1. See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/268146979
A study of the effects of road tunnel on driver behavior and road safety using
driving simulator
Article in Advances in Transportation Studies · January 2013
DOI: 10.4399/97888548611764
CITATIONS
23
READS
766
2 authors:
Some of the authors of this publication are also working on these related projects:
Assessment and Health Monitoring of Railway Ballast using Non-destructive Testing Methods View project
Alessandro Calvi
Università Degli Studi Roma Tre
66 PUBLICATIONS 457 CITATIONS
SEE PROFILE
Fabrizio D'Amico
Università Degli Studi Roma Tre
45 PUBLICATIONS 232 CITATIONS
SEE PROFILE
All content following this page was uploaded by Alessandro Calvi on 12 February 2016.
The user has requested enhancement of the downloaded file.
2. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 59 -
A study of the effects of road tunnel on driver behavior
and road safety using driving simulator
A. Calvi F. D’Amico
Department of Engineering, Roma Tre University,
Via Vito Volterra, 62, 00146 Rome, Italy
email: alessandro.calvi@uniroma3.it
subm. 7th
November 2012 approv. after rev. 15th
April 2013
Abstract
The present paper wants to contribute to the knowledge of the tunnel effects on driving performance and
safety using the advanced technology of driving simulator. Specifically this study presents the first results of
a wider research aimed at establishing how drivers behave inside road tunnel as well as approaching it and
exiting from it. Moreover the study verifies a correlation between accident rates and an advanced indicator of
simulation computed inside tunnel sections.
A highway scenario with eight existing tunnels is reproduced in CRISS driving simulator and several driving
parameters are recorded among a sample of twenty-five drivers. Tunnel scenario (TS) data are processed and
compared with those of a control scenario (CS), characterized by the same road geometries and alignment of
the first one, but without any tunnels. Results confirm previous findings of naturalistic and simulator driving
studies about drivers performance inside road tunnels, with significant differences of longitudinal speeds,
acceleration and lateral position recorded along the TS and the CS.
Moreover the literature safety indicator of driving simulation Pathologic Discomfort (PD) is computed in
order to 1) assess the length of approaching and exiting sections of road tunnel and 2) verify PD correlation
with the accident rate recorded inside each tunnel.
Simulator limitations and future directions of the research are discussed in order to provide guidelines for
practical application to road tunnel design and safety measures, taking in account driving performance.
Keywords – driving simulation, road tunnel, driving performance, tunnel effects
1. Introduction
Several safety analysis of road tunnels performed among years have demonstrated that the
number of road accidents inside tunnels is lower than elsewhere, while the severity associated
with a tunnel crash is absolutely higher (e.g. [1, 2]). In the last decades, above all in Europe,
several tragic fires in tunnels have prompted to change profoundly the safety procedures in
underground constructions, raising the profile of tunnel safety in applying the risk theories to the
design of new galleries or proposing safety measures for existing tunnels [3].
A wide body of research has examined different factors affecting road crash risk inside road
tunnel both from crash prevention perspective, analysing tunnel design (e.g. [4]), traffic
regulations (e.g. [5, 6]), appropriate facilities (e.g. [7, 8]) and maintenance (e.g. [9]), and from
mitigation of crash causes, studying for example proper emergency facilities and fire-resistant
3. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 60 -
structures (e.g. [10]). These previous studies focused generally on crash risk evaluation and
analysis of the relationship between tunnel safety and tunnel facilities with particular attention to
the management of emergency situations.
On the contrary there are few studies in literature that examined the impact of road tunnels on
driving performance with the aim of defining if and how tunnels influence driving operating and
safety and which tunnels features have the highest impact on drivers decisions and performances.
According to the increasing needs for interdisciplinary study to solve safety problems, the role of
human factors in tunnel safety is surely a crucial issue that should be studied more in depth to
provide a better understanding of user behavior in road tunnels in both normal and critical
situations (e.g. [11]).
The present study wants to contribute to enhance the actual knowledge of human factors
investigating the effects of road tunnels on driving performance and road safety using the driving
simulator tool that, in the last decades, has become a consolidated technology for assisting
geometrical road design and studying the driver performance under different traffic and
environmental conditions (e.g. [12-15]), evaluating the interactions between the driver, the
vehicle and the road environment through an interdisciplinary approach.
Specifically in this paper the authors present the main results of a study on the effects of road
tunnels on driving behavior and road safety, developed in the virtual reality environment of
CRISS (Interuniversity Research Centre of Road Safety) driving simulator.
The specific objectives of the study consist in evaluating the effects of tunnels on driving
performances and safety inside road tunnel, as well as approaching it and exiting from it,
following a procedure already presented and discussed in a previous study [16] to confirm and
enlarge previous findings. Moreover a literature safety indicator of driving simulation, the
Pathologic Discomfort (PD) [16-18]), is computed to verify its correlation with the real accident
rate recorded inside each tunnel that is implemented in simulation tests.
