Abstract— The purpose of this study is to model, simulate and equations are then simulated in MATLAB-SIMULINK
verify the entire vehicle handling dynamics of UNS Electric software under throttle, steering and braking inputs. The
vehicle (Semar-T) using Matlab-Simulink. The modeling of simulation results are then verified using other vehicle
Semar-T electric vehicle begins from the derivation equation of dynamics software namely CarSimEd. Five types of vehicle
motion and Dugoff tire model. The equations are derived using dynamics test are performed for the purpose of model
Second Newton’s Law and then simulated in Matlab-Simulink validation namely step steer test, double lane change test,
environment. Five handling condition tests are performed for slalom test, sudden braking test and sudden acceleration test.
exploring the handling performance namely step steer test,
double-lane change test, slalom test, sudden braking test and This paper is organized as follows: the first section contains
sudden acceleration test. The disturbances input come from the an introduction on handling model. The details of the equations
lateral and longitudinal direction. Disturbance in the lateral for handling and Dugoff’s tire model are described in Section
direction is steer input while disturbances in the longitudinal II. The modeling block diagram Simulink is introduced and
direction are braking and throttling. The steering, braking and validated with CarSimEd software in section III. While section
throttling input in this study is taken from the verification IV discusses the results obtained from the proposed model, and
software namely CarSimEd. The results of the validation show the last section presents the conclusions.
that the vehicle model is valid based on verification using
CarSimEd with the error magnitude between 5-15 %. Therefore, II. GOVERNING EQUATION
the results of simulation from Matlab-Simulink can be applied to
studies the behavior handling dynamics of Semar-T electric A. Handling Model
vehicle. Figure 1 shows the free body diagram of vehicle handling
model. In vehicle handling model, it is assumed that the vehicle
Keywords— vehicle handling; CarSimEd; handling dynamics; handling is moving on a flat road. The handling model
Dugoff tire model
employed in this study is a 7-DOF system. It takes into account
I. INTRODUCTION three degrees of freedom for the vehicle body in lateral and
longitudinal motions as well as yaw motion (r) and one degree
Handling is a vehicle response occurred due to the presence of freedom from the rotational motion of each tire. The vehicle
of an input from the driver in which it belongs to the major part experiences motion along the longitudinal x-axis and the lateral
of active vehicle safety [1], [2]. Handling includes the balance y-axis, and the angular motions of yaw around the vertical z-
of the vehicle when performing the maneuver and when axis. The motion in the horizontal plane can be characterized
moving on the straight path [3]. The handling dynamic by the longitudinal and lateral accelerations, denoted by ax and
behavior of the vehicle is determined by the forces imposed on ay respectively, and the velocities in longitudinal and lateral
the vehicle by the tire, gravity, and aerodynamics. The vehicle direction, denoted by vx and vy, respectively [7], [8].
and its components are studied to determine what force will be
produced by each of these sources at a particular maneuver and
trim condition, and how the vehicle will respond to these forces
[4]. The behavior of vehicle handling dynamics can be learned
from modeling and simulation to safe the vehicle development
cost [5], [6].
This study deals with the investigation on the development
of full DOF vehicle ride, tire and handling models. The
modeling begins from the building tire model in which the
Dugoff Tire Model is used in this study. The tire models are
then integrated with the derived handling model. All governed
equations are derived using second Newton’s Law. The derived Fig. 1. Free body diagram of vehicle handling
where, ߙ and ߙ correspond to the side slip angles at front and ܶ and ܶ are the applied throttling torques for the front and
rear tires, lf and lr are the distance between front and rear tire to rear wheels [9], [10].
the body center of gravity, respectively. To calculate the C. Dugoff’s Tire model
longitudinal slip, longitudinal component of the tire velocity
should be derived. The front and rear longitudinal velocity Tire model considered in this study is Dugoff model [11].
component is given by Eq. (7) and Eq. (9), Tire model can describe the behavior of tire in any driving
conditions such us input steering, braking, throttling by the
ݒ௪௫ ൌ ܸ௧
ߙ (7) driver [12]. The longitudinal and lateral forces generated by a
tire are a function of the slip angle and longitudinal slip of tire
ଶ
ܸ௧ ൌ ට൫ݒ௬ ݈ ݎ൯ ݒ௫ଶ (8) about the road. Dugoff’s model provides for the calculation of
forces under combined lateral and longitudinal tire force
ݒ௪௫ ൌ ܸ௧
ߙ (9) generation. It is assumed that a uniform vertical pressure
distributes on tire contact path [4], [13]. The Dugoff’s tire
ଶ model equations are shown as follows (Eq. (18) and Eq. (19)),
ܸ௧ ൌ ට൫ݒ௬ ݈ ݎ൯ ݒ௫ଶ (10)
The longitudinal tire force can be determined from Eq.
