You are on page 1of 16

Accident Analysis and Prevention 176 (2022) 106792

Contents lists available at ScienceDirect

Accident Analysis and Prevention


journal homepage: www.elsevier.com/locate/aap

A computational study on the basis for a safe speed limit for bicycles on
shared paths considering the severity of pedestrian head injuries in
bicyclist-pedestrian collisions
Milan Paudel a, b, *, Fook Fah Yap a, b, Tantyana Binte Mohamed Rosli b, c, Kai Hou Tan b,
Hong Xu b, c, Nader Vahdati d, Haider Butt d, Oleg Shiryayev e
a
School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore
b
Transport Research Center @ NTU, Singapore
c
School of Social Sciences, Nanyang Technological University, Singapore
d
Department of Mechanical Engineering, Khalifa University of Science and Technology, Healthcare Engineering Innovation Center, SAN Campus, Abu Dhabi P.O. Box
127788, United Arab Emirates
e
Department of Mechanical Engineering, University of Alaska Anchorage, 3211 Providence Dr., ECB 301, Anchorage, AK 99508, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Bicyclists and pedestrians are two large vulnerable groups of road users. Many cities have allowed cyclists to
Bicycle accidents share space with pedestrians on footpaths and off-road paths to reduce conflict with motor vehicles. The risk of
Pedestrian accidents bicyclist-pedestrian accidents is also increasing accordingly. Therefore, there is a need to understand the factors
Pedestrian safety
that affect the risk of injury in such accidents, especially to pedestrians who are considered more vulnerable. This
Throw distance
HIC value
paper presents a detailed investigation of bicyclist-pedestrian collisions and possible injury outcomes. The study
Injury severity has considered five levels of collision speed ranging from 10 km/h to 30 km/h, three pedestrian profiles (adult,
Collision speed child, and elderly) differentiated by their weight and height, three bicycles with different masses, and five impact
Accident mechanism directions. The bicyclist-pedestrian collision simulations have been analyzed based on four metrics: throw dis­
tance, peak head velocity on impact with the ground, head injury criterion (HIC) value, and the probability of
severe head injury. For each simulation, the throw distance and peak head velocity on impact with the ground
are extracted. Following that, the HIC and the probability of severe head injury to pedestrians are computed. The
results show a significant effect of collision speed (p < 0.05) on all four metrics. The analysis has been further
extended to study the effect of height and weight profile, bicycle mass, and impact directions on bicyclist-
pedestrian collisions. According to the results, the impact directions largely influence the outcome of bicycle-
pedestrian collisions. In general, direct impacts on pedestrian body center have been found to yield higher
HIC values and probability of severe head injury to pedestrians than off-center impacts. Also, video analysis of
simulated collisions has suggested that the accident mechanism depends on weight and height profiles (corre­
lated with different age groups) and impact directions. Finally, recommendations have been proposed based on
the study, including a speed limit of not more than 12 km/h for bicyclists on narrow shared paths and footpaths
where risks of collisions with pedestrians are high. The results and analysis presented could be helpful for
developing legislation to minimize conflicts between bicyclists and pedestrians on shared paths and to reduce
potential injury to pedestrians.

1. Introduction part of urban transportation and recreation (Tan, 2020; Biggar and
Filho, 2019). Many countries like Singapore, Netherland, U.K, Germany,
The benefits of walking and cycling to health and the environment and other densely populated cities like Victoria (Australia) and Seattle
are well recognized (Appleby and Webster, 2016; Komanoff, 1993; (USA) are constantly redeveloping their transportation infrastructure to
Simms and Wood, 2009). Active mobility is also becoming an integral promote healthier lifestyles through walking and bicycling (Appleby

* Corresponding author at: School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore.
E-mail address: milan002@e.ntu.edu.sg (M. Paudel).

https://doi.org/10.1016/j.aap.2022.106792
Received 8 March 2022; Received in revised form 25 July 2022; Accepted 2 August 2022
Available online 8 August 2022
0001-4575/© 2022 Elsevier Ltd. All rights reserved.
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

and Webster, 2016; Laing, 2020; LTA Singapore, 2018; Urbanczyk, Macmichael, 2020; Colon, 2019). Moreover, the actual number of
2010). However, with a considerable difference in speed and mass bicycle-pedestrian collision could be even more than the published data
compared to motor vehicles, pedestrians and bicyclists are two large as the underreporting of collision has been found to be quite prevalent
groups of vulnerable road users (Chong, 2010; SWOV, 2012). With the for bicycle collision (Gildea et al., 2021).
increasing walking and bicycling population, there is a greater proba­ Tuckel et al. (Tuckel et al., 2014) reported 35.4 % and 37.5 % of total
bility of conflicts and accidents between pedestrians, bicyclists, and pedestrian injuries in a bicyclist-pedestrian collision were head and neck
different modes of transportation including motor vehicle users injuries in New York and California, respectively. Head and neck injuries
(Haworth et al., 2014; Kiyota et al., 2000; Grzebieta et al., 2011). were reported to be followed by injuries to the shoulder and upper ex­
Therefore, there is an urgent need to develop infrastructures that pro­ tremity (Tuckel et al., 2014). Interestingly, it has been found that the
vide safe use of shared spaces for walking and bicycling (O’Hern and mechanism of injuries in a bicyclist-pedestrian collision is quite different
Oxley, 2019; Stevenson, 2016). from motor vehicle–pedestrian collisions (De Rome, 2014). In a motor
One of the proposed solutions to reduce the bicyclist-pedestrian vehicle–pedestrian collision, both the primary contact with the vehicle
conflicts is to separate the traffic flow that differs in speed, direction, and secondary contact with the ground could result in serious injuries to
and mass (Wegman and Aarts, 2006) by categorically defining different pedestrians (Shang, 2018). It has been reported that severity of the pe­
paths like footpaths, shared paths, cycling paths, and dedicated paths for destrian’s injury is higher during primary impact than during secondary
pedestrians and cyclists. However, in small and densely populated cities, contact for collisions between motor vehicles and pedestrians (De Rome,
shared paths are often recommended (Grzebieta et al., 2011). Consid­ 2014; Short et al., 2007). On the other hand, for a bicyclist-pedestrian
ering the vulnerability of the bicyclist on the road with motor vehicles, collision, severe injuries occur due to the secondary impact when the
both bicyclists and pedestrians are aften allowed on shared the path. pedestrian head strikes the ground (Sikic, 2009). In addition, the post-
Allowing bicyclists on shared paths and footpaths reduces the in­ crash entanglement between the pedestrian lying on the ground, fall­
teractions between the bicyclist and motor vehicles, but at the same ing bicyclist, and the bicycle itself could increase the severity of the
time, it increases the interactions between bicyclists and pedestrians. injury (Muggenthaler, 2017).
There has been a growing debate about pedestrians’ and bicyclists’ Most of the research is focused on pedestrians’ and bicyclists’ acci­
safety on shared paths (Shinkle, 2012) and pedestrians’ concerns about dents with motor vehicles. Nonetheless, research so far has identified
the risk or threats they perceive from cyclists (Shaw et al., 2012; Mon­ various methods and metrics to study accidents and the severity of in­
sere et al., 2017). The conflicts and accidents between bicyclists and juries. Some popular methods include available data from hospitals or
pedestrians have been reported in some studies (Haworth et al., 2014; agencies to analyze the prevalence of the accidents (Fergus, 2019; Air­
O’Hern and Oxley, 2019; Haworth and Schramm, 2011; Tuckel et al., aksinen et al., 2010; Charters et al., 2018; Chong, 2018), usage of
2014). Pedestrians are likely to sustain more severe injuries in bicyclist- headforms or dummies of pedestrians to analyze the injury severity and
pedestrian conflicts and accidents due to the speed differential between mechanisms (Xu, 2018; Matsui and Oikawa, 2018), and usage of soft­
bicyclists and pedestrians (O’Hern and Oxley, 2019). In fact, researchers ware to reconstruct the accidents (Muggenthaler, 2017; Wang, 2020;
have reported that pedestrians are a more vulnerable group of people Meng, 2017). The severity of the injury has been found to be a function
than bicyclists (Appleby and Webster, 2016; Graw and König, 2002) and of vehicle type, vehicle speed, vehicle geometry, and road surface
have the danger of suffering from severe injuries leading to death in (Ravani et al., 1981). Furthermore, various metrics like throw distance,
pedestrian-bicyclist collisions (Graw and König, 2002; Pucher and head injury criterion (HIC), and various severity indices such as
Dijkstra, 2003). Abbreviated Injury Score (AIS), Injury severity score (ISS), maximum
Although the number of pedestrian’s fatalities in bicyclist and abbreviated injury score (MAIS) have been used to study the injury se­
pedestrian collisions are relatively low compared to the pedestrians- verities (Short et al., 2007; Baker, 1974; Henn, 1998; Mariotti, 2019;
motor vehicle accidents and bicyclist-motor vehicle accidents (Chong, Mehmood, 2019).
2010; Graw and König, 2002; NYC, 2019), the possibilities for non-fatal Despite increasing concern from the pedestrians regarding the
severe injuries in bicyclist-pedestrian accidents are reported to be vulnerability they perceive from sharing the same space with bicyclists,
similar as compared to the accidents involving motor vehicles (Chong, there are very limited published articles on bicyclist-pedestrian acci­
2010). Grzebieta et al. (Grzebieta et al., 2011) highlighted that the dents data, injury mechanisms, and pedestrians’ injury severity. With
difference in kinetic energy between a bicycle at 30 km/h and a the increasing population of pedestrians and bicyclists sharing the
pedestrian at 5 km/h is similar to a car traveling at 60 km/hr and a pathways, there is a current need to reduce the bicyclist-pedestrian
bicycle at 30 km/h. In addition, research has shown that a large number conflicts and explore solutions to decrease the possibility of severe in­
of bicyclist-pedestrian accidents occur on shared paths and footpaths. juries to the pedestrians. Furthermore, in a context where both bicyclists
Haworth et al. (Haworth et al., 2014; Haworth and Schramm, 2011) and pedestrians are allowed on shared paths and footpaths (Speed
have reported that about 68 % of bicycle path accidents are bicyclist- management on shared paths, 2014; Insider Guides, 2020); it is rec­
pedestrian crashes. Rome et al. (De Rome, 2014) reported that the ma­ ommended that the shifting of bicyclists to shared space along with
jority of bicycle accidents happen in the cycle lanes, shared paths, and pedestrians needs to be approached with caution (Chong, 2010). In fact,
footpaths (60 % of all bicycle related accidents). Furthermore, about the most concerning topic currently is the maximum allowed riding
10.7 % of pedestrians have been found to be involved in all bicycle-only speed of bicyclists on footpaths and shared paths. Therefore, this paper
crashes, and 20 % of those have occurred on shared paths (De Rome, aims.
2014). Another research in Melbourne reported that majorities of the The main aim of this paper is to establish a scientific rationale for
cyclist-pedestrian collisions happen at midblock locations (53.5 %) and setting a cycling speed limit on crowded shared paths by relating the
intersections (45.8 %) (Oxley and O’Hern, 2018). In addition, they also severity of pedestrian head injuries to the speed of cyclist-pedestrian
reported that 35.4 % of pedestrians sustained serious or fatal injuries, collisions. We have formulated a new method of relating the bicycle-
which is almost 5 times higher compared to the bicyclist sustaining in­ pedestrian collision speed to the probability of severe head injury to
juries that required hospital treatments. In Sweden, about 11 % of bi­ the pedestrian. The method allows us to recommend a speed limit for
cycle collisions involve either pedestrians or another bicyclist (Ohlin bicyclists on shared paths. In addition to the bicycling speed, the paper
et al., 2019). Cecile (Nseya, 2018) reported that out of total bicycle also presents detailed analysis of the accident mechanisms and severity
accidents, around 12 % are bicyclist-pedestrian accidents during of injury in bicyclist-pedestrian collisions considering different age
2010–2016 in Stockholm. Similarly, many encounters between bi­ groups, impact directions, and bicycle weight. The throw distance, head
cyclists and pedestrians, which led to severe injuries and deaths, can also impact velocity on the ground, HIC, and the probability of severe head
be found published in media worldwide (Alkhatib, 2020; Chan, 2019; injury are used as indices to analyze the bicyclist-pedestrian impact. The

