Development of a low-cost motorcycle riding simulator for emergency scenarios involving swerving

Article (PDF Available)inProceedings of the Institution of Mechanical Engineers Part D Journal of Automobile Engineering 230(14) · January 2016with 470 Reads
DOI: 10.1177/0954407015624998
Cite this publication
Abstract
The development of advanced riding assistance systems requires the analysis of user reactions in emergency situations. Motorcycle riding simulators are an alternative to ‘on-road’ testing so that virtual environment dangerous scenarios can be investigated without risks for the participants. In this paper, we propose a process for validation of a low-cost motorcycle simulator characterized by, first, an elastic resistance on the steering input and, second, a counter-steering strategy. For this, 16 riders tested the simulator in different manoeuvres, including a cornering manouvre in a non-urban environment, a slalom manoeuvre and a lane-change manoeuvre. Objective evaluations and subjective evaluations showed that the simulator was realistic, in particular for investigating lateral avoidance scenarios. The development of suitable motorcycle simulators will significantly advance the field of motorcycle safety research.
Development of a low-cost motorcycle riding simulator for
emergency scenarios involving swerving
Authors
Giovanni Savino, PhD
Department of Industrial Engineering, University of Florence, Italy;
Monash Injury Research Institute, Monash University, Clayton, Victoria, Australia
A/Prof Marco Pierini, PhD
Department of Industrial Engineering, University of Florence, Italy
Prof Michael G Lenné, PhD
Monash Injury Research Institute, Monash University, Clayton, Victoria, Australia
Corresponding author:
Giovanni Savino, Department of Industrial Engineering, University of Florence, Via S. Marta
3, 50139 Firenze, Italy. Email: giovanni.savino@unifi.it. Mb: +39 347 667 3526
Keywords
Motorcycle, riding simulator, validation, emergency manoeuvre, steering input
This is an Author's Original Manuscript of an article submitted for consideration in
Proc IMechE Part D: J Automobile Engineering. The published paper is available online at
http://pid.sagepub.com/content/early/2016/01/19/0954407015624998.abstract
Abstract
The development of advanced riding assistance systems requires the analysis of user
reactions in emergency situations. Motorcycle riding simulators are an alternative to on road
testing as in the virtual environment dangerous scenarios can be investigated without risks for
the participants. In this paper, we proposed the validation process of a low-cost motorcycle
simulator characterised by: (i) elastic resistance on the steer input; and (ii) counter steer
strategy. Sixteen riders tested the simulator in different manoeuvres, including cornering in
non-urban environment, slalom and lane change. Objective and subjective evaluation
showed good realism of the simulator, in particular for investigating lateral avoidance
scenarios.
Introduction
Motorcycles and mopeds, often indicated with the name of powered two wheelers (PTWs),
are now more than 300 million around the world and the number is likely to increase.[1]
PTWs can play an important role for the current challenges of personal mobility at a global
level,[2] despite the higher risk of death and serious injuries for PTW users when compared
to other motorised vehicle users.[3] In the last two decades, new technologies were proposed
to improve PTW safety, including primary safety systems such as antilock braking, traction
control, collision warning, and curve warning.[4-7] The effectiveness of some of these
technologies, for example the warning systems, depends on the correct human-machine
interaction, which needs to be developed taking account of the users. In some cases,
experiments with users have been carried out in the real world, especially in low risk
activities.[7-9] Another approach used for practical testing involving users is via driving
simulators, which allows evaluating the human reactions in demanding conditions with low
risk for the participants.[6] The involvement of simulator experiments in the development of
safety technologies is documented for passenger cars,[10-12] and for trucks.[13] Driving
simulator studies were also conducted to investigate unexpected, highly emergency
situations.[14, 15]
A critical aspect of a simulator is the level of fidelity that it can achieve in the driving
context. In particular one aspect, functional fidelity (how the simulator behaves compared to
how the user expects it to behave), is important more important in fact than physical fidelity
(how it looks).[16] In this perspective, the use of a simulator to inform the vehicle design
process requires adequate fidelity in the inputs provided to the user during the simulation in
the specific test situation. If so, the feedback obtained from the user in the simulation is then
assumed to be compatible with the feedback the same user would provide in the
corresponding real-world situation.[17]
Producing appropriate functional fidelity is particularly challenging for motorcycle riding
simulators, as documented in several validation studies available in the recent literaturesee
[18] for a review. A remarkable example of motorcycle simulator is the one developed by
Cossalter, et al. [19]. It consisted in a five-degree of freedom motorcycle rig equipped with
sensors measuring several inputs from the rider, including throttle and brake controls, steer
torque input, gear shift, and lateral body position. The custom-built motion system was able
to produce lateral shifts, roll, yaw, and pitch rotations, and active steer feedback. The
dynamical engine of the simulator was a self-developed 14-degree-of-freedom multibody
model with high physical fidelity.[20] Cossalter et al. presented a subjective and objective
validation of their simulator in standard manoeuvres: acceleration and braking, steady
cornering, lane change and slalom. Despite the high degree of physical agreement and the
overall good level of satisfaction of the users, the subjective feedback provided by the riders
indicated an incomplete agreement between simulated and real riding experience. In fact,
most of the average ratings provided by the test users were up to 4 in a scale from 0 to 5. A
different approach is to start from a simple simulator setup in order to identify the most
important improvements to enhance fidelity. In this perspective, a static-rig simulator based
on a passenger car dynamical model was tested in BMW [21], showing the importance of
realistic steering feedback. Concerning the steering input, two are the common approaches:
positive steering (i.e. clock-wise steer angle to turn right, and vice versa), and counter
steering (counter clock-wise steer torque to turn right and vice versa). Both strategies have a
physical rationale: in steady state cornering the handlebar is typically rotated towards the
inner of the curve (as in the positive steering), whereas counter steer torque is the typical
strategy applied in a wide range of riding conditions.[22-24] These strategies for steering
input were analysed in a specific simulator study [25]. Positive steering appeared more
intuitive and allowed higher accuracy in the vehicle control, resulting more desirable than a
counter steer strategy. As a drawback, positive steering does not allow to measure realistic
steering torques during the simulation. A combination of the two strategies is also possible,
resulting in a more realistic simulation at the cost of a higher complexity [18].
This paper presents a subjective and objective validation of a simple motorcycle riding
simulator adopting counter steer strategy. The aim of the simulator was to capture steer input
timing, sign and magnitude in standard manoeuvres. The simulator was built within the EC
funded project ABRAM to investigate the reactions of rider facing unexpected collision
scenarios for the development of advanced safety systems for motorcycles.
