- Email: [email protected]
The impact of tunnel design and lighting on the performance of attentive and visually distracted drivers
Accident Analysis and Prevention 47 (2012) 153–161
Contents lists available at SciVerse ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
The impact of tunnel design and lighting on the performance of attentive and visually distracted drivers Katja Kircher ∗ , Christer Ahlstrom Swedish National Road and Transport Research Institute (VTI), S-581 95 Linköping, Sweden
a r t i c l e
i n f o
Article history: Received 28 October 2011 Received in revised form 13 January 2012 Accepted 15 January 2012 Keywords: Tunnel Driving performance Distraction Illumination Design Visual load Secondary task
a b s t r a c t The crash risk in tunnels is lower than on the open road network, but the consequences of a crash are often severe. Proper tunnel design is one measure to reduce the likelihood of crashes, and the objective of this work is to investigate how driving performance is influenced by design factors, and whether there is an interaction with secondary task load. Twenty-eight drivers participated in the simulator study. A full factorial within subject design was used to investigate the tunnel wall colour (dark or light-coloured walls), illumination (three different levels) and task load (with or without a visual secondary task). The results show that tunnel design and illumination have some influence on the drivers’ behaviour, but visual attention given to the driving task is the most crucial factor, giving rise to significant changes in both driving behaviour and visual behaviour. The results also indicate that light-coloured tunnel walls are more important than strong illumination to keep the drivers’ visual attention focused forward. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction The crash risk in tunnels is lower than on the open road network, but the consequences of a crash are often severe (Leitner, 2001; Kirkland, 2002; Carvel and Marlair, 2005). It is therefore of high importance to ensure high safety standards in tunnels. This is achieved partly by reducing the probability of crashes and partly by reducing the consequences of crashes or fires. The former involves tunnel design, traffic regulations, appropriate facilities (ventilation system, lighting and interior) and maintenance, whereas the latter involves proper emergency facilities and fire-resistant structures (Mashimo, 2002). In this study we focus on crash prevention and investigate in which way different levels of illumination and brightness of the tunnel walls influence the behaviour of attentive and visually distracted drivers. Tunnel lighting regulations are complex with different lighting levels in different zones of the tunnel and with different requirements in different countries. The requirements on lighting also depend on the traffic intensity, the speed limit and the outdoor conditions (day/night). Most countries use CIE Report 088:2004 (CIE, 2004) as a base. Common for all regulations is that there is a requirement on the luminance of the road surface, but not for the walls or the roof. For example in Norway and Sweden the regula-
∗ Corresponding author. Tel.: +46 13 204118; fax: +46 13 14 1436. E-mail addresses: [email protected] (K. Kircher), [email protected] (C. Ahlstrom). 0001-4575/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.aap.2012.01.019
tions recommend the use of white walls up to a certain height, but without mentioning any figure. The illumination of a tunnel can substantially change the impression on the driver, but, to the best of our knowledge, there is no previous research that investigates how tunnel illumination affects driver behaviour. In fact, very limited research has been devoted to the analysis of driver behaviour in tunnels at all. The possibility to use the tunnel walls to convey information to the drivers was suggested by Carmody (1997), who proposed vertical stripes for speed information and horizontal stripes for gradient information. Manser and Hancock (2007) investigated in a simulator how speed can be influenced with the help of patterns and different textures on the roadside. They found that the presence of texture had an attenuating effect on speed in general. Patterns with decreasing width led to gradually decreased speed, while the opposite was true for patterns with increasing width. Most research efforts have been allocated to accident analysis. Both in Norway (Amundsen, 1994; Amundsen and Engelbrektsen, 2009) and in Austria (Nussbaumer, 2007) it has been found that slightly fewer crashes occur in tunnels compared to normal roads, but the fatality rate is substantially higher. Generally the crash risk is higher in the entrance zone of the tunnel than further into the tunnel (Amundsen, 1994; Amundsen and Engelbrektsen, 2009), and the probability of being injured or killed is 19% higher in tunnels with bi-directional traffic compared to tunnels with unidirectional traffic (Robatsch and Nussbaumer, 2004). Nussbaumer (2007) also mentions that according to police reports, the main reasons for tunnel crashes are lacking vigilance (e.g. fatigue,
154
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
distraction, inattention) and behavioural aspects (e.g. safety distance, lane keeping, overtaking). Both constant high levels and constant low levels of activation can be detrimental to driving; performance breaks down when the workload becomes too high (Young et al., 2003; Horberry et al., 2006; Mccartt et al., 2006) whereas longer periods of monotonous driving can lead to boredom and fatigue (Thiffault and Bergeron, 2003; Ting et al., 2008). Depending on the traffic intensity as well as the length of the tunnel and the tunnel design, the resulting level of driver activation can be anywhere between high and low. As long tunnels become more common, the issue of monotony and a resulting propensity of the driver to take his eyes off the road when looking for activation is a concern. When planning the 24.5 km long Laerdal Tunnel in Norway, simulator studies were conducted to investigate how to combat monotony and distraction (Jenssen, 1999; Kvaale and Lotsberg, 2001; Flø and Jenssen, 2007), finding that different types of lightings and constructions influence both driving behaviour and gaze behaviour. While these studies focussed on different colours and lighting designs, it may also be the case that brightness levels influence behaviour. In this study the precondition was to examine drivers both during an imposed visual distraction and while not visually distracted, to determine possible interactions of visual distraction with tunnel brightness. In simulator studies it is common to induce visual distraction by presenting the driver with an additional task that requires the driver to take his eyes off the road (e.g. Zhang et al., 2006; Donmez et al., 2007). In the current study, the driver is considered visually distracted, when the driver interacts with a standardised visuo-manual divided attention (Merat et al., 2005). This task includes target identification by visual search, is system paced, and was developed specifically to represent interaction with an object requiring high visual engagement. The aim of this study was to investigate the effects of illumination in tunnels on attentive and visually distracted drivers. Three levels of illumination were investigated in combination with lightcoloured versus dark tunnel walls and attentive versus visually distracted drivers. To ensure a high level of control over the situation and to enable strict reproducibility, a simulator study was conducted. The specific research questions were: Does the illumination level in a tunnel influence driver behaviour? Does the brightness of the wall in a tunnel influence driver behaviour? Is there an interaction with distraction, in the sense that brightness might get crucial only when drivers are loaded with a visual task?
2.2. Design A full factorial 2 × 3 × 2 (tunnel design × light intensity × task load) within subject design was used. The tunnel could either have dark or light-coloured walls in combination with three different levels of illumination. These six combinations of tunnel designs were driven twice, once with a secondary task present and once without. This means that each participant drove through twelve tunnels in total. The order of the tunnel sections was randomised with certain restrictions. The same lighting levels were only allowed to occur twice in consecutive order, the same wall colour was not allowed to occur more than three times in consecutive order, and the same secondary task condition was not allowed to occur more than three times in consecutive order. 2.3. Driving simulator The experiment was conducted in the VTI Driving Simulator III, an advanced moving base simulator. The moving base generates forces with three different systems: a large linear motion to simulate lateral motion, a tilt motion to simulate long term accelerations such as driving in a curve or longitudinal acceleration and deceleration, and a vibration table to simulate road roughness. The visual system consists of 3 digital projectors providing a 120◦ forward field of view and 3 liquid crystal displays for the rear view mirrors. For gaze tracking the eye tracker SmartEye Pro 5.6 with a set of three cameras was used. One camera was installed next to the A pillar on the driver’s side, one in front of the driver on the dash board, and one further to the right on the dash board, above the centre console. A predecessor of this driving simulator has been validated for tunnel scenarios in terms of speed and lateral position (Törnros, 1998). The speed profile was similar in the simulator and in a field setting except for an offset with consistently higher speeds in the simulator. The same effect has been observed for validation studies of open road driving (Harms, 1994; Alm, 1995; Törnros et al., 1997). Regarding lateral position, drivers tended to position the car in the lane such that they kept some distance to the nearest wall. This effect was greater when the wall was located to the left of the driver, that is, closer to the driver, as left-hand steering was used. The effect could be observed both in the simulator and in the real tunnel, with a slightly greater effect in the latter (Törnros, 1998). 2.4. Driving scenario and environment
2. Materials and methods 2.1. Participants Twenty-eight drivers participated in the study, 10 women and 18 men. The mean age was 41.3 years, with a standard deviation of 7.6 years. Their driving experience ranged from below 5000 km yearly (3 participants) over 5000–10,000 km yearly (4 participants) and 10,000–20,000 km yearly (12 participants) to above 20,000 km yearly (9 participants), that is, 75% of the participants drove a relatively high yearly mileage. Most participants did not have much experience with driving in tunnels. Of the 28 drivers 20 stated that they drove through tunnels a few times per year at most. Six participants drove through tunnels at least once a month, one reported to drive through tunnels at least once a week, and the remaining participant used tunnels on a daily basis. None of the participants indicated more than a slight anxiety while driving in tunnels – 18 stated that they were not afraid at all, while 10 participants said that they felt a slight anxiety, but used tunnels anyway.
The trial started with a training section, which consisted of a 6 km long segment of open motorway without tunnel. After the training section the experimental trial began. Here the road consisted of 2 km long sections of motorway without tunnel and 4 km long sections of motorway in a tunnel. These two section types alternated, starting and ending with an open motorway section. The road had two lanes in each direction, and road shoulders. The road was 10 m wide altogether, the lane width was 3.5 m, and the shoulders were 1.5 m wide each. The open motorway was completely straight, while the first 400 m of the tunnel section consisted of a left curve with radius 2000 m, and the last 400 m of the tunnel section consisted of a right curve with the same radius. This prevented the drivers from seeing the daylight in the end of the tunnel too early. The posted speed limit was 90 km/h. Tunnel design was varied on two levels. The walls of a tunnel could either be dark or light-coloured. In case of light-coloured walls, the light part went from the road surface level up to 2 m above road surface level. In all cases the walls were rendered with a slight pattern, simulating the reflections of the lamps mounted under the tunnel roof (Fig. 1).
