European NCAP Driver State Monitoring Protocols: Prevalence of Distraction in Naturalistic Driving

Abstract

Objective

examine the prevalence of driver distraction in naturalistic driving when implementing European New Car Assessment Program (Euro NCAP)-defined distraction behaviours.

Background

The 2023 introduction of Occupant Status monitoring (OSM) into Euro NCAP will accelerate uptake of Driver State Monitoring (DSM). Euro NCAP outlines distraction behaviours that DSM must detect to earn maximum safety points. Distraction behaviour prevalence and driver alerting and intervention frequency have yet to be examined in naturalistic driving.

Method

Twenty healthcare workers were provided with an instrumented vehicle for approximately two weeks. Data were continuously monitored with automotive grade DSM during daily work commutes, resulting in 168.8 hours of driver head, eye and gaze tracking.

Results

Single long distraction events were the most prevalent, with .89 events/hour. Implementing different thresholds for driving-related and driving-unrelated glance regions impacts alerting rates. Lizard glances (primarily gaze movement) occurred more frequently than owl glances (primarily head movement). Visual time-sharing events occurred at a rate of .21 events/hour.

Conclusion

Euro NCAP-described driver distraction occurs naturalistically. Lizard glances, requiring gaze tracking, occurred in high frequency relative to owl glances, which only require head tracking, indicating that less sophisticated DSM will miss a substantial amount of distraction events.

Application

This work informs OEMs, DSM manufacturers and regulators of the expected alerting rate of Euro NCAP defined distraction behaviours. Alerting rates will vary with protocol implementation, technology capability, and HMI strategies adopted by the OEMs, in turn impacting safety outcomes, user experience and acceptance of DSM technology.

Keywords

Euro NCAP distraction driver state monitoring visual time sharing naturalistic driving

BACKGROUND

Driver distraction is a key contributor to motor vehicle crashes worldwide, with distraction and inattention a contributing factor to 29%–48% of fatal and serious injuries (Fitzharris et al., 2020; Sundfør et al., 2019). Mobile phones and increasingly complex infotainment systems, like those that feature touch screens rather than traditional tactile buttons, are competing for drivers’ attention more than ever before. Increasingly smart vehicles with safety and convenience features such as lane assist may reduce drivers’ perceived need to engage with the forward roadway and consequently increase their likelihood of engaging in secondary behaviours, even when they are required to monitor vehicle performance. Attempts to manage driver distraction have typically been through regulation and policing (e.g. outlawing phone use while driving), as real time distraction management has not been technologically possible. Recent developments in driver state monitoring (DSM) now offer a technical solution for a very human problem, utilizing driver monitoring cameras to detect driver distraction (Kuo et al., 2018; Yang et al., 2018). The recent increases in technology capabilities allowing for robust and continuous real-time driver tracking in combination with developments in the understanding of driver distraction and its relationship to risk has accelerated the maturation of DSM technologies. Despite the maturity of DSM technologies, widespread uptake of the technology in the automotive industry has been relatively slow and limited to higher-end luxury vehicles (e.g. General Motors Cadillac’s Super Cruise system).

Strong regulation and consumer advocacy is a powerful prerequisite to drive the widespread implementation of new technologies, particularly in the entry level and mid-tier price range. The European New Car Assessment Programme (Euro NCAP) provides consumer information about the safety of most new cars in Europe. Based on current technology trends and maturity, the distraction behaviours identified as high risk in the Safety Assist protocols will be reliant on camera-based systems which assess driver state by directly monitoring the driver (Euro NCAP, 2022). Driver distraction is a core component of the Euro NCAP testing protocol, accounting for roughly half of the available points for DSM (Euro NCAP, 2022). In recognition of real-world behaviours and risk, distraction is classified in two broad categories within the Euro NCAP protocol; single glance long distraction and shorter, multi-glance distraction (Euro NCAP, 2022; Fredriksson et al., 2021). The requirements for DSM technologies are discussed along two dimensions: detection difficulty and behavioural complexity (Fredriksson et al., 2021). More sophisticated DSM solutions will be able to provide higher levels of system availability – the proportion of time a system can provide protection to the driver – and will be able to capture a greater proportion of complex real-world driver behaviour. The testing approach initially proposes testing using a dossier of evidence provided by the Original Equipment Manufacturer (OEM) as well as track testing. The assessment protocol encourages capable systems to not rely only on warning strategies but to also include intervention strategies when a driver is not attentive (Euro NCAP, 2022). Euro NCAP’s adoption of DSM will be a catalyst for widespread industry adoption of DSM technologies. The protocol describes a collection of distraction behaviours that are supported by the literature as relating to driver safety and will set the standard across the automotive industry for the minimal tracking requirements.

Long glances (>2 seconds) to off-road regions have an established relationship to driver risk, doubling crash risk (Klauer et al., 2006). This behaviour has been operationalised by Euro NCAP as a single long glance away from the road, lasting 3 seconds or longer (Euro NCAP, 2022). Long distraction is then further sub-categorised into glances to driving-related (e.g. instrument cluster, rear mirror) or nondriving-related (e.g. infotainment system, drivers lap) region. While research clearly demonstrates that long periods of time with eyes off the forward roadway is a risky behaviour, other evidence suggests that rear mirror glances longer than 2 seconds may actually be protective instead of increasing crash risk (Klauer et al., 2006).