Method
Apparatus
The motorcycle riding simulator was a low-cost upgrade of a simulator available at the
Monash University Accident Research Centre (MUARC).[26] The principal characteristics of
the new motorcycle simulator were the following: (i) steering input with elastic feedback and
implementing counter steer strategy; (ii) vehicle dynamics computed with the dynamical
model adapted from a passenger car; (iii) motorcycle rig mounted on a motion base
implementing inverted roll angle for lateral motion cues; (iv) no tilting horizon in the visual
cue. The simulator rig consisted in a sports motorcycle (Honda NSR150), which provided
realistic geometries for the riding position (Figure 1 a). The motorcycle was lightened by
removing the engine, the rear wheel, the swing arm, and other ancillary components. The
standard steering assembly was connected to the frame via two helical springs attached to a
support mounted on the rim (Figure 1 b), to obtain an elastic torque in response to steer inputs
in the form of rotations of the handlebar along the steering axis (elastic coefficient: 3.43
Nm/degree). The steer, throttle and brake inputs were measured using the hardware of a
commercial steering wheel system for gaming application. The brake lever and pedal were
connected with their original independent hydraulic systems to preserve a realistic feeling.
The two brake controls were connected via Bowden cables to a single potentiometer, the
output of which was used as brake input for the vehicle model. Different leverages on the pot
side were adopted to mimic the different effectiveness of the front and rear brakes. The
motorcycle frame was mounted on a commercial motion base consisting in three actuators
(two in the front and one in the rear). The motion base produced bounce, pitch, and roll cues
computed by the simulation software. The signal for the roll movement computed based on
the passenger car model was inverted and amplified to account for the opposite tilting
directions between a four- and two-wheeled vehicle. A commercial electromagnetic shaker
for home theatre application was connected to the frame under the original saddle, with
vertical axis. The shaker was controlled via dedicated amplifier connected to the audio
channel of the simulator in order to produce a vibration correlated with the engine sound.
The simulator rig was controlled by a desktop pc running the Eca Faros driving simulation
software integrated with Carsim for the computation of the vehicle dynamics. The reference
vehicle was a 3 Series BMW, 3 dm
3
capacity diesel engine, rear wheel drive, and automatic
transmission. The simulator software computed the motion, auditory, and visual cues. The
latter were provided to the user via three Nec Multisync X-series screens (1.01 m x 0.58 m
each), with total resolution of 5760x1080 and refresh frequency of 60 Hz, positioned 1.20 m
off the user (horizontal field of view: 120 degrees). Standard, 3-channel desktop speakers
delivered the audio cues.
Figure 1. (a) Motorcycle rig. (b) Detail of the elastic connection linking the front rim to the
motion base.
Participants
Participation in this study involved attending the Advanced Driving Simulator facilities of
MUARC for a single 1.5 hour testing session, with a reimbursement of AU$ 30. Riders in the
age 20-65 years, holding a motorcycle licence, and riding at least once a week, were eligible
for recruitment. Fifty-two people were identified: (i) from an existing database of participants
who took part in previous road safety studies; (ii) from the colleagues; and (iii) from the
University newsletter. Forty-five people were contacted by the investigators and response rate
was 44%. One person refused the invitation. Finally, 16 participants took part in the study (15
males, 1 female). Details of the participants are provided in Table 1. The age was in the
range 22-63 years (mean 39.5, SD 14.5). Almost a third of the sample declared a daily use of
their motorcycle at the time of the tests. Despite the specific criterion for inclusion, three
participants declared less than one ride per week, one of which was not an active rider at the
time of the tests. Concerning the mileage, the majority of the participants declared between
1000 and 5000 km per year. Sports bikes were the most common type of motorcycles owned
by the sample. The sample included a former police motorcyclist and a former professional
motorcycle tour guide (both still riding daily their motorcycles at the time of the tests).
Table 1. Summary of the participant riders involved in the validation
Age
n
%
Type of bike
owned
n
km travelled per
year
n
%
Frequency
of riding
n
%
21-30
6
37.50
adventure
2
<1,000
2
12.50
less than
once a week
3
18.75
31-40
4
25.00
cruiser
2
1,000 - 5,000
6
37.50
once a week
4
25.00
41-50
2
12.50
off-road
2
5,000 - 10,000
3
18.75
2-3 times per
week
3
18.75
>50
4
25.00
sports
5
10,000 - 15,000
3
18.75
4-6 times per
week
1
6.25
unknown
0
0.00
sports tourer
1
15,000 - 20,000
1
6.25
daily
5
31.25
standard
1
> 20,000
0
0.0
unknown
0.00
touring
3
unknown
1
6.25
unknown
0
Total
16
100.00
16
16
100.0
16
100.00
Procedure
Ethics approval for this study was granted by the Monash University Human Research Ethics
Committee (project n. CF15/180 -2015000084). All participants received an explanatory
statement with details of the study and expressed their informed consensus.
a
b
Before using the simulator, participants filled in the ‘demographic and riding’ questionnaire
collecting demographic data, riding attitudes, and opinions about motorcycle safety
technologies.
Warm-up phase. Before starting, participants were instructed on the counter-steer control
strategy for lateral control of the simulator. Participants familiarised with the motorcycle rig
and its controls in a country road environment, free from obstacles in the carriageway, in a
speed range between 40 km/h and 80 km/h (see Figure 2). The warm-up phase consisted in
two runs of 5 minutes each, with 2 minute break in between. A representation of the track
used for the tests is plotted in Figure 3. Except for the first run of the warm up, participants
wore helmet and gloves during all the tests runs.
Figure 2. Screenshot from the visual output of the simulator showing the road environment
used for the tests.
