary sensor technology—This section
identifies the sensor(s) primarily used by the
system. The sensor may or may not currently be
used to track fatigue. However, most of these
sensors have been used to track fatigue to some
degree in different environments.
3.1 Machine Vision (Digital Video)
3.2 Infrared (IR) camera
3.3 Visible light camera
3.4 IR illumination and sensors
3.5 Equipment / Electrodes attached to Body
3.6 GPS
3.7 Laser Scanning
3.8 Accelerometry
3.9 Motion detection (gyro sensor)
3.10 Timer Switch / button
4.0 Primary measures (eye)—This is one of several
sections that identify the primary measure of the
system. In this case, the primary measure is the
eye. All the subcategories are generally accepted
measures for the eye that can be linked to alertness
and fatigue.
Figure 1: Technologies Reviewed by Type
Muscle Tone Skin
Conductance
Activity
Monitor
Mental
Reaction Time
Head Nod Lane Deviation/
Steering Analysis
Eye Feature
Analysis
Fitness
for Duty
NumberofProducts
7
6
5
4
3
2
1
0

4.1 Gaze tracking
4.2 PERCLOS/AVECLOS
4.3 Blink rate
4.4 Prolonged eye closure
4.5 Eye Features (pupil, eyelids)
4.6 Pupil Reactivity
5.0 Primary measures (behavior/physiology)—
Primary measure for behavior and physiology including
heart rate, body movements and brain activity.
5.1 Head Movement (Head nodding)
5.2 Facial Features
(yawning, muscle movements, muscle tone)
5.3 Heart rate
5.4 Electroencephalogram (EEG)
5.5 Electromyography (EMG)
5.6 Electroocculogram (EOG)
5.7 Grip force (Steering Wheel)
5.8 Skin Resistance
5.9 Body Movement (Posture)
6.0 Primary measures (operator performance)—
Primary measure based on operator performance.
This category includes characteristics about the
operator’s driving performance and quality of work
that can be measured to indicate fatigue.
6.1 Microsteering corrections
6.2 Variation Steering Angle
6.3 Variation Steering Angular Velocity
6.4 Lane Deviation
6.5 Distance to right/left lane
6.6 Time to Line Crossing
6.7 Operator reaction time to artificial stimulus
(mental reaction time)
6.8 Position relative to Objects (GPS, radar, laser)
7.0 Primary system characteristics—These are
capabilities of the system for detecting fatigue and
alertness or characteristics that indicate fatigue.
7.1 Sensor Fusion
(using multiple sensors and hardware)
7.2 Use of Alertness Models in Algorithm
(CIRCADIAN™ model, sleep model)
7.3 Ability to detect operator state in real time
7.4 Automated feature extraction (Identifying
portions of data for immediate analysis and storage,
e.g. capturing the analysis of hundreds of images
rather than each individual image to speed up the
decision making process)
7.5 Ability to store and retrieve data wirelessly
(wireless communication with device
to/from dispatch)
7.6 Ability to store and retrieve data locally (device
stores data locally on the truck)
8.0 System integration requirements
8.1 Ease of installation in vehicle
8.2 Takes status of vehicle into account
(transmission, speed, state, etc.)
8.3 Permanent integration ability (how easy it would
be to integrate the system into the dashboard and/
or with other machine systems (radio dispatch, GPS,
Machine health monitor, etc.))
8.4 Installation and integration.
9.0 Fatigue countermeasure
9.1 Countermeasure (is there any kind of
countermeasure that is initiated when a certain
condition is detected)
9.2 Multiple Countermeasures (does the system use
multiple countermeasures like lights, audible alarms,
seat vibrations, vehicle interventions, etc)
9.3 Adaptive Countermeasures (does the
countermeasure change in frequency, intensity, etc.
when certain conditions occur, e.g. alarm frequency
increases or decreases depending on predicted
fatigue level or change in reaction to stimulus)
9.4 Online Feedback about Alertness Level (i.e.
the operator is given information about his current
fatigue state)
9.5 External alarm (Alarms surrounding operators
about operator state)

9.6 Dispatcher Interaction (does the system
communicate with both the operator and dispatcher,
e.g. notifies dispatcher about operator state)
9.7 Identifying fatigue countermeasures that are
used in the system.
10.0 Environmental requirements for technology—
Are there any environmental conditions that might
interfere with the system?
10.1 Dust
10.2 Vibration
10.3 Weather conditions
10.4 Lighting
10.5 Road Conditions (rough bumpy roads, muddy/
slick roads, etc.)
11.0 Data evaluation, recording, reporting methods—
Data analysis methods used in signal analysis.
11.1 Conventional Statistics (threshold)
11.2 Intelligent Adaptive Data Analysis Methods
(fuzzy logic, neural nets)
11.3 System Reporting (the system provides
an easy to use interface that provides
meaningful reports)
12.0 Validation and system accuracy—Has the
system been tested in different environments
to ensure accuracy?
12.1 Objective Validation (the system has
been validated using scientifically accepted
objective measures)
12.2 Subjective Validation (the system has been
validated using subjective measures such as
opinion surveys, sleepiness scales, etc)
12.3 Validation in the Laboratory
12.4 Validation in the Field (On-highway)
12.5 Validation in the Field (Mining Environment)
12.6 Avoids False Positives (How often does the
system trigger an alarm that was incorrect?)
12.7 Avoids False Negatives (How often does the
system not trigger an alarm when one is required?)
13.0 Technologies integration ability—Integration
with the mining industry as well as other fatigue
detection systems.
13.1 Ease of integration of other measures into
the data analysis (e.g. if a system is collecting
eye feature data, how easily could it also assess
eye gaze?)
13.2 Ease of Integration with other alertness/fatigue
products (How easy would it be to integrate this
system with other systems to complement it?)
13.3 Long-term future of the system (the device is
funded well and has great support from investors or
have a long term life)
13.4 Compatibility with future safety technologies
(is system capable of integration with other safety
technologies?)
14.0 Operator acceptance—Issues around interface
with the operator/user and whether or not the
technology will likely be accepted.
14.1 General User Acceptability
(How well does the user accept the system?)
14.2 Mining User Acceptability
(How well does the mining user accept the system?)
14.3 Union Acceptance of technology
(field trial reports)
14.4 Robustness to Operator Manipulation
(how easily can the operator manipulate the system?
E.g. turn it off, avoid detection, etc)
14.5 Robustness to Operator destruction
(how easily can the operator physically damage
the system to make it inoperative?)
14.6 Robustness to individual differences
(The system handles individual differences easily
and with little time requirement)
14.7 Not mentally invasive
(does not require additional work for the
operator e.g. no response required?)
14.8 Not physically invasive
(does not require contact with the operator)
14.9 Easy integration into Mining Culture