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
155
Fig. 1. Pictures of the individual tunnels. The factor “tunnel design” was varied on two levels (columns), the factor “light intensity” was varied on three levels (rows). Each of the tunnels was driven twice, once under secondary task load and once without secondary task.
The factor “light intensity” was varied on three levels (Fig. 1). In order to simulate the transitional phase in the tunnel entry and exit where natural light and tunnel illumination mix, the ambient light level was reduced in a linear fashion. The illumination range in reality differs in several orders of magnitude from what can be rendered in a simulator. By taking photographs of and measuring the road surface luminance levels in real tunnels, the projectors in the simulator were calibrated to obtain relative validity. In real tunnels road surface luminance ranged from 3 cd/m2 to 10 cd/m2 . The illumination measured lay between 50 and 1700 lux. When photographs of the tunnels were projected on the simulator screen, road surface luminance ranged from 1.2 cd/m2 to 3.3 cd/m2 in corresponding sections of the image. In the simulator, illumination cannot be measured directly, as the visible light sources only are images of lamps projected on a screen. The luminance of the projected road surfaces was 1.1 cd/m2 in the darkest tunnel, 1.7 cd/m2 in the medium tunnel and 2.5 cd/m2 in the brightest tunnel. The complete procedure for how the brightness levels were calibrated is described in Kircher and Lundkvist (2011). In each tunnel a traffic event occurred (Fig. 2). A short while after entering the tunnel, two vehicles (Vehicle A and Vehicle B) became visible in front of the own vehicle. The own vehicle, Vehicle A and Vehicle B were located in the right lane. Another vehicle (Vehicle C) came up from behind, driving in the passing lane. The three surrounding vehicles were coupled to the own vehicle’s speed. When the own vehicle approached Vehicle B, Vehicle A braked slightly.
When the distance headway from the own vehicle to Vehicle A reached 170 m, the brake lights of Vehicle A were turned off, and Vehicle B started to brake. During this process the participant had to decide whether to brake and remain behind the two lead vehicles or whether to change lanes and overtake the two lead vehicles before Vehicle C had passed. Some additional traffic was present in the periphery for increased realism, but no interactions with these vehicles occurred. In half of the tunnel passages the participants were put under additional visual load, which was accomplished by the so-called arrow task (Jamson and Merat, 2005). This visual loading task occurred twice per tunnel, first during the overtaking event and then when the participant drove freely. In both cases the road was completely straight. To solve the task the participants had to determine whether in a 4 × 4 matrix of arrows there was an arrow pointing upward or not. The task was presented on an LCD touch screen attached to the centre console. Each matrix was present until an answer was given by pressing “yes” or “no” on the touch screen, or until 5 s had passed. Each new matrix was announced with an auditory signal. During the overtaking event the task was presented on a road stretch of 900 m, and further down the tunnel it was presented on a stretch of 800 m in a free driving situation. The difficulty level of the task had been adjusted in pre-trials by manipulating the matrix size and non-target arrow directions such that the task completion time in the absence of driving was approximately 2 s. This difficulty level entails glance durations away from the road which have been shown to be relatively dangerous (Klauer et al., 2006). In
156
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
Fig. 2. Schematic overview of the event. Vehicle X is the own vehicle driven by the participant. When Vehicle A and then Vehicle B brake, the participant has to decide whether to overtake before or after Vehicle C has passed.
the present study the participants could choose whether to use one or several glances in order to make the assessment of whether the target was present or not. For each matrix the correctness of the answer (correct, incorrect, no answer) and the reaction time were scored. 2.5. Assessment of behaviour Performance indicators reflecting both driving behaviour, that is, how the vehicle moves in relation to the road, and driver behaviour, that is, what the drivers do inside of the car, were selected as dependent variables. Driver behaviour as defined above was assessed in terms of visual behaviour. The performance indicators used were: Number of glances away from the road exceeding 2 s, based on the fact that long glances are detrimental to safety (Klauer et al., 2006), percentage of time with empty AttenD buffer (explanation of AttenD see below), and percentage of time with full AttenD buffer. AttenD is a real-time algorithm which estimates how visually attentive a driver currently is. The value is computed based on actual gaze direction in combination with the recent glance history. Slightly simplified, the driver has a “buffer” of 2 s, which is filled up to 2 s in real time when the driver looks at the forward roadway, and which is depleted in real time when the driver looks away from that area. When the buffer value is above 1.8 s the driver is considered to be fully attentive to the forward roadway, while an empty buffer (0 s) indicates a distracted driver. Values in between indicate a transition zone. Special rules apply for glances to the mirrors and the speedometer, for which a latency of 1 s is built in before the buffer is depleted. Special rules apply for intermittent loss of eye tracking and head tracking. For a full description of the algorithm see Kircher and Ahlstrom (2009). Driving behaviour can be assessed on different levels, a common subdivision is into skill-based, rule-based and knowledge based levels (Rasmussen, 1983). Skill-based behaviour is often automatised, such as lane tracking. Here skill-based behaviour is assessed in terms of mean speed, standard deviation of speed, mean lateral position, standard deviation of lateral position (Engström et al., 2005) and percentage of time speeding. Here, speeding is defined as the posted speed limit plus 3 km/h. Skill-based and rule-based behaviour is assessed in relation to the overtaking event, in which the driver has to judge whether to overtake or not, using the performance indicators minimum time to collision and minimum time headway (Vogel, 2003). 2.6. Subjective ratings In the end of each overtaking event as well as later on in the tunnel, while driving as a free vehicle, an auditory signal prompted the driver to make a subjective judgement about the experienced demand of the last driving scenario. The judgement scale was based on the method described by Schweitzer and Green (2007), in which
participants received two anchor values of “demand level 2” and “demand level 6”. These anchor values were represented by short video clips, which showed traffic scenes on an American motorway, with lower and higher traffic density, respectively, and lower and higher demands on the interaction with other vehicles. The same videos were used to calibrate the judgement of the participants in the present study. They watched the videos before the experimental drive and made practice judgements, and during the experimental drive the videos were played non-stop on a laptop computer on the passenger’s seat for reference. Participants were allowed to make judgements below “demand level 0”, and they could also use fractions. 2.7. Procedure Upon arrival the participants read through the instructions and were asked to sign an informed consent form. They also practiced the arrow task until they felt comfortable with the task. Once in the simulator, the participants were given a scenario description so that everyone had the same goal with the upcoming trip. This included a task description which made the arrow task less artificial. Participants were informed that they should drive as they normally would have driven in the described situation, which included that they were meant to meet with a friend at a certain time. An average speed of 88.8 km/h was needed to be on time. They were also informed that they had to respond to important time critical work related assignments that were frequently sent to their tablet computer (represented by the arrow task). In addition, the participants were informed that the one who performed best on a weighted combination of driving performance, secondary task performance and arriving on time would win D200. After the experiment, all participants received a reimbursement of D30. 2.8. Data processing and statistical analysis The analyses were done with a repeated measures generalised linear model (GLM), as the design was completely within subjects. The factors analysed within the GLM were task load, tunnel design and illumination. The overtaking event factor was added into the analysis as two event levels, because not all dependent variables were meaningful to be analysed in both event and non-event situations, and the intention was to keep analyses as equal as possible across the dependent factors. The analyses were made at a significance level of ˛ = 0.05, but for the overview tables presented below, significance levels of ˛ = 0.01 and ˛ = 0.10 are also indicated to give a more complete picture. The performance indicator calculations were based on data from 500 m long road stretches surrounding the arrow task. For the occasions when the secondary task was not performed, the same 500 m road stretch was selected. When the arrow task was performed during the overtaking event, the data collection ended either when the vehicle started to overtake or at the end of the scenario in cases where the driver chose not to overtake. Data collection always started 500 m before the determined end point. As the tunnel in the area in question was completely straight, this small variance in
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
sampling location was assumed not to influence the performance indicator values systematically.
3. Results For six of the participants the eye tracking system did not deliver reliable data, therefore those participants were excluded from further analysis in the driver behaviour category. Data for driving behaviour were not affected. On average the participants answered more than 80% (83–97% for different tunnel types) of all secondary tasks. Of those, 86% were answered correctly, that is, either the target arrow was detected correctly, or it was confirmed correctly that no target arrow was present. The mean reaction time was 2.8 s, and there were no significant differences for either illumination or tunnel design in the percentage of answered stimuli, the percentage of correctly answered stimuli, and reaction time. In Table 1 an overview of the main and interaction effects for each factor is given per driving performance indicator both for the event and the non-event situation. The factor with the biggest influence on driving behaviour was task load, which had a significant effect on mean speed with higher speeds for drivers without load and on the standard deviation of lateral position (SDLP) with higher values for visually loaded drivers both in the event and in the non-event situation. In the non-event situation there was an additional main effect of task load on the standard deviation of speed, with speed varying more (sd = 3.5 km/h) when the drivers were distracted than when the drivers were attentive (sd = 2.5; F(1,27) = 17.0, p < .05). There was one significant interaction effect on the standard deviation of speed for illumination and task load in the event situation (F(2, 54) = 4.4, p < .05), which did not carry over to the non-event situation. During the event the standard deviation of speed was larger for attentive than inattentive driving in the darkest condition (6 km/h vs. 5 km/h), while it was the other way round in the two brighter conditions (5.1 km/h vs. 6.1 km/h). In the nonevent situation the difference between attentive and distracted driving with respect to SDLP was intensified for light-coloured walls (Fig. 3). In the non-event situation illumination had a small but significant effect on mean lateral position (F(2, 54) = 3.4; p < .05). With brighter levels of illumination the drivers positioned themselves 3 cm and 5 cm,respectively, closer to the tunnel wall than in the darkest condition. With regard to driver behaviour again the task load factor had the biggest influence on the performance indicators under evaluation (Table 2). The number of glances with a duration of more than 2 s was increased, and the percentage of time during which drivers were fully attentive according to AttenD was decreased when under visual load. This held up both for the event and the non-event situation, even though for long glances there was an interaction with tunnel design in the event situation, with dark walls exaggerating the effect of visual load. Furthermore, in the non-event situation a larger number of long glances and a reduced attentional level were found for dark walls. It was planned to analyse the drivers’ tactical behaviour with respect to overtaking, but it turned out that the lead vehicles were overtaken only 11 times of 336 possible times before the vehicle from behind had passed the participant’s vehicle. These 11 overtaking manoeuvres stem from four different participants. More overtaking manoeuvres occurred in tunnels with light-coloured walls, under bright lighting conditions, and when the driver was attentive, but there were overtaking manoeuvres in the other conditions, too. It had been planned to analyse different aspects of the manoeuvres, but they were too few and came from too few participants to allow a more detailed analysis.