Test phase A. Participants performed four runs of 2-3 minutes each, riding the motorcycle
simulator along the same three-lane road setting used in the warm-up phase. In the first three
runs, participants were instructed to keep the vehicle centred in the middle lane, at the
constant speed of respectively 60 km/h, 80 km/h, and 100 km/h. In the fourth run,
participants were instructed to maintain the vehicle at a constant target speed, and to change
lane in a given sequence when indicated by the investigator, at intervals of 20 seconds. At the
end of each run, participants provided an evaluation of the handling of the simulator with
respect of the given task (i.e. tracking speed and steering). Handling qualities were rated in a
scale from 1 (excellent) to 10 (major deficiencies in the system) based on a rating scale
algorithm designed to evaluate the handling qualities of aircrafts. (The handling quality rating
chart used in the tests is provided in the Appendix.) At the end of the set of runs, participants
filled in a questionnaire with closed-ended questions on the realism of the following riding
conditions: (i) constant speed; (ii) braking and accelerating; (iii) steady cornering; and (iv)
lane change manoeuvres. Questionnaires involved five-option, Likert scale ratings ranging
from ‘strongly disagree’ to ‘strongly agree’ to capture participants’ agreement with a set of
statements. An example statement is the following: “While braking and accelerating, the
perception of speed change was realistic.” The Likert scale was then converted into a scale
from 0 (‘not realistic at all’) to 5 (‘highly realistic’). Additional questions addressed the
response of the control inputs, with options ranging from ‘too little’ to ‘too much’.
Figure 3. Representation of the test track adopted in phase A tests.
Test phase B. In this phase, participants performed slalom and lane change manoeuvres along
a straight road. The former manoeuvre consisted in a slalom around street cones aligned at a
distance of 21 m from each other (Figure 4a). The latter manoeuvre consisted in nominal
lateral deviations of 4 m in a longitudinal distance of 21 m, operated passing through
corridors of traffic cones, respectively 2 m and 4 m wide (Figure 4b). Participants performed
sets of three runs of slalom and lane change at each one of the following speeds: (i) 40 km/h;
(ii) 60 km/h; and (iii) 80 km/h. At the end of each set, participants evaluated the handling
qualities of the simulator for the specific manoeuvre at the given speed. At the end of this
phase B, participants filled in closed-end questionnaire addressing the realism of the
simulator during the slalom and during the lane change manoeuvres.
At the end of the test session, participants provided their subjective evaluation in writing via
open-ended questionnaire.
Figure 4. (a) Slalom test setting. (b) Lane change test setting.
Body lean strategy. Four participants constituted a first pilot group and did not perform phase
B. Indeed, at the end of phase A, they were instructed to implement small lean movements of
the body in the direction of the turn, in combination with the counter steer input. This will be
indicated with the name of ‘body lean strategy’. The four participants then repeated the whole
set of tests of phase A, and filled in again the evaluation forms.
x (m)
-1500 -1000
-500
0 500
y (m)
0
500
1000
1500
For the remaining twelve participants, the body lean strategy was introduced during the
warm-up phase as optional for the remainder of the tests. This second group of participants
performed both phase A and phase B tests.
Data analysis
The simulator was programmed to record the following parameters of the motorcycle during
the test: (i) Cartesian coordinates of the centre of gravity of the host vehicle in an absolute
reference system; (ii) lateral displacement of the host vehicle in a reference system aligned
with the road (natural coordinates); (iii) longitudinal speed of the host vehicle; (iv) handlebar
rotation; and (v) throttle and brake control values. For the handlebar mechanism, linear
regression of a static calibration was used to compute steering torque values based on steering
angles. This approximation was considered acceptable for the scope of the present study,
given the relatively low frequency of steer inputs operated by participants (main component
of the steer torque lower than 1 Hz in the tested manoeuvres).
The objective analysis focused on the following manoeuvres: (i) steady state cornering; (ii)
slalom; and (iii) lane change. For steady state cornering, steer torque actions applied by
participants in correspondence of a 40 m arc of 200 m radius curve in the first three runs of
phase A were considered. For slalom and lane change manoeuvres, successful attempts with
the lowest steering torque values were identified for each participant in each test condition.
For each selected run, mean speed (v
m
) and peak to peak values of the steering torque (τ
p-p
)
were computed. Results from the simulator tests were then compared with real-world tests
available in the literature and numerical simulations obtained with the software BikeSim, the
latter used as a surrogate of real world data.
Concerning subjective data, descriptive statistics were supported with statistical tests
performed using two-sample, unequal variances T-Student test.
Results
Sixteen and 12 participants completed respectively test phase A and B.
Objective data
Datasets from the tests of two participants (P06 and P13) were not available for the analysis
due to an unexpected fault in the recording script.
Steady state cornering
The values of steering torque and the mean speed for each participant along the constant
radius turn in phase A tests are presented in Table 2. As expected, the right hand side curve
(clock wise heading rotation) was negotiated while applying a counter-clock wise steer torque
(opposite to the heading rotation). Torque values ranged from 5.0 Nm to 6.7 Nm at the target
speed of 60 km/h, and from 6.9 Nm to 10.0 Nm at the target speed of 100 km/h. Average
torque values were typically higher at higher target speeds. The results were compared with
steady state torque values computed with Bikesim (Table 3). Bikesim results were consistent
with the values presented in the literature.[22] At 60 km/h, the torque measured in the
simulator was higher than the values obtained with Bikesim. At 80 km/h and 100 km/h the
torque values measured in the simulator were closer to those computed for a small sports
bike, and consistent with those of a large touring bike.
Table 2. Mean speed and steady state steer torque values adopted by participants during
phase A tests while negotiating a 200 m radius curve (single attempt at each one of the three
target speeds).
Particip
ant
Target speed: 60 km/h
Target speed: 80 km/h
Target speed: 100 km/h
Mean speed
(km/h)
Mean steer
torque
(Nm)
Mean speed
(km/h)
Mean steer
torque
(Nm)
Mean speed
(km/h)
Mean steer
torque
(Nm)
P00
na
na
na
na
na
na
P01
58.3
5.0
76.4
6.6
95.3
7.0
P02
56.8
5.2
74.0
5.9
96.8
6.6
P03
60.7
5.7
78.4
6.9
100.4
9.7
P04
59.2
5.7
78.4
7.4
99.9
6.9
P05
60.8
5.2
78.4
5.0
99.0
8.2
P06
na
na
na
na
na
na
P07
63.8
5.6
82.6
6.0
100.8
9.8
P08
63.5
5.6
80.8
6.5
99.3
9.0
P09
61.1
5.7
81.3
7.4
95.3
8.8
P10
53.7
6.7
82.2
6.1
97.7
7.0
P11
58.4
5.1
79.7
6.5
101.0
7.1
P12
60.6
5.7
80.9
7.0
na
na
P13
na
na
na
na
na
na
P14
61.3
5.8
82.4
7.4
100.5
10.0
P15
60.9
6.1
78.0
7.2
99.5
8.8
Mean
59.9
5.6
79.5
6.6
98.8
8.2
Table 3. Steer torques computed with baseline motorcycle models in BikeSim.
