To repeat a key point made in Chapter 3, there is no known biological substitute for sleep; the body cannot compensate for lost sleep through any other natural mechanism. Research cited in Chapter 3 indicates that many commercial motor vehicle (CMV) drivers may be obtaining less than 6 hours of sleep per 24-hour day, which is generally viewed as being insufficient to maintain adequate levels of alertness on the job. Accordingly, as discussed in Chapter 7, sleep insufficiency increases the risk of being involved in a crash. CMV drivers also must be cognizant of circadian, time-of-day influences on their levels of alertness and the associated increased crash risk. Even when a person has obtained sufficient sleep, he or she will experience two circadian-driven lulls that adversely affect alertness, bring about feelings of fatigue, and lead to degraded performance (e.g., slowed reaction times) (see Chapter 3). To address these concerns, both technological approaches and fatigue management protocols have been advanced that are designed to detect and manage fatigue among CMV drivers.
Conclusion 9: Acute and chronic sleep insufficiency produces fatigue in drivers, as do lengthy and irregular work schedules.
Conclusion 10: There is no biological substitute for sufficient sleep.
This chapter first reviews technological approaches for detecting and managing operator fatigue. It then describes infrastructure- and
vehicle-based systems designed to mitigate the effects of fatigue by alerting drivers that they are in a drowsy state. Next, the chapter briefly examines fatigue management programs. The final section addresses the importance of a safety culture.
Fatigue detection and management approaches are of three main types: (1) online operator fatigue detection technologies, (2) fitness-for-duty indicators, and (3) biomathematical scheduling models. In general, the effectiveness of all three of these approaches in detecting or predicting operator fatigue remains unclear. Driver drowsiness detection devices and biomathematical scheduling tools are in varying stages of development (see, e.g., Mallis and James, 2012), and many of them have not been tested in third-party, randomized controlled studies. In addition, the implementation of these approaches in the field poses a variety of challenges.
Fatigue detection technologies and predictive algorithms must be shown to meet a number of criteria prior to use. First, they need to be operationally feasible, working under a variety of environmental conditions such as during driving at nighttime, on different types of roadways, in traffic, and so on. Second, they need to have small errors both when indicating that a driver’s alertness is low or that he or she may be fatigued while driving and when not indicating fatigue. In other words, they need to have infrequent false positives (indicating a driver is fatigued when he or she is not) and false negatives (indicating a driver is alert when he or she is not). Third, such devices must work for a relatively heterogeneous group of drivers in a variety of driving situations, and they must designed to be user-friendly. Finally, all such devices need to be shown to measure a biobehavioral marker of hypovigilance due to drowsiness. (For details on sets of acceptance criteria for such devices, see Dinges and Mallis .)
With respect to false positives and false negatives, it is clear that a system with a high rate of false negatives would be a serious problem. However, a moderate rate of false positives can also be problematic. Providing a warning to a driver when he or she is not fatigued may produce annoyance, reduce trust in the system, and decrease operator compliance with future warnings. Thus it is important that research on new technologies be focused on improving the accuracy of these devices for those that both are and are not drowsy.
It should also be noted that an effective near-real-time drowsiness detection device would be extremely useful for preventing many future crashes. But so, too, would be a device that could help in establishing
Online Operator Fatigue Detection Technologies
Online fatigue detection technologies are intended to provide feedback to a driver on his or her alertness level and thereby to detect operator fatigue prior to its onset. Equipped with this real-time feedback, drivers can opt to take a break from driving or use some other validated fatigue countermeasure. Additionally, drivers can monitor their alertness levels across their work-rest duty cycles to determine whether they should implement schedule-related fatigue management and mitigation strategies. Such preventive strategies include scheduling sleep and nap periods prior to the start of a duty cycle; taking rest breaks during the duty day; maximizing sleep time on nonduty days, preferably in alignment with biological night; and if possible, negotiating with employers for more fixed, regular work duty scheduling. These preventive strategies can help reduce operational errors and crashes associated with insufficient alertness attributable to neurobehavioral deficits resulting from sleep loss, circadian misalignment, or simply boredom associated with sustained work.