157
The participants rated the demand of the current situation twice while driving in the tunnels, once during the event and once while driving freely. The ratings were analysed with respect to the three factors tunnel design, illumination and task load (Fig. 3). Generally ratings were higher during the event than in the non-event situation (F(2, 25) = 84.0, p < .05), which can also be seen in Fig. 3. For the ratings made during the event a significant difference in the ratings was found for the factor tunnel design, where dark walls received higher demand ratings (3.7) than light walls (3.4), indicating that dark walls were experienced as more demanding (F(1, 27) = 6.8, p < .05). Additionally, task load had a strong effect on the ratings, with the event situations being rated as 3.1 on average when the driver was attentive and 3.9 on average for inattentive drivers (F(1, 27) = 25.9, p < .05). 4. Discussion The results from the secondary task performance indicate that drivers put approximately equal effort into the secondary task across situations, independent of tunnel design or event status. The secondary task was attended to very frequently; more than 80% of all stimuli were answered, and more than 80% of those were answered correctly. The average reaction time was almost 3 s, which indicates that drivers often had to employ more than one glance in order to determine the answer. The fact that the number of very long glances, exceeding 2 s, was higher when the drivers were distracted by the arrow task than for normal driving indicates that the task engaged the drivers, and that drivers were reluctant to interrupt the task. They were willing to neglect driving substantially, even though in a tunnel and while interacting with other road users. The participants were requested to perform the secondary task when it showed up, therefore the results cannot be generalised to a normal driving setting where drivers can select whether to perform a secondary task or not. In real traffic it is likely, however, that those drivers who actually choose to perform a secondary task will have an intrinsic motivation to do so, which will prompt them to put an effort into the task. The results here can therefore be interpreted as an indication for what might happen if a driver actually chooses to engage in a secondary task. The tunnel design and illumination did not influence how well drivers performed on the secondary task, which does not necessarily mean that the invested resources were equal across tunnels. It could in theory also be possible that drivers put more effort into the task in more difficult tunnels. However, in the driving performance indicators there is no consistent pattern indicating such a trade-off. With respect to driving performance, the factor which had the highest influence on several performance indicators is the drivers’ visual task load status. Whether the drivers were loaded with the visual task or not influenced the driving performance indicators more than the tunnel lighting or the colour of the walls. Therefore, in the next section the drivers’ ability to perform a divided attention task will be investigated more closely. In the present study visual distraction was induced, such that conclusions about the likelihood for visual distraction to occur in such a scenario cannot be drawn. However, as mentioned in the introduction, monotony and distraction are a concern in tunnels. The results here show what may happen when the driver has become visually distracted, but not how likely it is for distraction to occur in the different scenarios. The mean speed decreased both during the event and especially while driving freely when the driver was distracted by the arrow task. In the latter case the speed was reduced by more than 4 km/h, which can either be a sign of degraded control or of compensatory behaviour. The standard deviation of speed increased during the distracted periods in comparison to the attentive periods, however. This could be a hint that speed control becomes more difficult under visual load. Still, some planned compensatory behaviour could be
158
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
Table 1 Overview of the significant main effects and interaction effects for the driving behaviour related performance indicators under investigation. Tunnel Event Mean speed sd speed Mean lat pos SDLP Percent speeding
Illum.
*
Task load
Tunnel × Illum.
Tunnel × Task load
Illum. × Task load
** *
(*) *
(*)
No event Mean speed sd speed Mean lat pos SDLP Percent speeding
**
(*)
** * **
**
*
p < .05. p < .01. (*)p < .10. **
involved, as drivers might reduce their mean speed on purpose, knowing that their speed will fluctuate more, and hoping that a general speed reduction will avoid periods of speeding while distracted. Lateral position is an important aspect to consider in tunnels. Driving too close to the wall can entail a safety risk, but in tunnels with oncoming traffic driving too close to the middle is not safe either. The present results for lateral control indicate a behaviour that is comparable to what was found for speed. The average lateral placement was not influenced during the event, but the SDLP was increased, just as the standard deviation of speed was increased, indicating a degradation of control. In a situation like the one present in the event it can be discussed whether it is safer to place the vehicle further to the middle or not. A placement further to the left would increase the distance to the tunnel wall, which was quite large to begin with (road shoulder = 1.5 m). There was traffic in the left lane, which in most cases passed the participant’s vehicle before the participant overtook the lead vehicles. A placement further to the right would increase the distance to the other traffic, while still keeping a substantial safety distance to the tunnel walls. The overtaking event was included in the study as it was reasoned that tunnel design, illumination and task load could have an influence on tactical driving behaviour. Changes in tactical driving behaviour might be observable before changes in control level behaviour occur, and they can have a large influence on traffic flow and efficiency (Cooper et al., 2009). In addition, how drivers make tactical decisions can give an indication of whether the driver’s understanding of the situation is reduced due to external circumstances. The hypothesis here was that drivers would be less likely to overtake in the situation when the tunnel was darker and when the participant was distracted. Unfortunately the overtaking event was designed too tight, such that most drivers chose not to overtake before the vehicle approaching from behind had passed. Therefore no systematic analysis of how the factors under investigation
influenced tactical driving behaviour could be made. It can only be stated that overtaking was more frequent for tunnels with lighter coloured walls and attentive drivers. For future studies it is recommended to improve the event such that overtaking becomes more frequent. Several aspects of eye gaze behaviour were analysed. As mentioned, single glances of more than 2 s are dangerous (Klauer et al., 2006). Not surprisingly, drivers showed a larger number of such very long glances when distracted by the arrow task than when not distracted. However, not only the distraction task, but also the tunnel walls had some influence on the number of very long glances away from the forward roadway. Generally, the dark tunnel walls produced more very long glances away from the forward roadway than the light-coloured tunnel walls. One possible explanation could be that the drivers used more foveal vision in order to track the road edges, as they were more difficult to see when the walls were dark. It appears unlikely that drivers put more effort into the arrow task in dark tunnels, which might have led to more long single glances. This notion is supported by the fact that they did not perform better on the arrow task when they used more long glances. The performance indicators based on the AttenD algorithm reflects when the driver is deemed to be distracted, AttenD output = 0, and when the driver is deemed to be fully attentive to the driving task, AttenD output >1.8. Intermediate values indicate a driver somewhere in between fully attentive and really distracted. Roughly speaking, drivers were visually distracted from the forward roadway for about 10% of the analysed time, and they were fully attentive to the forward roadway for around 70% of the analysed time, which leaves 20% for the middle ground. The duration of the distracted period did not change with the factors under investigation, but the duration of the attentive period did. Both during the event and in the non-event situation the drivers showed shorter percentages of full attention to the driving task when executing
Table 2 Overview of the significant main effects and interaction effects for the driver behaviour related performance indicators under investigation. Tunnel Event # of 2 s glances % AttenD empty % AttenD full No event # of 2 s glances % AttenD empty % AttenD full * **
p < .05. p < .01.
Illum.
Task load *
*
*
*
*
**
Tunnel × Illum.
Tunnel × Task load *
Illum. × Task load
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
the arrow task, which is not surprising. Drivers who perform the arrow task have to look away for a substantial amount of time. In cases where no secondary task was performed during the event, the percentage of time during which drivers were fully attentive was smaller than during the non-event situation. This can be an
a
159
indication that drivers exhibited a rather extended scanning behaviour during the event, checking for surrounding vehicles, which was interpreted by the algorithm as a decrease in attention to driving, as fewer glances were directed to the forward roadway. When the secondary task was added, the percentage with full
b
100
Task Load
100
Task Load
loaded not loaded
loaded not loaded 90
90
80
80
70
70
60
60
50
50 dark walls
dark walls
light-coloured walls
light-coloured walls
Tunnel Design
Tunnel Design
c
d Task Load
1,0
Task Load
1,0
loaded not loaded
loaded not loaded
0,8
0,8
0,6
0,6
0,4
0,4
0,2
0,2
0,0
0,0 dark walls
light-coloured walls
dark walls
Tunnel Design
light-coloured walls
Tunnel Design
e
f
Task Load
Task Load loaded not loaded
0,24 0,22
0,22
0,20
0,20
0,18
0,18
0,16
0,16
0,14
0,14
0,12
0,12
0,10
0,10 dark walls
light-coloured walls
Tunnel Design
loaded not loaded
0,24
dark walls
light-coloured walls
Tunnel Design
Fig. 3. Plots of tunnel design and distraction effects. Estimated means for several performance indicators for the factors attention and tunnel design in the event and the non-event situation. Only PI where significant differences were found are presented.