2
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

analysis could be helpful for governments or concerning authorities to 2.4. Software


make policies to improve pedestrian safety in shared space like footpaths
and shared paths. Virtual Crash (Virtual Crash, 2020) software has been used as a tool
to simulate the collision between bicyclists and pedestrians in different
2. Methods scenarios. Virtual Crash is a well-known impulse-moment based acci­
dent reconstruction software that has been widely used in motor vehicle
In this computational analysis, the bicyclist’s speed has been varied accident research as well as by law enforcement authorities in Europe
from 10 km/h to 30 km/h in increments of 5 km/h. The study has and America. The results from the Virtual Crash have been successfully
included three pedestrian profiles, three bicycle models, five bicycle- validated against international references such as Research Input for
pedestrian impact directions for each speed. The rider has been Computer Simulation of Automobile Collisions (RICSAC) and Japan
modeled as an adult of 75 kg weight. Furthermore, a plain concrete Automobile Research Institute (JARI), as well as other staged experi­
surfaced shared path has been modeled according to the 50 British ments, and have been found to have good agreement with experimental
Pendulum Number (BPN) standard for the simulation. results (Becker et al., 2015; Sovreski, et al., 2017; Hoxha et al., 2017;
Virtual CRASH. [cited, 2021; Coufal and Semeia, 2014). The software is
based on the Kudlich-Slibar rigid body impulse model and uses mixed
2.1. Pedestrian profiles
Euler and Runge-Kutta numerical integration methods (Virtual CRASH
LLC, 2020). Furthermore, different CAD models of vehicles, pedestrians,
Three pedestrian profiles based on height and mass have been
terrains, and obstacles can be imported for creating accident scenes.
selected for this study. The mass and height of pedestrian’s profiles
correspond to children (age: 7–12 years old), adults (age: 18–30 years
2.4.1. Human model
old), and the elderly (age: ≥ 65 years old) based on population profile of
Virtual Crash software uses 13 ellipsoid objects connected by joints
Singapore (Ministry of Health, 2013). Table 1 shows the population
to create a ragdoll model of riders and pedestrians where the shape, size,
mean of height and mass of each pedestrian profile. Although pedestrian
and weight of each body part are adjusted according to the body weight
profiles are chosen based on the age groups, it should be noted that the
and height. The ragdoll model with elliptical objects and corresponding
details of the biomechanical properties, like human tissue properties,
human model with an added layer of the exterior is shown in Fig. 2. The
have not been considered. This is because the main aim here is to un­
software allows for modification of mechanical properties like joint
derstand the influence of pedestrian height and mass on accident met­
stiffness, damping, and postures to create a realistic human model. The
rics. However, parameters related to the body joints, body-to-ground
human models were adjusted to replicate the bicyclists’ and pedestrians’
contacts, and coefficient of restitution have been considered and are
postures. The joints are defined to closely represent the human body
discussed in section 2.4.2.
movement degrees of freedom and constraints. For example, the neck
can snap forward and backward, whereas the knees and elbows cannot
2.2. Bicycle models bend backward. In addition, the software’s default coefficient of resti­
tution value of 0.1 for the human body is used in the simulation.
Three different bicycles with different masses have been considered
to represent a wide range of bicycles. The lightest bicycle has been 2.4.2. Accident scenarios
modeled to be 5.5 kg, followed by 17.75 kg and 30 kg bicycles and the The simulation scenarios consist of a bicycle, a bicyclist, and a
corresponding inertial properties of the bicycle models have also been pedestrian on a level road. In all simulations, the pedestrian is kept at
considered. Moreover, the height and the width of the handlebar have rest while the bicyclist moves at its respective speeds. No brake has been
been modeled to be 1.06 m and 0.58 m, respectively. applied during the collision. Applying these constraints makes it easy to
maintain a constant relative speed between the pedestrian and the
2.3. Impact directions bicyclist. Since we are trying to analyze the worst-case scenarios, the
simulation is designed in such a way that the pedestrian collapses when
Five impact directions have been considered in this study. The po­ crashed into by the bicyclist. In this way, we are trying to analyze the
sitions are 0◦ , 90◦ , 180◦ , 30 cm off-center at 0◦ , and 30 cm off-center at maximum speed with which the head will impact the ground resulting in
180◦ , as illustrated in Fig. 1. Here, the directions are defined from the maximum injury severity. This was achieved by choosing low values of
pedestrian reference. The 0◦ represents the front of the pedestrian. The spring stiffness and damping for human body joints. The parameter
impact directions have been varied to study if the kinematics of values relating to human and vehicle collisions used in the simulation
bicyclist-pedestrian collisions and severity of the injuries would differ in software are listed in Table 2.
different scenarios such as (i) collision between pedestrian body center It should be noted that the values of the coefficient of friction for
and bicycle front traveling in the opposite direction (0◦ ), (ii) same di­ pedestrian-ground, wheel-road adhesion coefficient, vehicle frame-
rection (180◦ ), (iii) collision at the junction (90◦ ) and (iv) collision from ground coefficient of friction, and handlebar height are selected
bicycle handlebar to pedestrian (30 cm off-center at 0◦ and 180◦ ). similar to the values reported in previously published articles (Batista,
2008; Leng, 2021; Lin et al., 2012; Paudel and Yap, 2020; Wicher, 2016;
Wood and Simms, 2000; Paudel, 2019).
Table 1
An example of the simulation scenario is shown in Fig. 3 (a). For each
Height and mass parameters for different pedestrian profiles based on the pop­
ulation mean of Singapore.
simulation, the throw distance and peak head velocity before impacting
the ground can be extracted, as shown in Fig. 3 (b). The throw distance is
Profiles Resemblence Height Mass Body Width COG height
defined as the distance over which the pedestrian’s head traveled from
(m) (kg) (m) (m)
the initial impact (Short et al., 2007). Similarly, peak head impact ve­
Profile Child 1.367 34.70 0.365 0.862 locity represents the maximum head velocity before it impacts the
1
Profile Adult 1.730 69.20 0.460 1.094
ground. Thereafter, head injury criterion (HIC) and abbreviated injury
2 scale (AIS) have been calculated using standard mathematical formulae.
Profile Elderly 1.647 63.20 0.440 1.035
3 2.5. Mathematical formulation
*center of gravity (COG) is based on the human model of the pedestrians in the
software. HIC estimates the severity of the head injuries by correlating the