Online real-time operator fatigue detection technologies include a variety of methods for directly measuring factors associated with driver fatigue, such as electroencephalographic (EEG) frequency band activity; heart rate variability; and ocular variables, including saccades, slow eye movements, blink rate, or eyelid closure (see Abe et al., 2011; Chua et al., 2012; Dinges et al., 1998). One ocular measure that has consistently shown promise is PERCLOS (percentage of eyelid closure), the proportion of a time interval during which the eyes were 80 to 100 percent closed (not including blinks) (Wierwille and Ellsworth, 1994; Wierwille et al., 1994). Such measurements have been experimentally validated as being sensitive to fatigue and circadian changes (Abe et al., 2011; Chua et al., 2012; Dinges et al., 1998, 2002; Mallis and Dinges, 2005; Mallis et al., 2007; Ong et al., 2013). A PERCLOS measurement system mounted in a vehicle cab entails video monitoring of the individual, and can effectively provide the driver with an alert that he or she is becoming fatigued and suggest that it is time to take a break from driving (Mallis et al., 1998).
Early research and on-the-road testing have shown, however, that implementing PERCLOS monitoring systems for truck drivers is not always operationally feasible. Early generations of PERCLOS devices had technological limitations, as they did not accurately track slow eyelid closure throughout all environmental conditions. Some of these inaccuracies were attributable either to interference due to sunlight reflections or to dif-
ficulty in measuring pupil opening because of low light conditions during nighttime driving. Additionally, PERCLOS systems were not able to distinguish between eyes being out of field of view of the PERCLOS camera because of drivers constantly turning their head, mainly to check their side-view mirror, and drivers closing their eyelids completely. PERCLOS monitoring devices continue to be improved. Some include information on other variables to address the above challenges. For example, efforts are now under way, using optical computer recognition, to develop the capability for continuous tracking of percentage of eyelid closure in real time (Dinges et al., 2005a, 2007).
Fitness-for-Duty Testing Devices
Fitness-for duty testing devices entail probed evaluation or temporally discrete sampling of neurobehavioral performance or aspects of physiological changes (primarily ocular and pupillary measures). They represent an attempt to assess fatigue risk, especially in safety-sensitive, around-the-clock operations in which the risk of errors can be elevated because of fatigue-inducing work schedules. Unlike online fatigue detection approaches, fitness-for-duty tests produce a snapshot in time of a worker’s level of alertness prior to a duty period that can predict performance capability later in the day during the duty period. The prediction is based on a comparison of baseline data specific to the individual, consisting of measurements taken when the individual was well rested or without a sleep debt.
Earlier generations of fitness-for-duty devices were relatively large, were not portable, and required relatively long sampling periods. Some were as large as the Truck Operator Proficiency System, a driving simulator installed at a truck terminal that measured psychomotor and divided-attention performance. The device was used in the 1990s in New Zealand and in Arizona. No formal predictive validity was established for the device; however, the test had high face validity, so it was believed that failing it meant the driver could not drive safety (Charlton and Ashton, 1998; Hartley et al., 2000).
Other, smaller devices used tabletop pursuit rotor tracking performance tasks, exemplified by the Factor 1000 system, which employed a critical tracking task requiring eye-hand coordination. These tabletop computerized test batteries included as many as 8 to 10 tests, including visual perception, reaction time, concentration, cognitive processing, and even personality assessment. (An example is ART 90, a fitness-for-duty test system used in several places in Europe; see Charlton and Ashton ).
Most such fitness-for-duty systems were performance-based and entailed substantial learning curves because they were affected by apti-
tude and experience levels and did not necessarily prove to be sensitive to fatigue. Some became more useful in detecting drug effects on performance at work. Over the years, technological advances have resulted in the development of fitness-for-duty testing devices that are portable, take less time to administer, and use measures that have been demonstrated to be sensitive to fatigue. Fitness-for-duty tests that can be administered on an electronic tablet or smartphone hold particular promise for being able to predict performance capability over a brief sampling timeframe (e.g., estimating whether it is safe for a driver to extend a duty period). One caution is that the tests may not be feasible in all operational environments.