Short multi-glance distraction, also referred to as visual attention time sharing (VATS), is another distraction behaviour that has been associated with driver risk. Typically, it involves a driver splitting time looking at the forward roadway, other driving-related regions and a secondary task. For example, when changing a song on the infotainment system drivers may look between the road and the infotainment system multiple times for short glances. VATS has been classified in multiple ways throughout the literature, with 30%–40% of time on the road centre (Percent Road Centre; PRC) a common metric used (Tivesten et al., 2019) and Klauer et al. (2006) classifying VATS as >2 seconds cumulative time off road over a 6-second window. Euro NCAP’s protocol defines VATS as a cumulative 10 seconds time looking away from the forward roadway within a 30-second window, where a driver does not return their gaze to the forward roadway for a minimum of 2 seconds (Euro NCAP, 2022). The proportion of time off road in this definition is roughly in alignment with the 30% PRC definition and the 2 seconds off-road in a 6 second window rule (Klauer et al., 2006; Tivesten et al., 2019).

Distraction is particularly risky where drivers are engaging in a cognitively demanding task, in particular mobile phone use. Mobile phone use is a high-risk activity and is over-represented in crash statistics (Dingus et al., 2016). Young et al. (2019) found that mobile phone use accounted for 23.2% of safety incidents, despite being only 7.4% of secondary task engagement. Given the high risk associated with phone use Euro NCAP has made a distinction between VATS where it involves glances towards a mobile phone (Euro NCAP, 2022). This is likely to incentivise the inclusion of phone use detection in DSM by OEMs, particularly given it is likely to be an unpopular inclusion with consumers.

A key delineator of technology capability will be the inclusion of head or gaze tracking. Drivers can engage different glance strategies when looking towards a target, including lizard (eye movement) and owl (head movement) glances (Fridman et al., 2016). Lizard glances describe glance behaviour which features little to no head movement and it is predominantly the driver’s eyes moving (Fridman et al., 2016). Owl glances describe behaviour where the driver’s head moves, followed by their eyes (Fridman et al., 2016). Lizard glances are typically to regions that are closer to the forward roadway, whereas owl behaviour tends to occur to regions at larger visual angles from the roadway (Fridman et al., 2016). DSM technologies may vary in terms of the methods used to track driver glance position, with some systems using head angle and others using eye gaze, and many systems using a combination of both. Tracking eye gaze is more technically difficult than tracking head position, and thus head tracking is used by many DSM technologies. However, DSM systems that only utilise head pose are unable to track lizard glances and will be less accurate at tracking mixed glances, where a combination of head and gaze movements are used (Fridman et al., 2016). This variation in detection capability will be captured by Euro NCAP through testing the extremes of both these glance strategies through distinguishing owl and lizard glance strategies within the testing protocol (Euro NCAP, 2022). In addition to distinguishing technologies based on tracking capabilities, the ability to detect lizard glances in particular may have implications for driver’s engaging in phone use (either hand-held in lap or behind steering wheel) which predominantly occurs using lizard glance strategies (Yang et al., 2021).

The implementation of Euro NCAP’s DSM protocol is imminent, however while these distraction behaviours are grounded in research, the expected prevalence of these behaviours in naturalistic driving is largely unknown. Euro NCAPs protocols are currently the most detailed guidelines surrounding DMS and will be a key driver in how these technologies are implemented. Thus it is critical to understand how the application of these guidelines will translate into naturalistic driving. This paper will investigate the prevalence of Euro NCAP-described distraction behaviours in naturalistic driving and explore the impact of different implementations of the distraction protocols on driver alerting rates.

METHODS

Participants

Shift workers from the Emergency Department and Intensive Care Units at a Melbourne metropolitan hospital were invited to take part in the study, with 7 doctors and 13 nurses consenting to participate. This research complied with Declaration of Helsinki the protocol was approved by the Monash University Human Research Ethics Committee (MUHREC) and the Austin Health Human Research Ethics Committee (HREC). All participants provided written informed consent. Participants were all regular shift workers and engaged in night shifts and day or evening shifts during their participation period, as part of their regular work duties. Participants were provided with the study vehicle at the beginning of their shift cycle and were instructed to use the vehicle for the period of the study. Participation periods varied depending on an individual’s shift schedule, with participation periods up to 2 weeks. Drive duration was an average of 36.12 minutes (SD = 27.32 minutes), however drives varied considerably in length from a few minutes (min = 3 minutes 30 seconds) to hours long (4 hours 54 minutes). Drives occurred across all times of day and throughout the night, with the majority of drives occurring between 06:00–09:00 and 18:00–20:00. This study took place across a range of road types varying from metropolitan and urban roads, highways, and winding mountain roads. There were 13 female and 7 male participants, all fully licensed with a mean of 14.1 years (SD = ±8.6) driving experience, mean age of 33.2 years (SD = ±9.0) and a mean drive duration of 40.4 minutes (SD = ±23.3). The analyses from the current paper were a secondary analysis of a study designed to monitor naturalistic driving in shift workers, with results from previous analyses of this dataset reported in Mulhall et al. (2020) and Kuo et al. (2018).

