On the simulation of electric scooter crash-test

with the hybrid human body model

T. Bońkowski*, J. Špička *, L. Hynčík *
* New Technologies – Research Centre, University of West Bohemia, Pilsen, Czech Republic
2022 20th International Conference on Emerging eLearning Technologies and Applications (ICETA) | 979-8-3503-2033-6/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICETA57911.2022.9974656

tomasz@[Link], spicka@[Link], hyncik@[Link],

Abstract transport can be seen in the highly developed cities,

mainly in Europe, Nord America and China [1][2][3][4].
This paper introduces the reader to the problem of traffic Micro mobility consists of the standard bicycles, scooters
accident reconstruction from a technical perspective and their electric versions. Recently the shared mobility
solution and simulation tools. The observed accident is not a
real accident, but an experimental crash test of a BMW E87,
or transport started to play an important role in the urban
118i, and an electric scooter. The rider of the electric scooter districts. More and more people are using such way of
in this experiment was replaced by a dummy Primus, transport in their everyday live. With the increasing
representing the average EuroNCAP male (175 cm, 78 kg). number of these powered two wheel vehicles (PTW), the
The dummy was placed on top of the electric scooter and number of accidents and crashed within standard
positioned in front of the vehicle so that it was knocked from vehicles, pedestrians or between other PTW has been
the left, approximately in the middle of the vehicle. The significantly growing [5] and [6]. Since the electric
vehicle was accelerated to a speed of approximately 50 km/h scooters might be used in the pedestrian urban areas as
and at this speed struck the rider of the electric scooter, who
well as on the road for vehicles, the crashes can result
was gradually struck on the bonnet and windscreen of the
vehicle, the scooter being thrown forward. The output data
serious injury for both involved. Driver on the e-scooter
from the experiment and the input data for the has higher kinetic energy then a pedestrian (mass and
reconstruction (simulation) are the initial and end positions velocity) and such can cause serious injury mainly for the
of the dummy, scooter, and vehicle, vehicle deformation, pedestrian. On the other hand, any crash of the e-scooter
and damage/injury to the dummy. The simulation tool used with the standard car can be fatal for the driver of scooter.
is finite element method software Virtual Performance Moreover, the travelling velocity of such PTW increase
Solution, in which the vehicle and scooter model was created an injury risk even more, even in the single-vehicle
and a hybrid human body model: Virthuman was used. The accidents. Several studies have been published related to
vehicle was modeled as partly rigid and partly deformable the crash of the vehicles and electric scooters
(deformable in the frontal part, where contact with the
driver appears) and rigid in the rest, in order to speed up
[7][8][9][10]. Como in his paper [11] suggested the
the computation time. The scooter was modeled as rigid in methodology for a crash test with anthropometric test
its entirety (to speed up the calculation and irrelevance of device (ATD). However, these scenarios are not generally
the deformation data - they do not affect the rider dynamics monitored and the safety guidance does not take them
and these data are not available). The Virthuman model is a into the account, yet.
numerical human body model, which could represent an
individual of a given height, weight, age, and sex. This Such accidents, however, introduce a new kinematics of
model is suitable specifically for simulations of dynamic the rider and consequently new patterns of injury to the
phenomena with potential impact. The main idea of this
human bodies. Generally, the system of driver (human)
study was to develop a kind of methodology for accident
reconstruction of scooters and electric scooters and to
and electric scooter has higher velocity (up to 50 km/h),
identify the relevant data needed for the simulation tools. higher mass (up to 100 kg – scooter itself, plus driver)
The principle is to vary unknown input data (rider position and centre of gravity is shifted up. All these significantly
relative to the vehicle, vehicle speed, etc.), and monitor the changed the impact kinematics and impact area location
simulation results. The aim is to keep the known input data on the vehicles.
and variations of those of the unknown, simulation results
as close as possible to the results of the experiment (vehicle Numerical models are playing an important role in the
damage, injury and final position of the rider, etc.). In this research and development of the new type of vehicles,
way, it is possible to determine the possible initial state and
with a great focused on safety. Global NCAP agency
progress of the accident and thus obtain information (data)
that could not be obtained at the accident scene and which
(Euro, US, Asia etc.) [12] has been introducing
may help to clarify or exclude a given collision scenario, or methodology for assessing safety rating for a vehicles,
could be used for the development of future safety systems. taking into the account drivers, passengers as well as
pedestrian. However, cyclist, PTW driver and other
vulnerable urban traffic user are considered only in the
I. INTRODUCTION assessment of the active systems of the cars.
Electro mobility as well as shared mobility has been
expanded world widely. Greatest increase of such way of