160
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
g
Task Load
100
h
Task Load
100
loaded not loaded
loaded not loaded 95
95
90
90
85
85
80
80
75
75 dark walls
dark walls
light-coloured walls
i
Task Load
6,0
loaded not loaded
5,5
light-coloured walls Tunnel Design
Tunnel Design
j
loaded not loaded
5,5
5,0
5,0
4,5
4,5
4,0
4,0
3,5
3,5
3,0
3,0
2,5
2,5
2,0
Task Load
6,0
2,0 dark walls
light-coloured walls
dark walls
Tunnel Design
light-coloured walls Tunnel Design
Fig. 3. (Continued ).
attention decreased and reached the same level both during the event and during the non-event situation. Possibly drivers do not want to exceed a certain amount of glances away from the forward roadway, which they either direct at other traffic or at the secondary task. Not only the arrow task influenced the percentage of time with full attention. In the non-event situation light-coloured walls led to somewhat higher percentages with full attention. An explanation could be that light-coloured walls help the drivers keep their gaze in the field relevant for driving, and they reduce the number of very long glances away from the forward roadway. For light-coloured walls the edge between road surface and wall is much easier to discern than for dark walls. Therefore, peripheral vision may be enough for edge tracking, while dark walls might lead the drivers to employ foveal vision more often to judge their position in relation to the tunnel walls. Not surprisingly, the results indicate that drivers become less attentive to the road when they are distracted by a secondary task. It is more difficult for the drivers to keep up their attention when the tunnel walls are dark, and there is an indication that very dark illumination can aggravate this. The results show that diminished attention also leads to less controlled driver behaviour. Light-coloured tunnel walls can compensate for these losses to some extent, but it is very important to consider that the visual load had the biggest influence on the performance indicators
evaluated here. It is therefore of major importance to ensure that drivers keep their eyes on the forward roadway when planning new tunnels and remodelling old ones. This has been done in newer long tunnels like the previously mentioned Laerdal Tunnel and the Chinese Zhonnanshan Tunnel by including lighting design dedicated to this purpose (Flø and Jenssen, 2007). Whether static designs are effective for commuters after a longer period of time has yet to be investigated, however, and dynamic activation might be considered. The behavioural results are supported by the subjective ratings, which indicate that the participants felt a higher task demand for dark than for light-coloured tunnel walls. As expected, task demand was experienced to be higher during the overtaking situation, and when loaded with the secondary task. Note that the participants were asked to rate task demand and not mental workload or effort per se. Illumination did not turn out to have a major influence on either driver behaviour, driving behaviour or the subjective ratings in the situation at hand. While this is encouraging with respect to the idea of saving energy by reducing the brightness requirements for tunnel illumination, it should still be investigated whether unexpected events and more complex traffic situations can be handled equally well under low and high illumination levels. The study was carried out in a high-end driving simulator. As already mentioned in the methods section, a predecessor of
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
the simulator was tested and relative validity was obtained with respect to speed and lateral position (Harms, 1994; Alm, 1995). Therefore we can have some confidence in the relative validity of the results with respect to speed and lateral positioning. A limitation in this study that should be kept in mind is that drivers were required to perform the secondary task. A different type of experimental design is necessary to evaluate the drivers’ actual willingness to engage in such tasks depending on the tunnel type. Furthermore, in the present study the same situation was repeated 12 times. It is likely that learning effects occurred due to this, even though counterbalancing was meant to mitigate this. Another limitation in this study is that most participants did not have much experience with driving in tunnels. This is an unfortunate consequence of the lack of tunnels in the proximity of Linköping, Sweden, where the simulator is located. The present study can be seen as a first step in finding out what affects behaviour and thereby safety in a tunnel. There is a multitude of factors that may play a role, and different combinations can yield different results. The approach taken here was to vary three factors, two of them situated in the environment, and one situated in the driver. No variation was made in the surrounding traffic, in the road layout, or in the “decoration” of the tunnels. These factors are candidates for further studies in the quest to make tunnels safer. 5. Conclusions While tunnel design did have some influence on the drivers’ behaviour, the drivers’ attention to the driving task was the most crucial factor. In the present study distraction was induced, but it is known that long tunnels with low traffic volumes are monotonous, which facilitates driver distraction. Therefore it is important to investigate whether tunnels can be designed in a way that keeps the drivers’ attention focused on the road. This might be achieved with the help of lighting effects, and other visual guidance in the environment. The results indicate that light-coloured tunnel walls are more important than strong illumination to keep the drivers’ visual attention focused forward. We are however not able to provide absolute values for “sufficient” illumination in a real tunnel since the absolute illumination range in a simulator is smaller than in the real world by several orders of magnitude. References Alm, H., 1995. Driving Simulators as Research Tools—A Validation Study Based on the VTI Driving Simulator. VTI, Linköping, Sweden. Amundsen, F.H., 1994. Studies of driver behaviour in Norwegian road tunnels. Tunnelling and Underground Space Technology 9 (1), 9–15. Amundsen, F.H., Engelbrektsen, A., 2009. Studies on Norwegian Road Tunnels II. An Analysis on Traffic Accidents in Road Tunnels 2001–2006. Vegdirektoratet, Roads and Traffic Department, Traffic Safety Section, Oslo. Carmody, J., 1997. Design Issues Related to Road Tunnels. University of Minnesota, Department of Architecture. Carvel, R., Marlair, G., 2005. A history of fire incidents in tunnels. In: Beard, A.N., Marlair, G. (Eds.), The Handbook of Tunnel Fire Safety. Thomas Telford Limited, London, pp. 3–41. CIE, 2004. Guide for the Lighting of Road Tunnels and Underpasses. CIE (International Commission on Illumination), Vienna, Austria. Cooper, J.M., Vladisavljevic, I., Medeiros-Ward, N., Martin, P.T., Strayer, D.L., 2009. An investigation of driver distraction near the tipping point of traffic flow stability. Human Factors 51 (2), 261–268.
161
Donmez, B., Boyle, L.N., Lee, J.D., 2007. Safety implications of providing realtime feedback to distracted drivers. Accident Analysis & Prevention 39 (3), 581–590. Engström, J., Johansson, E., Östlund, J., 2005. Effects of visual and cognitive load in real and simulated motorway driving. Transportation Research Part F: Traffic Psychology and Behaviour 8 (2), 97–120. Flø, M., Jenssen, G.D., 2007. Drivers’ perception of long tunnels. In: Proceedings of the 4th International Conference—Traffic and Safety in Road Tunnels, Hamburg, Germany, 25–27 April 2007. Harms, L., 1994. Driving performance on a real road and in a driving simulator: results of a validation study. In: Gale, A.G. (Ed.), Vision in Vehicles V. Elsevier, North Holland. Horberry, T., Anderson, J., Regan, M.A., Triggs, T.J., Brown, J., 2006. Driver distraction: the effects of concurrent in-vehicle tasks, road environment complexity and age on driving performance. Accident Analysis & Prevention 38, 185–191. Jamson, A.H., Merat, N., 2005. Surrogate in-vehicle information systems and driver behaviour: effects of visual and cognitive load in simulated rural driving. Transportation Research Part F 8, 79–96. Jenssen, G.D., 1999. Evaluation of interior design in the world’s longest road tunnel. In: First International Conference: Long Road and Rail Tunnels, Basel, Switzerland. Kircher, K., Ahlstrom, C., 2009. Issues related to the driver distraction detection algorithm attend. In: Regan, M., Victor, T. (Eds.), First International Conference on Driver Distraction and Inattention. Gothenburg, Sweden. Kircher, K., Lundkvist, S.-O., 2011. The Influence of Lighting, Wall Colour and Inattention on Traffic Safety in Tunnels. A Simulator Study. VTI, Linköping, Sweden. Kirkland, C.J., 2002. The fire in the channel tunnel. Tunnelling and Underground Space Technology 17 (2), 129–132. Klauer, S.G., Dingus, T.A., Neale, V.L., Sudweeks, J., Ramsey, D., 2006. The Impact of Driver Inattention on Near-Crash/Crash Risk: An Analysis using the 100-car Naturalistic Driving Study Data. NHTSA, Washington, DC, p. 226. Kvaale, J., Lotsberg, G., 2001. Measures against monotony and phobia in the 24.5 km long Laerdal tunnel in Norway. In: Krokeborg, J. (Ed.), Strait Crossings 2001: Proceedings of the Fourth Symposium on Strait Crossings, Bergen, Norway. Swets & Zeitlinger B.V., Lisse, The Netherlands. Leitner, A., 2001. The fire catastrophe in the Tauern tunnel: experience and conclusions for the Austrian guidelines. Tunnelling and Underground Space Technology 16 (3), 217–223. Manser, M.P., Hancock, P.A., 2007. The influence of perceptual speed regulation on speed perception, choice, and control: tunnel wall characteristics and influences. Accident Analysis & Prevention 39 (1), 69–78. Mashimo, H., 2002. State of the road tunnel safety technology in Japan. Tunnelling and Underground Space Technology 17 (2), 145–152. Mccartt, A.T., Hellinga, L.A., Braitman, K.A., 2006. Cell phones and driving: review of research. Traffic Injury Prevention 7 (2), 89–106. Merat, N., Anttila, V., Luoma, J., 2005. Comparing the driving performance of average and older drivers: the effect of surrogate in-vehicle information systems. Transportation Research Part F: Traffic Psychology and Behaviour 8 (2), 147–166. Nussbaumer, C., 2007. Comparative analysis of safety in tunnels. In: Young Researchers Seminar 2007, Brno. Rasmussen, J., 1983. Skills, rules, and knowledge; signals, signs, and symbols, and other distinctions in human performance models. IEEE Transactions on Systems, Man, and Cybernetics 13 (3), 257–266. Robatsch, K., Nussbaumer, C., 2004. Tunnels mit Gegenverkehr und Richtungsverkehr/tunnels with uni- and bi-directional traffic. Bundesministerium für Wissenschaft und Verkehr, Wien. Schweitzer, J., Green, P.A., 2007. Task Acceptability and Workload of Driving City Streets, Rural Roads and Expressways: Ratings from Video Clips. UMTRI, Ann Arbor, MI. Thiffault, P., Bergeron, J., 2003. Monotony of road environment and driver fatigue: a simulator study. Accident Analysis & Prevention 35 (3), 381–391. Ting, P.-H., Hwang, J.-R., Doong, J.-L., Jeng, M.-C., 2008. Driver fatigue and highway driving: a simulator study. Physiology & Behavior 94 (3), 448–453. Törnros, J., 1998. Driving behaviour in a real and simulated road tunnel—a validation study. Accident Analysis & Prevention 30 (4), 497–503. Törnros, J., Harms, L., Alm, H., 1997. The VTIDriving simulator—validation studies. In: Proceedings of the DSC 97—Driving Simulation Conference, Lyon, France. Vogel, K., 2003. A comparison of headway and time to collision as safety indicators. Accident Analysis and Prevention 35 (3), 427–433. Young, K.L., Regan, M.A., Hammer, M., 2003. Driver Distraction: A review of the Literature. Monash University Accident Research Centre, Clayton Campus, Victoria, Australia, p. 66. Zhang, H., Smith, M.R.H., Witt, G.J., 2006. Identification of real-time diagnostic measures of visual distraction with an automatic eye-tracking system. Human Factors 48 (4), 805–821.