The 10-minute psychomotor vigilance test (PVT) is an example of a probed-performance fitness-for-duty test. It measures the ability of the brain to sustain attention and repeatedly respond quickly to a simple visual (or auditory) stimulus. The stimuli occur at a predetermined interstimulus interval (ISI) range, and measurements are highly precise (Basner and Dinges, 2011). There is no lengthy learning curve, and the test is independent of aptitude, experience, and skill. Reaction time, response speed, and lapses in attention (longer response times or no response) are the primary outcome measures (Basner and Dinges, 2011; Dinges and Powell, 1985). The PVT is an accepted standard for measurement in sleep and circadian research because of its demonstrated sensitivity to acute total sleep deprivation, chronic sleep restriction, and circadian rhythmicity. Research protocols conducted in operational settings also have demonstrated its validity for identifying operator fatigue (Gander et al., 2008; Pack et al., 2006; Rosekind et al., 1994)
The 10-minute duration of the PVT is often considered impractical in most operational environments because it requires the individual to disengage completely from his or her operational duties (e.g., driving) to take the test. Consequently, two shorter-duration versions of the PVT have been developed: the adaptive PVT-A (about 3 minutes, depending on the individual’s performance [Basner and Dinges, 2012]) and the brief PVT-B (about 5 minutes). Both have modified algorithms for performance evaluation and have been extensively validated for their sensitivity to fatigue (Basner et al., 2011). Future studies are needed to show the feasibility and usefulness of employing PVT-like fitness-for-duty tests in safety-sensitive operational environments.
Biomathematical models of cognitive performance and alertness are used to calculate an estimate of alertness based on sleep-wake schedules and the timing and placement of duty shifts (Dawson et al., 2011; Mallis
et al., 2004). Most are based on the two-process model, as originally described by Borbély (1982). The algorithms model the known effects of the interaction between homeostatic sleep drive and the circadian system on alertness and cognitive performance. The interaction of these two physiological processes is dynamic and complex (see Chapter 3). In oversimplified terms, the sleep process is approximated to be an exponential saturation function during sleep itself, with linear degradation in performance with increasing hours of wakefulness (i.e., being awake 18 or more hours since the last sleeping period). The circadian process is of sinusoidal form and shows variations in performance on an approximately 24-hour cycle. The observed changes in performance are closely correlated with endogenous rhythms of core body temperature. These algorithms are operationally useful only when they are incorporated into a scheduling tool.
Each biomathematical scheduling tool uses specific algorithms that differ in various ways, including input variables, output measures, goals, and capabilities. It is also important to consider the operational environment for the model’s intended use. For example, some models use only data collected in tightly controlled laboratory studies, while others are adjusted based on measures collected in operational environments, including military, trucking, aviation, and railway operations. The U.S. Department of Defense, U.S. Department of Transportation, and National Aeronautics and Space Administration sponsored a modeling workshop in 2002. All seven of the models evaluated at the workshop showed promise in the prediction of alertness and performance, but each required further scientific validation in tightly controlled laboratory settings and improvements in reliability, sensitivity, and specificity before being transitioned to operational settings (Mallis et al., 2004). Once such a model has been deployed in an operational setting, data collected can be used to further refine the algorithm, making it more specific to its intended use.
Biomathematical models can be used to increase safety and minimize fatigue risk by enabling comparison of different work-shift or work-rest scheduling scenarios. The goal is to maximize safety by evaluating schedule parameters scientifically rather than in accordance with time-on-task theories (see Chapter 3). Schedule evaluations include (1) predicting times when performance is optimal, (2) identifying time frames in which recovery sleep will be most restorative, and (3) determining the impact of proposed work-rest schedules on overall neurobehavioral functioning (Mallis et al., 2004).
Government entities are showing increased interest in the use of biomathematical models in the development of regulations and comprehensive fatigue risk management programs. In 2014, the Federal Aviation
Administration (FAA) made changes to the Federal Aviation Regulations (FARs) that allow carriers to submit a fatigue risk management system (FRMS) application in requesting an exemption from prescriptive regulations (FRMSs are discussed further below). Currently, the FAA uses the Sleep, Activity and Fatigue Task Effectiveness (SAFTE) model (Hursh et al., 2004) as part of its process for evaluating FRMS applications. The SAFTE model was originally developed for military and industrial settings and is the foundation of the Fatigue Avoidance Scheduling Tool (FAST) (McCormick et al., 2013). And the U.S. Navy supported development of the OWL (Optimized Watchbill and Logistics) tool based on a published biomathematical model (McCauley et al., 2009, 2013) to generate duty schedules for sailors that minimize fatigue risk.
Currently, at least one biomathematical model has shown evidence of being able to predict cognitive deficits associated with chronic sleep loss (McCauley et al., 2009, 2013). Although improvements in biomathematical model development continue, however, limitations still exist. Scheduling tools that incorporate biomathematical models of fatigue and alertness are only one component of a comprehensive fatigue risk management program, and such models are suitable only for use as a tool to help inform final decisions about a schedule’s level of safety. Moreover, only one published model incorporates individual differences, providing more accurate estimates of fatigue (Van Dongen et al., 2007). Refinements in biomathematical models to account for individual differences would facilitate the use of individual countermeasures (e.g., recovery sleep, naps, or caffeine).