979-8-3503-2033-6/22/$31.00 ©2022 IEEE 59

Authorized licensed use limited to: BME OMIKK. Downloaded on March 01,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.
Here, the authors present the study, where the System (MBS) approaches to discretize the physical
reconstruction of the experimental crash test with the systems.
passenger’s vehicles and the ATD driver on the electric Due to the lack of the accessibility to the detailed FEM
scooter was performed. Goal of this study was to use models of the cars, a similar car model (BMW X1/X3)
available model of the human body, model of the was used. To adjust the used model to the global
particular vehicle and model and model of the e-scooter parameters of the car used in the crash-test, a scaling
and try to model the tested impact scenario to show method was used. Firstly, the ratio between such
validity of such numerical modelling also for these new dimensions as wheelbase, tire size, and the rooftop height
was calculated between the FEM model and the real E87
scenarios. Here model Virthuman was used as a e-scooter
BMW car. Secondly the FEM model tires were detached
driver, with the variation of the initial conditions from the main model and scaled locally by calculated ratio
(positions of the driver with respect to the vehicles and in the X and Z dimension (coordinate system according to
monitoring the injury to the e-scooter driver and the SAE). Thirdly the model was scaled in the Z
deformation of the car, such collision can be virtually dimension by the rooftop height ration and in the X
modelled (digital twin) and used for a further dimension by the wheelbase ratio. Finally, the model was
development of safety systems. once again assembled from the four scaled wheels and
scaled body of the car. This type of model adjustments
II. METHODS allows to meet the global dimensions of the real car from
the crash-test. Additionally, some parts of the FEM model
A. Electric scooter crash-test were set to be rigid to reduce the calculation time (those
parts did not experience the big deformations in real
The electric scooter crash-test was conducted by the crash-test). This simplification can be seen on Fig.2. The
CTS ([Link] GmbH) as a part of electric scooter was modelled as a rigid body by using the
PraxisConference Pedestrian Protection 2021 [13]. The geometry of the generic e-scooter from the [Link].
basic data and the test procedure as a series of the high-
speed picture sequence was obtained from the post-crash
database. The car used in the experiment was BMW (type
hatchback, 5 doors model, E87) with the test total weight
of 1365 kg. The electric scooter (Tretroller E-Roller
IOHAWK) which was used for this test has mass of 98 kg.
The PRIMUS BREAKABLE biofidelic dummy was
employed in this test as the driver of the e-scooter [14].
The e-scooter with the driver was suspended on the metal
scaffolding in front of the car with the electric release
mechanism. The impacting car was accelerated by the
volunteer driver equipped with the necessary safety
measures (e.g. helmet, safety gloves). Detailed
configuration could be seen on the Fig. 1. The aimed
impact speed was set to 50 km/h, however as the car was
driven by human volunteer driver, the real recorded Figure 2. Scaled FEM model with simplified parts (red colour).
impact velocity was 48.9 km/h.
To assure the realistic inertia the density of material
was set to 2.7 g/cm3 (aluminium), the rest of the mass
which does not comes from the structure was assume to be
battery mass and has been added to the bottom part of the
e-scooter deck.
C. Driver model

The driver (in the experiment represented by the

PRIMUS) in numerical environment was surrogated by
the Virthuman hybrid human body model, The Virthuman
is virtual model of a human body based on the hybrid
approach (combination of finite elements method – FEM
and multibody structures – MBS). Model is a fully
scalable (based on database of more than 1000 people in
Czechoslovakia [15]) and thanks to that and hybrid
Figure 1. The setup of the crash-test.
approach, it can be scaled based on the gender, age, size,
mass and percentile of the population to generate various
B. Simulation setup humans model [16]. Model was fully validated to ensure
its biofidelity within the segments validation tests as well
as full body tests [17],[18]. Virthuman model was
The simulation was conducted in the Virtual successfully used as a pedestrian [19], [20] motorcycle
Performance Solution (VPS) environment which employ driver [21], [22], [23], for the aviation purposes [24] or in
both the Finite Element Method (FEM) and Multi-Body

979-8-3503-2033-6/22/$31.00 ©2022 IEEE 60

Authorized licensed use limited to: BME OMIKK. Downloaded on March 01,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.
the forensic analysis [25]. Example of the various size of
Virthuman model is shown in Figure 3.