3
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Fig. 1. Bicyclist-pedestrian impact directions.

33.93 based on the experimental HIC values reported in reference (Fanta


et al., 2012). In addition, calculated values of HIC can be used to
compute the probability of severe injuries using the abbreviated injury
scale AIS. The probability of severe head injury in the AIS scale can be
computed using Eq. (3).

Probability of severe head injuryn = 1/[1 + exp ((4.9 + 200/HIC) − 0.00351


× HIC)]
(3)
The overall analysis approarch we have adopted in this analysis is
presented in Fig. 4. The proposed method uses the maximum head to
ground impact velocity from crash simulations to estimate the HIC
values. The HIC values are then used to estimate the probability of se­
Fig. 2. (a) Ragdoll model with elliptical objects to represent human and (b)
vere head injury based on the Abbreviated Injury Scale (AIS 4). Thus,
human model after adding realistic-looking exterior to ragdoll model.
this establishes a relationship between bicyclist’s collision speed and
probability of severe head injury.
Table 2
Human and vehicle-related parameters used in Virtual Crash simulation. 2.6. Simulation results vs Experimental results
Parameters Values Parameters Values

Stiffness of joints 2 N/m Height handlebar (Paudel 1.06


In order to ensure that the simulation results represent real-life
and Yap, 2020; Paudel, m scenarios, the results from Virtual Crash are compared with the exper­
2019) imental data and models from previously published articles and reports
Damping of joints 0.1 Ns/ Adhesion (wheel-road) ( 0.78 (Hajiaghamemar, 2015; Moure-Guardiola, 2020; Saczalski, 1976; Urban
m Leng, 2021)
Turf Solutions, 2016; Zhang et al., 2009). The data were collected from
Coefficient of friction(human- 0.8 Coefficient of friction 0.5
ground (Wood and Simms, (vehicle-ground) (Batista, two different sources: (i) published articles based on human dummies
2000; Han and Brach, 2001) 2008; Lin et al., 2012) and head-forms, and (ii) published articles based on real accidents and
proposed mathematical models for accident reconstructions. Due to the
lack of bicyclist-pedestrians crash data in published articles, the human
head’s effective deceleration during impact (Hutchinson et al., 1998). crash data from relevant experiments using the full human dummy and
The HIC is mathematically defined as: headforms drop tests were also considered for comparison. For example,
⎡ (∫ t2 )2.5 ⎤ we dropped an instrumented adult headform from different heights that
a(t)dt

HIC = max (t2 − t1 ).
t1 ⎦ (1) resulted in different head-ground impact velocities and HIC values.
t2 − t1 Besides the headform drop tests conducted at NTU, other results from
similar experiments which studied head injury in a standing fall
where, t2 , and t1 represent the time interval that maximizes the average (Hajiaghamemar, 2015) or headform blunt impacts (Moure-Guardiola,
deceleration. In general, a higher HIC index is associated with a more 2020) have been considered. Some examples of the relevant experiments
severe head injury. are shown in Fig. 5.
As the secondary impact between the pedestrian head and ground is Fig. 6 (a) shows the comparison of peak head impact velocity and
known to be the major cause of severe injuries (Sikic, 2009), the severity corresponding Head Injury Criterion (HIC) values obtained using Virtual
of the injury depends on the amount of energy interchanged between the Crash simulation results and previously published accident, dummy
head and the ground during the collision (Khorasani-Zavareh, 2013). Eq. collision, and head-form experimental data. On the other hand, owing to
(1) could be further modified, considering a simple mass-spring model of the fact that there are not much published data on pedestrian throw
impact (Gao and Wampler, 2009). The modified equation can be written distance and impact velocity for bicyclist-pedestrian collision, the re­
as: sults from Virtual Crash simulations were compared with the various
mathematical models, which are validated against the crash tests and
HIC = K*v2.5 (2)
1 real-world accident-related data (Toor and Araszewski, 2003; Searle,
Eq. (2) relates the impact velocity (v1) to HIC with proportionality 1993; Cheng, 2015). Fig. 6 (b) shows the comparison between impact
constant K. The proportionality constant value was estimated to be velocity and throw distance obtained from Virtual Crash and some of the
widely used mathematical models. As the mathematical models are

4
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Fig. 3. (a) Accident simulation using Virtual Crash software (b) throw distance and head velocity after collision extracted from software.

mostly based on direct head-on collisions between vehicles and pedes­ has a strong influence on accident metrics like the throw distance, peak
trians, Virtual Crash results for 0◦ center impact and 90◦ center impact head velocity, HIC value, and probability of severe head injuries. Fig. 6
between the bicyclist and adult pedestrians have been compared against shows the overall effect of the speed on these accident metrics. It should
the mathematical model. It can be seen from the figures that there is a be noted that for each speed, these metrics are computed based on the
good agreement between the Virtual Crash results and data from pre­ three levels of pedestrian, three levels of bicycle mass, and five levels of
viously published articles. The peak impact velocity and corresponding impact directions.
HIC values are similar to the real-world test data. Similarly, most of the Fig. 7 (a) shows that the mean throw distance generally increases
results for pedestrian throw distance and impact velocity obtained from with the vehicle’s speed. The throw distance represents the final resting
Virtual Crash simulations are within the minimum and maximum range position of the pedestrian’s head after the collision. The results have
specified by Searle’s model (Searle, 1993), and closer minimum value of shown an increase in pedestrian throw distance from 2.82 m to 4.02 m
Wood’s model (Wood and Simms, 2000) and Toor model (Toor and when bicycle speed upon collision increased from 10 km/h to 30 km/h.
Araszewski, 2003). The error bar represents a 95 % confidence interval (CI) for each colli­
sion speed. A single factor ANOVA test has suggested a significant effect
3. Results and discussion of collision speed on throw distance (F(df1, df2) = 14.25, p = 0.000).
Similarly, Fig. 7 (b), (c), and (d) show the significant effect of speed on
This section presents results obtained from a total of 225 simulations the pedestrian’s mean peak head velocity before ground impact (F(df1,
carried out in Virtual Crash software. This section first presents the df2) = 104.62, p = 0.000), mean HIC value (F(df1, df2) = 82.33, p =
parametric analysis of the bicyclist-pedestrian accident using major 0.000), and mean probability of severe head injury (F(df1, df2) = 77.55,
accident-related metrics such as the throw distance, head-impact ve­ p = 0.000).
locity, HIC value, and probability of severe head injury. After that, a When the bicyclist’s speed increased from 10 km/h to 30 km/h, the
detailed analysis of the accident mechanism has been presented based mean peak head-ground impact velocity increased from 4.3 m/s to 7.19
on the simulation video. The simulation video analysis tries to correlate m/s, the mean HIC value increased from 1349 to 4913, and the mean
the accident mechanisms with the results obtained for different pa­ probability of head injury increased from 42 % to 99 %, respectively. It
rameters and accident metrics. It should be noted that the results pre­ is also interesting to notice the sudden increase in HIC value when
sented in this section are the combined results that include multiple bicyclist-pedestrian collision speed is above 20 km/h. When the bicyclist
levels of different parameters. speed is increased, the momentum and kinetic energy (K.E.) of the
bicycle-rider system also increases significantly. As the bicyclist with
3.1. Parametric analysis such a high K.E. and momentum collides with a stationary or slow
walking pedestrian, much of the momentum and energy is transferred to
This section presents a detailed study on the effect of four major the pedestrian, leading to higher head-to-ground impact velocity for
parameters: collision speeds, age profiles, bicycle masses, and impact pedestrians after the collision. As a result, the HIC value and probability
directions on pedestrians during the collision on the accident metrics. of getting severe head injury increase significantly with the bicyclist’s
speed.
3.1.1. Effect of speed It is interesting to note that there is a steep increase in the probability
Speed is directly related to the momentum and the kinetic energy (K. of severe head injury when the pre-collision bicycle speed increases from
E) of the bicyclist. Therefore, speed is a crucial factor in accidents and 10 km/h to 15 km/h (Fig. 7d). The observation can be explained by the

5
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Fig. 4. (a) Flow chart for calculating the probability of severe head injury (b) methodology we adopted for relating collision speed with probability of severe
head injury.