15.0 System data integration, calibration,
maintenance and infrastructure costs - What
additional effort is required to implement the
system and maintain it?
15.1 Integration Ability to Mining Operations
15.2 (The system tracks data that is objective and
can be correlated to issues that mine managers are
concerned with)
15.3 Ease of System Calibration (Is it easy to
calibrate for individual operators or for each truck)
15.4 Ease of System Maintenance (how much
maintenance is required for general upkeep)
15.5 Does the System avoid additional expensive
Infrastructure (GPS, lane markers, radio frequency
identification, etc.)?
15.6 Ratio Accident Reduction/System Costs (Cost
efficiency is defined as the ratio between ability of the
system to reduce fatigue related incidents costs and
the overall cost of the system and its maintenance.)
16.0 Technology readiness—Is the system commercially
available and timeframe for optimal usefulness.
16.1 Is the product commercially available and used?
16.2 Has the product potential to be used in the
short-term (6 mos-1 yr) in mining operation?
16.3 Has the product potential to be used in the
middle-term (1-4 years) in mining operation with
some modifications?
16.4 Has the product potential long-term (4-6 years)
use in mining operation when directly integrated in
the mining truck?
Development of weighting and
scoring system
A weighting system was developed to allow users of
the matrix to rate each feature and feature category on
a 0-10 level of importance of that feature or category.
Each category’s weightings were then normalized
across all categories to ensure categories were equally
represented regardless of the number of features in
each category. Scoring was based on a 6-point scale
(none, potential, possible, low, medium, high) to allow a
score of each technology feature based on the degree
to which that feature was applicable to the technology
or the likelihood that a feature could be incorporated
into the technology through additional development.
Numerical values were assigned to each score (0.00,
0.10, 0.20, 0.25, 0.75, 1.00, respectively) for purposes of
calculating the overall scores for each technology.
Scoring and weighting of each fatigue/
alertness technology
To ensure that the scores generated for these
technologies were as objective as possible, ratings for
technologies were conducted both by the authors as
well as several experts from throughout the fatigue
research community from disciplines including optics,
occupational medicine, human factors, mining, and
transportation research. Experts had no known conflicts
of interest with any of the technologies included in the
matrix. In addition to providing actual product ratings,
the experts were given the opportunity to provide their
weightings to the features and feature categories, as
they deemed appropriate. In addition to the weightings
provided by the fatigue experts, input from a major
global mining company was provided to produce a set
of weightings for the features and feature categories
based on mining specific applications. Each expert was
also given the opportunity to subjectively score each
technology on a simple 1-10 scale, with 10 being the
highest rating. Table 2 shows the category weights for
the team, all experts, and the mining representatives.

Of interest were the distinct differences in the
category and feature weights given by the fatigue
experts compared to the weights provide by the
mining industry representatives. Average weights
from fatigue experts remained fairly consistent
between 4 and 8. Mining weights differed by in large
showing more importance on system capabilities,
operator acceptance and technology readiness.
Lowest scoring categories were: Focus of Technology
and Primary Measures (Eye, Behavior, and Operator
Performance). This difference is a reflection of the
mining representatives’ not placing much value on
how the technology detects fatigue, but more so on
the system’s operator acceptance and availability.
Table 2: Summary of all expert and mining weights
FMT Product Evaluation
Focus of Technology 8 5 3 5 1 10 5.3 0
System Capabilities 10 10 6 7 1 10 7.3 10
Primary Sensor Technology 6 6 4 5 1 5 4.5 3
Primary Measures (Eye) 0 8 8 7 1 8 5.3 1
Primary Measures 0 8 6 2 1 10 4.5 1
(Behavior/Physiology)
Primary Measures 0 6 4 6 1 10 4.5 2
(Operator Performance)
Primary System 8 7 8 6 1 7 6.2 6
Characteristics
System Integration 6 3 4 3 1 7 4.0 6
Requirements
Fatigue Countermeasure 10 10 8 7 1 10 7.7 5
Environmental 7 8 8 10 1 10 7.3 8
Data Evaluation, 8 4 8 5 1 10 6.0 4
Recording, Reporting
System Accuracy 8 8 8 10 1 10 7.5 4
Technologies Integration 8 5 6 5 1 7 5.3 4
Operator Acceptance 7 10 5 10 1 10 7.2 10
Data Int., Cal., Maint. 7 3 4 6 1 10 5.2 4
Infra. Costs
Readiness 9 4 6 5 1 7 5.3 9
Team
weights
Expert 1 Expert 2 Expert 3 Expert 4 Expert 5
Average
expert
weights
Mining
weights

F. Matrix Results
The completed matrix included final composite scores
for each technology along with subjective technology
ratings (Figure 3 and 4). Product composite scores were
influenced by the differing weights applied by fatigue
and mining experts. To simplify the review of the scores
technologies were divided into three tiers. Using the
fatigue expert weights, top tiered products consisted
of FaceLab, ASTiD™, Optalert™, HaulCheck, Delphi’s
Driver State Monitor and SmartEye. Of the top six
scoring products, four are eye feature detection systems
and two (ASTiD™ and Haulcheck) are vehicle-monitoring
systems. The second tier consists of 12 products. Of
these 12, two are eye feature detection systems, four
are physiology/behavioral devices (DVTCS, SleepWatch,
NovAlert, and PASS) and the three are mental reaction
time tests (Voice Commander, ARRB, Muirhead/RCT)
and three are vehicle-monitoring systems (SafeTrac,
MobileEye, and AutoVue). The bottom tier of products
consists of two pre-shift fitness-for-duty tests (Ospat and
Eyecheck) and two head worn head nod sensors (Sleep
Helmet and NapZapper).
When mining weights are applied the top tier of 6
technologies remain unchanged, however the ordering
Figure 2: fatigue expert weights versus mining weights
Focus
ofTechnology
System
Capabilities
Prim
ary
SensorTechnology
Prim
ary
M
easures
(Eye)
Prim
ary
M
easures
(Behavior/Physiology)
Prim
ary
M
easures
(O
peratorPerform
ance)
Prim
ary
System
CharacteristicsFatigue
Counterm
easureEnvironm
ental
D
ata
Evaluation,Recording,Reporting
System
Accuracy
Technologies
Integration
O
peratorAcceptance
D
ata
Int.,Cal.,M
aint.and
Infra.Costs
Readiness
System
Integration
Requirem
ents
10.0
9.0
8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
Weights
Average Expert Weights
Mining Weights

differs slightly. The second tier sees significantly
more shifting of technologies and finds 2 technologies
dropping into the third tier (both lane deviation systems)
and 2 tier three technologies climbing into tier two (both
fitness-for-duty tests). These findings are understandable
as the lane deviation technologies relied heavily on
image recognition of painted roadway markings, which
by definition are non-existent in off-road environments.
Further support for the technology shift is apparent in
that fitness-for-duty tests increased in value under the
mining weights. Mines typically have controlled access,
which could allow for pre-shift testing; something
far less practical in more general applications like
automotive driving. The head-worn, head-nod sensor
scores remained in the bottom tier regardless of which
system of weights was applied.
Subjective scoring of the technologies demonstrates
a clear trend for the expert reviewers towards eye
feature detection systems for both the general and
mining industry weights. Due to the popularity and
preponderance of research funding on methods for
automatically tracking percent eye closure (PERCLOS)
these subjective scores are not surprising. Further
evaluation of the general fatigue subjective scores
show that next most highly rated technologies were
those that monitored lane deviation utilizing image
recognition. The lowest general fatigue subjective
scores were for the mental reaction time and head nod
sensing technologies. In contrast, when considering
mining industry requirements, following the automatic
PERCLOS detection systems, the same expert
reviewers subjectively scored the two lane deviation/
steering deviation devices that do not depend on
image recognition (ASTiD™ and Haulcheck) followed
by mental reaction time technologies and head nod
sensors. Image recognition lane deviation technologies,
based on mining industry requirements, were
subjectively scored the lowest.
The highest ranking products displayed the following
characteristics: (1) Multiple sensors or ability to process
multiple features; (2) Multiple means of alerting
the operator of impending fatigue and signaling the
supervisors and/or dispatchers; (3) Previous validation
tests in the field, particularly in rough environments;
(4) The capability to be customized to the individual;
and (5) Required little or no operator input. With regard
to the features of user/operator acceptance, devices
that focused primarily on the vehicle tended to score
higher. Some of the technologies that did not score in
the upper rankings were nonetheless promising for the
long term. These included products like NOVAlert and
SleepWatch. The ideal solution to managing fatigue in
the mining industry will likely be comprised of several
different technologies working together. These will likely
but not necessarily include pre-shift assessments such
as Ospat or EyeCheck, vehicle-monitoring systems like
HaulCheck or ASTiD™, and operator monitoring systems
like Optalert™.
The usage of the scoring matrix is critical to providing an
unbiased and equitable evaluation of all the technologies.
Comparing the subjective and objective scores using the
two weighting systems emphasizes the importance of
such a matrix. Were these experts to merely provide their
professional opinion on which technologies they were to
recommend, their response could be largely biased on
the knowledge of their own particular industry and field of
study. However, when industries other than the on-highway
transportation industry seek advice from fatigue experts
or the scientific literature, not having a way to account for
the experts’ inherent bias towards their industry’s particular
needs could lead to drastically different and potentially
inappropriate recommendations. This matrix provides each
industry a way to leverage the knowledge of the fatigue
research community by tailoring it to each industry’s specific
needs through the use of the matrix weighting system.
The technology review project is significant in that it
brought together a wide range of experts from various
backgrounds and used the same methodology to
objectively and subjectively assess several commercially
available and emerging fatigue management technologies.
The end result of this collaboration and methodology was
not only an objective assessment of the currently available
technologies, but it also assessed the merits of emerging
technologies that may become available in the near future.