Figure 3. Virhuman family (small child, large male and average

female).

Virthuman model includes also algorithm for automatic

evaluation of an injury risk based on EuroNCAP injury
rating. The post processing algorithm automatically
evaluate mechanical quantities such as deflections,
velocities, accelerations, forces or moment and based on
the EuroNCAP data calculate an injury risk based on the
defined human tolerance thresholds. The visualisation of
the injury risk is presented via colouring of the particular
segments, where red colour stands for poor condition of
survivability, orange represent marginal injury risk, Figure 5. The analysis of the configuration for the numerical setup
yellow is acceptable risk and green represent good preparation.
condition of survivability and low injury risk, see Figure 4

Figure 6. Numerical setup.

III. RESULTS

Figure 4: Virthuman coloured in the injury risk assessment. A. Experimental results – kinematics

D. Simulation setup The kinematics of the accident could be described as

follows. The e-scooter driver was hit from his left side in
Critical issue of the simulation was to reproduce initial the tibia region, followed by tight and hip. He lay onto
position in order to follow the experimental results. The the bonnet. He hit windscreen with the elbow, follow by
initial position was not very well described and thus, the head. The windscreen got broken and dummy intruded
photos were analysed to defined initial position (initial the car with head and his upper part of the body. Due to
guess). Later on, initial position of the driver with respect the inertia, he flipped around the horizontal axis and
to the vehicle was modified, to get desire kinematics. The impact the car roof with his right leg. By the end of the
main goal was to reproduce initial contact of the body to test, the dummy was stacked inside the car with his head
the vehicle (hips and thighs to the bonnet and elbow and and upper torso respectively, only the breaking of the
heads to the windscreen). The Fig. 5 depict the basic impacting car results in releasing of ATD from the
analysis of the distances of the experiment confirmation, interior of the vehicle. The entire motion could be seen in
which allows to make the proper representation of test in the test procedure as high speed picture sequence
the numerical environment (Fig. 6.). Besides the proper depicted in Annex I
configuration, necessary contacts were assigned.
B. Experimental results – dummy/ vehicle damage

The experimental test results in the injury (damage) of

the PRIMUS dummy, damage of the car and of the e-

979-8-3503-2033-6/22/$31.00 ©2022 IEEE 61

Authorized licensed use limited to: BME OMIKK. Downloaded on March 01,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.
scooter. However, damage of the scooter was not C. Numerical results
monitored and described in the test protocol, and thus, it The 3 variants of the configurations has been proposed
is not assessed here. Moreover, the scooter in the for the correlation of the simulation with the real crash
simulation was modelled as a rigid body, since it does not test. Those configurations only differentiates in minimal
affect the overall results. change in position Z axis. Variants (I-III) has the COG of
PRIMUS dummy suffered following damage: the Virthuman placed 1087 mm, 1085 mm and 1086
above the ground respectively. Besides this variation the
 Right humerus: fracture configuration of the Virthuman comes from the analysis
 Right tibia: fracture of the Fig. 5. and could be described as follows:
 Skull: fracture of the frontal superior bone
 Right wrist: fracture  Angle of the knee set-up based on the frontal
 Right and left ankle: fracture picture, in order to best represent the Fig. 5.;
 Left hip: rupture of the straps (50° left knee and 46° right knee)
 Right knee: rupture of the straps  Elbows placed further from the body (to be more
physiological comfortable)
Generally, the damage of the plastic represents bone  Feet placed closer to each other (to fit in the
fractures (i.e. humerus, wrist’s bones, skull and tibia), scooter base deck).
while rupture of the straps stands for the soft tissue
deformation – ligaments or tendons (i.e. hip, knee and
ankles). Photos of the damaged segments of the body are
presented in Annex II at Fig. Annex II a-g.

Figure 9. Variant I kinematics.

Variant I (Fig. 9) reproduces the initial impact as well

as impact of the elbow to the windscreen. However, since
the windscreen is shorter then in experiment, the head
does not impact the windscreen glass, but the upper roof
pillar, that’s changed the further kinematics of the body.

Figure 7: Regions of the dummy with damage.