Manoeuvre 200 m radius curve Slalom
Lane change (lateral
displacement 4 m)
Target speed
(km/h)
60 80 100 60 40 60 80
Steer torque type
Steady state
τ (Nm)
Peak to peak
τ
p-p
(Nm)
Peak to peak
τ
p-p
(Nm)
Big cruiser
1.1
1.7
2.3
81.9
7.5
20
32.9
Big touring
motorcycle
3.2 4.4 5.8 88.9 10.6 25.7 42.5
Small sports
motorcycle
2.9 5.2 8.1 71.5 8.2 17.6 29
Slalom manoeuvres
Results from a subset of the successful slalom manoeuvres performed in phase B tests are
provided in Table 4. For each participant, we focused on the runs requiring the minimum
effort to accomplish the task. The inter-participant variability in the τ
p-p
was remarkable.
Even when restricting the analysis to the runs performed with a deviation from the target
speed within the range ±10%, the maximum value was almost double the minimum value in
all the three target speeds. However, the inter-participant mean values of τ
p-p
were similar for
the three target speeds, ranging from 30.4 Nm to 36.9 Nm respectively at 80 km/h and 60
km/h. A representative example of a slalom manoeuvre executed with the simulator is plotted
in Figure 5. The magnitude and phase of the steer torque signal were compared with the
results of the on road testing presented by Cossalter, et al. [27] (vehicle: Aprilia Mana 850;
cone distance: 21 m; mean speed: 68.8 km/h; peak torque: 45.3 Nm; τ
p-p
phase: 3.4 rad). The
steer torque was applied approximately in phase-opposition both in the real motorcycle and in
the simulator, but the peak torque measured in the real motorcycle was almost double the
values measured in the simulator. BikeSim simulations were consistent with Cossalter’s tests
(see Table 3).
Figure 5. Participant P07 performing the slalom with target speed 60 km/h (mean speed 62.1
km/h).
Table 4. Mean speed and peak to peak steer torque values adopted by participants during the
slalom tests in phase B.
Participant
Target speed: 40 km/h
Target speed: 60 km/h
Target speed: 80 km/h
v
m
(km/h)
τ
p-p
(Nm)
v
m
(km/h)
τ
p-p
(Nm)
v
m
(km/h)
τ
p-p
(Nm)
P00
na
na
na
na
na
na
P01
na
na
na
na
na
na
P02
na
na
na
na
na
na
P03
na
na
na
na
na
na
P04
40.1
35.6
61.4
49.3
86.9
45.3
P05
39.7
28.3
58.1
23.4
na
na
P06
na
na
na
na
na
na
P07
35.6
28.3
62.1
24.2
83.3
38.0
P08
52.2
53.4
53.2
46.1
na
na
P09
45.0
46.1
67.1
18.6
87.9
17.8
P10
41.6
25.9
57.6
31.5
76.8
27.5
P11
41.5
20.2
65.2
38.0
82.6
23.4
P12
44.6
27.5
68.0
38.0
82.7
20.2
P13
na
na
na
na
na
na
P14
40.0
32.3
60.9
28.3
82.4
44.5
P15
32.6
37.2
57.2
71.2
77.5
26.7
Mean
41.3
33.5
61.1
36.9
82.5
30.4
Lane change manoeuvres
Results from a subset of the successful lane change manoeuvres performed in phase B tests
are provided in Table 5. For each participant, we focused on the runs requiring the minimum
effort to accomplish the task. For this manoeuvre, the inter-participant variability was much
smaller than for the slalom, and the mean values of τ
p-p
were generally higher at higher speed.
A representative example of a lane change manoeuvre executed with the simulator is plotted
in Figure 6. Participant P07 performing a lane change with target speed 40 km/h (mean speed
Time (s)
15 20 25
-15
-10
-5
0
5
10
15
Steer torque (Nm)
Lateral position x10 (m)
37.9 km/h). Also for lane change, the magnitude of the steer torque signal was compared with
the results of the on road testing presented by Cossalter, et al. [27] (vehicle: Aprilia Mana
850; lateral displacement: 3 m; mean speed: 55.3 km/h; τ
p-p
: 84.0 Nm). Consistently with the
slalom manoeuvre, the peak torque measured in the real motorcycle was higher than the
typical values measured during the simulated manoeuvres. BikeSim simulations showed
lower steer torque inputs needed to perform the lane change compared to the real motorcycle.
BikeSim values obtained with the three vehicles at target speeds of 60 km/h and 80 km/h
were consistent with the steering torque inputs measured in the simulator at the same speeds
(see Table 3).
Figure 6. Participant P07 performing a lane change with target speed 40 km/h (mean speed
37.9 km/h).
Table 5. Mean speed and peak to peak steer torque values adopted by participants during the
lane change tests in phase B.
Participant
Target speed: 40 km/h
Target speed: 60 km/h
Target speed: 80 km/h
v
m
(km/h)
τ
p-p
(Nm)
v
m
(km/h)
τ
p-p
(Nm)
v
m
(km/h)
τ
p-p
(Nm)
P00
na
na
na
na
na
na
P01
na
na
na
na
na
na
P02
na
na
na
na
na
na
P03
na
na
na
na
na
na
P04
36.1
29.1
62.0
36.4
75.6
43.7
P05
39.3
29.9
56.3
59.9
79.9
30.7
P06
na
na
na
na
na
na
P07
37.9
26.7
57.2
28.3
81.2
29.9
P08
40.1
27.5
59.4
45.3
80.4
57.4
P09
41.5
21.0
56.6
13.7
84.3
16.9
P10
45.2
14.5
64.2
19.4
79.0
22.6
P11
39.9
17.8
61.1
16.1
78.4
29.9
P12
42.5
21.8
68.8
27.5
88.2
40.4
P13
na
na
na
na
na
na
P14
42.6
20.2
59.9
22.6
80.0
29.9
P15
37.4
22.6
55.5
30.7
77.5
55.0
Mean
40.3
23.1
60.1
30.0
80.5
35.7
Time (s)
27
28 29
30 31 32 33
34
-20
-10
0
10
20
Steer torque (Nm)
Lateral position x10 (m)
Subjective data
Body lean strategy
In the pilot study focusing on the body lean strategy, three out of four participants repeated
phase A runs after introducing this strategy (one participant withdrew after phase A due to
discomfort). The responses from this subset of participants, supported by handling ratings and
questionnaire results, indicated that counter steer inputs can be more intuitive implementing
the body lean strategy. Consequently, body lean strategy was introduced as optional during
the warm-up phase for the following participants. Finally, all participants implemented this
strategy during their tests.