FaceLab
ASTiD
Optalert
Delphi
SmartEye
SMI
DVTCS
SleepWatch
CoPilot
Voice Comm.
ARRB
PASS
NovAlert
SafeTrac
MobileEye
Autovue
Muirhead
Ospat
EyeCheck
Sleep Helmet
Nap Zapper
Figure 3: Product and technology ratings with fatigue experts’ weights
1
General Objective
Tier 1
Tier 2
Tier 3
General Subjective
0.90.5 0.70.3 0.80.4 0.60.20.10
ASTiD
FaceLab
HaulCheck
Optalert
Delphi
SmartEye
DVTCS
PASS
SleepWatch
Voice Comm.
ARRB
SMI
CoPilot
NovAlert
Ospat
Muirhead
EyeCheck
SafeTrac
Autovue
Sleep Helmet
MobileEye
Nap Zapper
Figure 4: Product and technology ratings with mining weights
1
General Objective
Tier 1
Tier 2
Tier 3
General Subjective
0.90.5 0.70.3 0.80.4 0.60.20.10

G. Gap-Analysis
The best in class technologies were selected from the
matrix scores using a Pareto plot to identify the top 25%
technologies (Figure 5). Based on this information the
team recommended a strategy for addressing the gaps in
both the available and emerging technologies and what if
any action should be taken to help move the best in class
technologies forward. Technologies identified as needing
additional investigation included: FaceLab (Seeing Machines),
HaulCheck (Accumine), Optalert™ (Sleep Diagnostics) and
the Driver State Monitor (Delphi). Additional technologies
were included in the gap analysis discussion because of their
extremely low scores on the matrix and the fact that they
are currently in use in the mining industry (Sleep Helmet
and NapZapper). The EDVTCS (Neurocom) system also was
included as a technology worth further investigation simply
because it used a less intrusive sensor (wrist watch) that
might make for an easier mining implementation were it
found to be effective.
ASTiD
Promising Predictive Technology
FaceLabH
aulcheck
O
ptalert
D
elphiD
SM
Sm
artEyeED
VTCS
PASS
ARRB
SM
I
CopilotN
ovalert
O
SPATM
uirheadEyecheckSafeTracAuto-Vue
Sleep
H
elm
etM
obileEye
Sleepw
atch
N
app
Zapper
Voice
Com
m
ander
Figure 5: Pareto Plot of Technologies
(top 25% of technologies to left of orange line)
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
No Further Activities RequiredAdditional Validation Required

Of the technologies identified as “promising” or
“requiring additional validation”, the technologies were
plotted on a time to delivery axis by their matrix score
(Figure 6). This activity demonstrates graphically what
the relative value would be when additional time and
money is invested to bring future promising technologies
to market versus those that were already immediately
available. A summary of each of these products’ current
state, long-term possibility, development time and
estimated cost can be seen in Table 3 and Table 4 for
reactive and predictive technologies. To conclude the
gap-analysis activities, a list of action items was created
to provide any additional information or justification to
enable the team to formulate its final recommendation.
One action item was clearly identified with needing
additional effort. The two head nod sensors included
in the evaluation scored extremely low by the fatigue
industry experts. Of concern to the team was that these
products have in the past been used at some mine sites
and could potentially be used by sites in the future. Due
to the fact that these devices were being used in the
field and their significantly low scores, it was determined
that the team should conduct a validation study of these
products to determine if the lower scores are justified.
This validation study was added to the original project
scope. The details and results of this study are discussed
in the following section.
Now
ASTiD
Haulcheck
Optalert
Delphi
EDVTCS
Sleep Helmet
Nap Zapper
Facelab
Time to Product Availability
3yrs+1.5-3yrs
Figure 6: Matrix score versus time to availability
Table 3: Available Reactive Technologies
0
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
System
Haulcheck
Nap Zapper
Sleep Helmet
Available in
mining industry
Available in
mining industry
Effective collision
prevention
through “berm
sensing”
Incremental
improvements
Now
Now
$50,000
AU per truck.
(High on-going
maintenance costs)
$10/$1000 per truck
and/or operator
Current State Long term possibility Time Cost

Table 4: Promising Predictive Technologies
System
ASTiD™
Optalert™
Facelab
Delphi
EDVTCS
Available in
mining industry
Available
Favorable results
in lab setting
Favorable results
in lab setting
Favorable results
in railroad and
trucking
Incremental
improvements
Accurate JDS
readings in mine
setting
Readings
analogous to EEG
from wristband
Accurate PERCLOS
readings alarming
from camera
Accurate PERCLOS
readings alarming
from camera
Now
Now
3-5 yrs
1/2-3 yrs
1/2-3 yrs
$6,000
AU per truck.
(High on-going
maintenance costs)
$16,000 AU per unit
-includes 1 pair of
glasses (Volume
pricing possible)
Unknown
$500-$3000 per OHT
(based on volume)
$8,000 per OHT
Current State Long term possibility Time system Cost

Iv. driving simulation study
A. Objectives
This study represents the continuation of the technology
review study conducted in 2006, which evaluated
Fatigue Management Technologies (FMT) using a
comprehensive matrix. The objective of this study is to
perform an in-depth evaluation of the performance of
head nod sensors compared with the tier 1 technologies
identified with the matrix. The selected FMT’s were
evaluated using a driving simulator in addition to a
host of physiological, behavioral and performance
measurements during an overnight driving protocol.
B. Team Members
David Edwards, Caterpillar
Acacia Aguirre, CIRCADIAN™
Bill Sirois, CIRCADIAN™.
Udo Trutschel, CIRCADIAN™
Dave Sommer, University of Applied Science, Schmalkalden
Martin Golz, University of Applied Science, Schmalkalden
C. Method
The study was conducted in the Department of
Adaptive Biosignal Analysis at the University of Applied
Sciences, Schmalkalden. The department has a
sophisticated driving simulator and ample experience
conducting this type of study. Figure 7 shows the
layout of the lab, with driving simulator and the
observation and data collection areas. The lab is fully
controlled by specialized software, allowing regulation
of light levels, car environment (temperature, noise
level, humidity), and of controls and instruments. The
communication between volunteers and experimenters
is conducted by interphone. The driving simulator
includes an Opel Corsa cab to provide a realistic driving
experience. The driving scenarios software is also very
realistic. Figure 8 shows an example of the driving
simulator driving scenario. Drives were conducted
on a rural road with no other traffic on the roads. The
monotony of the driving task ensured that drivers
would become fatigued throughout the driving session.
Figure 7: Lab layout; (1) Landscape Generation, (2) Video
Capture,(3) Driver State Sensor,(4) Experimental Control,(5) Video
Capture,(6) Electrophysiological Recording, (7) Car Hardware
Control,(8) Eye-Tracking Recording,(9) Video Projector,(10) Digital
Video Cameras, (11) HD Video Eye Cam, (12) Projection Screen, (13)
Car Opel Corsa
Figure 8: Simulator driving environment:
top down (left) and screen image (right)