Contents lists available at SciVerse ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
The impact of tunnel design and lighting on the performance of attentive and visually distracted drivers Katja Kircher ∗ , Christer Ahlstrom Swedish National Road and Transport Research Institute (VTI), S-581 95 Linköping, Sweden
a r t i c l e
i n f o
Article history: Received 28 October 2011 Received in revised form 13 January 2012 Accepted 15 January 2012 Keywords: Tunnel Driving performance Distraction Illumination Design Visual load Secondary task
a b s t r a c t The crash risk in tunnels is lower than on the open road network, but the consequences of a crash are often severe. Proper tunnel design is one measure to reduce the likelihood of crashes, and the objective of this work is to investigate how driving performance is influenced by design factors, and whether there is an interaction with secondary task load. Twenty-eight drivers participated in the simulator study. A full factorial within subject design was used to investigate the tunnel wall colour (dark or light-coloured walls), illumination (three different levels) and task load (with or without a visual secondary task). The results show that tunnel design and illumination have some influence on the drivers’ behaviour, but visual attention given to the driving task is the most crucial factor, giving rise to significant changes in both driving behaviour and visual behaviour. The results also indicate that light-coloured tunnel walls are more important than strong illumination to keep the drivers’ visual attention focused forward. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction The crash risk in tunnels is lower than on the open road network, but the consequences of a crash are often severe (Leitner, 2001; Kirkland, 2002; Carvel and Marlair, 2005). It is therefore of high importance to ensure high safety standards in tunnels. This is achieved partly by reducing the probability of crashes and partly by reducing the consequences of crashes or fires. The former involves tunnel design, traffic regulations, appropriate facilities (ventilation system, lighting and interior) and maintenance, whereas the latter involves proper emergency facilities and fire-resistant structures (Mashimo, 2002). In this study we focus on crash prevention and investigate in which way different levels of illumination and brightness of the tunnel walls influence the behaviour of attentive and visually distracted drivers. Tunnel lighting regulations are complex with different lighting levels in different zones of the tunnel and with different requirements in different countries. The requirements on lighting also depend on the traffic intensity, the speed limit and the outdoor conditions (day/night). Most countries use CIE Report 088:2004 (CIE, 2004) as a base. Common for all regulations is that there is a requirement on the luminance of the road surface, but not for the walls or the roof. For example in Norway and Sweden the regula-
∗ Corresponding author. Tel.: +46 13 204118; fax: +46 13 14 1436. E-mail addresses: [email protected] (K. Kircher), [email protected] (C. Ahlstrom). 0001-4575/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.aap.2012.01.019
tions recommend the use of white walls up to a certain height, but without mentioning any figure. The illumination of a tunnel can substantially change the impression on the driver, but, to the best of our knowledge, there is no previous research that investigates how tunnel illumination affects driver behaviour. In fact, very limited research has been devoted to the analysis of driver behaviour in tunnels at all. The possibility to use the tunnel walls to convey information to the drivers was suggested by Carmody (1997), who proposed vertical stripes for speed information and horizontal stripes for gradient information. Manser and Hancock (2007) investigated in a simulator how speed can be influenced with the help of patterns and different textures on the roadside. They found that the presence of texture had an attenuating effect on speed in general. Patterns with decreasing width led to gradually decreased speed, while the opposite was true for patterns with increasing width. Most research efforts have been allocated to accident analysis. Both in Norway (Amundsen, 1994; Amundsen and Engelbrektsen, 2009) and in Austria (Nussbaumer, 2007) it has been found that slightly fewer crashes occur in tunnels compared to normal roads, but the fatality rate is substantially higher. Generally the crash risk is higher in the entrance zone of the tunnel than further into the tunnel (Amundsen, 1994; Amundsen and Engelbrektsen, 2009), and the probability of being injured or killed is 19% higher in tunnels with bi-directional traffic compared to tunnels with unidirectional traffic (Robatsch and Nussbaumer, 2004). Nussbaumer (2007) also mentions that according to police reports, the main reasons for tunnel crashes are lacking vigilance (e.g. fatigue,
154
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
distraction, inattention) and behavioural aspects (e.g. safety distance, lane keeping, overtaking). Both constant high levels and constant low levels of activation can be detrimental to driving; performance breaks down when the workload becomes too high (Young et al., 2003; Horberry et al., 2006; Mccartt et al., 2006) whereas longer periods of monotonous driving can lead to boredom and fatigue (Thiffault and Bergeron, 2003; Ting et al., 2008). Depending on the traffic intensity as well as the length of the tunnel and the tunnel design, the resulting level of driver activation can be anywhere between high and low. As long tunnels become more common, the issue of monotony and a resulting propensity of the driver to take his eyes off the road when looking for activation is a concern. When planning the 24.5 km long Laerdal Tunnel in Norway, simulator studies were conducted to investigate how to combat monotony and distraction (Jenssen, 1999; Kvaale and Lotsberg, 2001; Flø and Jenssen, 2007), finding that different types of lightings and constructions influence both driving behaviour and gaze behaviour. While these studies focussed on different colours and lighting designs, it may also be the case that brightness levels influence behaviour. In this study the precondition was to examine drivers both during an imposed visual distraction and while not visually distracted, to determine possible interactions of visual distraction with tunnel brightness. In simulator studies it is common to induce visual distraction by presenting the driver with an additional task that requires the driver to take his eyes off the road (e.g. Zhang et al., 2006; Donmez et al., 2007). In the current study, the driver is considered visually distracted, when the driver interacts with a standardised visuo-manual divided attention (Merat et al., 2005). This task includes target identification by visual search, is system paced, and was developed specifically to represent interaction with an object requiring high visual engagement. The aim of this study was to investigate the effects of illumination in tunnels on attentive and visually distracted drivers. Three levels of illumination were investigated in combination with lightcoloured versus dark tunnel walls and attentive versus visually distracted drivers. To ensure a high level of control over the situation and to enable strict reproducibility, a simulator study was conducted. The specific research questions were: Does the illumination level in a tunnel influence driver behaviour? Does the brightness of the wall in a tunnel influence driver behaviour? Is there an interaction with distraction, in the sense that brightness might get crucial only when drivers are loaded with a visual task?
2.2. Design A full factorial 2 × 3 × 2 (tunnel design × light intensity × task load) within subject design was used. The tunnel could either have dark or light-coloured walls in combination with three different levels of illumination. These six combinations of tunnel designs were driven twice, once with a secondary task present and once without. This means that each participant drove through twelve tunnels in total. The order of the tunnel sections was randomised with certain restrictions. The same lighting levels were only allowed to occur twice in consecutive order, the same wall colour was not allowed to occur more than three times in consecutive order, and the same secondary task condition was not allowed to occur more than three times in consecutive order. 2.3. Driving simulator The experiment was conducted in the VTI Driving Simulator III, an advanced moving base simulator. The moving base generates forces with three different systems: a large linear motion to simulate lateral motion, a tilt motion to simulate long term accelerations such as driving in a curve or longitudinal acceleration and deceleration, and a vibration table to simulate road roughness. The visual system consists of 3 digital projectors providing a 120◦ forward field of view and 3 liquid crystal displays for the rear view mirrors. For gaze tracking the eye tracker SmartEye Pro 5.6 with a set of three cameras was used. One camera was installed next to the A pillar on the driver’s side, one in front of the driver on the dash board, and one further to the right on the dash board, above the centre console. A predecessor of this driving simulator has been validated for tunnel scenarios in terms of speed and lateral position (Törnros, 1998). The speed profile was similar in the simulator and in a field setting except for an offset with consistently higher speeds in the simulator. The same effect has been observed for validation studies of open road driving (Harms, 1994; Alm, 1995; Törnros et al., 1997). Regarding lateral position, drivers tended to position the car in the lane such that they kept some distance to the nearest wall. This effect was greater when the wall was located to the left of the driver, that is, closer to the driver, as left-hand steering was used. The effect could be observed both in the simulator and in the real tunnel, with a slightly greater effect in the latter (Törnros, 1998). 2.4. Driving scenario and environment
2. Materials and methods 2.1. Participants Twenty-eight drivers participated in the study, 10 women and 18 men. The mean age was 41.3 years, with a standard deviation of 7.6 years. Their driving experience ranged from below 5000 km yearly (3 participants) over 5000–10,000 km yearly (4 participants) and 10,000–20,000 km yearly (12 participants) to above 20,000 km yearly (9 participants), that is, 75% of the participants drove a relatively high yearly mileage. Most participants did not have much experience with driving in tunnels. Of the 28 drivers 20 stated that they drove through tunnels a few times per year at most. Six participants drove through tunnels at least once a month, one reported to drive through tunnels at least once a week, and the remaining participant used tunnels on a daily basis. None of the participants indicated more than a slight anxiety while driving in tunnels – 18 stated that they were not afraid at all, while 10 participants said that they felt a slight anxiety, but used tunnels anyway.
The trial started with a training section, which consisted of a 6 km long segment of open motorway without tunnel. After the training section the experimental trial began. Here the road consisted of 2 km long sections of motorway without tunnel and 4 km long sections of motorway in a tunnel. These two section types alternated, starting and ending with an open motorway section. The road had two lanes in each direction, and road shoulders. The road was 10 m wide altogether, the lane width was 3.5 m, and the shoulders were 1.5 m wide each. The open motorway was completely straight, while the first 400 m of the tunnel section consisted of a left curve with radius 2000 m, and the last 400 m of the tunnel section consisted of a right curve with the same radius. This prevented the drivers from seeing the daylight in the end of the tunnel too early. The posted speed limit was 90 km/h. Tunnel design was varied on two levels. The walls of a tunnel could either be dark or light-coloured. In case of light-coloured walls, the light part went from the road surface level up to 2 m above road surface level. In all cases the walls were rendered with a slight pattern, simulating the reflections of the lamps mounted under the tunnel roof (Fig. 1).
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
155
Fig. 1. Pictures of the individual tunnels. The factor “tunnel design” was varied on two levels (columns), the factor “light intensity” was varied on three levels (rows). Each of the tunnels was driven twice, once under secondary task load and once without secondary task.