The vehicle sustained following deformation and damage

(Fig. 8.):
 Windscreen: broken on the right part of the car
 Bumper: deformation on the central right part
(from hit to the legs)
 Bonnet: minor deformation in the frontal part
Figure 10. Variant II kinematics.
from hit to the scooter and pedestrian
This configuration (Variant II) reproduces the initial
impact. The body was placed in such position, where the
head impacts windscreen, but because of the car
geometry, the elbow hits the bonnet not the windscreen as
appeared in the experiment.

Figure 8. Post-crash damage of the car. Figure 11. Variant III kinematics.

979-8-3503-2033-6/22/$31.00 ©2022 IEEE 62

Authorized licensed use limited to: BME OMIKK. Downloaded on March 01,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.
 In this initial configuration (Variant III) of the e-scooter REFERENCES
driver, neither the impact of the head or elbow does not [1] MATHEW, Jijo K.; LIU, Mingmin; BULLOCK, Darcy M. Impact
fit to the experimental data. Elbow impacts the transition of weather on shared electric scooter utilization. In: 2019 IEEE
between windscreen and bonnet (where the rigid Intelligent Transportation Systems Conference (ITSC). IEEE,
2019. p. 4512-4516.
reinforcement of the car is located) and the head impacts
[2] KAZEMZADEH, Khashayar; SPREI, Frances. Towards an
area between windscreen and roof, which is also electric scooter level of service: A review and framework. Travel
considered as a rigid (reinforced) part of the vehicle. behaviour and society, 2022, 29: 149-164.….
[3] RECHKEMMER, Sabrina K.; ZHANG, Weimin; SAWODNY,
Oliver. Modeling of a permanent magnet synchronous motor of an
IV. DISCUSSION e-scooter for simulation with battery aging model. IFAC-
PapersOnLine, 2017, 50.1: 4769-4774.…
[4] FARLEY, Kevin Xavier, et al. Estimated incidence of electric
This study describes the reconstruction of the scooter injuries in the US from 2014 to 2019. JAMA network
experimental crash test of the rider on electric scooter open, 2020, 3.8: e2014500-e2014500.
impacted with the passenger car. The human was [5] YANG, Hong, et al. Safety of micro-mobility: Analysis of E-
represented with the Primus dummy and car make and Scooter crashes by mining news reports. Accident Analysis &
model was BMW 118i E87. The car impacts the rider Prevention, 2020, 143: 105608.
[6] SHAH, Nitesh R., et al. Comparison of motor vehicle-involved e-
from his left side in the velocity of approximately 50 scooter and bicycle crashes using standardized crash typology.
km/h. The kinematic was monitored with the cameras and Journal of safety research, 2021, 77: 217-228.
damage of the dummy was presented. Simulation was [7] ULUK, Deniz, et al. E-scooter incidents in Berlin: an evaluation of
performed in the software VPS where virtual model of a risk factors and injury patterns. Emergency medicine journal,
human Virthuman is available. Model of this particular 2022, 39.4: 295-300.
vehicle (geometry and FEM model) was not available and [8] KLEINERTZ, Holger, et al. Accident Mechanisms and Injury
Patterns in E-Scooter Users: A Retrospective Analysis and
thus, BMW X1 was used and scaled to as closed as BMW Comparison with Cyclists. Deutsches Ärzteblatt International,
118i in its geometry. Model of the scooter was used from 2021, 118.8: 117.
public available data (TRETROLLER E-ROLLER [9] PTAK, Mariusz, et al. Analysis of electric scooter user kinematics
IOHAWK). Since the initial position of the human with after a crash against SUV. PLoS one, 2022, 17.1: e0262682.
respect to the vehicle was not fully described (in [10] STÖRMANN, Philipp, et al. Characteristics and injury patterns in
electric-scooter related accidents—a prospective two-center report
experiment), authors used some geometrical tools to fit the from Germany. Journal of clinical medicine, 2020, 9.5: 1569.
simulation setting to the experimental one. [11] COMO, Steven Gerard, et al. Crash Test Methodology for Electric
Scooters with Anthropomorphic Test Device (ATD) Riders. 2022.
In total, three configurations were presented, where [12] Global NCAP; Available at: [Link]
