Simulation Modelling Practice and Theory 19 (2011) 2131–2150

Contents lists available at ScienceDirect

Simulation Modelling Practice and Theory

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

Evaluation of antilock braking system with an integrated model of full

vehicle system dynamics
T.K. Bera, K. Bhattacharya, A.K. Samantaray ⇑
Department of Mechanical Engineering, Indian Institute of Technology, 721 302 Kharagpur, India

a r t i c l e i n f o a b s t r a c t

Article history: Antilock braking system (ABS), traction control system, etc. are used in modern automo-
Received 12 December 2010 biles for enhanced safety and reliability. Autonomous ABS system can take over the trac-
Received in revised form 6 July 2011 tion control of the vehicle either completely or partially. An antilock braking system
Accepted 8 July 2011
using an on–off control strategy to maintain the wheel slip within a predefined range is
Available online 5 August 2011
studied here. The controller design needs integration with the vehicle dynamics model.
A single wheel or a bicycle vehicle model considers only constant normal loading on the
Keywords:
wheels. On the other hand, a four wheel vehicle model that accounts for dynamic normal
Bond graph
Antilock braking system
loading on the wheels and generates correct lateral forces is suitable for reliable brake sys-
Vehicle dynamics tem design. This paper describes an integrated vehicle braking system dynamics and con-
trol modeling procedure for a four wheel vehicle. The vehicle system comprises several
energy domains. The interdisciplinary modeling technique called bond graph is used to
integrate models in different energy domains and control systems. The bond graph model
of the integrated vehicle dynamic system is developed in a modular and hierarchical mod-
eling environment and is simulated to evaluate the performance of the ABS system under
various operating conditions.
Ó 2011 Elsevier B.V. All rights reserved.

1. Introduction

The main motivating factor behind this work on vehicle dynamics and antilock braking systems are the conditions that
prevail in Indian roads, especially in sub-urban, semi-urban and rural areas, where road infrastructure is poor and traffic is
chaotic. The vehicle performance is severely challenged in these conditions and optimizing the performance of the mecha-
tronic systems require controllers to be tuned by trial and error through exhaustive simulations and field testing. This is why
Indian automobile industry has been very conservative in adopting modern technology. The ABS is to be designed for diverse
driving conditions, especially where frequent braking and acceleration is required. Thus, increasing the life of the braking
system in these conditions becomes important.
Modeling and simulation of physical phenomena and physical systems plays a very important role in understanding the
underlying science. The necessary steps involved in modeling are writing the equations for each elementary physical system,
sorting these equations and implementing them in a solver. This approach becomes time-consuming when the system
becomes complex and multidisciplinary in nature. A common tool enabling a unified approach to the physical modeling
of various disciplines is bond graph [1–4]. Bond graph technique is also well suited for modular modeling of large physical
systems. A vehicle dynamic system is a multi-energy domain system which involves mechanical, hydraulic, pneumatic, elec-
tronic, electrical, chemical (batteries or fuel cells), thermal domains, to name a few. Bond graph modeling is an ideal tool to

⇑ Corresponding author. Tel.: +91 3222 282998/282999; fax: +91 3222 282277.
E-mail addresses: samantaray@lycos.com, ak_samantaray@yahoo.com, samantaray@mech.iitkgp.ernet.in (A.K. Samantaray).

1569-190X/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.simpat.2011.07.002
2132 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150

Nomenclature

a distance of front axle from the vehicle cg

A area
b distance of rear axle from vehicle cg
B stiffness factor
c half of track width
C shape factor
C1 maximum value of friction curve
C2 friction curve shape
C3 friction coefficient difference between the maximum value and value at rx = 1
C4 wetness characteristic value
D peak value
E curvature factor
F force
h height of vehicle cg from suspension reference point
J polar moment of inertia
K stiffness
kg discontinuous controller gain
l length
m mass
M moment
r effective radius
R damping
V volume
v, V velocity, velocity vector
x, y, z displacements in three directions
_ y;
x; _ z_ velocities in three directions
€; €z
€x; y accelerations in three directions
x0 output variable
yi input variable
a lateral slip angle
c camber angle
d steering angle
_ €
h; h angular velocity, acceleration
s torque
r slip ratio
l coefficient of friction
lm motor torque constant