2. Background
Road tunnel is a topic widely discussed in literature under different perspectives, from
accident analysis to tunnel design and management (e.g. [1, 2, 19, 20]). Some studies evaluated
the tunnel features that mainly affect the driving safety, using crash analysis. In example some
researchers [1, 21] found that the length of road tunnels was a prevalent factor for crash
occurrence in road tunnels: specifically longer the tunnel lower the accident rate recorded.
Other studies found that in the nearness of short road tunnels higher accident rates were
recorded [19, 20], demonstrating the attention that should be given to entrance and exit sections of
tunnels. In example Amundsen et al. [19] demonstrated that a considerable percentage of
accidents in tunnels occurred in a zone from 50 m ahead of the tunnel portal to 50 m past the
portal. A more recent accident analysis in 22 Norwegian tunnels [20] showed that the zone just
before each tunnel was four times as dangerous as the middle of the tunnel.
The effects of tunnel entrance on driving was also studied in terms of drivers physiological
performance. In example some studies [22, 23] demonstrated that since 150 meters before
entering the tunnel, driver attention was focused on the tunnel entrance, almost neglecting all the
information on signs placed closely at the portal.
Despite a great interest in road tunnel driving and safety among the academic and professional
world, there are few studies in literature that investigate the effects of road tunnels on driving
performance and specifically aimed at understanding if and how drivers change their behavior
approaching, through and exiting the tunnel and which are the main effects on driving safety
4. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 61 -
conditions. Some of these few studies used an advanced technology that overcomes the problems
(e.g. safety, cost, experimental control) of on field studies: the driving simulation. The main
reason of the increasing interest in driving simulator is that several studies demonstrated that this
tool provides the driver with enough visual information to allow him to correctly perceive speeds
and distances [24] and, consequently, the driver behaves in virtual reality as in the real world.
Some studies on road tunnels used driving simulators to achieve different objectives, from the
analysis of the effects of road tunnel on driver performance to the evaluation of the design of road
tunnel in terms of driver perception and action. Lidstrom [4] implemented in a driving simulator
some tunnel designs with the main objective of showing the design in advance and using the
driving simulator as a platform for future tunnel research projects. Results demonstrated the great
advantages in driving in a simulator, with all senses experiencing the design, rather than
evaluating a design passively only by watching a video animation. Recently Manser and Hancock
[8] studied the type of visual pattern and texture applied to road tunnel walls that mostly affect
driving performance. The authors found that, when compared to baseline condition (no visual
pattern), the drivers gradually decreased speed when exposed to the decreasing width visual
pattern and increased speed with the increasing width of visual pattern. Kircher and Ahlstrom [9],
investigating the influence of tunnel design factors on driving performance, demonstrated that,
although tunnel design and illumination had some influences on drivers’ behavior, the visual
attention given to the driving task was the most crucial factor.
The driving simulation was also used for assessing driver’s behavior in case of emergency in a
road tunnel. In example, a driving simulator study developed under the UPTUN project [25]
demonstrated that drivers underestimated the distance travelled inside a road tunnel. The authors
argued that it could affect drivers’ behavior especially in emergency case, when drivers have
rapidly to choose the nearer exit of the tunnel for escaping.
Other simulator studies focused on the evaluation of the effectiveness of signs and information
for driving exercise inside road tunnels. An interesting study developed by Upchurch et al. [5]
demonstrated that driving simulator allowed not only to individuate road design or safety
problems but also to evaluate the possible alternative remedies and measures. Same results were
found by Lorentzen et al. [6] that studied driver’s reactions during simulations tests to assist the
design of road signs in a tunnel.
Some literature studies provided a validation of driving simulation for road tunnels analysis.
Tornos [26], using a medium-cost driving simulator similar to the one used in this study,
demonstrated a relative validity of speeds and lateral positions by measuring and comparing them
in a real tunnel and in the same tunnel implemented in the VTI driving simulator. Hirata et al. [27]
validated their driving simulator for tunnel analysis in terms of perceived speed, distance
headway, and physiological data. Similar results were obtained by Akamatsu et al. [7] in terms of
driver’s accelerator pressure measured inside real and simulated tunnels.
The present research starts from the findings of a previous pilot study [18], where the effects
of tunnels on driving performance using CRISS driving simulator were evaluated for the first
time. Findings showed that drivers moved away laterally from right tunnel wall when they drove
inside the road tunnel and that they slightly slowed down. Moreover inside the tunnel the amount
of driver’s trajectory corrections was definitely lower as if the driver had paid more attention
when driving inside a tunnel. More recently Calvi et al. [16] proposed a new procedure, based on
driving simulation data, aimed at evaluating the lengths of road sections just before the first tunnel
portal and after the second one, where driving performance could be still influenced by the tunnel
and where several accident analysis have shown high crash rates.
5. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 62 -
3. Methodology
The research is organized into the following steps:
− driving simulator study (data collection of different tunnel sections of existing highways
and implementation of the highway alignment and environment in the driving simulator,
creation and implementation of road tunnels in the simulated scenario, driving tests and
data collection);
− post processing of data and calculation of longitudinal speeds, lateral positions and other
simulation outcomes both in the scenario without road tunnels (control scenario) and the
same road alignment and environment with road tunnels (tunnel scenario);
− statistical analysis of data aimed at evaluating the significant effects of road tunnels on
driving performance;
− correlation between an advanced indicator of simulation and the accident rate recorded
inside the existing tunnel sections.