Conclusion 11: Operator fatigue has been singled out for its negative safety implications for all workers, including commercial motor vehicle drivers. Such concerns have motivated a variety of applied research projects on detecting, preventing, and managing fatigue.
Conclusion 12: Despite almost three decades of research on the topic, technological innovations for detecting driver fatigue in near real time and operational strategies for their use are still in the early phases of understanding and application.
Conclusion 13: Biomathematical models can be useful for the development of general work-rest schedules. However, existing models do not account for individual variation, so care must be taken in applying them to address likely impacts of irregular work schedules.
Both road infrastructure-based and vehicle-based systems have been developed to mitigate the effects of fatigue by awakening drivers or warning them that they are in a compromised state and at risk of an accident.
Roadway rumble strips are an infrastructure-based system that does not specifically target driver fatigue, but may be effective in helping to prevent a range of crashes related to taking one’s eyes off the road as a result of either driver fatigue or distracted driving. Rumble strips may alert a sleepy driver as well as redirect the attention of an alert but distracted driver.
Studies of the safety effect of rumble strips consistently have shown a reduction in run-off-the-road crashes, head-on collisions, and sideswipes, crash types thought to be related to fatigue and distraction. One study on rumble strips in Maine directly measured the impact on crashes identified as fatigue-related in crash reports, estimated as a reduction of 58 percent in fatigue-related run-off-the-road crashes on rural freeways (Garder and Davies, 2006).
Most studies of rumble strips have focused on crashes in specific states. Most have been before-and-after studies with comparison groups to control for exposure and various other potentially confounding factors, such as shoulder width and roadway curvature. Resulting estimates of effectiveness cover a broad range, from about 12 percent on freeways with speed limits under 65 mph to more than 50 percent for run-off-the-road fatal crashes. While the range of effectiveness estimates may be broad, it is important to note that all the estimates are positive and show a significant reduction in lane- and road-departure crashes (Datta et al., 2015; Khan et al., 2015; Patel et al., 2007; Sayed et al., 2010).
One study of Connecticut roads specifically addressed the possibility of fatigued drivers using the rumble strips to alert them repeatedly to their own drowsiness and redirect their attention, thus encouraging them to keep driving until they drifted off the road along a stretch with no rumble strips. The study uncovered some evidence for this phenomenon, finding that some stretches of road with no rumble strips experienced an increase in run-off-the-road crashes relative to stretches of road along the same route with rumble strips installed (Smith and Ivan, 2005). A simulator study of fatigued drivers showed how this might occur. Thirty-five night shift workers were tested immediately after finishing work with a simulated driving task along a two-lane road. Objective measures of sleepiness included EEG reading and eye-closure duration. All drivers showed an increase in sleepiness immediately prior to hitting the (simu-
lated) rumble strips. The rumble strips had the effect of alerting the drivers, as indicated by the objective measures, but the drivers returned to their baseline fatigue within about 5 minutes (Anund et al., 2008).
Thus, one drawback of rumble strips is that they may provide drivers with a false sense of security. Even if drivers feel drowsy, they may not pull over and stop to take a rest break, thinking that rumble strips will continue to alert them, and so they become increasingly at risk the longer they continue to drive. As a practical matter, installing roadway rumble strips requires a costly capital investment in highway infrastructure and sustained maintenance of the system.
Conclusion 14: Roadway rumble strips serve to help prevent driver fatigue-related accidents. At the same time, there is the danger that drivers will treat rumble strips as if they provide repeated emergency alarm protection from falling asleep, and therefore will postpone taking other valid fatigue countermeasures, such as stopping for a rest.
When CMV drivers reach the limit of a duty period in accordance with hours-of-service (HOS) regulations, they are required to stop driving for a 10-hour stretch and spend at least 8 hours in the sleeper berth. Designated public rest areas are constructed along Interstate highways to provide both car and truck/bus drivers with a safe location to pull over and rest. Many such public rest areas have parking lots specifically allocated for trucks and buses. On many routes, moreover, numerous commercially operated truck refueling rest stops are available to accommodate many of the needs of long-haul truck drivers, including restaurant food, convenience store purchases, lounges, bath and shower rooms, and so on. These rest stops also provide spaces for extended truck parking to permit drivers to sleep in their truck-mounted sleeper berth.