The factor “light intensity” was varied on three levels (Fig. 1). In order to simulate the transitional phase in the tunnel entry and exit where natural light and tunnel illumination mix, the ambient light level was reduced in a linear fashion. The illumination range in reality differs in several orders of magnitude from what can be rendered in a simulator. By taking photographs of and measuring the road surface luminance levels in real tunnels, the projectors in the simulator were calibrated to obtain relative validity. In real tunnels road surface luminance ranged from 3 cd/m2 to 10 cd/m2 . The illumination measured lay between 50 and 1700 lux. When photographs of the tunnels were projected on the simulator screen, road surface luminance ranged from 1.2 cd/m2 to 3.3 cd/m2 in corresponding sections of the image. In the simulator, illumination cannot be measured directly, as the visible light sources only are images of lamps projected on a screen. The luminance of the projected road surfaces was 1.1 cd/m2 in the darkest tunnel, 1.7 cd/m2 in the medium tunnel and 2.5 cd/m2 in the brightest tunnel. The complete procedure for how the brightness levels were calibrated is described in Kircher and Lundkvist (2011). In each tunnel a traffic event occurred (Fig. 2). A short while after entering the tunnel, two vehicles (Vehicle A and Vehicle B) became visible in front of the own vehicle. The own vehicle, Vehicle A and Vehicle B were located in the right lane. Another vehicle (Vehicle C) came up from behind, driving in the passing lane. The three surrounding vehicles were coupled to the own vehicle’s speed. When the own vehicle approached Vehicle B, Vehicle A braked slightly.
When the distance headway from the own vehicle to Vehicle A reached 170 m, the brake lights of Vehicle A were turned off, and Vehicle B started to brake. During this process the participant had to decide whether to brake and remain behind the two lead vehicles or whether to change lanes and overtake the two lead vehicles before Vehicle C had passed. Some additional traffic was present in the periphery for increased realism, but no interactions with these vehicles occurred. In half of the tunnel passages the participants were put under additional visual load, which was accomplished by the so-called arrow task (Jamson and Merat, 2005). This visual loading task occurred twice per tunnel, first during the overtaking event and then when the participant drove freely. In both cases the road was completely straight. To solve the task the participants had to determine whether in a 4 × 4 matrix of arrows there was an arrow pointing upward or not. The task was presented on an LCD touch screen attached to the centre console. Each matrix was present until an answer was given by pressing “yes” or “no” on the touch screen, or until 5 s had passed. Each new matrix was announced with an auditory signal. During the overtaking event the task was presented on a road stretch of 900 m, and further down the tunnel it was presented on a stretch of 800 m in a free driving situation. The difficulty level of the task had been adjusted in pre-trials by manipulating the matrix size and non-target arrow directions such that the task completion time in the absence of driving was approximately 2 s. This difficulty level entails glance durations away from the road which have been shown to be relatively dangerous (Klauer et al., 2006). In
156
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
Fig. 2. Schematic overview of the event. Vehicle X is the own vehicle driven by the participant. When Vehicle A and then Vehicle B brake, the participant has to decide whether to overtake before or after Vehicle C has passed.
the present study the participants could choose whether to use one or several glances in order to make the assessment of whether the target was present or not. For each matrix the correctness of the answer (correct, incorrect, no answer) and the reaction time were scored. 2.5. Assessment of behaviour Performance indicators reflecting both driving behaviour, that is, how the vehicle moves in relation to the road, and driver behaviour, that is, what the drivers do inside of the car, were selected as dependent variables. Driver behaviour as defined above was assessed in terms of visual behaviour. The performance indicators used were: Number of glances away from the road exceeding 2 s, based on the fact that long glances are detrimental to safety (Klauer et al., 2006), percentage of time with empty AttenD buffer (explanation of AttenD see below), and percentage of time with full AttenD buffer. AttenD is a real-time algorithm which estimates how visually attentive a driver currently is. The value is computed based on actual gaze direction in combination with the recent glance history. Slightly simplified, the driver has a “buffer” of 2 s, which is filled up to 2 s in real time when the driver looks at the forward roadway, and which is depleted in real time when the driver looks away from that area. When the buffer value is above 1.8 s the driver is considered to be fully attentive to the forward roadway, while an empty buffer (0 s) indicates a distracted driver. Values in between indicate a transition zone. Special rules apply for glances to the mirrors and the speedometer, for which a latency of 1 s is built in before the buffer is depleted. Special rules apply for intermittent loss of eye tracking and head tracking. For a full description of the algorithm see Kircher and Ahlstrom (2009). Driving behaviour can be assessed on different levels, a common subdivision is into skill-based, rule-based and knowledge based levels (Rasmussen, 1983). Skill-based behaviour is often automatised, such as lane tracking. Here skill-based behaviour is assessed in terms of mean speed, standard deviation of speed, mean lateral position, standard deviation of lateral position (Engström et al., 2005) and percentage of time speeding. Here, speeding is defined as the posted speed limit plus 3 km/h. Skill-based and rule-based behaviour is assessed in relation to the overtaking event, in which the driver has to judge whether to overtake or not, using the performance indicators minimum time to collision and minimum time headway (Vogel, 2003). 2.6. Subjective ratings In the end of each overtaking event as well as later on in the tunnel, while driving as a free vehicle, an auditory signal prompted the driver to make a subjective judgement about the experienced demand of the last driving scenario. The judgement scale was based on the method described by Schweitzer and Green (2007), in which
participants received two anchor values of “demand level 2” and “demand level 6”. These anchor values were represented by short video clips, which showed traffic scenes on an American motorway, with lower and higher traffic density, respectively, and lower and higher demands on the interaction with other vehicles. The same videos were used to calibrate the judgement of the participants in the present study. They watched the videos before the experimental drive and made practice judgements, and during the experimental drive the videos were played non-stop on a laptop computer on the passenger’s seat for reference. Participants were allowed to make judgements below “demand level 0”, and they could also use fractions. 2.7. Procedure Upon arrival the participants read through the instructions and were asked to sign an informed consent form. They also practiced the arrow task until they felt comfortable with the task. Once in the simulator, the participants were given a scenario description so that everyone had the same goal with the upcoming trip. This included a task description which made the arrow task less artificial. Participants were informed that they should drive as they normally would have driven in the described situation, which included that they were meant to meet with a friend at a certain time. An average speed of 88.8 km/h was needed to be on time. They were also informed that they had to respond to important time critical work related assignments that were frequently sent to their tablet computer (represented by the arrow task). In addition, the participants were informed that the one who performed best on a weighted combination of driving performance, secondary task performance and arriving on time would win D200. After the experiment, all participants received a reimbursement of D30. 2.8. Data processing and statistical analysis The analyses were done with a repeated measures generalised linear model (GLM), as the design was completely within subjects. The factors analysed within the GLM were task load, tunnel design and illumination. The overtaking event factor was added into the analysis as two event levels, because not all dependent variables were meaningful to be analysed in both event and non-event situations, and the intention was to keep analyses as equal as possible across the dependent factors. The analyses were made at a significance level of ˛ = 0.05, but for the overview tables presented below, significance levels of ˛ = 0.01 and ˛ = 0.10 are also indicated to give a more complete picture. The performance indicator calculations were based on data from 500 m long road stretches surrounding the arrow task. For the occasions when the secondary task was not performed, the same 500 m road stretch was selected. When the arrow task was performed during the overtaking event, the data collection ended either when the vehicle started to overtake or at the end of the scenario in cases where the driver chose not to overtake. Data collection always started 500 m before the determined end point. As the tunnel in the area in question was completely straight, this small variance in
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
sampling location was assumed not to influence the performance indicator values systematically.
3. Results For six of the participants the eye tracking system did not deliver reliable data, therefore those participants were excluded from further analysis in the driver behaviour category. Data for driving behaviour were not affected. On average the participants answered more than 80% (83–97% for different tunnel types) of all secondary tasks. Of those, 86% were answered correctly, that is, either the target arrow was detected correctly, or it was confirmed correctly that no target arrow was present. The mean reaction time was 2.8 s, and there were no significant differences for either illumination or tunnel design in the percentage of answered stimuli, the percentage of correctly answered stimuli, and reaction time. In Table 1 an overview of the main and interaction effects for each factor is given per driving performance indicator both for the event and the non-event situation. The factor with the biggest influence on driving behaviour was task load, which had a significant effect on mean speed with higher speeds for drivers without load and on the standard deviation of lateral position (SDLP) with higher values for visually loaded drivers both in the event and in the non-event situation. In the non-event situation there was an additional main effect of task load on the standard deviation of speed, with speed varying more (sd = 3.5 km/h) when the drivers were distracted than when the drivers were attentive (sd = 2.5; F(1,27) = 17.0, p < .05). There was one significant interaction effect on the standard deviation of speed for illumination and task load in the event situation (F(2, 54) = 4.4, p < .05), which did not carry over to the non-event situation. During the event the standard deviation of speed was larger for attentive than inattentive driving in the darkest condition (6 km/h vs. 5 km/h), while it was the other way round in the two brighter conditions (5.1 km/h vs. 6.1 km/h). In the nonevent situation the difference between attentive and distracted driving with respect to SDLP was intensified for light-coloured walls (Fig. 3). In the non-event situation illumination had a small but significant effect on mean lateral position (F(2, 54) = 3.4; p < .05). With brighter levels of illumination the drivers positioned themselves 3 cm and 5 cm,respectively, closer to the tunnel wall than in the darkest condition. With regard to driver behaviour again the task load factor had the biggest influence on the performance indicators under evaluation (Table 2). The number of glances with a duration of more than 2 s was increased, and the percentage of time during which drivers were fully attentive according to AttenD was decreased when under visual load. This held up both for the event and the non-event situation, even though for long glances there was an interaction with tunnel design in the event situation, with dark walls exaggerating the effect of visual load. Furthermore, in the non-event situation a larger number of long glances and a reduced attentional level were found for dark walls. It was planned to analyse the drivers’ tactical behaviour with respect to overtaking, but it turned out that the lead vehicles were overtaken only 11 times of 336 possible times before the vehicle from behind had passed the participant’s vehicle. These 11 overtaking manoeuvres stem from four different participants. More overtaking manoeuvres occurred in tunnels with light-coloured walls, under bright lighting conditions, and when the driver was attentive, but there were overtaking manoeuvres in the other conditions, too. It had been planned to analyse different aspects of the manoeuvres, but they were too few and came from too few participants to allow a more detailed analysis.