4. Driving simulator study
4.1. Apparatus
Driving simulation tests are performed at the STI driving simulator system, located at the
laboratory of the Interuniversity Research Centre of Road Safety, CRISS (Figure 1) of Roma TRE
University. The system is an interactive fixed-base driving simulator including a complete vehicle
dynamics model based on the computer simulation Vehicle Dynamics Analysis Nonlinear. The
model has been adapted to run in real time, it has been validated extensively [28, 29] and used for
evaluating driving performance in terms of speed, acceleration and trajectory under different
driving conditions [13-16, 30-32].
The hardware consists of four networked computers and three hardware interfaces (the
steering systems, the pedals and the manual gearshift). The road scenario is projected onto three
big screens providing a 135 degree field of view. The resolution of the visual scene is 1024 × 768
pixels with a refresh rate of 30 to 60 Hz depending on scene complexity and traveling conditions
of the vehicle that depend on the actions of the driver on the pedals and the steering wheel. The
simulator allows to model the road in accordance with the traditional roads engineering
constraints. The data recording system acquires all the parameters at spatial intervals of 4 meters.
Fig. 1 - Driving simulator at CRISS laboratory
6. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 63 -
4.2. Simulated scenario
4.2.1. Test alignment
A highway scenario is reproduced in the driving simulator, composed by eight sections of
three different existing Italian highways. Specifically eight twin tube tunnel sections are
reproduced in the driving simulator with all the geometries that agree with the real ones. The
authors decided to reproduce existing road tunnel sections in order to verify a correlation between
safety indicator of simulation and accident rates recorded inside the tunnels and to allow also the
development of further validation studies to ascertain the CRISS driving simulator validity for
tunnel scenario. The road cross section is approximately the same for all the eight tunnels,
composed by a dual carriageway with two lanes (3.75 m wide). The shoulders are 2.50 m wide
and the median 3.0 m. The total length of the scenario is 10100 meter.
Figure 2 shows the road alignment of the simulated scenario. For each tunnel section, 700
meters of existing highway are exactly reproduced in the virtual environment (section of
investigation) and between two consecutive sections a straight of 500 meters (section of
transition) is added to prevent driving performance along a section of investigation being biased
by the previous section of the scenario, as demonstrated in another study [33]. At the beginning of
the scenario a straight of 1000 meter is added to allow the driver to reach a reasonable
approaching speed in the first tunnel section.
Fig. 2 – Horizontal alignment of simulated scenario
Fig. 3 - Tunnels case studies: frames of simulation and tunnels characteristics
7. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 64 -
The simulated tunnels have a length ranged between 210 and 325 meters. The total length of
the eight tunnels is 2145 meters (approximately the 20% of the scenario length). Five tunnels are
on curve section (tunnel 1, 3 and 8 placed along right curve, tunnel 5 and 6 along left curve),
while tunnels 2, 4 and 7 are placed along straight sections. Figure 3 summarizes the
characteristics of the tunnels.
4.2.2. Creation of the tunnel scenario
Tunnels and roadside features are reconstructed using a three dimensional software and then
introduced in the simulator scene as illustrated in Figure 4. All markings and signs are exactly
reproduced in the simulator. To make the built scene as similar as possible to the real one, the
background images are composed by photos of the real environment.
Moreover to reproduce the lighting condition inside road tunnel, some adjustments to the
hardware/software system were necessary. The simulator software allows setting the amount of
ambient and diffuse lighting affecting the overall scene or a specific road section. The ambient
component of the lighting determines how intense or bright the lighting will be, and the diffuse
component determines how much shadowing there will on unlit surfaces. By undertaking a trial-
and-error approach, it is obtained satisfactory results in reproducing the real lighting condition of
road tunnel in the simulator (Figure 5). All of these parameters allowed reproduction of an
effective lighting condition inside the road tunnel simulated.
However, a limitation of the present study is that the CRISS has not yet been validated for
tunnel driving conditions, which is a topic for future research, as discussed later in the paper.
Fig. 4 - Example of reconstruction of real tunnel in simulated environment
Fig. 5 - View of tunnel scenario inside road tunnels
8. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 65 -
4.3. Procedure
The driving tests follow a strict procedure that begins with the communication to the driver
about the duration of the driving and the use of the steering wheel, pedals and automatic gear.
Then, in accordance with other experimental protocols [13, 14] participants were required to
complete familiarization training that entailed driving the simulator vehicle for at least 15 min on
a specific alignment.
Later the test subject drives along the first scenario (control scenario CS or tunnel scenario TS)
and, after it, the driver takes about 1 hour of rest to re-establish psychophysical conditions similar
to those at the beginning of the test; during this time the driver fills an evaluation questionnaire
about type and entity of the discomfort perceived during the driving.
After the rest period the driver completes the full test driving along the second scenario (TS or
CS) and filling again the same questionnaire about this second test. Specifically in the CS, tunnels
are removed from simulation. It is considered as the baseline condition for the analysis of the
effects of road tunnel on driver performance.