Handling quality ratings
Participants rated phase A runs in the range from 1 (excellent) to 5 (moderately objectionable
deficiencies), with overall mean value of 2.96 (standard deviation 0.87). In phase A (country
road riding), the mean ratings were consistent across the speed range considered. When
involving lane change, the handling score in the same country road environment was poorer
than the basic scenario without lane change. In order to test the effects of adaptation to the
simulator with respect to the handling perception, six participants repeated the final lane
change test of phase A after completing phase B. The ratings for the lane change task in
country road environment performed at the end of the test session were slightly lower (mean
ratings at first and second attempt respectively 3.17 and 2.33). This suggested that the
simulator achieved good levels of handling quality (ratings around 3) in short time, with
slight improvement as participants got more used to it.
Concerning phase B, lane change manoeuvres reported better handling ratings than slalom. In
fact, the mean ratings for slalom and lane change were respectively 4.31 and 2.94 (T-Student
p<0.001). For the slalom, handling ratings were poorer at 80 km/h than at 40 km/h (T-
Student, p<0.005).
Mean handling quality ratings provided by the subgroup of twelve participants are given in
Table 6.
Table 6. Mean handling quality ratings in the different test sets in the range 1 (excellent) to
10 (major deficiencies in the system).
Phase A
Phase B
Target speed
(km/h)
Country road
Country road,
lane change
Slalom
Lane change
40
3.33 (1.23)
2.67 (1.07)
60
2.92 (0.80)
4.17 (1.27)
2.58 (1.00)
80
2.83 (0.83)
3.17 (0.94)
5.41 (1.88)
3.58 (1.83)
100
2.92 (0.90)
(Standard deviation in brackets)
Questionnaires
The results of the questionnaires for the sixteen participants were synthesised in the form of
radar-type graphs in Figure 7. This representation allowed for comparisons with previous
studies, in particular with the reference validation study.[27]
Figure 7. Mean scores for phase A and B tests in the range from 0 (‘not realistic at all’) to 5
(‘highly realistic’)
Open-ended questions addressed the following aspects: opinion about the simulator; opinion
about the steering control; and likes/dislikes. Concerning the overall opinion, statements
declaring general appreciation for the simulator and its good realism were frequent
(respectively seven and six instances). Three participants also highlighted the good quality of
the visual cues. Four statements indicated an initial disconcert with the steer control and four
participants declared that steering was counter-intuitive or not completely realistic during the
tests. Five statements expressed negative opinion about the motion cues (not enough or not
well correlated with the steering). When asked directly about the steering control, two
participants also indicated that it was difficult to get used to the steer input. Specific
deficiencies of the steer input were reported, with not full agreement: too much or not enough
sensitive (respectively two and three statements), and slow in its response (two statements).
Four statements highlighted the fact that the steering control became natural after some
practice, and in seven instances, participants expressed good appreciation for the steering
system. Participants liked the realism of the simulator, the visual and auditory cues, the
vibration cues, the motion cues during longitudinal accelerations, and the fact that a real bike
was used for the rig. Participants disliked the roll cues, the pitch cues, the throttle response,
and the steering response at low speeds. Only one participant expressed explicit dislike for
the counter steer approach adopted in the simulator. None of the participants expressed
negative opinions about the non-tilting horizon in the visual cues. Two participants noticed
that the dynamical behaviour of the bike resembled a passenger car while negotiating a curve,
due to the speed reduction produced by the turning manoeuvre. A synthesis of the responses
is provided in Table 7.
Constant speed
(phase A)
Acceleration and
braking (phase A)
Cornering (phase A)
(phase B)
Table 7. Synthesis of the responses provided by participants in the open-ended questions.
Positive opinions
Negative opinions
Question
Statements
Frequency
N.
Statements
Frequency
N.
Overall
opinion
Good/very
good/excellent
7
Disconcerting at start
4
Realistic
6
Counter
intuitive/unrealistic
steering
4
Good visual cues
3
Not enough physical
lean
3
Controllable
2
Slow steer response
2
Behaves like a car
2
Not enough pitch
1
Motion cues are
confusing
1
Throttle response
1
Steering
control
Quite good/good/very
good steering
7
Sensitiveness: Too
much/not enough
5
User was able to adapt
to steering
4
Not enough physical
lean
3
Realistic
1
Counter intuitive/weird
3
Intuitive
1
Disconcerting at start
3
Difficult to adapt
2
Slow response
2
A little demanding
1
Too soft feedback
1
Likes/Dislikes
Realism
5
Motion cues were
confusing
3
Auditory cues
5
Not enough pitch cues
3
Visual cues
5
Brakes
3
Pitch cues while
braking/accelerating
4
Difficult to maintain
target speed
4
Vibrations
4
Weird turning
behaviour at low speeds
2
Real motorcycle rig
3
Not enough lean cues
2
Brakes
1
Slow steer response
2
Speed sensation
1
Too much lean cues
1
Poor visual textures
1
Counter steer
1
Discussion
The aim of this study was to validate a motorcycle simulator that implemented counter steer
input strategy with realistic feedback on the handlebar, obtained via simple elastic
mechanism. This low-cost upgrade of an existing simulator was designed to investigate
realistic steering inputs of the rider for the purposes of the development of rider assistance
systems such as MAEB.
The validation process presented in this paper produced encouraging results both from
objective and subjective viewpoints. Considering the objective validation in standard
manoeuvres, general agreement was found between steer inputs applied in the simulator and
those applied in reference tests involving real and simulated motorcycles.
In steady state cornering, the magnitude of steering torques was generally higher in the
simulator. However, the sign of the inputs and the trend with speed variations were consistent
with riding a real motorcycle. It is worth noticing that realism of steady state cornering in
country road setting is important for the scopes of the simulator. In fact, an essential
condition for investigating steering reactions of the rider in unexpected, critical events is that
participants are subjected to a realistic virtual ride in normal conditions involving steer inputs
such as negotiating curves in a country road environment.
Slalom and lane change tests were challenging for participants, in particular at higher speed,
especially because of the absence of a specific warm-up session for these manoeuvres.
However, the best attempt was often achieved in the first run (30% of the tests in phase B).
This is particularly interesting in the perspective of investigating the rider behaviour when
facing unexpected events.