Rest stops play a crucial role in contributing to highway safety by allowing fatigued drivers to pull off the road. McArthur and colleagues (2013) investigated the safety implications of rest areas by analyzing the relationship between the proximity of a road segment to a rest area and the frequency of fatigue-related crashes occurring on that road segment. Between 2006 and 2010, the authors collected crash data on all road segments that fell within a 20-mile radius of rest areas in the state of Michigan. Their spatial analysis provided evidence for the safety benefit of rest areas in reducing the frequency of fatigue-related crashes. Another finding of the study was diminishing returns from rest areas appearing as the 20-mile radius was expanded, pointing to the importance of spacing of rest areas. State-specific studies (Banerjee et al., 2009; Carson et al.,
2011; SRF Consulting Group, Inc., 2007; Taylor et al., 1999) likewise have shown a positive relationship between spacing of rest areas and fatigue-related crashes. All these studies found that rest area spacing of more than 30 miles led to increased crash risk.
Two of the studies also found that overcrowded rest areas or those with insufficient parking spaces for trucks were positively associated with increased crash rates (Banerjee et al., 2009; Carson et al., 2011). In 1996, the Federal Highway Administration (FHWA) commissioned a study to evaluate truck driver rest and parking needs along the National Interstate System. The study identified a shortfall of 28,400 truck parking spaces in public rest areas nationwide and predicted the shortage would grow to 39,000 in 10 years (U.S. Federal Highway Administration, 1996).
One way of addressing this problem is technology that can inform drivers where to anticipate a parking place where they can sleep. A large number of public/municipal parking garages (e.g., at airports or shopping centers) currently identify electronically at the front entrance not only how many parking spaces are available but also on which floor and even the specific spaces that are available. It would appear that related technological innovations could be applied to on-the-roadway parking facilities as well. The American Transportation Research Institute (ATRI), in collaboration with the Minnesota Department of Transportation, is developing a system that can identify available trucking spaces and communicate that information to drivers.1
It should be pointed out as well that constructing, maintaining, and policing the security of public parking for CMV driver rest areas has significant cost implications for federal, state, and municipal governments. Thus the trucking industry may need to take a leadership role in generating more parking at commercially operated truck refueling rest stops in geographic areas where it is needed the most.
Conclusion 15: Repeated surveys by trucking industry and other research organizations have revealed insufficient numbers of publically available rest areas where commercial motor vehicle drivers can safely take a lengthy rest. This issue has the potential to impact fatigue-related crash risks and needs to be addressed by the safety community.
Vehicle-to-infrastructure (V2I) technology allows for wireless communication between passenger vehicles and traffic and highway infra-
1 Truck Parking Availability Study: Demonstration Project. Available: http://atri-online.org/wp-content/uploads/2012/04/truckparkingonepager.pdf [March 2016].
structure to prevent collisions and manage traffic. Examples of V2I systems include monitors on bridges that communicate ice accumulation to approaching vehicles, traffic signals that warn vehicles of stopped traffic, and sensors warning of nearby emergency vehicles or work zones. A National Highway Traffic Safety Administration (NHTSA) analysis found that such technologies could potentially prevent 26 percent of all vehicle crashes (Najm et al., 2010). Even though they are not designed to address driver fatigue, such technologies that warn of traffic slowness/stoppages could also reduce fatigue-related crashes, thereby making them an effective fatigue countermeasure in certain circumstances. Whether V21 systems that warn of traffic slowness/stoppages reduce fatigue-related crashes is a topic for further research.
In addition to the systems installed on the vehicle to detect drowsiness/fatigue in drivers discussed earlier, vehicle-based systems for reducing fatigue-related crash risk include forward collision warning, automatic emergency braking, lane departure warning systems (LDWS), blind-spot object detection, and adaptive cruise control. Other systems monitor the driver’s use of controls, such as steering and braking, to detect degraded performance. Changes in the pattern of steering wheel adjustments, for example, have been used to detect degraded driver vigilance.
LDWS were developed for use on heavy commercial trucks as a type of driver alertness monitoring system. These systems are designed to warn drivers when they drift from their driving lane unintentionally, perhaps as a result of driver fatigue or distraction.
LDWS include several types of sensors installed on vehicles to monitor lane-tracking/lane-keeping performance, and then provide warnings when drivers deviate noticeably from the center of the lane over time (i.e., the past several minutes). Video sensors—usually camera-like devices mounted behind the windshield and aimed at the roadway in front of the vehicle or integrated beside the rear view mirrors—detect visible roadway lane and edge painted markings. These sensor data are fed into onboard computerized image recognition software to track a driver’s lane travel performance. Infrared sensors (either behind the windshield or under the vehicle) may be part of the system as well. The video sensors also may be accompanied by laser sensors mounted on the front of the vehicle.