Data Collection

A 2015 Honda Jazz with an automatic transmission was instrumented with an automotive grade Driver Monitoring System (DMS) with a steering wheel mounted infrared camera (Seeing Machines Ltd, ACT, Australia). The DMS utilises proprietary algorithms to derive head position, facial feature detection, eyelid position and pupil and gaze tracking to monitor the drivers’ visual attention. The system monitors driver’s gaze direction and identifies gaze targets towards the forward roadway or towards other regions described in Figure 1. The DMS was monitoring drivers continuously throughout the study; however, drivers did not receive any alerts or information about their state.

Figure 1.

Gaze regions included in analysis, with driving-related regions in blue and nondriving-related regions in red. The remaining cabin regions were classified as off-road.

Data Analysis

There was a total of 168.6 hours of DMS data where gaze and head data were available to include in the analysis. Driver head and gaze data was used to determine driver fixations and fixation location. Consecutive fixations towards the same glance region were aggregated to understand glance duration towards the glance region. This was used to generate information on driver visual attention, and then the long and short distraction guidelines were applied according to the Euro NCAP protocol (Euro NCAP, 2022). Data was then filtered to include driving at speeds >10 km/h (6.2 mph) and glance regions were divided into driving-related and nondriving-related regions, with the included glance regions outlined in Figure 1.

Single Long Glance Away (LGA) events included glances ≥3 seconds to a single off-road glance region, with an increase to the threshold of 1 second where there is compelling evidence for implementation (Euro NCAP, 2022). The protocol specifies glance targets to both driving-related regions and nondriving regions. Research examining LGA events demonstrates increased risk primarily to driving unrelated regions, with driving-related LGA events even considered protective (Klauer et al., 2006). While ‘compelling evidence for implementation’ is open to interpretation, one likely application of a longer glance threshold is to differentiate driving-related and unrelated glances, given their different risk profiles. Thus, we examined driving-related glance regions using two different thresholds (3 vs. 4 seconds) to determine how each of these thresholds may affect the incidence of driving-related LGA events in real world driving. Another distinction made in the Euro NCAP protocol is driver glance strategy, owl glances primarily consisting of head movement and lizard glances primarily consisting of eye movement. LGA events were classified either as lizard, owl or mixed glance strategies to determine the proportion of each glance strategy in naturalistic driving, by glance region. Head and gaze position plots were created using the approach described in Yang et al. (2021), see Figure 2 for example plots. Each long distraction event plot was visually reviewed in combination with a review of the event video to determine whether the event primarily featured head movement (owl), eye movement (lizard) or a combination of both glance strategies where both head and gaze moved away from the road (mixed). A second researcher reviewed half of all events (50.67% events) and interrater reliability was calculated to be .88.

Figure 2.

Head and gaze fixation and yaw plots for a typical lizard glance (top) and a typical owl glance (bottom). In a lizard glance head pitch and yaw remain relatively stable while gaze moves, whilst in an owl glance head pitch and yaw change.

Short distraction events (Visual Attention Time Sharing; VATS) were determined using the recommended protocol for VATS events suggested by Euro NCAP, a cumulative amount of time of 10 seconds off-road within a 30-second time period, where the driver does not return their attention on road for ≥2 seconds. Glances off-road during VATS events were limited to <3 seconds, as this would be considered an LGA event. VATS events to a single glance location (time sharing between the road and one other location) include both driving-related and unrelated regions. Euro NCAP requirements do not specify different time requirements for driving-related versus unrelated regions for VATS events. Multi-glance VATS, or time sharing between the road and multiple other locations, are only required to include nondriving-related gaze regions. VATS events were classified based on the above criteria, with the addition of also classifying multi glance events with driving-related regions to determine the impact it has on implementation. VATS events were then reviewed to determine the driver’s glance strategy using the same method described for long distraction events. Determining whether the glance strategy was lizard or owl is less straightforward for multiple glances however, particularly when multiple glance regions are involved as different glance regions lend themselves to different glance strategies due to the variation in distances and angles of the glance regions (Fridman et al., 2016). As such, the majority of VATS events were found to include a combination of glance strategies.

Phone use is a secondary category within VATS that has the same timing requirements (a cumulative 10 seconds off road within 30 seconds) but where the driver’s gaze is located towards a mobile phone. Camera placement within the current study limited the visibility of some potential phone-use locations as the steering mounted camera limits frame view to the driver’s face and shoulders. A driver’s lap is the primary location for driver’s to hold their phone when using a handheld phone while driving (Faulks, 2020; Oviedo-Trespalacios et al., 2020; Roady et al., 2020) and is used as a proxy for phone use in this study. The driver’s lap glance region includes approximately half of the NCAP basic phone locations, as phones held in driver lap or on either driver knee would be captured with this glance region. The remaining basic phone-use regions (dashboard mounted phone, phone held at top of wheel and phone at centre wheel) cannot be examined in the current dataset, and the advanced phone use category is also not examined (phone in front of windscreen, phone held at instrument cluster). Phone events were then visually reviewed for visible phone use (where phones were directly visible to the camera, given camera placement limitations) and to examine driver glance strategy. All VATS events that included at least one glance to the driver’s lap was considered a phone use case in the current study, as it is not specified whether phone use VATS events should include other glance locations.