Concerning the slalom, it is worth noting that driving simulator studies typically avoid rapid
and repeated cornering due to difficulties in achieving decent fidelity. In our tests, despite the
magnitude of steer inputs was not fully coherent with real world and simulated data, shape,
signs, and phase of the inputs were consistent with real riding.
Lane change is highly relevant for the development of assistance systems that operate vehicle
control actions in the pre-crash phase (such as MAEB, which applies autonomous braking)
that may interfere with the rider’s steer inputs. In fact, this type of manoeuver can be
considered an approximation of an emergency lateral avoidance manoeuvre; see for example
Giovannini, et al. [28]. Results from the experiments indicated consistency between the
inputs for lane change recorded in the simulator, those measured in real world data and those
simulated with detailed motorcycle models. Furthermore, steering input magnitudes in the
simulator were comparable with BikeSim data. In conclusion, lateral control in demanding
conditions appeared realistic from a quantitative perspective.
Regarding the subjective assessment, ratings provided by participants indicated good
handling qualities and realism of the simulator in country road setting (phase A). As said, this
aspect is important to allow participants immerse in the virtual environment prior to
presenting unexpected events, in the perspective of investigating emergency reactions.
Participants also indicated good handling ratings for the slalom and the lane change tests.
Furthermore, results suggested that these handling properties were achieved quickly.
Concerning realism, the overall results of the subjective evaluation were comparable with
those of more sophisticated and complex simulators presented in the literature.[27] Responses
to the open-ended questions highlighted the good level of visual and auditory realism. These
aspects play an important role as they contribute in the process of adaptation to the simulated
environment. Some participants’ responses also indicated that counter-steering was
occasionally perceived as counter-intuitive, confirming the results of previous studies.[25]
This must be taken into account when designing future experiments with the simulator.
Finally, further investigations could try to clarify the contribution of what we called “body
lean strategy”, which in our study seemed to improve the perceived realism of our simple
motorcycle simulators.
Limitations
This validation study focused on medium-high speeds. At speeds lower than 40 km/h, the
behaviour of the passenger car model used in the physical engine of the simulator deviates
remarkably from a motorcycle model. The present setup is expected to achieve poor levels of
realism at lower speeds. Simulating a motorcycle at low speed is particularly challenging
even when using a detailed motorcycle model, as shown in previous studies.[29] Other
studies recommended avoiding counter-steer strategies at low speeds.[25] Considering that
MAEB is relevant typically from 30 km/h,[30] further consideration should be given to
identify low-cost options for low-speed, realistic riding simulations.
Conclusions
This process of validation showed that a low-cost motorcycle simulator could achieve a level
of realism that from a point of view of the subjective assessment is comparable with much
more sophisticated solutions. Despite the vehicle dynamics was based on a passenger car
model, participants generally accepted the steering assembly equipped with elastic resistance
and implementing counter-steer strategy. Recommending leaning the body while steering was
a simple way to improve realism. From a quantitative point of view, results indicated that this
low-cost simulator can be adopted to measure realistic steer inputs in lateral avoidance
scenarios at medium-high speeds. In order to fully validate the use of this simulator as a tool
for the development of rider assistance system, further work should investigate also
the realism of participants’ reactions when simulating unexpected, emergency situations.
Acknowledgements
The research leading to these results has received funding from the European Union Seventh
Framework Programme FP7/2007-2013 under grant agreement n. 328067 (ABRAM project).
The authors acknowledge Mr Nebojsa Tomasevic for his technical support.
The authors also acknowledge the Monash Injury Research Institute, Monash University, for
providing funding and equipment.
References
[1] IMMA, "Motorcycle Safety: IMMA's contribution to the Decade of Action for Road
Safety 2010-2020," 2010.
[2] N. Haworth, "Powered two wheelers in a changing world-challenges and
opportunities," Accid Anal Prev, vol. 44, pp. 12-8, Jan 2012.
[3] R. A. Blackman and N. L. Haworth, "Comparison of moped, scooter and motorcycle
crash risk and crash severity," Accid Anal Prev, vol. 57, pp. 1-9, Aug 2013.
[4] E. C. Yeh, G. K. Roan, and I. H. Yun, "Development of an Anti-Lock Brake System
for Motorcycle," Vehicle System Dynamics, vol. 24, pp. 427-444, 1995/06/01 1995.
[5] M. Corno and G. Panzani, "Traction Control Systems Design: A Systematic
Approach," in Modelling, Simulation and Control of Two-Wheeled Vehicles, ed: John
Wiley & Sons, Ltd, 2014, pp. 198-220.
[6] F. Biral, R. Lot, S. Rota, M. Fontana, and V. Huth, "Intersection Support System for
Powered Two-Wheeled Vehicles: Threat Assessment Based on a Receding Horizon
Approach," Ieee Transactions on Intelligent Transportation Systems, vol. 13, pp. 805-
816, Jun 2012.
[7] F. Biral, P. Bosetti, and R. Lot, "Experimental evaluation of a system for assisting
motorcyclists to safely ride road bends," European Transport Research Review, vol.
6, pp. 411-423, 2014/12/01 2014.
[8] D. Ruscio, M. R. Ciceri, and F. Biassoni, "How does a collision warning system shape
driver's brake response time? The influence of expectancy and automation
complacency on real-life emergency braking," Accident Analysis & Prevention, vol.
77, pp. 72-81, 4// 2015.
[9] G. Savino, F. Giovannini, N. Baldanzini, and M. Pierini, "Training system for optimal
braking on a powered two-wheeler," in Driver Behaviour and Training. vol. 5, ed,
2012, pp. 101-125.
[10] F. Kobiela and A. Engeln, "Autonomous emergency braking studies on driver
behaviour," ATZ worldwide, vol. 112, pp. 4-8, 2010.
[11] D. Miller, A. Sun, and W. Ju, "Situation awareness with different levels of
automation," in Systems, Man and Cybernetics (SMC), 2014 IEEE International
Conference on, 2014, pp. 688-693.
[12] M. Keller, C. Hass, A. Seewald, and T. Bertram, "Driving simulator study on an
emergency steering assist," in Systems, Man and Cybernetics (SMC), 2014 IEEE
International Conference on, 2014, pp. 3008-3013.
[13] G. Markkula, O. Benderius, K. Wolff, and M. Wahde, "Effects of experience and
electronic stability control on low friction collision avoidance in a truck driving
simulator," Accident Analysis & Prevention, vol. 50, pp. 1266-1277, 2013.