Measurements
Data collected for this study included the following:
• Driving parameters: lane deviation, steering angle,
driving errors (defined as all wheels out of the lane)
• Driver behavior: video recordings of driver’s eyes, face
and head position
• Physiological parameters: EEG (9 channels), EOG (2
channels), EMG, ECG
• Driver subjective alertness: Karolinska Sleepiness Scale
(KSS), Samn-Perelli, alertness and micro sleeps self
assessment
• Performance tests: Continuous Tracking Tasks (CTT)
and PVT
The following FMT suppliers made their products
available for testing: Seeing Machines, Delphi, SMI
and NapZapper. SeeingMachine’s DSSR records the
following parameters: head position, head rotation, and
eye closure (PERCLOS). Delphi Driver State Monitor
(DSM) records head movement, distraction, eye closure
(AVECLOS). The system also provides eye closure and
eye closure duration warnings. SMI Insight Eye Tracker
records head movement, point of gaze, eye closure
(PERCLOS 70, PERCLOS 80). NapZapper records head
nodding. The following figures show the devices and
data collection screens of Seeing Machines and Delphi
(Figure 9 and 10).
Volunteers
Volunteers were recruited among Schmalkalden
University students. Students were informed about the
study by information posted in the University website.
All volunteers were interviewed before the experimental
night. They were informed about the study protocol
and signed an informed consent. Sixteen volunteers
participated in the study, ten men and six women. The
average age was 22 (range 18 – 31). They were all
healthy and had regular sleep/wake schedules.
Protocol
The study consisted of two overnight driving
simulation sessions. Before the experimental
nights, volunteers were trained in the different
tests, and wore an activity monitor and completed
a sleep/wake log for at least 24 hours prior to the
experiment. On both nights, volunteers arrived at
the lab at 10 PM. After wire-up, checking logs and
activity monitors, and retraining, the experimental
sessions started at roughly 11:30 PM. There were
eight experimental sessions, each one lasting one-
hour with the last session being finished at 8:30 AM.
Volunteers had a 1-h break at 3:30 AM. Each session
included: 40-minutes driving session, 10-minutes CCT
performance test, and 10-minutes PVT. Alertness
self-assessment and Samn-Perelli questionnaires
were performed at the end of the driving task. KSS
and brief alertness assessments were performed at
regular intervals during the driving task, as well as
before and after the task. To minimize distractions
during the driving task, the experimenter would ask
volunteers about their alertness using both KSS and
the self-assessment questionnaire. Figure 11 shows
the experimental protocol.
Volunteers were asked to complete two experimental
nights, several weeks apart, so that they would fully
Figure 9: Seeing Machines (DSSR)
Figure 10: delphi (DSM)

recover from sleep deprivation. The first night, FMT
were used without alarms, and the second night with
alarms. To study included twelve volunteers completing
both experimental nights, and six others completed only
the first experimental night.
D. Results
In this report, the following analyses are presented:
• Fatigue assessment using Delphi, Seeing Machines
and NapZapper
• Correlation between Delphi and Seeing Machines
fatigue assessment and subjective alertness (KSS) and
driving performance
• Volunteers evaluation of Delphi, Seeing Machines and
NapZapper (operator acceptance)
Alertness, fatigue and driving
performance progression throughout
the night
As expected, sleepiness as measured by the KSS, DSM,
DSSR increased, and driving performance deteriorated
progressively throughout the night. Driving performance
and KSS showed a strong correlation (91%) overall
throughout the entire night, meaning that as the driver
became more fatigued the number of driving errors
increased throughout the night.
10:00pm
1:30am 1:40am
(Re)calibration
Self-Assessment
(Questionnaire)
Samn Perelli
(Questionnaire)
CTT: Continuous Tracking Task
PVT: Psychomotoric Vigilance Test
KSS+Awareness
(Questionnaire)
1:50am 2:00am 2:10am 2:20am 2:30am
Preparation 1st
Session 2nd
Session 3rd
Session 4th
Session 5th
Session 6th
Session 7th
Session 8th
SessionBreak
CTT PVT Break
11:30pm 0:30am 1:30am 2:30am 3:30am 4:30am 5:30am 6:30am 7:30am 8:30am
Driving in the Simulator
Figure 11: Experimental Protocol
Figure 12: Average session values during experimental night
(Left: Seeing Machines[DSSR] Right: Delphi[DSM])
Session Session
ScaledParameters
ScaledParameters
Mean Driving Errors (yellow), Mean KSS (black) and DSSR PERCLOS (gray) Mean Driving Errors (yellow), Mean KSS (black) and DSM AVERCLOS (gray)
Drive DriveKSS KSSDSSR DSM
1 12 23 34 45 56 67 78 8
0.25
0.2
0.15
0.1
0.05
0
0.25
0.2
0.15
0.1
0.05
0

36%58% 11% 66% 48% 73% 47% 74%
Fatigue assessment using Delphi and Seeing
Machines
Since subjective alertness (KSS) and driving performance
show a well-known pattern of increased deterioration
throughout the night, these values were used as
benchmarks to evaluate Delphi (DSM) and Seeing
Machines (DSSR) fatigue measurements. Correlations
between subjective alertness (KSS) and driving
performance and Delphi (DSM) and Seeing Machines
(DSSR) as an overall correlation across all volunteers and
all sessions.
Individual correlations were calculated for each system,
(DSSR and DSM) and each driving session throughout
the night. The overall correlation for all sessions are
shown in Table 5. Both Delphi (DSM) and Seeing
Machines (DSSR) showed strong correlations with (KSS)
subjective alertness (90% and 89% respectively). The
correlation with driving performance (Table 5) was also
extremely high for the DSM (98%) and the DSSR (84%).
Time
Time
Figure 14: Session by session correlations for Seeing Machines (DSSR) and driving performance
12pm
12pm
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.4
0.3
0.2
0.1
0
1am
1am
2am
2am
3am
3am
4am
4am
5am
5am
6am
6am
7am
7am
8am
8am
ScaledParametersScaledParameters
Mean KSS (green) and Mean DSSR PERCLOS (red) for 14 Subjects; Numbers show Correlations
Mean Driving Errors (blue) and Mean DSSR PERCLOS (red) for 14 Subjects; Numbers show Correlations
sess 8sess 1 sess 2 sess 3 sess 4 sess 5 sess 6 sess 7
Kss (subjective
fatigue)
DSM
DSSR
90%
89%
98%
84%
driving
performance
Table 5: OVERALL CORRELATIONS
Figure 13: Session by session correlations for Seeing Machines (DSSR) and subjective alertness