RESULTS

Single Long Glance Distraction

Fourteen participants engaged in LGA behaviours (average of 10.7 alerts per person), with a total of 150 LGA events that met the Euro NCAP threshold of >3 seconds in duration, resulting in drivers engaging in LGA behaviour every 1.1 hours (.89 events/hour). When considering alerts at an individual level, drivers had an event range of .07–4.55 events/hour. Glances to nondriving-related regions accounted for 57.3% of all LGA glances (n = 86), with glances towards the console the most frequent (Figure 3). Glances towards a driver’s lap, a common location for drivers to hold a phone, accounted for 8.6% of all LGA glances (15.1% of driving unrelated glances).

Figure 3.

Number of long distraction events to driving unrelated regions (red) and driving-related regions with a >3-second threshold in translucent blue and >4 seconds in solid blue. Circle diameter is proportional to number of events. Off road other event location includes all glances not to a defined off road gaze region.

Implementing a 3-second threshold for long distraction to driving-related regions would result in an LGA event every 2.6 hours (.38 alerts/hour), accounting for 42.7% of all LGA events (n = 64). Increasing the event threshold to 4 seconds would reduce the frequency of driving-related LGA events to an event every 8 hours (.12 alerts/hour, n = 21). The majority of driving-related glances were to the instrument cluster and the rear mirror, regardless of the threshold.

Current NCAP protocols suggest that points will be assigned separately for both owl and lizard glances for both driving-related and driving unrelated glance regions (Euro NCAP, 2022). LGA events were assessed for whether they were lizard (primarily gaze movement) or owl (primarily head movement) glances or a combination of both gaze and head movement. The majority of driving unrelated long distraction events were lizard glances (65.1%, n = 56), followed by mixed glances (19.8%, n = 17) and owl glances (15%, n = 13). The contribution of lizard versus owl glances varied by gaze region, with some regions like the console and off-road-right split between lizard, owl and mixed glances, and other regions such as the driver’s lap entirely represented by lizard glances (Figure 4). Driving-related long distraction demonstrated a similar pattern for a 3-second threshold (lizard n = 39, owl n = 14, mixed n = 11) and a 4-second threshold (lizard n = 13, owl n = 4, mixed n = 4), with instrument cluster and rear mirror glances having a combination of glance strategies (Figure 4).

Figure 4.

Percentage of long glance away distraction events by glance strategy. Off road other event location includes all glances not to a defined off road gaze region.

Visual Attention Time Sharing

A total of 35 short distraction VATS events were detected within the dataset (13 participants, average 2.69 events each), which equates to a VATS event every 4.8 hours (.21 events per hour). Event rates at an individual level ranged from .07 events/hour to .69 events/hour. Twelve events had visual time sharing that occurred between the road and only a single glance region. Of these, eleven were to driving unrelated regions, including the centre console (9 events), driver side window (1 event) and the driver lap (1 event), with the remaining event to the instrument cluster (a driving-related region). The remaining 23 events each included multiple glance regions (i.e., more than one off-road region). Multi-glance VATS events are not required to include driving-related glance regions in the 10 second in 30 second threshold, although many VATS events also included glances to these regions, resulting in the total proportion of time off-road being longer than 10 seconds for many VATS events (Figure 5). Differentiating driving related and nondriving regions requires high precision gaze tracking, which is a technically difficult task, particularly where glance regions are small or in close proximity such as the side windows and mirrors. As a result, some systems will need to implement VATS algorithms that include driving-related regions in their VATS alerts to achieve maximum points for VATS, which would increase the number of multi-glance VATS events from 23 (an alert every 7.3 hours) to 68 (an alert every 2.5 hours).

Figure 5.

Examples of glance behaviour during VTS events for (a) VTS events to a single glance location, (b) VTS events to multiple glance locations and (c) VTS events defined as phone use. Vertical line marks 10 second VTS threshold.

Separate criteria exist within the Euro NCAP protocol for visual time sharing where the driver’s gaze is repeated towards their mobile phone (Euro NCAP, 2022). Using glances to a driver’s lap as a proxy, due to camera placement in the current study not allowing for visibility of most phone use locations, 9 VATS events could be classified as driver phone use based on the NCAP protocol (visible phone use was able to be confirmed in 2 events, as the driver briefly moved the phone from their lap and into view of the camera during VATS events).

Overall Distraction Events

Overall, the distraction alerting rates tested within this protocol could result in a range of alerting rates depending on system capabilities and chosen implementation; lizard versus owl capabilities; 3- or 4-second thresholds for driving unrelated long glances; and whether the system can distinguish driving-related from unrelated regions in VATS events. Distraction rates range from an alert every 44 minutes (1.36 alerts/h, n = 230) to an alert every 1.2 hours (.84 alerts/h, n = 142).