[14] M. Itoh, M.-P. P. Lemoine, F. Robache, and H. Morvan, "An Analysis of Driver's
Avoiding Maneuver in a Highly Emergency Situation," SICE Journal of Control,
Measurement, and System Integration, vol. 8, pp. 27-33, 2015.
[15] A. Schieben, S. Griesche, T. Hesse, N. Fricke, and M. Baumann, "Evaluation of three
different interaction designs for an automatic steering intervention," Transportation
research part F: traffic psychology and behaviour, vol. 27, pp. 238-251, 2014.
[16] D. L. Fisher, M. Rizzo, J. Caird, and J. D. Lee, Handbook of driving simulation for
engineering, medicine, and psychology: CRC Press, 2011.
[17] J. A. Greenberg and M. Blommer, "Physical fidelity of driving simulators," 2011.
[18] S. Benedetto, R. Lobjois, V. Faure, N.-T. Dang, M. Pedrotti, and S. Caro, "A
comparison of immersive and interactive motorcycle simulator configurations,"
Transportation research part F: traffic psychology and behaviour, vol. 23, pp. 88-
100, 2014.
[19] V. Cossalter, R. Lot, and S. Rota, "Objective and subjective evaluation of an
advanced motorcycle riding simulator," European transport research review, vol. 2,
pp. 223-233, 2010.
[20] V. Cossalter, R. Lot, and M. Massaro, "An advanced multibody code for handling and
stability analysis of motorcycles," Meccanica, vol. 46, pp. 943-958, 2011.
[21] BMW, "Motorcycle riding simulation to assess instrument and operation concepts and
informing riding assistance systems," presented at the International Motorcycle
Conference IFZ, Cologne,Germany, 2014.
[22] V. Cossalter, Motorcycle dynamics: Lulu. com, 2006.
[23] A. Popov, S. Rowell, and J. Meijaard, "A review on motorcycle and rider modelling
for steering control," Vehicle System Dynamics, vol. 48, pp. 775-792, 2010.
[24] J. Kooijman and A. Schwab, "A review on bicycle and motorcycle rider control with a
perspective on handling qualities," Vehicle System Dynamics, vol. 51, pp. 1722-1764,
2013.
[25] A. Stedmon, E. Brickell, M. Hancox, J. Noble, and D. Rice, "MotorcycleSim: A user-
centred approach in developing a simulator for motorcycle ergonomics and rider
human factors research," Advances in Transportation Studies, vol. 27, 2012.
[26] A. J. Filtness, C. Rudin-Brown, C. Mulvihill, and M. G. Lenné, "Impairment of
simulated motorcycle riding performance under low dose alcohol," Accident Analysis
& Prevention, vol. 50, pp. 608-615, 2013.
[27] V. Cossalter, R. Lot, M. Massaro, and R. Sartori, "Development and validation of an
advanced motorcycle riding simulator," Proceedings of the Institution of Mechanical
Engineers, Part D: Journal of Automobile Engineering, vol. 225, pp. 705-720, 2011.
[28] F. Giovannini, G. Savino, M. Pierini, and N. Baldanzini, "Analysis of the minimum
swerving distance for the development of a motorcycle autonomous braking system,"
Accident Analysis and Prevention, vol. 59, pp. 170-184, 2013.
[29] J. F. Lenkeit, B. K. Hagoski, and A. I. Bakker, "A Study of Motorcycle Rider Braking
Control Behavior," U.S. Department of Transportation2011.
[30] G. Savino, M. Rizzi, J. Brown, S. Piantini, L. Meredith, B. Albanese, et al., "Further
Development of Motorcycle Autonomous Emergency Braking (MAEB), What Can
In-Depth Studies Tell Us? A Multinational Study," Traffic Injury Prevention, vol. 15,
pp. S165-S172, 2014.
Appendix
Figure 8. Handling qualities rating scale
This research hasn't been cited in any other publications.
  • Chapter
    The Problem. Driving simulators are often used to explore important facets of the transportation landscape. They have been used to measure driver response to new technologies like crash avoidance systems or to assess the effect of distracting tasks on situation awareness. Both the authors and the readers of these studies hope that the results obtained in the simulator can be generalized to the real world. But in deciding how or even if such a generalization is warranted we are confronted with questions about the fidelity of the simulation. Did the simulator accurately represent the critical aspects of the driving experience? In addition to the visual scene, did the driver receive other cues for audio, motion, or tactile perception? How accurately were these cues delivered? And most critically, how might the known errors in the simulation affect the results? A Framework for Understanding Fidelity. An important way to begin answering these questions is to gain an understanding of what contributes to the physical fidelity of a driving simulator. We introduce a four-step framework that describes how any physical aspect of the driving experience (e.g., the acceleration that occurs when a vehicle changes lanes) is transformed from an abstract representation into a sensation. The framework allows us to examine the types of errors and distortions that occur at each stage of the cue-rendering process. Deciding When Fidelity Matters. Assessing the importance of these errors for any given experimental (or training) goal requires an estimation of the experimental artifacts that may result. If the artifacts are controllable, the fidelity of the simulator is sufficient. This viewpoint turns the unanswerable question “Is my simulator valid?” into a more practical one: “Do I understand the errors in my simulator well enough to explain their effect on my experimental data?” The treatment of cueing systems in this chapter is not meant to be comprehensive. Specific aspects of the motion, steering, and audio systems are explored in some depth in order that simulator users can carry out the four-step process for their particular applications. In spite of its obvious importance, the visual system is not considered in detail here but has been left to more knowledgeable authors.
  • Article
    This contribution is concerned with an emergency steering assist and its evaluation through subject testing in a driving simulator. In an emergency traffic situation where a rear end collision is imminent a swerving maneuver is often too difficult for most drivers. Therefore an assistance system is usefull that supports the driver by steering torque overlay. The paper presents the algorithm used for the assistance. The system was prototypically implemented in a driving simulator and testet with subjects to evaluate the benefits and challenges. The results show that the collision avoidance behaviour of the driver can be improved by the emergency steering assist.
  • Article
    What effect will periods of automated driving will have on driver performance after transfer of control? In our driving simulator experiment (N = 48) participants in four different automation conditions (fully autonomous vehicle, autonomous steering, autonomous speed control, no automation) were evaluated based on their post-transition accident avoidance, situational awareness, and feelings of trust in and comfort with autonomous or partially autonomous driving. Preliminary results from behavioral data show significant differences in time to initiate evasive action across conditions. Participants in the fully autonomous condition showed greater trust and comfort with the car's autonomous features than those in the autonomous speed control condition.