3.1.2. Effect of pedestrian profiles


All age groups use a shared path. It should be noted that the models
of three pedestrian profiles that have been included in this study only
consider the heights and masses of different age groups. The results
evaluate the severity of the injuries on pedestrians’ having different
heights and mass. Such analysis based on age groups could help legis­
lators, especially to decide whether to allow adult bicyclists on
footpaths.
Fig. 8 shows the mean throw distance, mean peak head-ground
impact velocity, mean HIC value, and mean probability of head in­
juries for each age group. The results have not shown a significant effect
on throw distance (F(df1, df2) = 1.94, p = 0.306), peak head-ground
impact velocity of pedestrian (F(df1, df2) = 0.62, p = 0.539), HIC
value (F(df1, df2) = 0.91, p = 0.404), and the probability of severe head
injuries (F(df1, df2) = 0.23, p = 0.792). However, it should be noted that
the results present a combined effect of multiple levels of different pa­
rameters. The mean value presented in Fig. 8 includes 3 levels of bicycle
Fig. 5. Some examples of head impact studies (a) doing a human head form masses, 5 levels of speeds, and 5 levels of impact directions.
drop test at NTU (b) using human dummies (Hajiaghamemar, 2015).
The analysis has been extended to obtain further insight into the
effect of age profiles on the bicyclist-pedestrian accidents for different
“S” shaped AIS curve shown in Fig. 4. There is a sharp increase in AIS collision speeds. Although overall results for age profiles have not sug­
score for severe head injury when the HIC value increases from 1200 to gested a significant effect on performance metrics, Fig. 9 suggests that
1800. This range corresponds to a collision speed from 10 km/h to 15 children have generally higher head-ground impact velocity, HIC
km/h. Therefore, the speed range between 10 km/h and 15 km/h is values, and the probability of severe injury with increasing collision
crucial for the threshold safe speed limit. speed. However, somewhat counterintuitively, adult pedestrians have
slightly higher mean throw distances compared to the elderly and the

6
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Fig. 6. Comparison between Virtual Crash results with published experimental data (a) impact velocity vs HIC (b) pedestrian throw distance vs impact velocity.

Fig. 7. Effect of collision speed on accidents metrics. (a) mean throw distance (b) mean peak head-ground impact velocity (c) mean HIC value, and (d) mean
probability of severe head injury (45 simulations per collision speed).

child. The results seem to contradict Short et al. (Short et al., 2007) that bicyclist’s speeds ranging from 10 km/h to 30 km/h.
higher mass generally has lower throw distance than lighter mass. The overall results have suggested that the bicycle mass has a sig­
However, further analysis has shown that the throw distance also de­ nificant effect (F(df1, df2) = 3.66, p = 0.027) on the pedestrian throw
pends on the impact direction. For the 0◦ impact angle as of Short et al. distance. As the bicycle mass increased from 5.5 kg to 30 kg, the mean
(Short et al., 2007), the child has a longer throw distance than elderly throw distance increased from 3.20 m to 3.63 m. However, the differ­
and adult pedestrians. Although the simulation results have shown a ence in bicycle mass has not shown any significant effect on overall
slightly higher probability of severity of the head injury on adults mean peak head-ground impact velocity (F(df1, df2) = 0.22, p = 0.80),
compared to the elderly, the other aging-related factors such as gradual mean HIC value (F(df1, df2) = 0.30, p = 0.74), and mean probability of
losses in hearing, vision, reflexes, and flexibility put elderly pedestrians severe head injury (F(df1, df2) = 0.25, p = 0.78).
at risk for pedestrian collisions (Chakravarthy et al., 2007) and severity Although no significant effect has been observed from the overall
compared to adults. mean analysis for different mass bicycles, the values are slightly higher
for higher bicycle mass. This is further illustrated in Fig. 11. For each
3.1.3. Effect of bicycle mass collision speed, the throw distance increased with increasing bicycle
The mass of the bicycle varies with the type of bicycle. Generally, mass. Furthermore, the difference in throw distance for each bicycle
racing bicycles are the lightest, weighing about 5–6 kg, most shared mass has become more distinct with the increased bicycle speed.
bicycles weigh up to 20 kg, and electric-assisted bicycles (E-bike) are the Notably, the results show that heavy bicycles such as E-bikes have a
heaviest, weighing up to 30 kg. Fig. 10 shows the overall effect of the higher probability of causing severe head injury compared to light­
bicycle mass on bicyclist-pedestrian collision. The results are based on weight bicycles.
the 3 levels of pedestrian profiles, 5 levels of impact directions, and 5

7
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Fig. 8. Overall effect of age profile on accident metrics: (a) mean throw distance (b) mean peak head-ground impact velocity, (c) mean HIC value, and (d) mean
probability of severe head injury (75 simulations per age profile).

3.1.4. Effect of impact directions before hitting the ground is higher for 90◦ impacts. The trend is the same
The impact direction in bicyclist-pedestrian collisions depends on the for each bicyclist speed except below 15 km/h.
direction of travel. When pedestrians and bicyclists traveling in the same Besides the abovementioned parameters, the authors would also like
direction, there is a high chance of 180◦ or 180◦ off collisions. Similarly, to highlight other factors such as environmental factors, human ability,
if the traveling direction is opposite, the collision may occur at 0◦ or vehicle’s dynamic performance that can directly or indirectly influence
0◦ off directions. Furthermore, there is a high chance of 90◦ collisions accidents and their outcomes. For example, there is a high chance of
during pedestrian crossings or bicycle turning. This section evaluates the accidents because of poor visibility and a slippery road surface.
nature of the impact and severity of the injuries due to different collision Furthermore, the capacity to endure an impact for age groups and
scenarios. Fig. 12 shows the overall mean results for each of 5 impact genders is not similar (European Commission, 2020), resulting in
directions. The results here presented are based on the 3 levels of bicycle different levels of severity for different pedestrian groups for the same
masses, 5 levels of speeds, and 3 levels of the pedestrians. impact force. In addition, poor eyesight, loss of hearing and reflex, and
The results show a significant effect of impact direction on throw delay in response could influence the occurrence of accidents (Chakra­
distance (F(df1, df2) = 42.22, p = 0.000), peak head-ground impact varthy et al., 2007). Similarly, the curious behavior of a child can in­
velocity (F(df1, df2) = 2.68, p = 0.033), and HIC (F(df1, df2) = 4.31, p fluence the occurrence of bicyclist-child accidents.
= 0.002). The mean throw distance of pedestrians after collision ranges It has been observed that a lower bicycle riding speed is associated
from 2.8 m to 4.21 m for the impact direction of 0◦ off-center and 180◦ , with a lower head to ground impact speed and a lower probability of
respectively. The mean peak head velocity before hitting the ground severe head injury. Thus, reducing the bicycle riding speed on the shared
ranges from 5.3 m/s to 6.15 m/s for 180◦ off-center and 90◦ impact path could be one approach to minimizing the probability of severe
directions, respectively. Similarly, the mean HIC value ranges between injury. However, the rideability and laterally unstable dynamics of these
2063 and 3018 for 180◦ off-center and 90◦ , respectively. However, no single-track vehicles at low speed should also be considered. It has been
significant effect on the probability of head injury (F(df1, df2) = 0.63, p found that single-track mobility devices like bicycles, especially with
= 0.64) has been observed. smaller wheels, are less stable, twitchy or wobbly in steering, and more
Fig. 13 further illustrates how the bicyclist-pedestrian collision difficult to control at speed below 15 km/h (Paudel and Yap, 2020). The
metrics vary with the bicyclist’s speed. It is evident from the figure that rideability consistently decreases as the speed decreases. Moreover,
0◦ and 180◦ impact have similar characteristics. Similarly, there is no reducing the riding speed limit to a very low speed could discourage
significant difference in results between the impact directions of 0◦ off- people from cycling. Therefore, the best practical solution for shared
center and 180◦ off-center. The 0◦ and 180◦ impact has resulted in a space is to find a compromise between the possibility of severe injuries
much further throw distance compared to the remaining impact angles. and rideability of these vehicles.
In contrast, 0◦ off-center and 180◦ off-center have the shortest throw It should also be noted that accident metrics calculated in this
distance. Although the impact direction at 90◦ has a significantly lower analysis are closer to the worst possible cases where a pedestrian head
mean throw distance than 0◦ and 180◦ , the mean peak head velocity impacts the road with a maximum velocity enduring the maximum

8
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Fig. 9. Effect of age profile on accidents metrics for different collision speed. (a) mean throw distance (b) mean peak head-ground impact velocity (c) mean HIC value
and (d) mean probability of severe head injury (15 simulations per age profile per collision speed).