Then, the longitudinal slip ratio of front and rear tires are (18),
determined from Eq. (11) and Eq. (12), ఙೣ
ܨ௫ ൌ ܥఙ ݂ሺߣሻ (18)
ଵାఙೣ
௩ೢೣ ିఠ ோೢ
ݏ ൌ (11)
௩ೢೣ Meanwhile, the lateral tire force is obtained from Eq. (19),
௩ೢೣೝିఠೝ ோೢ
ݏ ൌ (12) ୲ୟ୬ሺఈሻ
௩ೢೣೝ ܨ௬ ൌ ܥఈ ݂ሺߣሻ (19)
ଵାఙೣ
where, Ȧr, Ȧf, and ܴ௪ are angular velocities of rear and front
tirer, and the wheel radius, respectively. The yaw motion is where, the longitudinal slip ratio (ߪ௫ ), slip angle (ߙ),
also dependent on the tire forces ܨ௫ and ܨ௬ as stated below longitudinal tire stiffness (ܥఙ ), and cornering stiffness (ܥఈ ), ߣ
(Eq. (13)), are given by (Eq. (20) until Eq. (22)),
281
5B1-4 Joint International Conference on Electric Vehicular Technology and Industrial,
Mechanical, Electrical and Chemical Engineering (ICEVT & IMECE) 2015
ஜி ሺଵାఙೣ ሻ comparison between Simulink model and verification software,
ߣ ൌ మ భΤ మ
(20)
ଶቄሺ ఙೣ ሻమ ା൫ഀ ௧ሺఈሻ൯ ቅ the steering, braking and throttling inputs in this simulation are
the disturbances taken from the CarSimEd software. The model
݂ሺߣሻ ൌ ሺʹ െ ߣሻߣ if ߣ ൏ ͳ (21) is said to be true or acceptable if the error between the model
and the comparison software ranges between 2-15% [14].
݂ሺߣሻ ൌ ͳ if ߣ ͳ (22)
The notation ܨ௭ is the vertical force on the tire, while Ɋ is the IV. SIMULATION RESULTS
tire-road friction coefficient. A. Results in Lateral Direction
III. SIMULATION AND VERIFICATION Simulation procedures employed in lateral direction involve
step steer test at speed 50 km/h, double-lane change at speed 80
A. Simulation km/h, and slalom test at 30 km/h. Figure 2 includes the
The governing equations described in section II are then responses of Semar-T electric vehicle including body roll
adopted into Simulink block diagrams. The simulation angle, acceleration, and yaw rate in a lateral direction under
parameter values of the Semar-T electric vehicle which will be step steer disturbance. The responses of the electric vehicle in
used as input in the simulation of vehicle handling models is double-lane change condition were captured in Figure 3.
listed in Table 1. Meanwhile, Figure 4 corresponds to the response of the electric
vehicle under slalom test. The vehicle responses resulted by the
TABLE 1. SIMULATION PARAMETERS Simulink model and CarSimEd at the lateral direction shows
good agreements, and the simulation results closely follow the
Symbol Description Unit trend obtained from the CarSimEd.
݉௦ Sprung masses 1400 kg The slight difference in magnitude may be due to the
ܫ௬௬ Pitch moment inertia 1263.5 kgm2 body flexibility that was ignored in the simulation model. The
ܫ௫௫ Roll moment inertia 450.1 kgm2 slight difference in the responses may also be because the roll
ܫ௭௭ Yaw moment inertia 1263.5 kgm2 center in CarSimEd which assumed to be fixed on the ground,
݉௨ Front left wheel unsprung masses 40 kg whereas, in the developed model, the roll center is assumed to
݉௨ Rear left wheel unsprung masses 35 kg be at fixed distance from the sprung mass center of gravity. It
݉௨ Rear right wheel unsprung masses 35 kg is also be caused by one of the modeling assumptions namely
݉௨ Front right wheel unsprung masses 40 kg the effects of anti-roll bar which were completely ignored in
ܭ௦ Spring stiffness at the front left 18000 N/m the simulation model.
ܭ௦ Spring stiffness at the rear left 18000 N/m
ܭ௦ Spring stiffness at the rear right 18000 N/m matlab carsim
1
ܭ௦ Spring stiffness at the front right 18000 N/m
ܥ௦ Damping stiffness at the front left 1500 Ns/m 0
ܥ௦ Damping stiffness at the rear left 1500 Ns/m -1 0 1 2 3 4 5 6 7 8 9 10
Roll (deg)
-1 0 1 2 3 4 5 6 7 8 9 10
B. Verification -3
(m/s2)
282
5B1-4 Joint International Conference on Electric Vehicular Technology and Industrial,
Mechanical, Electrical and Chemical Engineering (ICEVT & IMECE) 2015
matlab carsim matlab carsim
5
8
-5 0 1 2 3 4 5 6 7 8 9 10
4
-15 0
-25 -4 0 1 2 3 4 5 6 7 8 9 10
-35 -8
-12
-45 Time (s)
Time (s)
(c) Yaw rate
(c) Yaw rate
Fig. 3. Vehicle response under double-lane change tests at 80 km/h
Fig. 2. Vehicle responses under the step steer input at 50 km/h
The verification is also performed for slalom maneuver
The next maneuver for verification process of the vehicle
with the speed of 30 km/h. Figure 3 shows the responses
model is the double lane change maneuver with the speed of between developed model and CarSimEd regarding roll angle,
80 km/h. Double lane change is one of the handling
lateral acceleration, and yaw rate. As can be seen in the figure,
maneuvers used to evaluate the road holding of the vehicle.