Lane drift can be the result of anything from drowsiness and distraction to adverse weather conditions (e.g., snow, rain) that obscure the roadway paint markings. The LDWS continually monitor a vehicle’s position
and detects when the vehicle begins to drift toward an unintended lane change (e.g., perhaps approaching the roadway edge near the shoulder). Drivers inform the LDWS of plans to make intentional lane changes, such as to pass another vehicle, by first activating the turn signal device. Upon detection of lane drifting, the LDWS may present the driver with a visual display of his or her lane-tracking performance and/or present an audio warning. In the case of the AutoVue® LDWS (developed by Iteris Corporation), when a driver drifts out of a lane, the system emits a distinctive “rumble strip” noise from right or left door speakers. The LDWS may present other audible warnings to alert the driver to make a course correction to stay within the lane. False alarms are minimized by disabling the warnings when the vehicle’s speed is low.
A few versions of LDWS, particularly early prototype systems, were subjected to independent on-the-road testing in heavy trucks (see, e.g., Dinges et al., 2005a). Although LDWS were not completely validated as a driver fatigue predictor in such road testing (it would be unethical to induce driver fatigue for such testing), such a system was shown to “sharpen lane position awareness.” Moreover, LDWS gained acceptance by drivers as alertness monitors, and it was noted that in some cases, they even helped to improve driving skills.
The Iteris AutoVue LDWS was fielded on Mercedes Actros commercial trucks in Europe as early as 2000. In 2002, Freightliner Trucks’ North American vehicles made the Iteris system available as an “after market” option. Before selling the AutoVue system to Bendix CVS in 2011, Iteris reported that the system was in use in thousands of trucks, sold as an original equipment manufacturer option on new class 8 trucks. The system is widely available in newer automobile models as part of special option order electronic monitoring packages. It is clear that such LDWS technology could eventually allow truck fleet owners to analyze near-real-time safety information transmitted wirelessly from their vehicles using existing fleet communication systems.
Partially Autonomous Vehicles
In the relatively near future, partially autonomous vehicles may be used widely. Such vehicles are unlikely to fully replace the driver in the near term because driving remains a complex task. Nonetheless, partially autonomous vehicles have the potential to reduce the likelihood of fatigued driving by performing a majority of driving tasks (without direct involvement of the driver), thus reducing the driver’s cognitive workload and attention-demanding tasks.
On the other hand, partially autonomous vehicles could have a fatigue-related negative impact on safety, especially in driving situations
in which the hand-off of control to the human driver went beyond the design parameters of the system. Basically, as soon as the driving task became complicated, such as when the vehicle was entering a dense traffic area or something unexpected happened, the driver would need to take back control from the vehicle. How this would be done safely is both a design and operational challenge. Currently, autonomous driving systems do the easy part—driving vehicles down the road where the main challenge is lane keeping. But should a challenge arise that required the driver to assume control of the vehicle, that driver could have been lulled into disengagement by the system and could, for example, be using a smartphone or a laptop computer.
The success or failure of autonomous driving systems relies on effective human-system integration in design and practice. For such automation to be successful, the human user must be aware of the automation and react to it appropriately (see Shaikh and Krishnan, 2012). The type of warning that is most effective in attracting a driver’s attention and at times alerting the driver to the need to retake control of the vehicle needs to be determined.
Given that most of these systems may have sensors that can relay information from the vehicle to dispatchers or fleet managers, carriers might like to have the ability to force a truck or bus driver to pull over before a situation resulted in a crash. In any case, since the decision-making abilities of a fatigued driver are compromised, relying on the driver’s decision to pull over might not be sufficient in many cases. This is one of the many reasons why fatigue detection and mitigation technologies and autonomous driving systems need to be carefully designed, then thoroughly tested, evaluated, and validated. Moreover, given the various incentives that affect a driver’s decision making concerning pulling over to take a rest when tired, there is likely an important benefit to testing such systems in less controlled settings, such as in naturalistic driving studies.