Fig. 10. Effect of bicycle mass on accident metrics. (a) mean throw distance, (b) mean peak head-ground impact velocity, (c) mean HIC value, and (d) mean
probability of severe head injury (75 simulations per bicycle mass).

9
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Fig. 11. Effect of bicycle mass on accidents metrics for different collision speed. (a) mean throw distance (b) mean peak head-ground impact velocity (c) mean HIC
value, and (d) mean probability of severe head injury (15 simulations per bicycle mass per collision speed).

Fig. 12. Effect of impact direction on accident metrics. (a) mean throw distance, (b) mean peak head-ground impact velocity, (c) mean HIC value, and (d) mean
probability of severe head injury (45 simulations per impact direction).

injury severity. However, in real scenarios, when a person falls to the Because of this, the severity of the head injury, in most cases of the real
ground, they would often reflexively break a fall and avoid hitting their bicyclist-pedestrian collisions, will be lower than the worst possible
head on the ground by using their hands and stiffening up the body. cases computed from the software. Both the rideability of the bicycle at

10
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Fig. 13. Effect of impact directions on accidents metrics for different collision speed. (a) throw distance (b) head-ground impact velocity, (c) HIC value, and (d)
probability of severe head injury (9 simulations per bicycle mass per collision speed).

low speed and the potential severity of the injury should be considered head, respectively. Similar results were observed for e-scooter and
for selecting an appropriate cycling limit. pedestrian accidents using human dummies (Guidedogs, 2021; RCAR,
Considering both rideability and potential for head injury, a speed 2022) Also, these impact directions cause pedestrians to fall forward in
limit of not more than 12 km/h could be recommended. This is because the bicyclist’s travel direction. The kinetic energy transferred to the
the speed limit of 12 km/h is closer to the normal riding speed, and most pedestrian depends on the impact location and direction (Saulić et al.,
importantly, the probability of severe head injury in the worst possible 2020). The impacts at 0◦ and 180◦ represent direct impacts, which result
case is less than 50 %. Such a speed limit can be implemented through in more energy transfer between a bicyclist and pedestrian and is
legislation. Policymakers could consider imposing the speed limit on responsible for longer throw distances. Furthermore, simulation video
pedestrian footpaths and mixed-use shared paths with high pedestrian analysis has suggested a high chance of pedestrians being dragged for­
traffic. As a safety measure, speed barriers or speed regulating strips ward along with the bicyclist when there is a direct impact between
could be introduced on shared paths that will reduce cyclist-pedestrian pedestrian and bicyclist.
conflicts. Pedestrian crossings are the regions associated with higher In 0◦ off-center and 180◦ off-center impacts, pedestrians are observed
conflicts and risks of injury (Angel-Domenech, 2014). These regions also to be turning while falling on the side of the bicyclist. In off-center
have a higher probability of 90-degree collisions between bicyclist and bicyclist-pedestrian collisions, the bicycle handlebar hits the pedes­
pedestrians. As 90-degree collision has the highest probability of severe trian and exerts a force on either the left or right side of the body, see
head injury, junctions between the pedestrian road crossing and the Fig. 1. This force on either right or left side of the body is responsible for
shared paths or curved paths could be broadened to reduce risk of generating a torque around the pedestrian body center, which causes
collision. In addition, public education campaigns and signages to pedestrians to turn or rotate while falling. Similar results were reported
inform bicyclists about the speed limit could be introduced. Moreover, when motor vehicle collided with pedestrian at an oblique angle away
mobility devices with improved designs for better stability and handling from the pedestrian’s CG (Chen, 2015). Furthermore, as the pedestrians
could increase safety (Paudel et al., 2018, 2020; Paudel and Yap, 2020; are likely to fall on the side of the bicyclists (Short et al., 2007), there is
Paudel, 2019). less chance of pedestrians being dragged after colliding with a bicyclist.
Therefore, the throw distance for 0◦ off-center and 180◦ off-center
impact directions are shorter than 0◦ and 180◦ impact directions. The
3.2. Accident mechanisms collision mechanisms for different impacts are illustrated in Fig. 14.
Pedestrian height and weight are important factors that influences
This section discusses some of the key differences in accident the injury mechanism and outcome (Trusca et al., 2010). Our analysis
mechanisms for different pedestrian profiles, impact directions, and also suggested differences in accident mechanisms between adult and
speeds. The discussion tries to correlate the accident mechanisms with child pedestrians for different impact location and impact direction. For
the results presented in Section 3.1. The simulation video analysis has direct impacts at 0◦ and 180◦ impacts, children generally have a longer
provided interesting insights into the accident mechanism for different throw distance than adults. However, for 0◦ off-center and 180◦ off-
impact directions and age profiles. center impact directions child has a much shorter throw distance than
The impact direction has a strong influence on how a pedestrian’s adults. In an off-center collision, pedestrians generally fall on the side of
head collides with the ground. Usually, in 0◦ , 180◦ , and 90◦ impacts, the the bicyclist and have less drag. So, the throw distance in off-center
first head-to-ground impact occurs on the back, front, and side of the

11
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Fig. 14. Effect of impact directions on accident mechanism. (a) 0◦ impact, (b) 90◦ impact, and (c) 0◦ off-center impact.

impacts depends on the pedestrian’s height. Thus, shorter throw dis­ impact have higher values for mean peak head velocity, HIC, and the
tances for the child in off-center impacts can be attributed to the height probability of severe head injury. The results can be tributed to a com­
difference between a child and an adult. bination of two factors: (i) less drag and (ii) rider-pedestrian entangle­
Furthermore, the pedestrian falling mechanism also depends on the ment during an accident.
relative location of vehicle impact from the center of gravity of the When a bicyclist riding at a speed less than 15 km/h collides with a
pedestrian (Bhalla, 2002). When a bicyclist collides with a child at pedestrian, simulations have shown slightly higher drag for 90◦
0◦ off-center or 180◦ off-center position, the handlebar hits the child at compared to 0◦ off-center and 180◦ off-center, but less than 0◦ and 180◦
the chest level, which is significantly above the child’s center of gravity impacts, which explains the higher peak head velocity, HIC, and the
(COG). In this case, the impact force causes the child to fall directly probability of severe injury for off-center positions. However, the
towards the ground. In contrast, adults have their COG at about the same simulation video analysis has shown a very high possibility of rider-
height as the bicycle handlebar. Thus, the impact force on an adult can pedestrian entanglement during 90◦ impacts causing the rider to land
result in a slight translational effect before falling onto the ground, as on top of pedestrians, as illustrated in Fig. 16. The bicyclist riding at a
illustrated in Fig. 15. Therefore, an adult pedestrian has a greater throw higher speed has a higher momentum. When bicyclist with higher mo­
distance compared to a child who is shorter. These observations explain mentum collides with pedestrians, most of the momentum is absorbed
the higher overall mean throw distance for the adult as seen in Fig. 8 (a). by pedestrians. In addition, the entanglement between riders having
It is interesting to note a higher mean peak head-ground impact higher momentum and pedestrians further increases the head velocity.
velocity, HIC value, and the probability of severe injury in 90◦ impact The situation further worsens if the rider lands on top of pedestrians
directions compared to other impact directions except collision speed causing higher HIC and probability of severe injuries. Thus, the com­
less than 15 km/h. Below 15 km/h, 0◦ off-center, and 180◦ off-center bined effect of less drag and rider pedestrian entanglement is responsible

Fig. 15. Difference in accident mechanism for child and adult pedestrians at 180◦ off-center impact.