  • Article
    This chapter presents a systematic approach to the design of traction control (TC) systems for motorcycles. Several aspects are considered: activation strategy, reference generation and gain-scheduling. Tests on instrumented motorcycles are used to show several critical aspects. The chapter presents a complete and systematic treatment of the issues that practitioners face when designing a TC system for a road motorcycle. It first discusses the most relevant concepts regarding wheel slip dynamics. Next, the TC system is designed, taking into account many aspects: activation, deactivation, wheel slip reference generation and robustness. Fine tuning and testing of the TC system are explained in the last section.
  • Article
    The evasive capabilities of motorcycles and riders are often an important consideration when analyzing a motorcycle crash. Specifically, the longitudinal distance or time required for a motorcycle to move laterally some distance can be of critical interest. Previous publications on this topic have not all measured the same thing and have often included limited test data so their results can be difficult to compare or apply. In addition to reviewing some of the literature on the topic, this paper will present the results of a series of tests conducted with four riders on four motorcycles swerving 2 m (6.5 ft) to their left after passing through a gate at speeds of 40 to 88 km/h (25 to 55 mi/h). The most recent testing involved relatively skilled riders who had faster transitions and greater willingness to lean than the "average" rider generally described in the literature. Separating the perception-reaction time from the evaluation of the turn-away maneuver itself simplifies the analysis, though wide individual performance variation still exists.
  • Chapter
    The Problem. To date, there has been no single convenient and comprehensive source of information on driving simulation research being conducted around the world. Nor has there been a single repository for information regarding the numerous challenges that confront new simulator users or the broader challenges that confront the whole community. The Handbook of Driving Simulation for Engineering, Medicine and Psychology strives to put much of this critical information in one easily accessible place. Role of Driving Simulators. Driving simulation is now a key element in regional, national, and international efforts to probe the: (1) efficacy of novice, commercial, and older driver training programs; (2) fitness to drive in patients with performance declines due to aging and mild cognitive impairment, traumatic brain injury, and neurodegenerative disorders such as Alzheimer’s and Parkinson’s diseases, and sleep disturbances; (3) acute and chronic effects of many medications such as analgesics, antidepressants, psychostimulants, antidepressants, and cancer chemotherapy agents; (4) impact of alternative signs, signals, and pavement markings on drivers’ behavior; and (5) relative advantages and/or disadvantages of technologies being used or introduced into the vehicle such as cell phones, left-turn and rear-end collision warning systems, advanced cruise control, navigation aids, and internet access. Key Results of Driving Simulator Studies. The chapters in the Handbook detail the many key results. In the broadest of terms, there has been a literal explosion in our understanding of the differences in vehicle and driver behaviors among different groups of people; of the impact of different traffic control devices, traffic scenarios, road geometries, and general lighting and weather conditions on vehicle and driver performance; of the effect of different types of distraction inside and outside the vehicle on various metrics of performance; and on the use of driving simulators as assessment and training tools. Scenarios and Dependent Variables. The chapters are replete with discussions of the scenarios and dependent variables that are critical for analysis purposes. Additionally, there are key sections specifically targeting scenario authoring and dependent variables. Limitations. There are still a number of “evergreen” issues. They include simulator adaptation syndrome, fidelity of the virtual environment, cross-platform comparisons among different simulators, development of standard scenarios for testing and training, and transfer of effects to the real world.
  • Article
    Full-text available
    This paper discusses driver's emergency avoiding behavior while he or she is driving a car in a rural area. Based on behavioral data obtained from an experiment conducted in France with a high-fidelity driving simulator, the authors analyzed the driver behavior just before the collision. The results showed that the drivers tended to choose a “rational” behavior to avoid the collision by using the available visual cues even though the situation is extremely time critical, i.e., the time to collision was only approximately 1.5 seconds.
  • Article
    Full-text available
    ‘MotorcycleSim’ is a new development in motorcycle simulation that is being built by the University of Nottingham for research into motorcycle ergonomics and rider human factors. It is the first motorcycle simulator to incorporate ‘STI-SIM Drive’ software that allows motorcyclists to ride a full size motorcycle and interact with a virtual riding environment. The riding scenarios are projected onto a large screen in front of the rider, the motorcycle has the potential to lean in a realistic manner and the ‘STI-SIM Drive’ software provides enhanced perceptions of acceleration, braking and leaning effects. To build a simulator that was both fit for the purpose of research and provides the desired levels of fidelity associated with real world riding, a user-centred design process was adopted from the outset (based in principles of ISO:13407). MotorcycleSim provides a usable tool for high impact research. This paper presents the background to the development of the simulator from initial concepts through to the design of the rig, integration of controls with simulation software and development of different riding modes. Key factors in the design process are presented which offer a basis for the future development of motorcycle simulators.
  • Article
    Traffic fatalities and injuries continue to demand the attention of researchers and governments across the world as they remain significant factors in public health and safety. Enhanced legislature, together with vehicle and roadway technology, has helped to reduce the impact of traffic crashes in many scenarios. However, one specifically troublesome area of traffic safety which persists is the run-off-road crash where a vehicle’s wheels leave the paved portion of the roadway and begin to travel on the shoulder or side of the road. Large percentages of fatal and injury traffic crashes are attributed to run-off-road events. One of the most critical reasons why run-off-road scenarios quickly evolve into serious crashes is poor driver performance. Drivers are unprepared to handle the situation safely and often execute dangerous maneuvers, such as overcorrection, which can lead to devastating results. One countermeasure which directly addresses the driver performance is driver education and training. In this article, a simulator-based driving environment is proposed specifically for run-off-road recovery training. A human subject study is used to validate the simulator as an effective tool for replicating the run-off-road experience with the additional benefit of receiving insight into driver reactions to run-off-road events. Analysis of variance (ANOVA) results of subjective questionnaire data and objective performance evaluation parameters show strong correlations to run-off-road crash data and previous run-off-road study conclusions. In particular, higher vehicle velocities, curved roads, and larger differences between the friction coefficient of the road and the friction coefficient of the shoulder all negatively impacted drivers’ recoveries from run-off-road scenarios. The only non-significant impact found was that of the roadway edge, indicating a possible limitation of the simulator system with respect to that particular environment variable. The validation study provides a foundation for further evaluation and development of a simulator-based run-off-road recovery training program to help to equip drivers with the skills to recognize and recover safely from this dangerous and often deadly scenario.