DISCUSSION

Long distraction events have the clearest relationship to risk relative to other distraction events within the NCAP distraction protocol, with glances to driving unrelated regions longer than 2 seconds doubling crash risk (Klauer et al., 2006). Despite the risk of LGA events they are a common driving behaviour, with a driving unrelated long distraction event occurring every 2 hours in the current dataset.

Long glances to driving-related regions has not been a core focus of research, thus the relationship with driver risk is less established and may even be protective, with long glances of >2 seconds towards rear view mirrors having a crash risk odds ratio of .45 (Klauer et al., 2006). The current findings highlight that implementing thresholds of 3 or 4 seconds may have a dramatic difference in terms of the number of alerts drivers will receive for driving-related LGAs. In order to implement a different time threshold for driving-related and nondriving-related long distraction though, a system must first be capable of distinguishing between these regions, requiring a higher degree of accuracy in driver tracking. This is particularly difficult as many driving-related regions are in direct proximity to driving unrelated regions (e.g. driver mirror and driver side window). Technologies incapable of distinguishing glance regions will effectively need to opt for a 3-second threshold out of necessity. This will enable those systems to achieve the same number of NCAP points as more sophisticated systems that can distinguish between driving-related and unrelated regions; however, this will very likely come at the expense of the user experience, with drivers likely to receive a driving-related alert every 2.6 hours, as opposed to every 8 hours with a 4-second threshold. Furthermore, driving-related LGA events may be perceived as unwarranted by the user, negatively impacting user acceptance and trust of DSM (Reagan et al., 2018; Reinmueller et al., 2018; Reinmueller & Steinhauser, 2019).

A key differentiator in the distraction behaviours outlined by Euro NCAP’s protocol concerns glance strategy, with systems needing to be capable of detecting both lizard and owl glances to achieve maximum scores for distraction (Euro NCAP, 2022). DSM systems that can only detect owl behaviours however will likely miss a large proportion of glances as 65% of all driving unrelated glances in the current study were lizard glances and only 15% classified as owl glances (with the remaining being mixed); however, as no statistical comparisons were made it is unclear whether this is a statistically significant difference. Mixed and owl glances were more common at extreme glance angles where it is more typical – often out of necessity – to have at least some head movement to view the glance region (Fridman et al., 2016). Fridman et al. (2016) demonstrated individual preference for glance strategies impacts lizard-owl glances, with lizard types and mixed types benefitting from the inclusion of eye tracking far more than owl types, indicating that DSM systems that use head pose only may perform well for owl types but be less accurate at monitoring mixed and lizard types. System capability to monitor lizard versus owl glances may also have real safety impacts, with high-risk behaviours such as phone use favouring lizard glance strategies (Yang et al., 2021). These findings are supported by our current results, with all LGA events to the driver’s lap, the most common location for handheld phone use, utilising lizard glances (Faulks, 2020; Oviedo-Trespalacios et al., 2020; Roady et al., 2020). While lizard-owl glances represent the extremes of glance behaviour strategies (primarily eyes vs. primarily head movement), 20% of driving unrelated glances were mixed glances, containing both head movement and gaze movement. It was observed in the current dataset that when drivers were making glances to gaze regions at large angles they would often employ a gaze strategy where they would make the minimum head movement required to accommodate their gaze to meet the glance target. It is unclear whether owl only DSM systems will be able to capture these behaviours and accurately determine gaze location on glances with minimal head movement, or if eye gaze is required to capture this behaviour.

VATS to a single glance region was relatively rare, particularly to driving-related regions. Given the relationship between short distraction to driving-related glance regions is still an evolving area of research and the links to driver risk are not yet established the low prevalence of this behaviour indicates that few of these events should occur, reducing the chances of these behaviours strongly influencing user experience. In contrast to single glance region VATS, multi glance events are only required to include glances to nondriving task locations. DSM systems that are unable to distinguish between driving-related and unrelated glance regions may still be able to achieve full points for VATS distraction events by including driving-related regions in their implementation, however it will come at a significant cost to user experience, as including driving-related regions in multi glance VATS almost triples the alerting rate.

The placement of the DSM camera in the current study prevented the examination of all of the phone use regions proposed by Euro NCAP; however, the driver’s lap (one of 9 nominated phone use regions in the Euro NCAP protocol) was able to be used as a proxy for phone use regions. The driver’s lap is one of the most common positions for handheld phone use (Oviedo-Trespalacios et al., 2020; Roady et al., 2020). Nine potential phone use cases were detected, with drivers glancing between their lap and the road and phones visible in 2 cases, demonstrating that the Euro NCAP multi glance protocol is a valid approach for detecting at least some phone use while driving. Even where VATS events to the driver’s lap are not false positives for phone use, it is a high-risk glance behaviour regardless (Liang et al., 2014). While the expected prevalence and alerting rates for Euro NCAP phone use cases cannot be estimated based on the current data, we can expect that at least some phone use cases can be detected with this methodology. DSM technologies that are only capable of single glance region VATS events will be unlikely to detect many phone use cases, given only a single phone use case was to a single glance location. All of the remaining cases had glances to both other nondriving regions and driving-related regions. One explanation for the low incidence of single glance region phone use is that drivers may be compensating for a perceived risky behaviour through increasing their scanning behaviour.