12
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Fig. 16. Illustration of post-accident entanglement between pedestrian and bicyclist during 90◦ collision.

for higher mean values of peak head-ground impact velocity, HIC, and there is a higher possibility of body-to-body collision between pedes­
injury severity above 15 km/h collision speed, especially for 90◦ impact trian and cyclist as result we were able to observe possibility of rider-
direction. pedestrian entanglement after the collision.
Although the focus has been given to collision mechanisms causing On the other hand, the contact parameters such as friction between
head injuries, the simulations have also provided insights into the rubber wheels, bike frame and the road surface, and road surface and
possible injuries to other body parts. In frontal collisions (0◦ and 0◦ off- human were selected based on the limited information available. The
center impacts), the neck snaps forward after the impact, whereas in ground contact characteristics are different for different road surface
backward collisions (180◦ and 180◦ off-center impacts), the head has the type and clothing. Therefore, the results should be considered together
tendency to snap backward. The snapping of the head backward can with the limitations and uncertainty associated with the selected pa­
cause serious injury to the neck (Kappagantu et al., 2021). Moreover, rameters as not much experimental data is currently available for
0◦ and 90◦ impacts can cause knee injuries as it cannot bend backward. bicycle-pedestrian collision. We would also like to acknowledge that the
Therefore, the collision between bicycle and pedestrian often leads to simulations could be further improved if more experimental data related
multiple injuries to the pedestrians (O’Hern and Oxley, 2019). to pedestrian-bicycle collisions are available in the future.

4. Conclusions
3.3. Limitations
Allowing bicyclists to share footpaths with pedestrians needs to be
The study has some limitations given the challenging nature of considered with care. This paper has investigated the bicyclist-
studying vehicle collisions involving human subjects. The authors would pedestrian accidents considering four different parameters: collision
like to highlight some limitations of this study. The first limitation is the speeds, pedestrian profiles, bicycle masses, and impact directions. In this
lack of actual bicycle-pedestrian collision data. There has not been much study, five levels of collision speeds have been considered ranging from
data from published articles. Therefore, we have also considered rele­ 10 km/h to 30 km/h. Similarly, three pedestrian profiles (children,
vant data from using full human dummy falls and headform drop tests as adults, and elderly), three bicycles with different weight (5.5, 17.5, and
references to compare with the results from Virtual Crash. 30 kg) and five impact directions (0◦ , 180◦ , 90◦ , 0◦ , and 180◦ off-center)
The second limitation is that the study has not considered the have been considered. The bicyclist-pedestrian collisions have been
biomechanical properties of the human body, such as properties of skin analyzed based on four accident-related metrics: throw distance, peak
and tissues. Although the study has considered three different pedestrian head velocity before ground impact, HIC value, and the probability of
profiles based on the weight and height distribution of adults, elderly severe head injury.
and children, the biomechanical properties of the skin and tissues are The results suggest a significant effect of collision speed on the
not considered. This is because the scope of this study is not to investi­ collision metrics. As the bicyclist’s speed increases, the momentum and
gate the injury types or injury susceptibility but to analyze the head- kinetic energy also increase. When a bicyclist with a high momentum
ground impact velocity and throw distance of the pedestrians of collides with a pedestrian, a large portion of the momentum is trans­
different height and weight in bicyclist-pedestrian collisions. To achieve ferred to the pedestrian, leading to a longer throw distance and a higher
this, different pedestrian profiles are modeled according to their head-to-ground impact speed, potentially resulting in more severe head
anthropomorphic properties like their height, mass, and moment of injury. Similarly, the results have also shown that the outcome of
inertia properties. Moreover, the human models included in the Virtual bicyclist-pedestrian accidents depends on the impact direction. The
Crash software have been found to produce accurate reconstructions of impact direction has a significant effect on the overall mean values for
accidents (Virtual CRASH. [cited, 2021). Section 2.6 presents a valida­ throw distance, peak head velocity before ground impact, and HIC.
tion of simulation results by comparing it with data obtained from Among the five impact directions, the 90◦ impacts have resulted in the
previously published relevant articles. highest peak head velocity and HIC value.
Besides the abovementioned limitations, there are a few smaller On the effect of age profile, our results have not shown any signifi­
limitations in the model such as rider’s connection with the bicycle and cant effect on the accident metrics. However, further analysis of results
contact characteristics. For example, in real life, a rider sometimes holds has suggested slightly higher peak head velocity, HIC value, and the
on to the handlebar to try to avoid a collision. However, in our simu­ probability of severe injury in children followed by elderly and adults.
lations, the rider does not hold on to the handlebar after colliding with On bicycle mass, our results have also not shown any significant
the pedestrian. Instead, the cyclist moves forward due to the inertia and

13
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

effect on the accident metrics. In general, a heavier bicycle is more likely Innovation Fund, Singapore (Grant No: FY15/11E).
to cause slightly higher values in the accident metrics. However, for
speeds below 20 km/h, the metrics have comparable values for 5.5 kg CRediT authorship contribution statement
and 17.5 kg bicycles.
On the post-collision fall mechanism, our simulations have shown Milan Paudel: Conceptualization, Validation, Formal analysis,
that impact direction and age profile have a significant influence. For Writing – original draft. Fook Fah Yap: Conceptualization, Methodol­
direct impacts (0◦ , 180◦ , 90◦ ), the pedestrians tend to fall forward in the ogy, Funding acquisition, Writing – review & editing, Supervision.
bicyclist’s direction of travel, whereas off-center impacts cause pedes­ Tantyana Binte Mohamed Rosli: Software, Formal analysis, Data
trians to rotate while falling on the side of the bicyclist. Moreover, curation. Kai Hou Tan: Software, Formal analysis, Data curation. Hong
analysis has also shown a possibility of rider-pedestrian entanglement Xu: Conceptualization, Funding acquisition, Writing – review & editing,
during 90◦ impacts causing the rider to land on top of pedestrians, which Supervision. Nader Vahdati: Writing – review & editing. Haider Butt:
amplifies the peak head velocity, HIC, and probability of severe head Writing – review & editing. Oleg Shiryayev: Writing – review & editing.
injury to pedestrians.
Finally, it has been found that a collision speed below 12 km/h has Declaration of Competing Interest
less than 50 percent overall mean probability of severe injury to pe­
destrians. Considering the rideability of bicycles and the severity of The authors declare that they have no known competing financial
injury for the worst possible case in a bicyclist-pedestrian crash, it is interests or personal relationships that could have appeared to influence
recommended to concerned authorities to implement a cycling speed the work reported in this paper.
limit of not more than 12 km/h on narrow and crowded shared paths
and footpaths. To further minimize the accident severity to pedestrians, Data availability
regulations could be made only to allow children to ride on footpaths
with adult supervision. Furthermore, 90◦ impacts in the bicyclist- No data was used for the research described in the article.
pedestrian collision have the highest probability of severe head in­
juries on pedestrians. Therefore, crossing or turning areas should be Acknowledgment
designed carefully to minimize s potential 90◦ impacts between bi­
cyclists and pedestrians. In addition, E-bikes and power-assisted bicycles We thank Mr. Yang Ba and Mr. Steve Zhang Xianli for their technical
that can be ridden at higher speeds, should be regulated for speed on support, Dr. Yiik Diew Wong and Mr. Gopinath Menon for their advice
shared paths. on the project.

Funding

This work was supported by Land Transport Authority (LTA)

Appendix A: Pedestrian Mass Properties

Adult (69.2kg)

Mass (kg) Inertia-x (kgm2) Inertia-y (kgm2) Inertia-z (kgm2)

Head 4.436 0.077 0.077 0.063


Neck 2.662 0.019 0.019 0.014
Right Lower Arm 1.331 0.026 0.004 0.026
Right Upper Arm 1.952 0.039 0.007 0.039
Left Lower Arm 1.331 0.026 0.004 0.026
Left Upper Arm 1.952 0.039 0.007 0.039
Foot Right 1.153 0.003 0.019 0.021
Lower Leg Right 3.549 0.17 0.17 0.025
Femur Right 7.985 0.425 0.425 0.082
Foot Left 1.153 0.003 0.019 0.021
Lower Leg Left 3.549 0.17 0.17 0.025
Femur Left 7.985 0.425 0.425 0.082
Hip 10.646 0.308 0.185 0.341
Torso 19.518 1.236 0.929 0.799
Child (34.7kg)
Mass (kg) Inertia-x (kgm2) Inertia-y (kgm2) Inertia-z (kgm2)
Head 2.224 0.024 0.024 0.02
Neck 1.335 0.006 0.006 0.004
Right Lower Arm 0.667 0.008 0.001 0.008
Right Upper Arm 0.979 0.012 0.002 0.012
Left Lower Arm 0.667 0.008 0.001 0.008
Left Upper Arm 0.979 0.012 0.002 0.012
Foot Right 0.578 0.001 0.006 0.006
Lower Leg Right 1.779 0.053 0.053 0.008
Femur Right 4.004 0.133 0.133 0.026
Foot Left 0.578 0.001 0.006 0.006
Lower Leg Left 1.779 0.053 0.053 0.008
Femur Left 4.004 0.133 0.133 0.026
Hip 5.338 0.096 0.058 0.107
Torso 9.787 0.387 0.291 0.25
Elderly (63.2kg)
(continued on next page)