Subscripts
a arm
b braking
bc brake cylinder
bd brake drum
c vehicle body
cfr, crr front and rear cornering force
ca cable
cx, cy, cz x, y, z direction of vehicle body
e equivalent
E engine
fr front
l left
lm mechanical loss
m motor
nfr, nrr normal for front and rear wheel
r right
re return
rr rear
stx, sty, stz x, y, z direction of structure
st steering (front axle)
 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150 2133

stw steering wheel

sx, sy, sz x, y, z direction of suspension
t tyre
tfr, trr tangential (front and rear)
v vehicle
w wheel
wx, wy, wz x, y, z direction of wheel
1 left suspension reference point

develop an integrated model of such a system with coupling of several energy domains. Multi bond graph is suitable for com-
pact representation of complex multibody system models [5,6]. Moreover, efficient control algorithms can be derived from
analysis of the bond graph structure [7–10]. For the above reasons, bond graph models have been extensively used for design
of mechatronic systems [1,11,12]. Although this article concerns only a few energy domains, bond graph modeling has been
used here to keep open the possibilities for addition of further dynamical component models (e.g., IC engine, electric motor
drive, fuel cell, transmission system, etc.).
Bond graph modeling has been extensively used in vehicle dynamics studies [13,14]. A four-wheel, non-linear vehicle dy-
namic model with electrically controlled brakes and steering was developed by Margolis and Shim [15]. Three dimensional
dynamics of coupled bodies and the multiple transformations required to model multibody systems were developed by
Pacejka [16]. The engine model along with drive train and vehicle dynamics models with complexity was developed by Louca
et al. [17]. In this paper, initially, a bicycle model with steering is considered for basic brake system design. Then a full vehicle
model which has 6-DOF for the vehicle chassis and also 6-DOF for each wheel and includes tyre forces [18] is used to study
the vertical, longitudinal and lateral vehicle dynamics. The bond graph model for the braking system is then integrated with
the vehicle model for dynamic analysis.
Antilock braking system (ABS) is an electronically controlled braking system that maintains control over the directional
stability of the vehicle during emergency braking or braking on slippery roads by preventing wheel lock-up. Another advan-
tage of using ABS is reduction in the stopping distance during emergency braking or braking on slippery roads. This is
achieved by utilizing the maximum brake power available for which the wheel does not get locked [19]. However soft sur-
faces or surfaces made up of gravels can enhance stopping distances. Conventional ABS has some limitations in control and
performance. One main drawback of conventional ABS is the slip control strategy where the slip is maintained in an accept-
able range [20] rather than at the optimal value. Friction forces generated during vehicle acceleration or braking are propor-
tional to the normal load on the wheels of the vehicle. Studies show that the friction coefficient is a non-linear function of the
wheel slip [21]. ABS controller is developed to keep the vehicle slip in a particular range for which the road wheel friction
coefficient is highest, to achieve optimal performance. Obtaining an accurate mathematical model of ABS is very difficult as
the controller operates in an unstable equilibrium point. It is also very difficult to identify the actual road surface by using
any available sensor. Therefore, road condition information cannot be used in optimization of the ABS controller perfor-
mance. Various control algorithms such as sliding mode control [22], fuzzy logic [23], and neural networks are reported
in literature to optimize the ABS performance.
One finds several articles in literature dealing with quarter car ABS models. However, the quarter car model is unsuitable
to evaluate ABS performance during a curve negotiation. This necessitates the use of a bicycle model of the vehicle where
steering or handling model is included. Although the bicycle model offers improved response characteristics than the single
wheel model, it does not represent the true dynamics of the ABS equipped vehicle. When a vehicle brakes, the vehicle tilts
forward and the load shifts from the rear wheels to the front wheels. The reverse happens when the vehicle accelerates. Like-
wise, during a turn, the vehicle load is transferred from the inner wheels to the outer wheels. The peak ABS braking force
depends on the value of the coefficient of friction between the tire and the ground, which in turn depends on the normal
load on the tire and other factors. Thus, proper accounting of the load transfer is required to accurately predict the ABS sys-
tem performance. The load transfer mechanism during braking and turning cannot be represented in the bicycle model.
Therefore, a four wheel vehicle model should be used for final performance evaluation, although the single wheel and bicycle
models may be used in the initial design stage.
The structure of this paper is as follows: First, models of various subsystems of the full vehicle model are explained along
with the control algorithms for ABS. Then the detailed bond graphs of bi-cycle model and full vehicle model are constructed
based on the various kinematic relations. Finally, simulation results of vehicle dynamics along with antilock braking system
are provided and compared.