only initial position in Z-direction (up-down) was changed [13] [Link], PraxisConference Pedestrian Protection 2021; Available
and the kinematics of the human was monitored. The goal at: [Link]
[Link]
was to achieve such results, where both elbow and head [14] CTS dummy-solution, Available at: [Link]
impact the windscreen, was appeared in the experiment. [Link]/download-file?file_id=2112&file_code=4caf3f97db
However, none of the three configurations result in such [15] BLÁHA, P., Anthropometry of Czech and Slovak Population from
kinematics. Variant I (z = 1087 mm) results in the good 6 till 55 years, Ustr. stab. CS. Spartakiady 1985: Praha, Czech
kinematics of the elbow (impacting windscreen) but the Republic, 1987. Technical Report.
head hits the roof pillar (too high). In the Variant II (z = [16] HYNČÍK, L., et al. On scaling virtual human models, SAE
Technical Paper 2013-01-0074, 2013.
1085 mm) the head impacts the car in the desire position [17] VYCHYTIL, J., et al. Multi-directional virtual human model for
(windscreen) but the elbow hits the bonnet. Variant II (z = safety assessment. SAE Technical Paper, 2014, 01-0534.
1086 mm) didn’t meet the desire kinematics neither in the [18] HYNČÍK, L., et al. Stature Based Approach towards Vehicle
head, nor in the elbow. Both of those body’s sections Safety. Stature (No. 2015-26-0209); SAE Technical Paper; SAE:
impact the car pillar/reinforcement. Besides the Detroit, MI, USA, 2015
miscorrelation of local impact of the elbow to the [19] ŠPIČKA, Jan; VYCHYTIL, Jan; HYNČÍK, Luděk. Numerical
analysis of a pedestrian to car collision: Effect of variations in
windscreen, the simulation setup (Variant II) with the walk. Applied and Computational Mechanics, 2017, 10.2.
Virthuman could recreate the real-world event kinematics [20] VYCHYTIL, Jan, et al. Prediction of injury risk in pedestrian
of the e-scooter driver body (Fig. 10 vs. Annex I). To accidents using virtual human model VIRTHUMAN: real case and
reach high fidelity representation of the test, the detailed parametric study. SAE Technical Paper, 2016.
FEM model of the real impacting vehicle would be [21] LV, Wenle; HYNCIK, Ludek; BONKOWSKI, Tomasz. Rider
Stature Influence to Injury Risk in Motorcycle Rear Impact to Car.
needed. 2019.
[22] BONKOWSKI, Tomasz; HYNCIK, Ludek; LV, Wenle. PTW
ACKNOWLEDGMENT Passive Safety: Numerical Study of Standard Impact Scenarios
with Rider Injury Risk Assessment. SAE Technical Paper, 2020.
This work is supported by the John H. and Amy [23] HYNČÍK, Luděk; BOŃKOWSKI, Tomasz; LV, W. Numerical
Bowles Lawrence Foundation and by the European assessment of motorcyclist accident. Journal of the Society of
Regional Development Fund-Project “Application of Automotive Engineers Malaysia, 2019, 3.2: 210-217.
Modern Technologies in Medicine and Industry” (No. [24] LINDTSTEDT, L., et al. Numerical tests of the virtual human
CZ.02.1.01/0.0/0.0/17_048/0007280). model response under dynamic load conditions defined in Federal
Aviation Regulation Part 23.562 and 25.562–preliminary study.
Archive of Mechanical Engineering, 2016, 63.4: 511--530.

979-8-3503-2033-6/22/$31.00 ©2022 IEEE 63

Authorized licensed use limited to: BME OMIKK. Downloaded on March 01,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.
 [25] ČERMÁK, Martin. Forensic Analysis and a New Investigation.
Journal of Forensic Identification, 2022, 72.3: 245.

V. ANNEX I
Test procedure as high speed picture sequence

979-8-3503-2033-6/22/$31.00 ©2022 IEEE 64

Authorized licensed use limited to: BME OMIKK. Downloaded on March 01,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.
 VI. ANNEX II
Photos of the damages of the PRIMUS dummy

Figure Annex II a: injury to the right arm Figure Annex II b: injury to the right calf
Figure Annex II a: injury to the head

Figure Annex II b: injury to the right hand Figure Annex II c: injury to the right foot Figure Annex II d: injury to the pelvis

Figure Annex II e: injury to the right knee

979-8-3503-2033-6/22/$31.00 ©2022 IEEE 65

Authorized licensed use limited to: BME OMIKK. Downloaded on March 01,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.