In the TS tunnels are replaced exactly in the same sites of the real environment. In such a way,
comparing driving simulation outcomes of the two scenarios, it was possible to study the effects
of road tunnels on driving performance, limiting any biases due to the road alignment and
transversal section. The sequence of the two scenarios was counterbalanced in order to avoid
influence due to repetition of the same order in the experimental conditions.
Subjects are required to drive in the centre of the right lane. They can see their speed on the
speedometer and are free to choose the velocity they prefer (speed limit 130 km/h). The traffic
density used in the study is low. No other vehicles are travelling in the same lane as the test
driver. The low traffic density used is to ensure that vehicles interferences do not bias the
collection of data and driving performance.
4.4. Test drivers
Twenty five drivers (15 men and 10 women; mean age of 27 years, age range of 21-48 years)
take part to the experiments, selected among students and staff of Roma Tre University that had
no previous experience with driving simulator, had a driving license since at least 3 years and had
driven at least 2000 km on highway in the last year. All the test drivers are able to complete the
simulation without showing any problems during the driving. Also the questionnaires filled by the
drivers do not show any evidences for suggesting their exclusion from data elaboration.
4.5. Driving parameters
4.5.1. Longitudinal speed and lateral position
According to several literature studies, speed (e.g. [34, 35]) and lateral position (e.g. [36]) are
driving characteristics to be considered as useful indicators of safe driving. In this study the
authors record continuously these two parameters along each test, as illustrated in Figures 6 and 7,
that show respectively the speed and the lateral position profiles of a single test. The lateral
position is here considered as the distance between the driver’s vehicle centre of gravity and the
right side line of the right lane.
The average longitudinal speed within each tunnel is computed and compared with the
average longitudinal speed that the same driver adopted in the CS on the same road section. The
same procedure is applied for comparing the lateral position among the two scenarios.
9. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 66 -
Fig. 6 - Effects of tunnels on Driver 05 longitudinal speed profile
Fig. 7 - Effects of tunnels on Driver 05 lateral position profile
4.5.2. Pathologic Discomfort
The Pathologic Discomfort (PD) is an advanced indicator of simulation presented and
discussed in previous papers (for exhaustive explanation see [16-18]), where it was successfully
correlated with road accident rates.
The indicator is computed based on the local instantaneous variability of lateral acceleration,
that takes into account the corrections of trajectory that the driver assumes (Figure 8) to maintain
the curvature of the road axis.
If the driver corrects the vehicle’s trajectory more than what road curvature imposes, the road
is not self-explaining and, consequently, it can be unsafe.
Therefore the repeated local oscillations of lateral acceleration represent a violation to driver
expectancy. PD is here computed within each tunnel using Equation (1):
dx
x
xv
xaPD
Lx
x
t∫
=
=
−=
0
2
)(
)(
)(
ρ
(1)
where x is the road abscissa (longitudinal position or distance travelled by the driver), the at is
driver’s lateral acceleration (simulation output), v is the average speed of the driver along the
curve, ρ is the radius of the curve, and L is the length of the tunnel.
PD corresponds to the area between the diagram of driver’s lateral acceleration at and the
diagram of theoretical lateral acceleration v2
/ρ as illustrated in Figure 8.
10. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 67 -
Fig. 8 - Computation of PD
In this study, the authors evaluate the indicator within each tunnel in the TS and compared it
with the correspondent PD in the CS. The differences are statistically validated using the analysis
of variance (t-student tests).
Moreover in this study PD is also used to assess the approaching and exiting length of tunnel,
defined as the lengths of the road sections just outside tunnel, where driving performance could be
still influenced by the tunnel itself. For this aim it is followed a procedure proposed in a previous
paper [16].
Finally it is verified a correlation between PD computed inside the simulated tunnel and the
accident rate recorded inside the corresponding real one.
5. Results and Discussion
The post processing and elaboration of simulation data are here presented in two different
sections.
The first one concerns the analysis of driving performance, comparing simulation outcomes of
the two scenarios and verifying statistically their differences in order to evaluate the effects of
tunnel on driving performance; moreover the approaching and exiting lengths of road sections are
evaluated for each tunnel. The second part presents the results of the correlation between PD and
the accident rate recorded inside road tunnels.
5.1. Driving performance approaching, through and exiting road tunnels
Table 1 shows the results of the elaboration of data for each tunnel section in terms of lateral
position (LP), longitudinal speed (V), and Pathologic Discomfort (PD) averaged within the tunnel
section recorded both in the CS and in the TS. Data are averaged among the sample of tests
drivers.
Figure 9 shows the average driver’s speed and lateral position for each tunnel compared
between the CS and the TS. The light bars refer to the average longitudinal speed (at the top of the
figure) and lateral position (at the bottom of the figure) recorded in the CS. The dark bars show
the averages of the parameters inside the tunnels measured along the TS.
The differences of parameters recorded in the CS and in the TS for each tunnel are provided
too, as well as the results of the statistical analysis (t-student tests) performed for each tunnel and
parameter. Results are discussed in next paragraphs separately for each parameter.
11. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 68 -
Tab. 1 - Comparison of the average value of the parameters
Fig. 9 - CS vs TS: longitudinal speed and lateral position
5.1.1. Longitudinal speed
The average longitudinal speed (V) of each driver is computed within each road tunnel of the
TS and compared with the average longitudinal speed within the same road section of the CS.
Table 1 summarizes the results and provides the longitudinal speed of each tunnel averaged
among the sample of drivers. The standard deviation is provided, too. The average longitudinal
speed recorded inside each tunnel always reduces from the CS to the TS, on average, by 5.1 km/h,
with a minimum reduction of 3.2 km/h in tunnel 2 and a maximum reduction of 7.7 km/h in
tunnel 5.
In six tunnels out of eight (Figure 9) it is found that speed in TS is significantly lower than the
corresponding speed in the CS (p<0.05). However in tunnels 4 and 7 p=0.08, confirming the same
reduction of the other tunnels. Similar results are also found by previous researches [16, 18, 37].
12. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 69 -
5.1.2. Lateral position
The average lateral position (LP) of the driver’s vehicle with respect to the right line of the
main lane is computed within each road tunnel of the TS and compared with the average lateral
position within the same road section of the CS. Table 1 summarizes the results and provides the
lateral position of driver’s vehicle inside each tunnel averaged among the sample of drivers. The
standard deviation is provided, too.
In all the eight tunnels (see Figure 9) it is found that lateral position in the TS is significantly
higher than the corresponding lateral position in the CS (p<0.01), resulting for all the cases in a
lateral shift of driver’s vehicle towards the centre of the carriageway, probably caused by the
presence of the wall of the tunnel on the right side.
The average difference of lateral position among the eight tunnels is 29 cm with a minimum of
16 cm (tunnel 4) and a maximum of 48 cm (tunnel 8). Results confirm previous literature findings
on driver’s lateral position inside road tunnels [16, 18, 38].
5.1.3. Pathologic Discomfort
The average Pathologic Discomfort computed inside each tunnel along the TS is always lower
than the PD evaluated in the same road section along the CS as shown in the above part of Figure
11. This difference, averaged on the eight tunnels, is 5.0 m2
/s2
, with an average percentage
difference of 14.0%. Table 1 summarizes the results and provides the PD inside each tunnel
averaged among the sample of drivers. The standard deviation is provided, too.
In all the tunnels placed along a curve it is found that PD in the TS is significantly lower than
the corresponding PD in the CS (p<0.05). However also for the other three tunnels placed along a
straight it is confirmed that PD is lower in the TS than in the CS.
The lower PD inside road tunnel can be reasonably explained by a less need of drivers for
correcting their trajectories. It seems that the tunnel provides the road user with a kind of guide
for his trajectory, represented by the lateral walls of the tunnel itself. Results confirm previous
literature findings [16, 18] and are consistent with the outcomes of several studies (e.g. [1, 19, 20,
22, 23]) that demonstrated that the crash rate inside road tunnels is definitely lower than outside,
confirming the goodness of PD as a road safety indicator.
As discussed in the introduction and background sections, several literature studies [e.g. 19,
20, 22, 23] demonstrated the attention that should be given to entrance and exit zone of tunnels, as
their crash rates are often higher than inside the tunnel.
Length of approaching and exiting sections
In order to assess the road section length outside the tunnel (approaching and exiting)
influenced by the tunnel itself and according to a procedure presented and discussed in a previous
paper [16], the authors evaluate the integral function of the Pathologic Discomfort analysing the
increasing profile of the indicator PD along the longitudinal position travelling on TS and CS.
The integral function of PD (PD(x)) is provided in Equation (2):
dz
z
zv
zaxPD
x
t∫ −=
0
2
)(
)(
)()(
ρ
(2)
where x is the road abscissa (longitudinal position or distance travelled by the driver), at is the
driver’s acceleration (simulation output), v is the average speed of the driver on curve, ρ is the
radius of the curve.
13. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 70 -
The procedure used for evaluating the approaching length La (road section before the entrance
portal of the tunnel) and the exiting length Le (road section after the exit portal of the tunnel) is
based on the comparative analysis of the profiles of the function PD(x) along the distance
travelled in the CS and in the TS, as shown in an example for a single tunnel in Figure 10.
However such trends are representative for the most of the cases analysed.
The solid line represents the profile of the indicator for the TS while the dotted line stands for
the CS. It can be observed that the integral function PD(x) along the CS increases almost linearly
with the longitudinal position (distance travelled).
On the contrary PD(x) of the TS is characterized by three main trends. The first one is
observed in the approaching area, before the beginning of the tunnel: here the PD increases more
rapidly than in the case of the CS. The same trend is observed exiting the tunnel until the two
PD(x) functions return to coincide. Finally inside the tunnel section PD(x) in the TS increases
more slowly than in the CS, as it is already discussed earlier in the paper.