2. Brake system model

The wheels are modeled by their mass, rotary inertia, radius and tyre stiffness. The tyre is the most important among
wheel components because tyre forces and moments play an important role in vehicle dynamics. Tyre forces are necessary
to control the vehicle. As the tyres are the only means of contact between the road and the vehicle, they are the key factors
2134 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150

Fig. 1. Tyre forces and moments.

determining the vehicle handling performance. Tyre models are broadly classified as physical models, analytical models, and
empirical models. The physical models are constructed to predict tyre elastic deformation and tyre forces [25]. In such mod-
els, complex numerical methods are required to solve the equations of motion. Analytical models are not useful at large slip
and at combined slip. Empirical models based on experimental correlations are generally more accurate [26]. Pacejka’s magic
formula [18] is a widely used empirical model with which one can compute the longitudinal and cornering forces and self-
aligning moment.

2.1. Tyre slip forces and moments

The tyre forces and moments from the road surface act on the tyre as shown in Fig. 1. The forces acting along x, y and z
axes are longitudinal force Fx, lateral force Fy and normal force Fz, respectively. Similarly the moments acting along x, y and z
axes are overturning moment Mx, rolling resistance moment My and the self-aligning moment Mz, respectively [27].
In actual case, the wheel speed and linear acceleration [28] are measured and used to compute the longitudinal slip. The
longitudinal slip ratio is defined as the normalized difference between the circumferential velocity and the translational
velocity of the wheel [29]. It is expressed as
8_
< hwy_ rw x_ w ðduring traction; assuming h_ wy > 0Þ
hwy r w
rx ¼ ð1Þ
: x_ w h_ wy rw
x_ w
ðduring braking; assuming x_ w > 0Þ

Lateral wheel slip is the ratio of lateral velocity to the forward velocity of wheel [18]. It is given as

y_
ry ¼ tan a ¼ _ w ð2Þ
xw
For small slip ratios, the longitudinal force Fx and lateral force Fy can be approximated as Fx = rxCx and Fy = ryCy where, Cx
and Cy are longitudinal tyre stiffness (coefficient) and cornering coefficient, respectively, and rx and ry are the longitudinal
slip ratio and lateral slip ratio, respectively [30]. However, these linear relations are invalid for large slip ratios.
The empirical magic formula based on experimental data is adopted for the development of a tyre-road friction model.
The magic formula gives more accurate results for larger slip angles and is also applicable to wide range of operating con-
ditions. Longitudinal slip velocity (i.e., h_ wy r w  x_ w during traction and x_ w  h_ wy rw during braking) develops longitudinal force
Fx whereas lateral or side slip velocity y_ w and camber angle c generate side force Fy and self-aligning moment Mz. Pacejka’s
magic formula states that longitudinal force, side force and the self-aligning moment are functions of longitudinal slip and
side slip, respectively. It is given as
   