the results for the developed model are very close and
This handling maneuver can be considered as the
following the trend of CarSimEd software with acceptable
representation of some crash avoidance. Figure 2 shows the
error. In overall, similar responses and trend can be obtained
responses of model and vehicle behaviors obtained from
from the developed model compared to the behaviors obtained
CarSimEd regarding roll angle, lateral acceleration and yaw
with CarSimEd.
rate. Again, it can be seen that the trend between model matlab carsim
developed and CarSimEd results are almost similar but 4
slightly different in magnitude. This difference might be due
to the parameters assumption in the modeling of vehicle 2
Roll (deg)
0.5 6
Lateral Acceleration
-0.5 4
0 1 2 3 4 5 6 7 8 9 10
2
-1.5
(m/s2)
0
-2.5
Time (s) -2 0 1 2 3 4 5 6 7 8 9 10
-4
(a) Roll angle
-6
Time (s)
matlab carsim
5
(b) Lateral acceleration
Lateral Acceleration
3 matlab carsim
60
1
(m/s2)
40
Yaw Rate (deg/s)
-1
0 1 2 3 4 5 6 7 8 9 10
20
-3
0
-5
Time (s) 0 2 4 6 8 10
-20
(b) Lateral acceleration
-40
Time (s)
283
5B1-4 Joint International Conference on Electric Vehicular Technology and Industrial,
Mechanical, Electrical and Chemical Engineering (ICEVT & IMECE) 2015
B. Results in Longitudinal Direction matlab carsim
0.004
The simulation process employed in longitudinal direction
0.000
involves sudden braking at speed 60 km/h and sudden
Longitudinal Slip
acceleration at speed 20 km/h. Figure 5 and 6 correspond to -0.004 0 1 2 3 4 5 6 7 8 9
the responses of Semar-T vehicle include longitudinal -0.008
acceleration, pitch rate and longitudinal slip at each tire for -0.012
sudden braking and sudden acceleration, respectively. -0.016
The results of model verification for sudden braking test -0.020
at 60 km/h are shown in Figures 5(a) to 5(d). Regarding -0.024
longitudinal acceleration and pitch rate, it can be seen that Time (s)
there are quite good similarities during the initial transient (d) Longitudinal slip at rear tire
phase as well as during the following steady state phase as
Fig. 5. Responses of vehicle to the sudden braking condition at 60 km/h
shown in Figures 5(a) and 5(b). Longitudinal slip responses of
the tires also showed substantial agreement with CarSimEd The results of sudden acceleration test at a constant speed
Software for all recorded values and satisfying comparison of 20 km/h indicate that CarSimEd and the Simulink model
with only small deviation in the transition area between results agree with a relatively good accuracy as shown in
transient and steady state phases as shown in Figures 5(c) and Figures 6(a) to 6(d). Figure 6(a) shows the behavior of
5(d). longitudinal acceleration obtained from sudden acceleration
matlab carsim
test maneuver that is also used as the input for the simulation
0.5 model. Regarding pitch rate, it is clear that the simulation
results closely follow the measured data with a minor
Acceleration (m/s2)
-0.5
0 1 2 3 4 5 6 7 8 9 difference in magnitude as shown in Figure 6(b). The minor
Longitudinal
6
Longitudinal
2.5
4
2 1.5
0
0.5
-2 0 1 2 3 4 5 6 7 8 9
-4 -0.5 0
Time (s) 1 2 3 4 5 6 7 8 9 10
Time (s)
(b) Pitch rate
(a) Longitudinal acceleration
matlab carsim
0.004
matlab carsim
1
0.000
Longitudinal Slip
0
Pitch Rate (deg/s)
-0.004 0 1 2 3 4 5 6 7 8 9
-1 0 1 2 3 4 5 6 7 8 9 10
-0.008
-2
-0.012
-3
-0.016 -4
-0.020 -5
Time (s) Time(s)
284
5B1-4 Joint International Conference on Electric Vehicular Technology and Industrial,
Mechanical, Electrical and Chemical Engineering (ICEVT & IMECE) 2015
matlab carsim sudden braking tests are significantly similar with the
-0.0001 0 CarSimEd Software and prove that the model is effectiveness
1 2 3 4 5 6 7 8 9 10
to perform the validation. As long as the trend of the model
Longitudinal Slip
matlab carsim
0.0280
REFERENCES
Longitudinal Slip
285