NHTSA is exploring new on-board technologies for combatting drowsy driving. DrIIVE (Driver Monitoring of Inattention and Impairment using Vehicle Equipment) is an NHTSA project currently under way. Drowsiness often is evidenced by short episodes of degraded performance. Thus, the goal of the project is to use vehicle-based driver behavior data to predict impairment due to alcohol, drowsiness, and distraction (Brown et al., 2014). The research entails examining steering and pedal inputs and lane position and comparing this information with “signatures” of normal driving, when a person is awake and alert. Another goal
of the project is to demonstrate the potential to detect these states without the use of cameras to monitor drivers’ faces.2
In addition to the crash avoidance technologies available to truck drivers and fleets and work being pursued by NHTSA, car manufacturing companies have developed systems to warn drivers of inattentiveness or drowsiness. Mercedes-Benz’s Attention Assist system helps drivers recognize when they are drowsy or inattentive and advises them to take a break. When drivers are alert, they constantly, and subconsciously, monitor the position of their car and make continual small steering adjustments to keep the vehicle on a safe path. When drivers are fatigued, they experience periods of inattentiveness during which there is little steering input, followed by sudden and exaggerated corrections when the driver regains attention. Attention Assist uses a sensitive steering angle sensor to monitor the way in which the driver is controlling the car. At speeds between 80 and 180 km/h (55 to 110 mph), the system identifies a steering pattern that is characteristic of drowsy driving and combines this with other information, such as time of day and duration of journey. If a sequence of such events is identified, the system warns the driver to take a break by showing a coffee cup signal on the dashboard and emitting an audible tone. The driver may acknowledge the warning to make it disappear from the display. If the driver does not take a break and his or her driving style continues to indicate drowsiness or inattentiveness, the warning is repeated after 15 minutes.
Bosch is designing a system to evaluate driver microsleep, determine the level of drowsiness, and warn the driver if necessary. The system analyzes steering behavior to identify when the driver does not steer and then makes an abrupt steering correction. This system also makes use of the speed and duration of travel.
Finally, it is worth mentioning, as is the case for rumble strips, that safety issues can arise if crash avoidance technologies are considered countermeasures for fatigue because they provide protection against crashes, rather than addressing fatigue. These devices essentially help protect drivers from some of the consequences of fatigue, but they should not be used as a justification to continue to drive.
2 Previous contributions by NHTSA include detection of impairment from alcohol using vehicle measures (DOT HS 811 358), visual measures for detecting driver distraction (DOT HS 811 547A), and advanced countermeasures for multiple impairment (DOT HS 811 886). The agency is in the process of (1) developing and evaluating a system of algorithms for identifying signatures of alcohol-impaired, drowsy, and distracted driving; (2) assessing potential countermeasures for drowsy driving-associated lane departures; and (3) evaluating an initial proof of concept for the use of this system in the development of safety technologies such as driver feedback displays for drowsiness.
Conclusion 16: Additional research is needed on the effectiveness of all devices that may address reduced vigilance due to fatigue, including forward collision warning, automatic emergency braking, lane departure warning systems, blind-spot object detection, adaptive cruise control, and any other in-vehicle driver drowsiness/fatigue detection systems. This research needs to encompass not only the devices’ effectiveness but also the results of actual deployment, the impact of driver acceptance, and any negative consequences of using such devices inappropriately as countermeasures for fatigue.
Fatigue management programs define policies and procedures for managing and mitigating fatigue in safety-sensitive environments (Lerman et al., 2012). They often are implemented within health and wellness programs or safety management systems. Each freight- or passenger-moving operation is unique and presents its own fatigue challenges. A key feature of fatigue management programs is that they consider both physiological (see Chapter 3) and operational factors. Therefore, these programs most commonly are tailored to the operational needs and constraints of particular work settings (or companies) (Dinges and Mallis, 1998; Mallis and James, 2012).
Fatigue management programs can be classified into two broad categories: (1) fatigue risk management plans (FRMPs) and (2) fatigue risk management systems (FRMSs). FRMPs establish policies on managing and mitigating fatigue during operations. They typically include a requirement for fatigue awareness training for employees (e.g., drivers, fleet managers, dispatchers), as well as processes for reporting instances of fatigued driving. FRMSs take FRMPs one step further as they aim to manage operator fatigue at a more granular level. They include a continuous feedback loop that provides a means for continuous measurement and monitoring of an individual worker’s schedules using subjective and objective data collection (see Gander  for additional details, and Fourie et al. [2010a, 2010b] for discussions of the effectiveness of FRMPs and FRMSs in the trucking industry.)