There are a number of distraction behaviours specified within the Euro NCAP protocol that are not tested within the current study. Body lean behaviours for long distraction were unable to be tested as the camera position within the current study reduced the availability of body lean data. Thus, it remains unclear how prevalent this behaviour is in the wild. Camera position was also a limiting factor in the measurement of phone use, as the steering wheel mount limits the visibility within the cabin. This meant it was not possible to detect phone use for any occasions where the phone was held at the dashboard region, the centre of the steering wheel, and other specified locations within the protocol. Noise factors within the Euro NCAP protocol cover an extensive range of behavioural, environment and driver appearance variables that were not examined within the current study. The ability of OSM systems to cope with noise factors will likely impact a systems capability to detect distraction behaviours in their presence; for example, systems that cannot track driver gaze while sunglasses are worn will not be able to detect lizard glances in the presence of this noise factor. The impact of noise factors will likely be a key differentiator between DSM systems and have a significant impact on the user experience. This study was designed to capture naturalistic driving from shift workers engaging in night shifts, who are more likely to experience drowsiness. Drowsiness may lead to an increase in distractibility of drivers (Anderson & Horne, 2013; Kuo et al., 2018) which may limit the generalisability of findings.

The inclusion of DSM in Euro NCAP’s protocols will incentivise many OEMs to consider the uptake of OMS technology and provide new opportunities for future research as the technology becomes more available. A critical avenue for future research will be examining how DSM of drowsiness and distraction impact driver crash risk and other key safety metrics. The ability of DSM to modify driver behaviour must be a core area of research, as the recency of this technology has allowed little opportunity to collect empirical data on changes to driver behaviour following DSM implementation (Domnez et al., 2007; Fitzharris et al., 2017). Fitzharris et al. (2017) found real-time fatigue alerts reduced fatigue events by 66% in naturalistic driving and Donmez et al. (2007) demonstrated altered driver scanning behaviour following distraction alerts so while it is likely that DSM will alter driver distraction behaviour the safety benefits of the technology must be examined. This is critical to ensure that driver behaviour is being modified in a way that enhances safety, particularly given the DSM implementation and human machine interface (HMI) may vary considerably between vehicles. Where the relationship of distraction behaviour and driver risk is less established, such as long glances to driving-related glance regions, the impact of DSM on driver behaviour must be closely monitored as it could negatively impact both driver perception and acceptance of DSM technology (Reagan et al., 2018; Reinmueller et al., 2018; Reinmueller & Steinhauser, 2019). Furthermore, it could discourage drivers from engaging in some scanning behaviours that may enhance driver safety in an effort to avoid distraction alerts (Klauer et al., 2006).

Euro NCAP defined distraction behaviours are common in naturalistic driving and the findings support rewarding technologies that can track eye gaze, instead of head pose alone. This high prevalence of lizard glances suggests that less sophisticated technologies that utilise head pose only tracking will miss a substantial number of distraction events, particularly to gaze regions commonly associated with phone use. As these protocols are implemented across Europe, emerging evidence should be used to consider the safety implications for driving-related distractions with future iterations of the protocols revised to reflect emerging evidence of real-world safety outcomes.

Supplemental Material

Supplemental Material - European NCAP Driver State Monitoring Protocols: Prevalence of Distraction in Naturalistic Driving

Supplemental Material for European NCAP Driver State Monitoring Protocols: Prevalence of Distraction in Naturalistic Driving by Megan Mulhall, Kyle Wilson, Shiyan Yang, Jonny Kuo, Tracey Sletten, Clare Anderson, Mark E. Howard, Shantha Rajaratnam, Michelle Magee, Allison Collins and Michael G Lenné in Human Factors.

Footnotes

Acknowledgments

This research is supported by the CRC for Alertness, Safety and Productivity, Melbourne, Australia and Seeing Machines Ltd, Fyshwick, Australia. We thank Kaitlyn Crocker, Niamh McDonald and Dr Grace Vincent from the Sleep and Circadian Medicine Laboratory, School of Psychological Sciences and Turner Institute for Brain and Mental Health, Monash University, Melbourne, Australia for their assistance with data collection. We thank Dr Graeme Hart, Joanne Clarke, Helen Young, Simon Judkins, Melanie Stark and the staff of the Intensive Care Unit and Emergency Department at Austin Health, Melbourne, Australia for their support during data collection. We also thank the engineering and technical staff from Seeing Machines Ltd, Canberra, Australia.

KEY POINTS

Euro NCAP defined distraction behaviours occur in naturalistic driving, with long distraction events occurring once every 1.1 hours and short distraction occurring every 4.8 hours.

Drivers typically utilise lizard glance strategies (eye movement only) when engaging in long distraction (65.1% of long distraction events).

The implementation of Euro NCAP's protocol will result in varying numbers of alerts, depending on tracking capability (lizard and owl) and the ability to distinguish driving-related and nondriving-related glance regions.