14
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

(continued )
Adult (69.2kg)

Mass (kg) Inertia-x (kgm2) Inertia-y (kgm2) Inertia-z (kgm2)

Mass (kg) Inertia-x (kgm2) Inertia-y (kgm2) Inertia-z (kgm2)


Head 4.051 0.064 0.064 0.052
Neck 2.431 0.016 0.016 0.011
Right Lower Arm 1.215 0.022 0.003 0.022
Right Upper Arm 1.783 0.032 0.006 0.032
Left Lower Arm 1.215 0.022 0.003 0.022
Left Upper Arm 1.783 0.032 0.006 0.032
Foot Right 1.053 0.002 0.016 0.017
Lower Leg Right 3.241 0.141 0.141 0.02
Femur Right 7.292 0.352 0.352 0.068
Foot Left 1.053 0.002 0.016 0.017
Lower Leg Left 3.241 0.141 0.141 0.02
Femur Left 7.292 0.352 0.352 0.068
Hip 9.723 0.255 0.153 0.282
Torso 17.826 1.024 0.769 0.662

References Fanta, O., et al., 2012. Kinematic analysis of backward falls of pedestrian and figurine in
relation to head injury. Trans. Transp. Sci. 5 (4), 179.
Fergus, K.B., et al., 2019. Trends in bicycle-related injuries, hospital admissions, and
Airaksinen, N., Lüthje, P., Nurmi-Lüthje, I., 2010. Cyclist Injuries Treated in Emergency
deaths in the USA 1997–2013. Traffic Inj. Prev. 20 (5), 550–555.
Department (ED): Consequences and Costs in South-eastern Finland in an Area of
Gao, D., Wampler, C.W., 2009. Head injury criterion. IEEE Rob. Autom. Mag. 16 (4),
100 000 Inhabitants. Ann. Adv. Automot. Med. 54, 267–274.
71–74.
Playsafe-specialising in playground safety inspections, surface impact testing and training.
Gildea, K., Hall, D., Simms, C., 2021. Configurations of underreported cyclist-motorised
Urban Turf Solutions.
vehicle and single cyclist collisions: Analysis of a self-reported survey. Accid. Anal.
Alkhatib, S., Cyclist jailed for crash that killed pedestrian, in THE STRAITS TIMES. 2020:
Prev. 159, 106264.
Singapore.
Graw, M., König, H.G., 2002. Fatal pedestrian-bicycle collisions. Forensic Sci. Int. 126
Angel-Domenech, A., et al., 2014. Traffic conflict analysis by an instrumented bicycle on
(3), 241–247.
cycle tracks of Valencia. Proceedings of the 3rd International Cycling Safety Conference,
Grzebieta, R., McIntosh, A., Chong, S. Pedestrian-cyclist collisions: issues and risk. 2011.
Gothenburg, Sweden.
Hajiaghamemar, M., et al., 2015. Measurement of head impact due to standing fall in
Appleby, G., Webster, A., 2016. Cycling and the Law. UNSWLJ 39, 129.
adults using anthropomorphic test dummies. Ann. Biomed. Eng. 43 (9), 2143–2152.
Baker, S.P., et al., 1974. The injury severity score: a method for describing patients with
Han, I., Brach, R.M., 2001. Throw model for frontal pedestrian collisions. SAE Trans.
multiple injuries and evaluating emergency care. J. Trauma Acute Care Surg. 14 (3),
110, 1115–1128.
187–196.
Haworth, N.L., Schramm, A.J. Adults cycling on the footpath: what do the data show? 2011.
Batista, M., 2008. A simple throw model for frontal vehicle-pedestrian collisions. Promet-
Haworth, N., Schramm, A., Debnath, A.K., 2014. An observational study of conflicts
Traffic Transp. 20 (6), 357–368.
between cyclists and pedestrians in the city centre. J. Aust. College Road Saf. 25 (4),
Becker, T., Reade, M., Scurlock, B. Simulations of Pedestrian Impact Collisions with Virtual
31.
CRASH 3 and Comparisons with IPTM Staged Tests. arXiv preprint arXiv:1512.00790,
Henn, H.-W., 1998. Crash Tests and the Head Injury Criterion. Teach. Math. Appl. Int. J.
2015.
IMA 17 (4), 162–170.
Bhalla, K., et al., 2002. Vehicle impact velocity prediction from pedestrian throw
Hoxha, G., Shala, A., Likaj, R., 2017. Pedestrian crash model for vehicle speed calculation
distance: trade-offs between throw formulae, crash simulators, and detailed multi-
at road accident. Int. J. Civil Eng. Technol. 8 (9), 1093–1099.
body modeling. IRCOBI Conference.
Hutchinson, J., Kaiser, M.J., Lankarani, H.M., 1998. The Head Injury Criterion (HIC)
Biggar, M., 2019. Non-motorized Transport: Walking and Cycling. In: Leal Filho, W.
functional. Appl. Math. Comput. 96 (1), 1–16.
(Ed.), Sustainable Cities and Communities, W. Springer International Publishing,
Kappagantu, A., Yaremchuk, K., Tam, S., 2021. Head and neck injuries and electronic
Cham, pp. 1–10.
scooter use in the United States. Laryngoscope 131 (11), E2784–E2789.
Chakravarthy, B., Lotfipour, S., Vaca, F.E., 2007. Pedestrian injuries: emergency care
Khorasani-Zavareh, D., et al., 2013. Kinetic energy management in road traffic injury
considerations. California J. Emerg. Med. 8 (1), 15–21.
prevention: a call for action. J. Injury Violence Res.
Chan, E., Pedestrian dies from injuries 5 days after getting hit by bicycle, in Channel News
Kiyota, M., et al., 2000. Bicycle and pedestrian traffic conflicts on shared pavements.
Asia. 2019: Singapore.
Fourteenth Velo-city International Conference Proceedings, Munich.
Charters, K.E., Gabbe, B.J., Mitra, B., 2018. Pedestrian traffic injury in Victoria Australia.
Komanoff, C., 1993. The environmental benefits of bicycling and walking. Federal
Injury 49 (2), 256–260.
Highway Administration.
Chen, H., et al., 2015. Pedestrian response with different initial positions during impact
Laing, K. GREEN & HEALTHY STREETS : How C40 cities are implementing zero emission
with a mid-sized sedan. 24th International Technical Conference on the Enhanced Safety
areas, in C40 Cities. 2020, Transformative Urban Mobility Initiative.
of Vehicles (ESV).
Leng, B., et al., 2021. Estimation of tire-road peak adhesion coefficient for intelligent
Cheng, Y.-K., et al., 2015. Validation of pedestrian throw equations by video footage of
electric vehicles based on camera and tire dynamics information fusion. Mech. Syst.
real life pedestrian/vehicle collisions. Forensic Sci. Int. 257, 409–412.
Sig. Process. 150, 107275.
Chong, S., et al., 2010. Relative injury severity among vulnerable non-motorised road
Lin, C.-Y., Hsu, C.-C., Fung, C.-P., 2012. The study of coefficient of friction for light
users: comparative analysis of injury arising from bicycle–motor vehicle and
motorcycle sliding on asphalt road. Int. J. Phys. Sci. 7 (30), 5167–5174.
bicycle–pedestrian collisions. Accid. Anal. Prev. 42 (1), 290–296.
Macmichael, S. Pedestrian killed and cyclist seriously injured following crash on cycleway in
Chong, S.-L., et al., 2018. Epidemiology of Pedestrian-Motor Vehicle Fatalities and
Stevenage, in Road.cc. 2020, England.
Injuries, 2006–2015. Am. J. Prev. Med. 55 (1), 98–105.
Mariotti, G.V., et al., 2019. Head injury criterion: Mini review. Am. J. Biomed. Sci. Res. 5
Colon, D. Cops: Pedestrian Dies After Being Knocked Down by Hit-And-Run Cyclist, in
(5), 406–407.
Streetsblog NYC. 2019: USA.
Matsui, Y., Oikawa, S., 2018. Reduction in the head injury level due to junior high school
Commision, E. What forces can be tolerated the human body? Road Safety 2020 [cited 2020
cyclists wearing helmets. Int. J. Automot. Technol. 19 (6), 1023–1032.
December 22]; Available from: https://ec.europa.eu/transport/road_safety/specialis
Mehmood, A., et al., 2019. Performance of injury severity measures in trauma research: a
t/knowledge/vehicle/key_issues_for_vehicle_safety_design/what_forces_can_be_to
literature review and validation analysis of studies from low-income and middle-
lerated_the_human_body_en.
income countries. BMJ Open 9 (1), e023161–e.
Coufal, T., Semeia, M., 2014. Research of impact parameters for traffic accident analysis
Meng, Y., et al., 2017. A finite element model of a six-year-old child for simulating
in case of small overlap crash test. Proceedings of 2nd Annual International Conference
pedestrian accidents. Accid. Anal. Prev. 98, 206–213.
on Forensic Science-Criminalistics Research (FSCR). Singapore.
Ministry of Health (Singapore). Singapore National Health Surveillance Survey, 2013.
Crash, V. Virtual CRASH, LLC USA.
Monsere, C., 2017. Risk factors for pedestrian and bicycle crashes. Dept. of
Cycling and the Law: Bike Laws Across Australia. 2020 [cited 2020 October 25]; Available
Transportation Research Section, Oregon.
from: https://insiderguides.com.au/cycling-and-the-law/.
Moure-Guardiola, C., et al., 2020. Evaluation of combat helmet behavior under blunt
De Rome, L., et al., 2014. Bicycle crashes in different riding environments in the
impact. Appl. Sci. 10 (23), 8470.
Australian capital territory. Traffic Inj. Prev. 15 (1), 81–88.
Muggenthaler, H., et al., 2017. Fatal abdominal injuries in a bicycle-pedestrian
Dogs, G. New study reveals the danger of an e-scooter collision. 2021; Available from: https
collision–Reconstruction using multibody simulation. Forensic Sci. Med. Pathol. 13
://www.guidedogs.org.uk/news/study-reveals-danger-of-an-escooter-collision/.
(2), 230–233.