Both technologies appear to do a good job of tracking
the progression of fatigue throughout the night according
to individuals’ subjective assessment (KSS) and their
driving performance.
A more detailed analysis evaluated the systems’
performance session by session (all volunteers averaged).
Figures 13 and 14 show the results and correlations
between Seeing Machines (DSSR) and subjective
alertness and driving performance, and Figure 15 and 16
show the same plots for Delphi (DSM) respectively.
The two systems performed similarly when looking at
the subjective alertness and driving performance session
by session. In general it could be stated about both
technologies that in the early driving sessions (up until
3 AM) KSS increased dramatically but the fatigue measures
from DSSR and DSM showed very little change overall.
Similarly, the number of driving errors remained relatively
flat during the same time period, not beginning to increase
until after 3:00 AM. Therefore subjectively, volunteers were
“feeling” tired during the early sessions, but their body
wasn’t showing the fatigue related increases in percent eye
closure nor was their driving behavior getting much worse.
Time
Time
Figure 15: Session by session correlations between Delphi (DSM) and subjective alertness
Figure 16: Session by session correlations between Delphi (DSM) and driving performance
12pm
12pm
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.4
0.3
0.2
0.1
0
1am
1am
2am
2am
3am
3am
4am
4am
5am
5am
6am
6am
7am
7am
8am
8am
ScaledParametersScaledParameters
Mean KSS (green) and Mean DSM AVECLOS (red) for 15 Subjects; Numbers show Correlations
Mean Driving Errors (blue) and Mean DSM AVECLOS (red) for 15 Subjects; Numbers show Correlations

It was not until the fourth driving session (3:00 - 4:00 AM)
that the fatigue measures started to elevate and driving
performance started to decline. Driving performance values
did rise and fall from minute to minute, but in general,
throughout the early morning hours driving errors were
frequent. Likewise, similar to driving performance, the
fatigue metric was highly variable from moment to moment,
but there was an overall upward trend with the fatigue
values remaining elevated throughout the early morning
driving sessions. Throughout all the early morning sessions
it is clear that both the DSSR and DSM were correlating
well with driving performance. As the fatigue monitoring
devices were showing high levels of fatigue, the volunteers
were experiencing high numbers of driving errors.
Even though the overall correlations shown above were
strong, if you look at the data on a person-by-person basis
there are some interesting differences. Figure 17 and 18
show excellent data from a single individual during a single
session where the fatigue device correlated very well with
driving performance. Lane deviation in the driving simulator
Minutes
Figure 17: Example of Delphi DSM performance with numerous driving errors
0
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
5 10 15 20 25 30 35 40
ScaledParameters
threshhold DSM
Minutes
Figure 18: Example of Delphi DSM performance with no driving errors
0
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
5 10 15 20 25 30 35 40
ScaledParameters

is shown as the dark blue line with an average lane position
in light blue. Significant driving errors are marked as yellow
vertical lines. DSM and DSSR output are shown in red
with an average output in pink. The volunteer commits
many driving errors and fatigue is detected (Figure 17).
A volunteer commits no driving errors and no fatigue is
detected (Figure 18).
However, not all individuals’ data were this clear. Figures
19 and 20 show examples of individual volunteers and
sessions where the FMT did not work well, that is, the
volunteer committed errors and fatigue was not detected,
or the driver committed no errors and the FMT detected
fatigue (Figure 19 and 20 respectively). The present study
was not able to determine the direct cause of these broad
differences between volunteers, however the reasons
behind these major differences will be critically important
for technology developers to understand to ensure that
their systems are equally effective for all individuals.
Minutes
ScaledParameters
Figure 19: Example of Seeing Machines DSSR performance with many driving errors
0
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
5 10 15 20 25 30 35 40
Minutes
ScaledParameters
Figure 20: Example of Seeing Machines DSSR performance with no driving errors
0
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
5 10 15 20 25 30 35 40
threshhold DSM

In summary, both Seeing Machines (DSSR) and
Delphi (DSM) show, overall, strong correlations
with subjective alertness and driving performance.
However, correlation level varies by subject and
session (time of day). With both systems, there are
cases when the volunteer commits errors and fatigue
is not detected, or the volunteer commits no errors
and the FMT detects fatigue. This finding supports the
accepted belief in the fatigue research community that
there are individual differences in the population with
some individual’s performance being very tolerant to
the CIRCADIAN™ cycle and the physiological effects
of fatigue, while others are extremely susceptible to
both. There is room, however, for fatigue algorithms
to be improved for both systems so that they are less
affected by individual differences.
NapZapper and driving performance
That NapZapper is a small device worn over the ear that
triggers an alarm when head nodding is detected (Figure
21). It has been sold to numerous industries as an accident
prevention device. The NapZapper seldom was activated
during our driving simulation sessions. Below is an example
session from one individual who experienced numerous
alarms. In this example you see that the majority of the
driving error events show that the NapZapper was not
activated until after the car was already completely outside
of the lane after the driving error occurred (Figure 22).
The Nap Zapper performed poorly overall. In Figure 23 we
see the number or correct and false alarms. Correct alarms
are those instances where the NapZapper alarm went
off and a driving error was committed within the next few
seconds. False alarms were those were the alarm went off
but no driving errors occurred. For all drivers combined,
there were 1633 out of road errors. According to the data,
the NapZapper only alarmed 50 times for all subject for
all events.
Figure 21: NapZapper device.
Figure 22: Individual example of
NapZapper driving error detection

16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
0 42 6 91 5 83 7 10
Number of Alarms
Figure 23: NapZapper correct and false alarms by volunteer
Figure 24: Total number of driving errors per subject
Subject
15 Correct Alarms
35 False Alarms
16-2
16-1
15-1
14-2
14-1
13-2
13-1
12-1
11-2
11-1
10-2
10-1
9-2
9-1
8-2
8-1
7-2
7-1
6-1
5-2
5-1
4-2
4-1
3-2
3-1
2-2
2-1
1-2
1-1
NightofSubject
Number of 4 Weels Out
0 50 100 150 200 250
Sess1
Sess2
Sess3
Sess4
Sess5
Sess6
Sess7
Sess8

Of those 50 alarms, only 15 were authentic and associated
with a subsequent driving error and 35 were considered
false alarms. This leaves 1618 driving error events that
were completely missed by the device, less than a 1%
success rate
For example, looking at the number of driving errors by
subject (Figure 24) subject number 16 had over 200 driving
errors never triggering a single correct alarm from the
device. Other subjects showed this as well, but to a lesser
extent. The individuals that committed the most number
of driving errors (subject 12) with approximately 275 out-of-
road errors had the most number of head nods (8 correct
and 3 incorrect alarms). The accuracy for this individual was
around 2%. The only other individuals who experienced
numerous alarms (subjects 3, 5, 11, and 13) all had more
than twice the number of false alarms as correct alarms.
Using head-nod as measured by the NapZapper to predict
and alarm for driving errors is not effective based on the
evidence provided here. The correlation between alarms
and accidents for the NapZapper was less than
1% compared to similar overall correlations for the DSSR
and the DSM, which were above 80% correlated with
driving performance.