Fig. 10 - PD(x) for the assessment of the length of approaching (La) and exiting (Le) road section
Fig. 11 - CS vs TS: Pathologic Discomfort (above); length of approaching and exiting road section (below)
14. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 71 -
The results of this analysis are summarized in the bottom part of Figure 11, where the values
of La and Le, averaged among the sample of drivers, are reported for each tunnel. They are fully
consistent with previous findings [16], assessing a possible range for the length of these road
sections adjacent to tunnels portals: specifically both La and Le range between 79 meters and 272
meters. Future analysis will be carried out to generalize these results and provide useful guide
lines for the design and management of these external sections.
5.2. Road safety inside road tunnels
In previous studies [17, 18] PD was successfully correlated with road accident rates:
specifically it was noted that higher the PD higher the crash rate. The authors have here verified
such correlation inside road tunnels.
An in depth accidental analysis has been developed for each section of the present study. Time
history of accidents is extended over five years in order to collect a large number of collisions
(Na) to reach the best reliability [39]. The accident data are provided by the police collision
reports. Actually these data are recorded by police per each kilometre of highway with no
distinction about the carriageway (or tunnel tube) where the accident really occurred. However it
was possible to extract only those accidents located inside road tunnels.
Moreover for each section of the highways traffic flows data are collected in terms of Average
Annual Daily Traffic (AADT) for both directions of travel to compute the Accident Rate (AR)
inside each tunnel, defined as Na·108
/F, where F is the number of vehicles travelled on the road
section within the time period of investigation (AADT·365·5).
The average value of PD for each tunnel computed in the TS has been then correlated with the
AR recorded inside the same tunnel section within a period of 5 years of observation. The
correlation between AR and PD for road tunnel section is shown in the following Equation (3):
068.4242.00009.0 2
−⋅−⋅−= PDPDAR (3)
The results of this analysis are shown in Figure 12. It is clearly evident that very strong and
unexpected correlation has been obtained between the real accident rate AR and the indicator
derived from simulation. The graph shows that when PD increases the AR increase, confirming
previous results [17, 18].
Fig. 12 - Accident rate vs Pathologic Discomfort for the tunnel investigated
15. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 72 -
Data have been here interpolated using a quadratic curve of a continuous function and a linear
regression unlike the common safety models approach. In fact although the accidents are
obviously random, discrete and non-negative, the aim of this research concern the attribution of
an accidental level to a tunnel section of an infrastructure so as to qualify the road tunnel in terms
of safety.
Moreover, as PD and AR could be influenced by the length of the road section in which they
are computed (PD) or recorded (AR), they are divided by the length L (in km) of the
corresponding tunnel, obtaining the Pathologic Discomfort standardized (PDs) and the Accident
Rate standardized (ARs) every one kilometer of tunnel.
The Equation (4) reports the correlation so obtained:
413.5215.0001.0
2
−⋅−⋅−= sss PDPDAR (4)
Figure 13 shows the results of this analysis. It is confirmed that higher the PDs higher the ARs.
Also in this case the final interpolating curve is quadratic, following almost a linear trend.
Fig. 13 - Accident rate vs Pathologic Discomfort standardized by the length of road tunnel
Tab. 2 - Accident, traffic and PD data for the tunnel investigated
16. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 73 -
It is interesting to observe that the lower PD (or PDs) and AR (or ARs) values correspond to the
tunnels placed along a straight (number 2, 4 and 7), while tunnels placed along curve with smaller
radius (about 500 meters for tunnels 1, 5 and 6 and 300 meters for tunnels 3 and 8) are
characterized by higher PD and AR. Such promising results will be extended in future analysis
over a larger number of case studies in order to verify the influence of highway alignment or other
factors on tunnel safety.
In Table 2 all the data collected and computed for these analysis of correlation are
summarized.
Of course this unexpected and surprisingly promising result has to be validated for more
tunnels but for now it confirms the strict and evident correlation between lateral accelerations,
more specifically, the Pathologic Discomfort PD, and the accident rate AR.
6. Conclusions and future research
In this driving simulator study it is demonstrated that drivers change their way of driving when
they go through a tunnel. Specifically they move away laterally from right tunnel wall and slightly
slow down.
Moreover inside the tunnel the amount of trajectory corrections by the driver is definitely
lower, as the driver pay more attention when driving inside a tunnel.
The driving simulator data and specifically the Pathologic Discomfort allow to assess the
lengths of road sections just before the first tunnel portal and after the second one, where driving
performance are still influenced by the tunnel itself, confirming results of previous studies and
suggesting to direct the efforts of future research also in this topic.
Such belief is even stronger if we consider that actually there are no guidelines that could help
the designers to improve the safety of such road sections beyond the tunnel, that are characterized
by high accident rates.
Finally the results of the correlation analysis between simulation parameter and accident rate
of road tunnel are really promising, confirm the validity of the indicator for crash prediction also
inside road tunnels and seem to provide important indication about tunnel geometrical features,
such as the road alignment, that mostly affect tunnel safety.
Although the results presented in this paper are certainly promising, the limitations of driving
simulators should be addressed. The main limit of the simulation tests concerns the lower risk
perceived by drivers during the driving due to the possible occurrence of a virtual crash that does
not cause any kind of damages. Although the drivers are immersed in a simulated environment
that is very consistent with the real one, their behaviors can be different than that on a real road.