15
M. Paudel et al. Accident Analysis and Prevention 176 (2022) 106792

Nseya, C. Bicycle accidents: An analysis of the causes of single bicycle accidents in Stockholm. Shinkle, D., 2012. Bicycle and Pedestrian Safety. in National Conference of State Legisla
2018. tures. Transportation Review, Washinton, D.C.
O’Hern, S., Oxley, J., 2019. Pedestrian injuries due to collisions with cyclists Melbourne Short, A., Grzebieta, R., Arndt, N., 2007. Estimating bicyclist into pedestrian collision
Australia. Accid. Anal. Prev. 122, 295–300. speed. Int. J. Crashworthiness 12 (2), 127–135.
Ohlin, M., Algurén, B., Lie, A., 2019. Analysis of bicycle crashes in Sweden involving Sikic, M., et al., 2009. Bicycling injuries and mortality in Victoria, 2001–2006. Med. J.
injuries with high risk of health loss. Traffic Inj. Prev. 20 (6), 613–618. Aust. 190 (7), 353–356.
Oxley, J., O’Hern, S. Pedestrian injuries due to collisions with cyclist Melbourne, Australia. in Simms, C., Wood, D., 2009. Pedestrian and cyclist impact: a biomechanical perspective.
18th International Conference Road Safety on Five Continents (RS5C 2018), Jeju Island, Springer Science & Business Media 166.
South Korea, May 16-18, 2018. 2018. Statens väg-och transportforskningsinstitut. Sovreski, Z., et al., Application of mathematical modeling with the help of a specialized
Paudel, M., Yap, F.F., 2020. Development of an mproved design methodology and front licensed Virtual Crash 3.0 for a specific traffic accident in Republic of Macedonia. 2017.
steering design guideline for small-wheel bicycles for better stability and performance. Speed management on shared paths. 2014, Department of Transport and Main Roads:
Proceedings of the Institution of Mechanical Engineers, Part P: Journal of Sports Australia.
Engineering and Technology 234 (3), 227–244. Stevenson, M., et al., 2016. Land use, transport, and population health: estimating the
Paudel, M., Yap, F.F., Bastola, A.K., 2018. Dynamic stability assessment of different health benefits of compact cities. Lancet 388 (10062), 2925–2935.
wheel sized bicycles based on current frame design practice with ISO requirement Tan, C. More will commute by cycling, walking in 10 years: Study, in The Straitstimes. 2020:
for bicycle safety. Int. J. Sport Health Sci. 12 (9), 366–373. Singapore.
Paudel, M., Yap, F.F., Bastola, A.K., 2020. Safety assessment of personal mobility devices Toor, A., Araszewski, M., 2003. Theoretical vs. empirical solutions for vehicle/pedestrian
with different wheel size based on their dynamic stability performance. Int. J. collisions. SAE Trans. 853–865.
Sustain. Des. 3 (4), 227–250. TRUSCA, D., et al., INFLUENCE OF VARIOUS ANTHROPOMETRIC CONSTITUTION OF
Paudel, M. An investigation into the design for rideability of small wheel single‑track bicycles PEDESTRIAN IN CASE OF VEHICLE-PEDESTRIAN IMPACT. 2010.
and e‑scooters. 2019. Tuckel, P., Milczarski, W., Maisel, R., 2014. Pedestrian injuries due to collisions with
NYC Pedestrian Fatalities: Bicycle-Pedestrian and Motor Vehicle-Pedestrian Crashes. bicycles in New York and California. J. Saf. Res. 51, 7–13.
2019 [cited 2020 December 5, 2020]; Available from: https://www1.nyc.gov/ht Urbanczyk, R. Promoting Cycling for Everyone as a Daily Transport Mode 2010: Germany.
ml/dot/downloads/pdf/nycdot-pedestrian-fatalities-by-bike-motor-vehicle.pdf. Validation and study of Virtual CRASH. [cited 2021; Available from: https://www.vcrash
Pucher, J., Dijkstra, L., 2003. Promoting safe walking and cycling to improve public usa.com/vc-validation-vc#validation-vc-vc.
health: lessons from the Netherlands and Germany. Am. J. Public Health 93 (9), Virtual Crash User’s Guide. Virtual Crash LLC.
1509–1516. Vulnerable road users, in SWOV. 2012, Institute for Road Safety Research Netherland.
Ravani, B., Brougham, D., Mason, R., 1981. Pedestrian post-impact kinematics and injury Walking and Cycling Design Guide 2018, Land Transport Authority, singapore; Urban
patterns. SAE Trans. 3279–3292. Redevelopment Authority Singapore.
RCAR. Crash-tests of electric scooters, analysis by Fundación MAPFRE and CESVIMAP. [cited Wang, Q., et al., 2020. Reverse reconstruction of two-wheeled vehicle accident based on
2022 28 January ]; Available from: https://www.rcar.org/scooter-testing. Facet vehicle model and hybrid human model. Int. J. Crashworthiness 1–16.
Saczalski, K.J., et al., 1976. A critical assessment of the use of non-human responding Wegman, F., Aarts, L. Advancing Sustainable Safety; National Road Safety Outlook for 2005-
surrogates for safety system evaluation. SAE Trans. 2538–2550. 2020. 2006, SWOV Institute for Road Safety Research: Leidschendam. p. 215.
Saulić, N., Papić, Z., Ovcin, Z., 2020. Pedestrian throw distance prediction from vehicle Wicher, J., 2016. Uncertainty of the estimation of model parameter values in the analysis
damage intensity. Promet-Traffic Transp. 32 (3), 371–382. of pedestrian throw. Archiwum Motoryzacji 74 (4).
Searle, J.A. The physics of throw distance in accident reconstruction. 1993, SAE Technical Wood, D., Simms, C., 2000. Coefficient of friction in pedestrian throw. Impact 9 (1),
Paper. 12–15.
Shang, S., et al., 2018. Detailed assessment of pedestrian ground contact injuries Xu, T., et al., 2018. Development and Validation of Dummies and Human Models Used in
observed from in-depth accident data. Accid. Anal. Prev. 110, 9–17. Crash Test. Appl. Bionics Biomech. 2018, 3832850.
Shaw, L., Lennon, A.J., King, M.J. A qualitative investigation of older pedestrian views of Zhang, J., Yoganandan, N., Pintar, F.A. Dynamic biomechanics of the human head in lateral
influences on their road crossing safety. 2012. impacts. in Annals of Advances in Automotive Medicine/Annual Scientific Conference.
2009. Association for the Advancement of Automotive Medicine.

16

You might also like