157
The participants rated the demand of the current situation twice while driving in the tunnels, once during the event and once while driving freely. The ratings were analysed with respect to the three factors tunnel design, illumination and task load (Fig. 3). Generally ratings were higher during the event than in the non-event situation (F(2, 25) = 84.0, p < .05), which can also be seen in Fig. 3. For the ratings made during the event a significant difference in the ratings was found for the factor tunnel design, where dark walls received higher demand ratings (3.7) than light walls (3.4), indicating that dark walls were experienced as more demanding (F(1, 27) = 6.8, p < .05). Additionally, task load had a strong effect on the ratings, with the event situations being rated as 3.1 on average when the driver was attentive and 3.9 on average for inattentive drivers (F(1, 27) = 25.9, p < .05). 4. Discussion The results from the secondary task performance indicate that drivers put approximately equal effort into the secondary task across situations, independent of tunnel design or event status. The secondary task was attended to very frequently; more than 80% of all stimuli were answered, and more than 80% of those were answered correctly. The average reaction time was almost 3 s, which indicates that drivers often had to employ more than one glance in order to determine the answer. The fact that the number of very long glances, exceeding 2 s, was higher when the drivers were distracted by the arrow task than for normal driving indicates that the task engaged the drivers, and that drivers were reluctant to interrupt the task. They were willing to neglect driving substantially, even though in a tunnel and while interacting with other road users. The participants were requested to perform the secondary task when it showed up, therefore the results cannot be generalised to a normal driving setting where drivers can select whether to perform a secondary task or not. In real traffic it is likely, however, that those drivers who actually choose to perform a secondary task will have an intrinsic motivation to do so, which will prompt them to put an effort into the task. The results here can therefore be interpreted as an indication for what might happen if a driver actually chooses to engage in a secondary task. The tunnel design and illumination did not influence how well drivers performed on the secondary task, which does not necessarily mean that the invested resources were equal across tunnels. It could in theory also be possible that drivers put more effort into the task in more difficult tunnels. However, in the driving performance indicators there is no consistent pattern indicating such a trade-off. With respect to driving performance, the factor which had the highest influence on several performance indicators is the drivers’ visual task load status. Whether the drivers were loaded with the visual task or not influenced the driving performance indicators more than the tunnel lighting or the colour of the walls. Therefore, in the next section the drivers’ ability to perform a divided attention task will be investigated more closely. In the present study visual distraction was induced, such that conclusions about the likelihood for visual distraction to occur in such a scenario cannot be drawn. However, as mentioned in the introduction, monotony and distraction are a concern in tunnels. The results here show what may happen when the driver has become visually distracted, but not how likely it is for distraction to occur in the different scenarios. The mean speed decreased both during the event and especially while driving freely when the driver was distracted by the arrow task. In the latter case the speed was reduced by more than 4 km/h, which can either be a sign of degraded control or of compensatory behaviour. The standard deviation of speed increased during the distracted periods in comparison to the attentive periods, however. This could be a hint that speed control becomes more difficult under visual load. Still, some planned compensatory behaviour could be
158
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
Table 1 Overview of the significant main effects and interaction effects for the driving behaviour related performance indicators under investigation. Tunnel Event Mean speed sd speed Mean lat pos SDLP Percent speeding
Illum.
*
Task load
Tunnel × Illum.
Tunnel × Task load
Illum. × Task load
** *
(*) *
(*)
No event Mean speed sd speed Mean lat pos SDLP Percent speeding
**
(*)
** * **
**
*
p < .05. p < .01. (*)p < .10. **
involved, as drivers might reduce their mean speed on purpose, knowing that their speed will fluctuate more, and hoping that a general speed reduction will avoid periods of speeding while distracted. Lateral position is an important aspect to consider in tunnels. Driving too close to the wall can entail a safety risk, but in tunnels with oncoming traffic driving too close to the middle is not safe either. The present results for lateral control indicate a behaviour that is comparable to what was found for speed. The average lateral placement was not influenced during the event, but the SDLP was increased, just as the standard deviation of speed was increased, indicating a degradation of control. In a situation like the one present in the event it can be discussed whether it is safer to place the vehicle further to the middle or not. A placement further to the left would increase the distance to the tunnel wall, which was quite large to begin with (road shoulder = 1.5 m). There was traffic in the left lane, which in most cases passed the participant’s vehicle before the participant overtook the lead vehicles. A placement further to the right would increase the distance to the other traffic, while still keeping a substantial safety distance to the tunnel walls. The overtaking event was included in the study as it was reasoned that tunnel design, illumination and task load could have an influence on tactical driving behaviour. Changes in tactical driving behaviour might be observable before changes in control level behaviour occur, and they can have a large influence on traffic flow and efficiency (Cooper et al., 2009). In addition, how drivers make tactical decisions can give an indication of whether the driver’s understanding of the situation is reduced due to external circumstances. The hypothesis here was that drivers would be less likely to overtake in the situation when the tunnel was darker and when the participant was distracted. Unfortunately the overtaking event was designed too tight, such that most drivers chose not to overtake before the vehicle approaching from behind had passed. Therefore no systematic analysis of how the factors under investigation
influenced tactical driving behaviour could be made. It can only be stated that overtaking was more frequent for tunnels with lighter coloured walls and attentive drivers. For future studies it is recommended to improve the event such that overtaking becomes more frequent. Several aspects of eye gaze behaviour were analysed. As mentioned, single glances of more than 2 s are dangerous (Klauer et al., 2006). Not surprisingly, drivers showed a larger number of such very long glances when distracted by the arrow task than when not distracted. However, not only the distraction task, but also the tunnel walls had some influence on the number of very long glances away from the forward roadway. Generally, the dark tunnel walls produced more very long glances away from the forward roadway than the light-coloured tunnel walls. One possible explanation could be that the drivers used more foveal vision in order to track the road edges, as they were more difficult to see when the walls were dark. It appears unlikely that drivers put more effort into the arrow task in dark tunnels, which might have led to more long single glances. This notion is supported by the fact that they did not perform better on the arrow task when they used more long glances. The performance indicators based on the AttenD algorithm reflects when the driver is deemed to be distracted, AttenD output = 0, and when the driver is deemed to be fully attentive to the driving task, AttenD output >1.8. Intermediate values indicate a driver somewhere in between fully attentive and really distracted. Roughly speaking, drivers were visually distracted from the forward roadway for about 10% of the analysed time, and they were fully attentive to the forward roadway for around 70% of the analysed time, which leaves 20% for the middle ground. The duration of the distracted period did not change with the factors under investigation, but the duration of the attentive period did. Both during the event and in the non-event situation the drivers showed shorter percentages of full attention to the driving task when executing
Table 2 Overview of the significant main effects and interaction effects for the driver behaviour related performance indicators under investigation. Tunnel Event # of 2 s glances % AttenD empty % AttenD full No event # of 2 s glances % AttenD empty % AttenD full * **
p < .05. p < .01.
Illum.
Task load *
*
*
*
*
**
Tunnel × Illum.
Tunnel × Task load *
Illum. × Task load
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
the arrow task, which is not surprising. Drivers who perform the arrow task have to look away for a substantial amount of time. In cases where no secondary task was performed during the event, the percentage of time during which drivers were fully attentive was smaller than during the non-event situation. This can be an
a
159
indication that drivers exhibited a rather extended scanning behaviour during the event, checking for surrounding vehicles, which was interpreted by the algorithm as a decrease in attention to driving, as fewer glances were directed to the forward roadway. When the secondary task was added, the percentage with full
b
100
Task Load
100
Task Load
loaded not loaded
loaded not loaded 90
90
80
80
70
70
60
60
50
50 dark walls
dark walls
light-coloured walls
light-coloured walls
Tunnel Design
Tunnel Design
c
d Task Load
1,0
Task Load
1,0
loaded not loaded
loaded not loaded
0,8
0,8
0,6
0,6
0,4
0,4
0,2
0,2
0,0
0,0 dark walls
light-coloured walls
dark walls
Tunnel Design
light-coloured walls
Tunnel Design
e
f
Task Load
Task Load loaded not loaded
0,24 0,22
0,22
0,20
0,20
0,18
0,18
0,16
0,16
0,14
0,14
0,12
0,12
0,10
0,10 dark walls
light-coloured walls
Tunnel Design
loaded not loaded
0,24
dark walls
light-coloured walls
Tunnel Design
Fig. 3. Plots of tunnel design and distraction effects. Estimated means for several performance indicators for the factors attention and tunnel design in the event and the non-event situation. Only PI where significant differences were found are presented.