A. Short-term
Short-term recommendations are based on looking
at only technologies that are currently commercially
available for use. Three of the top 5 rated
technologies in the technology assessment matrix
are currently available: ASTiD™ (Pernix), Haulcheck
(Accumine) and Optalert™ (Sleep Diagnostics). All of
these technologies have been trialed at mine sites.
There are two fundamental differences between
these three technologies.
Haulcheck merely looks at the trucks lane position
and alarms solely when that deviation passes a
set threshold that is considered dangerous. This
system should help prevent accidents so long as
the operator reacts quickly and appropriately to the
alarm. Unfortunately, the operators’ state at the
time of the alarm cannot be known, hence their
reaction to the alarm may not be timely enough to
prevent accidents with other vehicles, machines, or
infrastructure. As a last line of defense this product
will assist in reducing the number of collisions at a
mine site, however, in Haulcheck’s current design, it
is unable to identify behavioral or physiological signs
of drowsiness at an early stage.
ASTiD™ has gone through very detailed on-site
testing accumulating almost 1000 hours of field
usage. Results from this trial demonstrated that
when operators are willing and cooperative the
system is capable of providing excellent feedback
throughout a shift (both day and night) on the
operators’ state of drowsiness. With ASTiD’s
new analysis techniques, drowsiness detection is
getting more accurate and predictive. The latest
generation of ASTiD™ now includes an integrated
dispatch system that provides up to the minute
information on all operators in all trucks in the
fleet with a simple Red, Yellow or Green rating.
This real-time feedback can allow the dispatcher to
strategically alter truck assignments or crib breaks
to accommodate operators who are showing signs
of increased drowsiness compared to the rest of the
shift. ASTiD’s major shortfall comes with its need
for operators to enter their sleep quality and quantity
for the previous 24-hour period. So long as operators
are cooperative this system allows the system to
fine tune its algorithm to increase its sensitivity and
reduce false alarms. However, if operators do not
understand how to correctly enter this information or
purposefully enter in incorrect information, then this
could result in either increasing or decreasing the
sensitivity of the system. Either of which could lead
to undesired results. The supplier has suggested that
future versions of the device will no longer require
this input for accurate detection.
The benefit of this system is that it requires no direct
contact with the operator and only evaluates the
operation of the truck. Operators tend to feel more
comfortable with technologies that focus on the
vehicle rather than themselves. With proper training
and implementation ASTiD™ could provide a very
effective supporting technology to a site’s fatigue
management program.
As with ASTiD™, Optalert™ has also undergone field
trials. During these field trials excellent information
was gathered which allowed for hardware and
software changes to be made to accommodate
the mining environment. Specifically the system
takes transmission and speed information to
dynamically activate or deactivate the system based
on whether the machine is parked or whether it is
in motion. Optalert™ has a very effective algorithm
that evaluates the operators’ state of drowsiness
on a continuous basis. The nature of this system
requires the operator to wear a pair of sensor
glasses, which provides the algorithm with the
drowsiness measures. The glasses are attractive
and can accommodate prescription lenses or tinted
lenses. Shortfalls for this system are that each
pair of glasses requires professional fitting and
adjustment to ensure consistent and accurate data
collection. As was found in the field trials, slippage
or misalignment of the glasses can negatively affect
the systems performance or cause the system not to
work entirely. As with the ASTiD™ system, success
v. Project Recommendations

v. Project Recommendations
of the system requires operator cooperation. If the
glasses are not worn correctly or are not worn at
all, the system is compromised. Optalert™ is also
developing dispatch software similar to the ASTiD™
that will inform a remote location the drowsiness
level of each operator.
The benefit of this system is that is uses an
extremely sensitive algorithm that is personalized
to each operator. This personalized calibration
leads to improved accuracy and better prediction of
drowsiness for the user.
For the purposes of immediate implementation,
customers only have three real options: Haulcheck,
ASTiD™ and Optalert™. Each of the technologies
has pros and cons that are not straightforward in
terms of identifying a clear leader. The strengths
and weaknesses of the technologies differ across
systems. The decision as to which technology
would provide the best solution for a particular
site will require assessing each site’s cultural and
organizational makeup. Based on the expert ratings
and the field trials ASTiD™ and Optalert™ have
consistently performed well and supplier interviews
have shown each company is dedicated to producing
and providing their customers with a drowsiness
detection solution geared specifically towards the
mining industries needs. It is the recommendation
of this team that Optalert™ and ASTiD™ both can
provide an immediate assistance in identifying and
warning operators of eminent drowsiness.
B. Intermediate-term
Other companies are developing detection systems that
do not require operators to wear glasses, as required
by the Optalert™ system. These newer technologies
will also not require input from the operator on sleep
quality or quantity. These have been the major concerns
with the technologies currently available to the market.
The industry should encourage and support companies
such as Sleep Diagnostics Pty Ltd, Pernix Ltd, Seeing
Machines and Delphi in their development of machine-
integrated systems in the hope that an equally effective
solution can be developed that doesn’t require the
up-front and on-going costs associated with providing
individualized glasses, placing PVC pipes along all haul
roads or operator input.
C. Long-term
The outlook long-term for fatigue technologies would
be to have sensors completely off of the operator using
both operator and machine performance variables
as part of a combined fatigue/drowsiness detection
system. This system could be dynamically linked with
the dispatch system and a collision warning system to
provide multiple layers of protection for the operator
regardless of their level of fatigue. It is recommended
that suppliers look at combining the best elements
of their respective technologies together. This would
provide the quickest and best chance for success
in moving the state of fatigue/drowsiness detection
technologies forward.

We would like to thank all the technology suppliers
for their willingness to share their products with our
team. Specifically, Delphi, Pernix Ltd, Sleep Diagnostics
Pty Ltd, SMI and Seeing Machines all provided
unprecedented access to their engineers and scientists
in support of this project.
vi. Acknowledgements

David Edwards, Ph.D.
Senior Engineering Specialist and
6 Sigma Black Belt, Caterpillar
David Edwards is a Sr. Engineering Specialist and a
former 6 Sigma Black Belt with Caterpillar Inc. David
holds a doctorate degree in Behavioral Neuroscience
from the University of Alabama at Birmingham. Dr.
Edwards specialized in cognitive ergonomics and
transportation safety examining age-related effects
on visual function, attention and driving performance.
Following his education, Dr. Edwards worked for
Hyperion Technologies, one of the world’s leading
makers of automobile simulators for research and driver
training as a product development engineer developing
simulation software for novice driver-training.
Dr. Edwards has been with Caterpillar’s Technology
Solutions Division in the Ergonomics Technology Group
for the past 6 years. During this time Dr. Edwards has
focused his research on the development of analysis
tools for studying operator mental and physical workload.
Highlights include implementing portable EEG and eye
tracking technology on earth-moving machines to analyze
operator mental workload and visual behavior during earth
moving applications. Dr. Edwards is now coordinator for
safety research and development with Caterpillar’s newly
formed Customer Safety Services division.
Dr. Edwards began studying operator fatigue in 2001
working on a joint study with the National Institute of
Occupational Safety and Health (NIOSH) to examine the
state of fatigue detection technology and it’s potential
use in mining operations. This research led to technology
trials in U.S. and Indonesian mines.
Dr. Edwards has been an invited speaker on the topics
of driver safety, fatigue and collision warning by mining
companies, NIOSH and professional research organizations.
Appendix 1: Team Member Bios