yo ¼ D sin C tan1 Bxi  E Bxi  tan1 ðBxi Þ ð3Þ
where output variable, yo:Fx, Fy or Mz and input variable, xi:rx or ry.
Ply-steer, rolling resistance, conicity effects may cause slight variation in the function in Eq. (3), but these variations may
be neglected. The constant parameters (B, C, D, E) can be determined by measuring the tyre forces and moments by sophis-
ticated equipments.
Pacejka’s magic formula is unsuitable when snow and ice start significantly affecting the vehicle performance. Moreover, it
does not consider the velocity dependence of friction, which is critical while designing brake systems. Therefore, another
formula developed by Burckhardt [21] is usually used in research on brake system design. Details of the Burckhardt formulae
are given in the next section.
 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150 2135

2.2. Antilock braking system

Antilock braking system (ABS) as a vehicle autonomous system can be used to improve stability and to reduce longitu-
dinal stopping distance while maneuvering under braking condition. In Ref. [31], it has been shown that antilock braking
system approaches the optimum braking performance. Antilock braking system is suitable for dangerous braking conditions
such as braking on icy or wet asphalt road or for panic braking situation. Upon initiation of braking, the longitudinal slip ratio
starts increasing because the angular velocity of the wheel decreases faster than the decrease in the vehicle’s linear velocity.
It is already known that at high value of slip ratio, the grip between the tyre and the road decreases and thus rate of decel-
eration of the vehicle reduces. Therefore, as the slip ratio reaches a maximum threshold value, the applied braking torque is
reduced. As a result, the angular speed of the wheel increases and the longitudinal slip ratio decreases. At low value of slip
ratio, like at high value of slip ratio, the grip between the tyre and the road reduces. Thus, when the slip ratio reaches a min-
imum threshold value, the braking torque is increased. This way, the slip ratio is continually maintained between an upper
and a lower threshold by modulating the braking torque [32]. The control of ABS is a combination of slip and wheel accel-
eration control. In wheel acceleration control, the wheel angular velocities are measured and slip is controlled indirectly by
changing the speed of the wheels. Slip cannot be kept at the exact optimum value in a conventional ABS. This is the main
drawback for this system. Lots of testing and tuning are required for every ABS controller designed for a specific vehicle.
The testing is usually performed through software-in-the-loop and hardware-in-the-loop simulations [33]. Hydraulic valve
control is usually used to regulate the brake pressure in antilock braking systems. Kuang et al. [34] have modeled hydraulic
brake system using the bond graph technique. In Ref. [35], sliding mode controllers have been proposed to improve the per-
formance of ABS. Corno et al. [36] have shown that it is necessary to take into account the model dependence on velocity and
wheel slip to devise a robust ABS controller. In Ref. [36], this dependence is modeled via a linear parameter varying (LPV)
system for the development of wheel slip controller of a two-wheeled vehicle under straight running conditions. Composite
ABS control technique has been used in Ref. [37] for a vehicle with four electrically driven wheels. The composite ABS control
combines the electric motor ABS control with the hydraulic ABS control. For lower braking torque requirement, only the elec-
tric motor ABS control works whereas both the electric and hydraulic ABS controls work for higher torque requirement [37].
The main components of ABS are brake servo, lever arm, cable, return spring, rod, cam, rotors and brake pads [20]. Schema
of an ABS is shown in Fig. 2a. The ABS controller controls the voltage that is fed to the motor. The resistance is in series with
the motor. The lever arm is connected to the motor. The cable is connected to the arm at one end and the other is connected
to the lever with cam. The return spring is used to get back to the initial stage.
The tyre friction models are generally non-linear in nature. The coefficient of friction is
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
F 2x þ F 2y
l¼ ð4Þ
Fz
where longitudinal force Fx and side force Fy can be determined for particular conditions of constant value of linear and angu-
lar velocity from the most widely used magic formula (Eq. (3)) given by Pacejka [18] and Fz is the normal force. However, one
main disadvantage of this model is its inability to describe low slip effects and large forward and side slip effects during
wheel lockup.
Another friction model often used to model tyre forces is given by Burckhardt as follows [21]

Fig. 2. (a) Schema of an ABS [20] and (b) the flow diagram for ABS algorithm.
2136 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150

Table 1
Tyre-road friction parameters [21].