160
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
g
Task Load
100
h
Task Load
100
loaded not loaded
loaded not loaded 95
95
90
90
85
85
80
80
75
75 dark walls
dark walls
light-coloured walls
i
Task Load
6,0
loaded not loaded
5,5
light-coloured walls Tunnel Design
Tunnel Design
j
loaded not loaded
5,5
5,0
5,0
4,5
4,5
4,0
4,0
3,5
3,5
3,0
3,0
2,5
2,5
2,0
Task Load
6,0
2,0 dark walls
light-coloured walls
dark walls
Tunnel Design
light-coloured walls Tunnel Design
Fig. 3. (Continued ).
attention decreased and reached the same level both during the event and during the non-event situation. Possibly drivers do not want to exceed a certain amount of glances away from the forward roadway, which they either direct at other traffic or at the secondary task. Not only the arrow task influenced the percentage of time with full attention. In the non-event situation light-coloured walls led to somewhat higher percentages with full attention. An explanation could be that light-coloured walls help the drivers keep their gaze in the field relevant for driving, and they reduce the number of very long glances away from the forward roadway. For light-coloured walls the edge between road surface and wall is much easier to discern than for dark walls. Therefore, peripheral vision may be enough for edge tracking, while dark walls might lead the drivers to employ foveal vision more often to judge their position in relation to the tunnel walls. Not surprisingly, the results indicate that drivers become less attentive to the road when they are distracted by a secondary task. It is more difficult for the drivers to keep up their attention when the tunnel walls are dark, and there is an indication that very dark illumination can aggravate this. The results show that diminished attention also leads to less controlled driver behaviour. Light-coloured tunnel walls can compensate for these losses to some extent, but it is very important to consider that the visual load had the biggest influence on the performance indicators
evaluated here. It is therefore of major importance to ensure that drivers keep their eyes on the forward roadway when planning new tunnels and remodelling old ones. This has been done in newer long tunnels like the previously mentioned Laerdal Tunnel and the Chinese Zhonnanshan Tunnel by including lighting design dedicated to this purpose (Flø and Jenssen, 2007). Whether static designs are effective for commuters after a longer period of time has yet to be investigated, however, and dynamic activation might be considered. The behavioural results are supported by the subjective ratings, which indicate that the participants felt a higher task demand for dark than for light-coloured tunnel walls. As expected, task demand was experienced to be higher during the overtaking situation, and when loaded with the secondary task. Note that the participants were asked to rate task demand and not mental workload or effort per se. Illumination did not turn out to have a major influence on either driver behaviour, driving behaviour or the subjective ratings in the situation at hand. While this is encouraging with respect to the idea of saving energy by reducing the brightness requirements for tunnel illumination, it should still be investigated whether unexpected events and more complex traffic situations can be handled equally well under low and high illumination levels. The study was carried out in a high-end driving simulator. As already mentioned in the methods section, a predecessor of
K. Kircher, C. Ahlstrom / Accident Analysis and Prevention 47 (2012) 153–161
the simulator was tested and relative validity was obtained with respect to speed and lateral position (Harms, 1994; Alm, 1995). Therefore we can have some confidence in the relative validity of the results with respect to speed and lateral positioning. A limitation in this study that should be kept in mind is that drivers were required to perform the secondary task. A different type of experimental design is necessary to evaluate the drivers’ actual willingness to engage in such tasks depending on the tunnel type. Furthermore, in the present study the same situation was repeated 12 times. It is likely that learning effects occurred due to this, even though counterbalancing was meant to mitigate this. Another limitation in this study is that most participants did not have much experience with driving in tunnels. This is an unfortunate consequence of the lack of tunnels in the proximity of Linköping, Sweden, where the simulator is located. The present study can be seen as a first step in finding out what affects behaviour and thereby safety in a tunnel. There is a multitude of factors that may play a role, and different combinations can yield different results. The approach taken here was to vary three factors, two of them situated in the environment, and one situated in the driver. No variation was made in the surrounding traffic, in the road layout, or in the “decoration” of the tunnels. These factors are candidates for further studies in the quest to make tunnels safer. 5. Conclusions While tunnel design did have some influence on the drivers’ behaviour, the drivers’ attention to the driving task was the most crucial factor. In the present study distraction was induced, but it is known that long tunnels with low traffic volumes are monotonous, which facilitates driver distraction. Therefore it is important to investigate whether tunnels can be designed in a way that keeps the drivers’ attention focused on the road. This might be achieved with the help of lighting effects, and other visual guidance in the environment. The results indicate that light-coloured tunnel walls are more important than strong illumination to keep the drivers’ visual attention focused forward. We are however not able to provide absolute values for “sufficient” illumination in a real tunnel since the absolute illumination range in a simulator is smaller than in the real world by several orders of magnitude. References Alm, H., 1995. Driving Simulators as Research Tools—A Validation Study Based on the VTI Driving Simulator. VTI, Linköping, Sweden. Amundsen, F.H., 1994. Studies of driver behaviour in Norwegian road tunnels. Tunnelling and Underground Space Technology 9 (1), 9–15. Amundsen, F.H., Engelbrektsen, A., 2009. Studies on Norwegian Road Tunnels II. An Analysis on Traffic Accidents in Road Tunnels 2001–2006. Vegdirektoratet, Roads and Traffic Department, Traffic Safety Section, Oslo. Carmody, J., 1997. Design Issues Related to Road Tunnels. University of Minnesota, Department of Architecture. Carvel, R., Marlair, G., 2005. A history of fire incidents in tunnels. In: Beard, A.N., Marlair, G. (Eds.), The Handbook of Tunnel Fire Safety. Thomas Telford Limited, London, pp. 3–41. CIE, 2004. Guide for the Lighting of Road Tunnels and Underpasses. CIE (International Commission on Illumination), Vienna, Austria. Cooper, J.M., Vladisavljevic, I., Medeiros-Ward, N., Martin, P.T., Strayer, D.L., 2009. An investigation of driver distraction near the tipping point of traffic flow stability. Human Factors 51 (2), 261–268.
161
Donmez, B., Boyle, L.N., Lee, J.D., 2007. Safety implications of providing realtime feedback to distracted drivers. Accident Analysis & Prevention 39 (3), 581–590. Engström, J., Johansson, E., Östlund, J., 2005. Effects of visual and cognitive load in real and simulated motorway driving. Transportation Research Part F: Traffic Psychology and Behaviour 8 (2), 97–120. Flø, M., Jenssen, G.D., 2007. Drivers’ perception of long tunnels. In: Proceedings of the 4th International Conference—Traffic and Safety in Road Tunnels, Hamburg, Germany, 25–27 April 2007. Harms, L., 1994. Driving performance on a real road and in a driving simulator: results of a validation study. In: Gale, A.G. (Ed.), Vision in Vehicles V. Elsevier, North Holland. Horberry, T., Anderson, J., Regan, M.A., Triggs, T.J., Brown, J., 2006. Driver distraction: the effects of concurrent in-vehicle tasks, road environment complexity and age on driving performance. Accident Analysis & Prevention 38, 185–191. Jamson, A.H., Merat, N., 2005. Surrogate in-vehicle information systems and driver behaviour: effects of visual and cognitive load in simulated rural driving. Transportation Research Part F 8, 79–96. Jenssen, G.D., 1999. Evaluation of interior design in the world’s longest road tunnel. In: First International Conference: Long Road and Rail Tunnels, Basel, Switzerland. Kircher, K., Ahlstrom, C., 2009. Issues related to the driver distraction detection algorithm attend. In: Regan, M., Victor, T. (Eds.), First International Conference on Driver Distraction and Inattention. Gothenburg, Sweden. Kircher, K., Lundkvist, S.-O., 2011. The Influence of Lighting, Wall Colour and Inattention on Traffic Safety in Tunnels. A Simulator Study. VTI, Linköping, Sweden. Kirkland, C.J., 2002. The fire in the channel tunnel. Tunnelling and Underground Space Technology 17 (2), 129–132. Klauer, S.G., Dingus, T.A., Neale, V.L., Sudweeks, J., Ramsey, D., 2006. The Impact of Driver Inattention on Near-Crash/Crash Risk: An Analysis using the 100-car Naturalistic Driving Study Data. NHTSA, Washington, DC, p. 226. Kvaale, J., Lotsberg, G., 2001. Measures against monotony and phobia in the 24.5 km long Laerdal tunnel in Norway. In: Krokeborg, J. (Ed.), Strait Crossings 2001: Proceedings of the Fourth Symposium on Strait Crossings, Bergen, Norway. Swets & Zeitlinger B.V., Lisse, The Netherlands. Leitner, A., 2001. The fire catastrophe in the Tauern tunnel: experience and conclusions for the Austrian guidelines. Tunnelling and Underground Space Technology 16 (3), 217–223. Manser, M.P., Hancock, P.A., 2007. The influence of perceptual speed regulation on speed perception, choice, and control: tunnel wall characteristics and influences. Accident Analysis & Prevention 39 (1), 69–78. Mashimo, H., 2002. State of the road tunnel safety technology in Japan. Tunnelling and Underground Space Technology 17 (2), 145–152. Mccartt, A.T., Hellinga, L.A., Braitman, K.A., 2006. Cell phones and driving: review of research. Traffic Injury Prevention 7 (2), 89–106. Merat, N., Anttila, V., Luoma, J., 2005. Comparing the driving performance of average and older drivers: the effect of surrogate in-vehicle information systems. Transportation Research Part F: Traffic Psychology and Behaviour 8 (2), 147–166. Nussbaumer, C., 2007. Comparative analysis of safety in tunnels. In: Young Researchers Seminar 2007, Brno. Rasmussen, J., 1983. Skills, rules, and knowledge; signals, signs, and symbols, and other distinctions in human performance models. IEEE Transactions on Systems, Man, and Cybernetics 13 (3), 257–266. Robatsch, K., Nussbaumer, C., 2004. Tunnels mit Gegenverkehr und Richtungsverkehr/tunnels with uni- and bi-directional traffic. Bundesministerium für Wissenschaft und Verkehr, Wien. Schweitzer, J., Green, P.A., 2007. Task Acceptability and Workload of Driving City Streets, Rural Roads and Expressways: Ratings from Video Clips. UMTRI, Ann Arbor, MI. Thiffault, P., Bergeron, J., 2003. Monotony of road environment and driver fatigue: a simulator study. Accident Analysis & Prevention 35 (3), 381–391. Ting, P.-H., Hwang, J.-R., Doong, J.-L., Jeng, M.-C., 2008. Driver fatigue and highway driving: a simulator study. Physiology & Behavior 94 (3), 448–453. Törnros, J., 1998. Driving behaviour in a real and simulated road tunnel—a validation study. Accident Analysis & Prevention 30 (4), 497–503. Törnros, J., Harms, L., Alm, H., 1997. The VTIDriving simulator—validation studies. In: Proceedings of the DSC 97—Driving Simulation Conference, Lyon, France. Vogel, K., 2003. A comparison of headway and time to collision as safety indicators. Accident Analysis and Prevention 35 (3), 427–433. Young, K.L., Regan, M.A., Hammer, M., 2003. Driver Distraction: A review of the Literature. Monash University Accident Research Centre, Clayton Campus, Victoria, Australia, p. 66. Zhang, H., Smith, M.R.H., Witt, G.J., 2006. Identification of real-time diagnostic measures of visual distraction with an automatic eye-tracking system. Human Factors 48 (4), 805–821.