Here is the importance of verifying the validation of the advanced instruments of simulation.
Although CRISS simulator has been already successfully validated for different driving situations,
a validation study inside road tunnel is needed to enable its use for in depth analysis of the driving
performance influenced by road tunnels features as well as for proposing guide lines for tunnel
safety in terms of accident prevention, both inside and outside tunnels.
Despite such limitations the results reported here are promising and show the effectiveness of
the driving simulator for the study of the tunnel effects on driving performance and safety.
Future research will involve:
− performing validation of the simulation results against data from the real-world in the
tunnel conditions. For this aim a research project is ongoing. The research will allow to
compare the drivers’ speeds and lateral positions adopted on site (using an instrumented
car equipped with GPS that allows to collect the speed profile of each participant), along
17. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 74 -
the highway sections reproduced in this simulation study, with the driving parameters
recorded along the same highways in the simulated environment;
− evaluating the effectiveness of the parameters used in this study. Further studies with
varying traffic volume and geometric features of road tunnels are planned in order to
confirm the findings and to strengthen and generalize the results. Particularly the analysis
should be extended to a larger sample of tunnels (varying the cross section, the number
of lanes, the number of tubes, the length, the vertical and horizontal alignment inside the
tunnel, the type of construction and other tunnel features and facilities) and the
investigation of drivers performance should be enlarged among different traffic
configurations (in terms of traffic volume, uni-directional or bi-directional, percentage of
heavy vehicles) and in other road categories, in order to promote the use of driving
simulators among the road design community and provide practical applications in traffic
engineering;
− enlarging the sample of case studies for proposing a consistent and robust correlation
between Pathologic Discomfort and accident rates inside road tunnels. Actually such
correlation seems to be very promising for accident prediction inside road tunnel also
related to the geometry of the road alignment and could lead to identify those tunnel
features that mostly affect driving behavior and road safety inside tunnels.
References
1. Lemke, K., 2000. Road safety in tunnels. Transportation Research Record: Journal of the
Transportation Research Board 1740, 170-174;
2. Carvel, R., Marlair, G., 2005. A history of fire incidents in tunnels, in: A.N. Beard, R. Carvel (Eds.), The
Handbook of Tunnel Fire Safety, Thomas Telford Limited, London, 3-4;
3. European Parliament and Council, Directive 2004/54/EC on minimum safety requirements for tunnels in
the Trans-European Road Network. 2004. Available at:
http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CELEX:32004L0054:en:NOT;
4. Lidstrom, M., 1998. Using advanced driving simulator as design tool in road tunnel design.
Transportation Research Record: Journal of the Transportation Research Board 1615, 51-55;
5. Upchurch, J., Fisher, D., Carpenter, R.A., Dutta, A., 2002. Freeway guide sign design with driving
simulator for Central Artery–Tunnel: Boston, Massachusetts. Transportation Research Record: Journal
of the Transportation Research Board 1801, 9-17;
6. Lorentzen, T., Ito, Y., Tazawa, S., Goto, H., 2011. Virtual driving trials to assist road sign design: a case
study of Ohashi junction. Advances in Transportation Studies, an International journal 24, 77-84;
7. Akamatsu, M., Imachou, N., Sasaki, Y., Ushiro-Oka, H., Hamanaka, T., Onuki, M., 2003. Simulator
Study on Driver’s Behavior While Driving through a Tunnel in a Rolling Area. Driving Simulation
Conference, North America (DSC-NA 2003), National Advanced Driving Simulator;
8. 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 and Prevention 39, 69-
78;
9. Kircher, K., Ahlstrom, C., 2012. The impact of tunnel design and lighting on the performance of
attentive and visually distracted drivers. Accident Analysis and Prevention 47, 153-161;
10. Mashimo, H., 2002. State of the road tunnel safety technology in Japan. Tunnelling and Underground
Space Technology, 17 (2), 145-152;
11. World Road Association-PIARC Technical Committee C3.3 Road Tunnel Operation, 2008. Human
factors and road tunnel safety regarding users. International Standard Book Number 2-84060-218-0.
Available at http://www.piarc.org;
12. Yan, X., Abdel-Aty, M., Radwan, E., Wang, X., Chilakapati, P., 2008. Validating a Driving Simulator
Using Surrogate Safety Measures. Accident Analysis and Prevention 40, 274-288;
18. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 75 -
13. Calvi, A., De Blasiis, M.R., 2011. Driver Behavior on Acceleration Lanes: Driving Simulator Study.
Transportation Research Record: Journal of the Transportation Research Board 2248, 2011, 96-103;
14. Calvi, A., Benedetto, A., De Blasiis, M.R., 2012. A driving simulator study of driver performance on
deceleration lanes. Accident Analysis and Prevention 45, 2012, 195-203;
15. Bella, F., Calvi, A., 2013. Effects of simulated day and night driving on the speed differential in tangent-
curve transition: a pilot study using driving simulator. Traffic Injury Prevention 14, 4, 2013, 413–423;
16. Calvi, A., De Blasiis, M.R., Guattari, C., 2012. An Empirical Study of the Effects of Road Tunnel on
Driving Performance. Procedia Social and Behavioral Sciences 53, 2012, 1099-1109;
17. Calvi, A., D’Amico, F., 2006. Quality control of road project: identification and validation of a safety
indicator. Advances in Transportation Studies, an International journal 9, 47-66;
18. Calvi, A., 2010. Analysis of Driver’s Behaviour in Road Tunnels: a Driving Simulation Study. Progress
in Safety Science and Technology 8, 1892-1904;
19. Amundsen, F.H., Melvear, P., Ranes, G., 1997. An analysis on Traffic Accidents and Car Fires in Road
Tunnels. Norwegian Public Roads Administration, Report no. TTS 15;
20. Amundsen, F.H., Roald, P.O., Engebretsen, A., Ragnoy, A., 2005. Traffic Accidents in Norwegian
Subsea Tunnels. Norwegian Public Roads Administration, Report no. TTS.;
21. Amundsen, F.H., Ranes, G., 2000. Studies on Traffic accidents in Norwegian Road Tunnels. Tunnelling
and Underground Space Technology 15(1), 3-11;
22. Narisada, K., Yoseoikawa, K., 1974. Tunnel Entrance Lighting Effect of Fixation Point and other Factors
on the Determination of Requirements. Lighting Research and Technology 6, 9-18;
23. Verwey, W.B., 1995. Effects of Tunnel Entrances on Driver’s Physiological condition and performance.
Report TM 1995 C-19, Soesterberg, The Netherlands: TNO Human Factors Research Institute;
24. Bella, F., 2009. Can Driving Simulators Contribute to Solving Critical Issues in Geometric Design?
Transportation Research Record: Journal of the Transportation Research Board 2138, 120-126;
25. Martens, M.H., 2005. Human Factors Aspects in Tunnels: Tunnel User Behaviour and Tunnel Operators.
Deliverable 3.3 in the frame of the European UPTUN project;
26. Tornos, J., 1998. Driving behaviour in a real and a simulated road tunnel: a validation study. Accident
Analysis and Prevention 30, 497-503;
27. Hirata, T., Yai, T., Tagakawa, T., 2007. Development of the driving simulation system MOVIC-T4 and
its validation using field driving data. Tsinghua Science & Technology 12, 141-150;
28. Bella, F., 2005. Validation of a driving simulator for work zone design. Transportation Research
Record: Journal of the Transportation Research Board 1937, 136-144.
29. Bella, F., 2008. Driving Simulator for Speed Research on Two-Lane Rural Roads. Accident Analysis and
Prevention 40, 1078-1087;
30. Calvi, A., Benedetto, A., D’Amico, F., 2012. Effects of mobile telephone tasks on driving performance:
a driving simulator study. Advances in Transportation Studies, an International journal 26, 2012, 29-44;
31. Benedetto, A., Calvi, A., Messina, M., 2012. Potentialities of driving simulator for engineering
applications to Formula. Advances in Transportation Studies, an International journal, RSS2011 Special
Issue, 2012, 127-138;
32. Calvi, A., De Blasiis, M.R., Guattari, C., 2012. The effectiveness of Variable Message Signs
information: A driving simulation study. Procedia Social and Behavioral Sciences 53, 2012, 693-702;
33. Benedetto, A., Calvi, A., D’Amico, F., Zakowska, L., 2009. The effect of curve characteristics on
driving behavior: a driving simulator study. TRB 88th Annual Meeting Compendium of Papers DVD,
11-15 January 2009, Washington DC, USA;
34. Elvik, R., Christensen, P., Amundsen, A, 2004. Speed and Road Accidents: An Evaluation of the Power
Model. The Institute of Transport Economics (TOI), Oslo;
35. Aarts, L., van Schagen, I., 2006. Driving speed and the risk of road crashes: a review. Accident Analysis
and Prevention 38 (2), 215-224;
36. de Ridder, S.N., Thomson, R., van der Horst, A.R.A., 2006. Identify Envelope of Vehicle and Driver
Response Prior to Collisions. RISER Deliverable D04. Chalmers University of Technology, Göteborg,
Sweden;
19. Advances in Transportation Studies an international Journal Section B 30 (2013)
- 76 -
37. Theeuwes, J., Horst, A.R., Van Der, A, Hoekstra, W., Kaptein, N.A., 1995. A Simulator Study on Choice
and Driving Behaviour in the Second Beneluxtunnel. Phase I; the Effects of Tunnel Design and Expected
Likelihood of a Traffic Jam. Report TM 195 C-12, Soesterberg, The Netherlands: TNO Human Factors
Research Institute;
38. Blaauw, G. J., Horst, A.R., Van Der, A. Lateral Positioning Behaviour of Car Drivers Near Tunnels,
1982. Final Report IZF 1982 C-30. Soesterberg: The Netherlands: TNO Institute for Perception;
39. Benedetto, A., Benedetto, C., Calvi, A., De Blasiis, M.R. 2009. Model for the Individuation of Road
Dependent Collisions. Progress in Safety Science and Technology 7, 2009, 1933-1941.
View publication statsView publication stats