Disclosure Statements

The primary authors who have driven this work are employed by Seeing Machines Ltd, a company that designs and sells driver monitoring systems. Furthermore, Dr Kuo and Dr Wilson are also shareholders in the company. As Seeing Machines Ltd sells driver monitoring systems, the company stands to gain business through the introduction of DMS into regulation such as Euro NCAPs. However, given this regulation is already introduced and the protocols have already been published the company will not gain further business through the findings of this paper. While the results of this study utilizes Seeing Machines tracking algorithms, the glance behaviours are not calculated using Seeing Machines proprietary algorithms, and thus results are not a commentary on the performance of Seeing Machines algorithms against Euro NCAPs standards but are more general in nature. Given this potential conflict of interest we have attempted to limit the results of this paper to their naturalistic prevalence rates. Our position as a DMS technology company uniquely qualifies the authors to discuss the impacts of the regulation and understand the intricacies around this topic from a technology perspective. It also places us currently as some of the few researchers with the data required to answer this question, and we believe this paper answers some critical questions that are important to distribute to the wider scientific community.

Dr Mulhall is employed as a research scientist by Seeing Machines Ltd, one of the funders of this research. Dr Wilson, Dr Yang and Dr Kuo are employed as senior research scientists, and Dr Wilson and Dr Kuo hold shares in Seeing Machines Ltd, one of the funders of this research. Dr Sletten served as a Project Leader in the CRC for Alertness, Safety and Productivity which was the primary funder of this work. Dr Anderson has received a research award/prize from Sanofi-Aventis; contract research support from VicRoads, Rio Tinto Coal Australia, National Transport Commission, Tontine/Pacific Brands, and an ARC Linkage with Seeing Machines Ltd; lecturing fees from Brown Medical School/Rhode Island Hospital, Ausmed, Healthmed and TEVA Pharmaceuticals; and reimbursements for conference travel expenses from Philips Healthcare. In addition, she has served as a consultant through her institution to the Rail, Bus and Tram Union, the Transport Accident Commission (TAC) and the National Transportation Committee (NTC) and VicRoads. She has also served as an expert witness and/or consultant in relation to fatigue and drowsy driving. Dr Anderson is a theme leader in the CRC for Alertness, Safety and Productivity which was the primary funder of this work. Dr Howard served as a theme leader in the CRC for Alertness, Safety and Productivity Melbourne, Australia, which was the primary funder of this work; and has received grants from ResMed Foundation, Prevention Express and TEVA, and equipment support from Optalert, Seeing Machines Ltd., and Philips Respironics for Research which are not related to the work reported in this paper. Dr Rajaratnam reports grants from Vanda Pharmaceuticals, Philips Respironics, Cephalon, Rio Tinto and Shell, and has received equipment support and consultancy fees through his institution from Optalert, Tyco Healthcare, Compumedics, Mental Health Professionals Network, and TEVA Pharmaceuticals, which are not related to this paper. Dr Rajaratnam is also a program leader for the CRC for Alertness, Safety and Productivity, Melbourne, Australia, which was the primary funder of this work. Dr Magee served as a project leader in the CRC for Alertness, Safety and Productivity. Ms Collins has no conflicts to declare. Dr Lenné is employed as chief science and innovation officer and holds shares in Seeing Machines Ltd, one of the funders of this research.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Cooperative Research Centre for Alertness, Safety and Productivity.

ORCID iDs

Megan Mulhall

Shiyan Yang

Mark E. Howard

Supplemental Material

Supplemental material for this article is available online.

References

Anderson

Horne

J. A.

(2013). Driving drowsy also worsens driver distraction. Sleep Medicine, 14(5), 466–468. https://doi.org/10.1016/j.sleep.2012.11.014

Dingus

T. A.

Guo

Lee

Antin

J. F.

Perez

Buchanan-King

Hankey

(2016). Driver crash risk factors and prevalence evaluation using naturalistic driving data. Proceedings of the National Academy of Sciences, 201513271. https://doi.org/10.1073/pnas.1513271113

Donmez

Boyle

L. N.

Lee

J. D.

(2007). Safety implications of providing real-time feedback to distracted drivers. Accident Analysis & Prevention, 39(3), 581–590. https://doi.org/10.1016/j.aap.2006.10.003

Euro NCAP 2025 Roadmap- In pursuit of Vision Zero

. (2022). European new car assessment programme (Euro NCAP). https://cdn.euroncap.com/media/70315/euro-ncap-assessment-protocol-sa-safe-driving-v101.pdf

Faulks

I. J.

(2020). Caught red-handed: Automatic cameras will spot mobile-using motorists, but at what cost? The Conversation. https://theconversation.com/caught-red-handed-automatic-cameras-will-spot-mobile-using-motorists-but-at-what-cost-125638

Fitzharris

Lenné

M. G.

Corben

Arundell

T. P.

Peiris

Liu

Stephens

Fitzgerald

Judson

Bowman

Gabler

Morris

Tingvall

(2020). ECIS report 1. Overview and analysis of crash type, injury outcomes and contributing factors. (No. 343; p. 232). Monash University Accident Research Centre.