Finally, the Federal Motor Carrier Safety Administration (FMCSA) and Transport Canada have worked over the past decade to develop the North America Fatigue Management Program (NAFMP), discussed in detail in Chapter 8. The purpose of this online program is to present effective ways to manage and mitigate fatigue in trucking operations. The information provided encompasses prescriptive fatigue-related regulations (i.e., HOS rules), fatigue awareness, the nature of sleep and sleep disorders, work-rest schedule development, and known fatigue countermeasures.
Conclusion 17: Fatigue risk management plans and fatigue risk management systems used in aviation, the rail industry, and the pipeline industry need to be studied further since they may provide models that can be applied to commercial motor vehicle driving.
This chapter has described various measures and technologies for dealing with CMV driver fatigue, all of which are aimed at the driver. Drivers play a major role in safe CMV operations, and if they are more aware of their degree of fatigue and how best to counter it, the risk of crashes should be reduced. Barriers to entering the CMV driver profession are somewhat low, since all one must do is obtain a commercial driver’s license (CDL) (see Chapter 2 for description of different licensing requirements) and pass the U.S. Department of Transportation’s (DOT) physical exam.3 Trucking companies do prefer a clean driving record when hiring drivers, and many of them hire from CDL training schools. Safety training is a significant part of the curriculum of these schools (e.g., how to drive on ice or snow or in the rain). However, the curriculum does not always include sufficient coverage of driver fatigue awareness. Therefore, one cannot be certain that a qualified truck driver is fully aware of the risks of driver fatigue and its consequences. Big trucking companies usually have orientation and on-the-job training programs that educate drivers on fatigue and how to manage it. Such companies are aiming to ensure that their employees are safe drivers, thereby maintaining their safety records and controlling one of the most obvious costs—from accidents.
Focusing on corporate safety records has its benefits, as safety and economic gains are linked—unsafe drivers are bad for business. Both economic gains and a company’s approach to safety are determined by the company’s organizational culture—values and norms held and shared by workers on the aims of the company and how the work should be done. FMCSA requested that the Transportation Research Board’s Commercial Truck and Bus Safety Synthesis Program (CTBSSP) broaden the understanding of “safety culture” and synthesize best practices and guidelines on the development of such a culture among motor carriers. CTBSSP Synthesis 14 (Short et al., 2007) highlights the importance of treating safety as the responsibility not only of drivers but also of dispatchers and fleet managers, as they are responsible for scheduling loads and are aware of how many hours drivers have been on duty. CTBSSP Synthesis 14 cites
3 DOT Medical Exam and Commercial Motor Vehicle Certification. See https://www.fmcsa.dot.gov/medical/driver-medical-requirements/dot-medical-exam-and-commercial-motor-vehicle-certification [March 2016].
examples of trucking companies undertaking initiatives on various fronts that mirror these best practices and guidelines.4 In essence, it is necessary to educate dispatchers, fleet managers, and safety managers about the fatigue-related challenges faced by drivers.5 Fatigue management programs, in combination with training for drivers and trucking officials, can help drivers make optimal use of their off-duty hours.
As discussed in Chapter 2, the trucking industry is highly heterogeneous in terms of operational structure; certain populations of drivers are not formally taught a wide array of safe driving practices, and driver fatigue management may or may not be part of that training. When independent owner-operators engage in sustained contract work for larger carriers, they occasionally are expected to engage in the same training received by the company’s employees. Generally, however, independent owner-operators lack ready access to the same depth of education and training available to drivers working for large carriers.6 However, they do have ready access to the NAFMP online. In the end, however, implementation of fatigue management practices depends mainly on personal initiative by the independent driver.
The above discussion relates to the concept of safety culture, which is achieved when shared values and beliefs interact with a carrier’s structures to produce behavioral norms. It is important to study how the different approaches to safety culture of various carriers relate to the decisions made by CMV drivers about whether to continue driving when they feel fatigued. A separate but related concept of “safety climate” is also worthy of study for its impact on driver behavior. Safety climate is defined as “shared perceptions of the organization’s policies, procedures and practices as they relate to the value and importance of safety within the organization. In short, safety climate is the measurable aspect of safety culture” (Huang et al., 2011).
Conclusion 18: Further research is needed on the impact of safety culture on driver decision making with respect to countering fatigued driving and on crash frequency.
5 There is a precedent for educating motor carriers and their drivers about fatigue—the Mastering Alertness and Managing Driver Fatigue Program run by FMCSA and ATRI, the research arm of the American Trucking Associations.
6 Some independent owner-operators who provide services to large trucking fleets that have fatigue management programs may have access to the fleet’s training and education programs.
This page intentionally left blank.