William G. Sirois
William G. Sirois
Senior Vice President COO,
CIRCADIAN™
Bill Sirois is Vice President and Chief Operating
Officer for CIRCADIAN™. He is responsible for all
CIRCADIAN™ services in North America and Europe,
including development of Industrial Shift work
Strategies, Alertness Assurance programs, Human
Alertness Technologies, Ergonomics programs, Industrial
Engineering, Pre-employment Screening, Behavioral
Safety Development, and Bio-compatible Shift work
Scheduling and support training on Managing a Shift
work Lifestyle.
By addressing human limitations and capabilities from a
holistic perspective (i.e., operational, physiological, and
sociological), Mr. Sirois has demonstrated that a new
frontier of opportunity exists for human asset utilization and
continuous improvement in overall employee health, safety,
and operational performance for all types of business.
Mr. Sirois has also published and lectured extensively
as a featured speaker at numerous corporate meetings
and international conferences, including the National
Association of Manufacturers, The Society of Plastics
Engineers, National Ergonomics Conference, The
American Petroleum Institute, The American Shipping
Club, International Semiconductor Safety Association,
Canadian Electric Association, the National Food
Processors Association, the National Transportation
Safety Board, the Puerto Rico Health and Safety
Conference, and the Institute of Mining Health, Safety
and Research. Mr. Sirois holds a degree in Chemical
Engineering from the University of New Hampshire.
William Davis
Vice President of Operations, CIRCADIAN™
Bill Davis joined CIRCADIAN™ as a former client and
now serves as Vice President of Operations for CTI.
He is an industrial safety manager with a broad-based and
unique operational background that spans nearly 20 years.
This has included production experience at the facility,
divisional and corporate levels at International Paper and
other leading pulp paper companies.
Beginning as a shift worker in the Pennsylvanian steel
mills, Bill has held both plant management and corporate
safety positions in the paper and specialty board
industries. He has extensive experience working with
a variety of unions and governmental safety and health
regulatory agencies, as well as first-hand experience with
high performance self-directed work environments. His
real-world industrial background affords a natural rapport
with managers, union representatives and employees at
all organizational levels.
Todd A. Dawson, M.S.
Director of Research, Grants Special Projects,
CIRCADIAN™
Mr. Dawson graduated from Harvard University with
a BA in biological anthropology. While at Harvard, he
focused his studies on the biological rhythms of human
hormones with a special focus on cortisol. After joining
CIRCADIAN™ in 1994, Mr. Dawson was part of several
Fatigue Risk Assessments in which he investigated the
sources of fatigue and proposed fatigue countermeasures
for industries including commuter rail operations,
manufacturing, and marine transport.
Mr. Dawson spent nearly three years as a project manager
for the Canadian National Rail While on the Canadian
National Rail project, his focus was strictly on the train
crews. The combination of these two projects has provided
him with excellent understanding of the freight rail
operation. Mr. Dawson has managed projects at companies
including ChevronTexaco, Roadway Express, Tidewater
Marine, Hutchison Port Holdings, and GO Transit.
As the Director of Research, Grants and Special Projects
at CIRCADIAN™, Mr. Dawson leads CIRCADIAN’s world-
class research team in developing assessments and
solutions for the wide range of challenges confronting

companies with extended hours operations. Current
research focus areas include employee health, employee
demographics, and operations management best
practices. Mr. Dawson also oversees the development
of new technologies, including the Flexible Workforce
Management System (FWMS), Freight Rail Crew
Optimization System (FRCOS), CIRCADIAN™ Alertness
Simulation (CAS), Shift work Adaptation Testing System
(SATs), and microsleep prediction systems (including
alertness monitoring technologies).
Udo Trutschel, PH.D.
Senior Research Scientist, CIRCADIAN™
Udo Trutschel graduated with a degree in theoretical physics
from the Institute for Solid State Physics and Theoretical
Optics from Friedrich-Schiller-University (Germany). He
received his doctoral degree in applied physics from the
Physical Institute, Technical University Ilmenau (Germany).
After leaving Germany in 1991, Dr. Trutschel worked for 18
months as research assistant at Tufts University, Boston in
the Electro-Optics Technology Center. Afterwards he took
a position as visiting Professor at the Electrical Engineering
Department, Laval University, Quebec for 3 months.
Dr. Trutschel joined CIRCADIAN™ in 1995. He pursued
research on automatic detection of microsleeps / drowsiness
from electrophysiological recordings, time series modeling,
and the development of algorithms for alertness simulation
and prediction, resulting in the development of the
CIRCADIAN™ Alertness Software (CAS).
More recently, Dr. Trutschel supervised the development
of the Optimization-Simulation System for biocompatible
crew scheduling and the Freight Rail Crew Optimization
System (FRCOS) software used in railroads. He supervised
several research projects focusing on the characterization
and detection of microsleeps based on EEG, EOG and eye-
parameter measures using supervised and unsupervised
neural network techniques.
Dr. Trutschel currently serves as Senior Research Consultant
and focuses on the development of software system based on
a Flexible Workforce Management Approach (FWMA). Other
current activities include the design of Shift work Adaptation
Testing-System (SATS), microsleep detection technologies and
knowledge-based systems for alertness prediction.
Dr. Trutschel has published his research in over 50 scientific
publications and currently holds 8 patents.
Dr. Acacia Aguirre
Medical Director, CIRCADIAN™
Dr. Aguirre has over fifteen years experience in sleep
and alertness research, focusing on factors affecting
shift workers alertness, safety and health, and the
development of fatigue countermeasures. She also
has extensive clinical experience in the field of sleep
medicine, having practiced as a sleep disorders specialist
at one of the major teaching hospitals in Paris, France.
After receiving her MD degree, Dr Aguirre completed
her D.M.Sc., which obtained the mention Summa cum
Laude. She completed her graduate research work in at
the University of Paris VI (France), where she received
her PhD in Neuroscience.
Dr. Aguirre’s work at CIRCADIAN™ includes providing
training and consulting support on major client
engagements, such as fatigue risk assessments,
workload analysis, evaluation of employees’ alertness
health and safety, scheduling and implementation of
fatigue countermeasures. She is also involved in the
design of educational materials for shift workers and
developed CIRCADIAN’s sleep disorders screening and
treatment program.
Dr. Aguirre is actively involved in the scientific
community and participates regularly in specialized
scientific meetings and symposia. She is member of
the European Sleep Research Society, and is also in

the Editorial Board of the Journal of the Spanish Sleep
Research Society. She has published over 50 scientific
articles and book chapters.
David Sommer
University of Applied Sciences, Schmalkalden,
Germany
David Sommer received his Master’s degree in
Computer Science in 1998 from University of Applied
Sciences, Schmalkalden, Germany. Since 1998, he
has been a scientific co-worker at the Department of
Computer Science and an Associate Lecturer in neural
networks and pattern recognition.
David has written over 50 publications on neural
networks, evolutionary algorithms, nonlinear signal
processing, data fusion and pattern recognition in
different areas of applications, such as driver fatigue,
posturography and sleep physiology.