Fitzharris

Liu

Stephens

A. N.

Lenné

M. G.

(2017). The relative importance of real-time in-cab and external feedback in managing fatigue in real-world commercial transport operations. Traffic Injury Prevention, 18(sup1), S71–S78. https://doi.org/10.1080/15389588.2017.1306855

Fredriksson

Lenné

M. G.

van Montfort

Grover

(2021). European NCAP program developments to address driver distraction, drowsiness and sudden sickness. Frontiers in Neuroergonomics, 2, 786674. https://doi.org/10.3389/fnrgo.2021.786674

Fridman

J. L.

Reimer

Victor

(2016). ‘Owl’ and ‘Lizard’: Patterns of head pose and eye pose in driver gaze classification. IET Computer Vision, 10(4), 308–313. https://doi.org/10.1049/iet-cvi.2015.0296

10.

Klauer

S. G.

Dingus

T. A.

Neale

V. L.

Sudweeks

J. D.

Ramsey

D. J.

(2006). The impact of driver inattention on near-crash/crash risk: An analysis using the 100-car naturalistic driving study data (Technical Report DOT-HS-810594). National Highway Traffic Safety Administration. https://www.nhtsa.gov/DOT/NHTSA/NRD/Multimedia/PDFs/CrashAvoidance/2006/DriverInattention.pdf

11.

Kuo

Lenné

M. G.

Mulhall

Sletten

Anderson

Howard

Rajaratnam

Magee

Collins

(2018). Continuous monitoring of visual distraction and drowsiness in shift-workers during naturalistic driving. Safety Science. https://doi.org/10.1016/j.ssci.2018.11.007

12.

Liang

Lee

J. D.

Horrey

W. J.

(2014). A looming crisis: The distribution of off-road glance duration in moments leading up to crashes/near-crashes in naturalistic driving. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 58(1), 2102–2106. https://doi.org/10.1177/1541931214581442

13.

Mulhall

M. D.

Cori

Sletten

T. L.

Kuo

Lenné

M. G.

Magee

Spina

M.-A.

Collins

Anderson

Rajaratnam

S. M. W.

Howard

M. E.

(2020). A pre-drive ocular assessment predicts alertness and driving impairment: A naturalistic driving study in shift workers. Accident Analysis & Prevention, 135, 105386. https://doi.org/10.1016/j.aap.2019.105386

14.

Oviedo-Trespalacios

Briant

Kaye

S.-A.

King

(2020). Assessing driver acceptance of technology that reduces mobile phone use while driving: The case of mobile phone applications. Accident Analysis & Prevention, 135, 105348. https://doi.org/10.1016/j.aap.2019.105348

15.

Reagan

I. J.

Cicchino

J. B.

Kerfoot

L. B.

Weast

R. A.

(2018). Crash avoidance and driver assistance technologies – Are they used? Transportation Research Part F: Traffic Psychology and Behaviour, 52, 176–190. https://doi.org/10.1016/j.trf.2017.11.015

16.

Reinmueller

Koehler

Steinhauser

(2018). Adaptive warning signals adjusted to driver passenger conversation: Impact of system awareness on behavioral adaptations. Transportation Research Part F: Traffic Psychology and Behaviour, 58, 242–252. https://doi.org/10.1016/j.trf.2018.06.013

17.

Reinmueller

Steinhauser

(2019). Adaptive forward collision warnings: The impact of imperfect technology on behavioral adaptation, warning effectiveness and acceptance. Accident Analysis & Prevention, 128, 217–229. https://doi.org/10.1016/j.aap.2019.04.012

18.

Roady

Wilson

K. M.

Kuo

Lenne

M. G.

(2020). How do drivers hold their phone? Age, prevalence, & handedness. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 64. https://doi.org/10.1177/1071181320641298

19.

Sundfør

H. B.

Sagberg

Høye

(2019). Inattention and distraction in fatal road crashes – Results from in-depth crash investigations in Norway. Accident Analysis & Prevention, 125, 152–157. https://doi.org/10.1016/j.aap.2019.02.004

20.

Tivesten

Victor

T. W.

Gustavsson

Johansson

Ljung Aust

(2019). Out-of-the-loop crash prediction: The automation expectation mismatch (AEM) algorithm. IET Intelligent Transport Systems, 13(8), 1231–1240. https://doi.org/10.1049/iet-its.2018.5555

21.

Yang

Kuo

Lenné

M. G.

(2018). Analysis of gaze behavior to measure cognitive distraction in real-world driving. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 62(1), 1944–1948. https://doi.org/10.1177/1541931218621441

22.

Yang

Shiferaw

Roady

Kuo

Lenné

M. G.

(2021). Drivers glance like lizards during cell phone distraction in assisted driving. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 65(1), 1410–1414. https://doi.org/10.1177/1071181321651147

23.

Young

K. L.

Osborne

Koppel

Charlton

Grzebieta

Williamson

Haworth

Senserrick

(2019). What are Australian drivers doing behind the wheel?: An overview of secondary task data from the Australian naturalistic driving study. Journal of the Australasian College of Road Safety, 30(1), 27–33. DOI:10.1145/1620509.1620531. https://search.informit.org/doi/abs/10.3316/INFORMIT.273295494260606

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.30 MB