Appendix 2: Expert Panel Reviewers
Dr. Larry Barr
U.S. Department of Transportation
Volpe National Transportation Systems Center
Advanced Safety Technology Division
Dr. Lawrence Barr is a senior research engineer in
the Advanced Safety Technology Division of the
U.S. Department of Transportation’s Volpe National
Transportation Systems Center. He has conducted
numerous safety-related programs and research studies
for the Federal Highway Administration, the National
Highway Traffic Safety Administration, the Federal
Motor Carrier Safety Administration, and the National
Aeronautics and Space Administration. These include
conducting motor vehicle crash causation studies as
well as a comprehensive benefit-cost study of crash
countermeasure systems for all major crash types and
vehicle platforms in support of the Intelligent Vehicle
Initiative, completing a detailed analysis of naturalistic
driving data to develop an understanding of the nature
and extent of driver fatigue and driver distraction
among truck drivers, providing technical support to
the evaluation plans for the Drowsy Driver Warning
System and Road Departure Crash Warning System
field operational tests, and developing investment
analysis and risk models of advanced aviation safety
and security technologies for the NASA Aviation Safety
and Security Program.
Dr. Barr recently completed a comprehensive survey
study on emerging vehicle-based driver drowsiness
detection and alertness monitoring technologies for
the Federal Motor Carrier Safety Administration. The
major objective of the study was to review and discuss
many of the activities currently underway to develop
unobtrusive, in-vehicle, real-time drowsy driver detection
and fatigue monitoring/alerting systems and evaluate
them against a set of proposed design guidelines and
user interface/acceptance criteria.
Dr. Barr received his bachelor’s degree in mechanical
engineering from the University of California at Davis,
a master’s degree in aerospace engineering from the
Pennsylvania State University, and a doctoral degree in
systems engineering from the University of New Hampshire.
Dr. Martin Golz
University of Applied Sciences, Schmalkalden,
Germany
Prof. Dr. Martin Golz graduated with a degree
in Electrical Engineering from the Institute of
Microelectronics from Technical University of Ilmenau
(Germany). He received his doctoral degree in applied
physics from the Physical Institute, Technical University
Ilme-nau. After graduation, Prof. Golz worked for four
years as research assistant at Central Research Hospital
of Neurology and Psychiatry “Wilhelm Griesinger”,
Berlin (Germany) in the Laboratory of Evoked Potentials
(1988-1990) and in the Sleep Polygraphy Lab (1990-
1992). Here he worked in biosignal analysis as well as in
specialized hardware development for performance test
technologies.
Afterwards he took a position as a Professor at the
Department of Computer Science at the University of
Applied Sciences Schmalkalden (Germany). From 1992
to 2003 he was Professor for Physics and Measurement
Engineering, and since 2004 he has been holding
the full time Professorship for Signal Processing and
Neuroinformatics at the same De-partment.
Prof. Golz pursues research on automatic detection
of microsleeps and of evaluation of drowsiness from
electrophysiological recordings utilizing numerous
data fusion algorithms from the field of Soft
Computing, especially Fuzzy, Neural, Neuro-Fuzzy
and Evolutionary Technology.
He supervised several Master and PhD thesis
as well as several research projects focusing on
occulography, posturography and on classification of
sleep composition. In 2001, he built a driving simulation
laboratory to further his fatigue research. Since then,
he has conducted more than ten research studies on
fatigue, microsleep and driver performance. Recently,
Prof. Golz established a gold standard for fatigue
prediction and detection based on a database of about
20,000 examples of microsleep events. His presentation
on this topic was awarded at the Sensation International
Conference “Monitoring Sleep and Sleepiness - From

Physiology to New Sensors” in Basel May 2006.
Prof. Golz is the organizer of the special invited session
“Signal Processing Techniques for Knowledge Extraction
and Information Fusion” at the 10th International
Conference on Knowledge-Based Intelligent
Information Engineering Systems, Bournemouth U.K.
which will be held in October 2006.
Dr. Rich Hanowski
Virginia Tech Transportation Institute,
Blacksburg Virginia
Richard Hanowski is the Leader of the Truck and Bus
Safety Group at the Virginia Tech Transportation Institute
(VTTI). Dr. Hanowski has been conducting transportation
safety research since 1992. Previously a Research
Scientist at the Battelle Memorial Institute, Dr. Hanowski
joined the Safety Human Factors Engineering Group
at VTTI in 1996. In 2003, after completing several
successful projects and building a sustainable program
in truck and bus safety, a new group at VTTI was formed
with Dr. Hanowski as leader.
Dr. Hanowski has formal training in human factors,
system design, safety, cognitive psychology,
human-computer interaction, training evaluation,
experimental design methods, and statistics. His
experience includes transportation human factors
with both light heavy vehicles, laboratory field-
testing, focus groups, real-time automobile heavy
vehicle simulation, human factors design guideline
development, older driver investigation, collision
warning, and Intelligent Transportation Systems. Dr.
Hanowski specializes in human factors engineering,
advanced product design/test/evaluation, and human
performance evaluation. He is skilled in all phases of
research including conceptual framing, research design,
data collection, data synthesis analysis, assessment
of results, and presentation of findings.
Dr. Hanowski is the author of over 70 scientific articles
and technical reports. He is an active member of the
Intelligent Transportation Society of America, and served
as Chairman of the ITS America Safety Human Factors
Committee (2000-2002). Dr. Hanowski is also active in
the Human Factors and Ergonomics Society, and serves
as a technical paper reviewer for various transportation-
related organizations and journals.
Todd Ruff
National Institute of Occupational Safety and
Health, Spokane Research Center
Todd Ruff obtained a Bachelor of Science in Electrical
Engineering from Gonzaga University in 1988 and a
Master degree in Electrical Engineering from Gonzaga
in 1993. He currently works for the National Institute
for Occupational Safety and Health, Spokane Research
Laboratory, in Spokane, WA as an electrical engineer and
research project manager.
The Spokane Research Laboratory (SRL) serves as the
second focal point for mine health and safety research.
While research programs touch most mining sectors, the
major program focus is on metal and nonmetal mining.
To prevent injuries and fatalities in both underground
and surface mines, SRL: (1) identifies and classifies
risk factors in mining; (2) evaluates recommendations
for strategies to prevent injuries and disease through
the use of effective control technologies; (3) studies
the design of mining equipment to assess the potential
risks involved in using it; and, (4) designs, builds, and
tests equipment that incorporates innovative control
technologies.
Mr. Ruff has been the technical team leader for the
development and evaluation of technologies to improve
the safety of mining and construction equipment,
particularly in the area of collision warning systems
and operator fatigue detection. He is registered as a
Professional Engineer in Washington State.
Dr. Mario Sandoval
Fulcrum Engineering, Partner and Director
Dr. Sandoval has 17 years of experience on mining and
work at high altitude. He is a partner and director of
Fulcrum Engineering; a company specialized on help

business in the areas of project engineering, human
resources (shift work) and strategic planning. One of
his main projects in Fulcrum has been as a director of
a project to develop software programs to evaluate
and design shift work systems for mining operations.
Besides holding a MD degree, Dr. Sandoval holds a
Master in Environmental Science (University of Chile)
and a Master in Ergonomics (Universidad Politécnica,
Catalonia, Spain). He also obtained a Diploma on
Evaluation and Preparation of Health Projects (University
of Chile).
Before joining Fulcrum, Dr. Sandoval has been Director
of the Department of Work in Altitude (Workers
Hospital, depending of the Chilean Safety Association),
Director of the Center of Ergonomics of Work in Altitude
(1997-2002), and Director of the RD Department of
the Aerospace Medical Center in the Chilean Air Force
(1999-2002). Since 2002, he has been Medical Advisor
for the Chilean Safety Association.
In addition to his consulting and research work, Dr.
Sandoval also participates as a instructor in the Master
on Public Health and Risk Prevention (Institute of Public
Health, University of Santiago, Chile) and Master on
Science of Exercise (University Andrés Bello, Chile),
where he is in charge of the courses on physiology of
extreme environments.

Operator Fatigue Detection Tech Review Finds ASTiDTM & OptalertTM Most Effective

Operator Fatigue Detection Tech Review Finds ASTiDTM & OptalertTM Most Effective

Recommended

Recommended

More Related Content

Similar to Operator Fatigue Detection Tech Review Finds ASTiDTM & OptalertTM Most Effective

Similar to Operator Fatigue Detection Tech Review Finds ASTiDTM & OptalertTM Most Effective (20)

Recently uploaded

Recently uploaded (20)

Operator Fatigue Detection Tech Review Finds ASTiDTM & OptalertTM Most Effective