Surface condition C1 C2 C3 C4
Asphalt, dry 1.029 17.16 0.523 0.03
Asphalt, wet 0.857 33.822 0.347 0.03
Cobblestones, dry 1.3713 6.4565 0.6691 0.03
Snow 0.1946 94.129 0.0646 0.03

lðrx ; x_ w Þ ¼ ½C 1 ð1  eC2 rx Þ  C 3 rx eC4 rx x_ w ð5Þ

where constant parameters C1, C2, C3 and C4 are determined from experiments. These parameters for various road conditions
[21] are given in Table 1. The slip–friction curve obtained from this model is shown in Fig. 4.
The flow chart for the ABS algorithm used in this article is shown in Fig. 2b. Here the longitudinal slip is switched between
maximum and minimum values and, accordingly, brake force signal changes and ultimately the braking torque varies as
follows:
8
< smax
> rx < rlow
si ¼ si1 rlow 6 rx 6 rhigh ð6Þ
>
:
0 rx > rhigh
The slip–friction curve for a wheel traveling with linear speed of 10 m/s for various road conditions is shown in Fig. 3. It is
seen from this figure that the coefficient of friction increases with the increase of slip to a maximum value and then it de-
creases monotonically. If the wheel gets locked (rx = 1) then value of the coefficient of friction is low and the wheel starts
slipping and the steering control may be lost, which is totally undesirable. To increase steerability and lateral stability of the
vehicle and to decrease the stopping distance during braking, the slip value must be maintained within a range to get the
high value of friction force. As the slip dynamics is very fast and at any value after the peak of the friction curve is open loop
unstable, the slip value is kept within a certain range which is also called sweet–spot (see the shaded area in Fig. 3). The
attainable maximum friction coefficient is different at different vehicle speeds (see Fig. 4). Brake systems are designed to give
maximum braking efficiency at higher speeds. Moreover, it is difficult to accurately predict the road condition from onboard
measurements. In ABS design, the optimal slip ratio range is thus taken to be 0.2–0.25. This range offers a good compromise
where near maximum friction coefficient is achieved for all kinds of road surfaces (thus, eliminating the requirement of road
condition recognition) and vehicle linear speed.

3. Bicycle vehicle model

A bicycle model of the vehicle [38] was used for the initial study. This model does not take into account the roll, pitch and
heave motions and the suspension dynamics is neglected. Thus, the load transfer during maneuvering and braking cannot be

Fig. 3. Friction coefficient versus slip ratio curves for different road surfaces at linear speed of 10 m/s with the tire properties given in Ref. [21].
 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150 2137

Fig. 4. Slip–friction curves for asphalt dry road condition for different vehicle speeds.

Fig. 5. (a) Schema of the bicycle vehicle model and its (b) word bond graph.

included in this model. Schematic of the considered vehicle is shown in Fig. 5a and its word bond graph is shown in Fig. 5b.
As the road is considered to be flat, the motion of the vehicle is planar.

3.1. Kinematic relations

The kinematic relations derived in Refs. [2,38] are used to construct a major part of the model. Thereafter, the wheel rota-
tions and longitudinal and lateral slip calculations are inserted into the model. It is assumed that only the front wheel is
steered (by steering angle d).
Velocities normal and tangential to the plane of rotation of the front wheel are

v nfr ¼ ðy_ þ h_ cz aÞ cos d  x_ sin d

ð7Þ
v tfr ¼ ðy_ þ h_ cz aÞ sin d þ x_ cos d
Likewise, velocities normal and tangential to the plane of rotation of the rear wheel are

v nrr ¼ ðy_  h_ cz bÞ; v trr ¼ x_ ð8Þ

From Newton–Euler equations with z ¼ h_ cx ¼ h_ cy ¼ 0, one obtains
X
mv €x ¼ mv h_ cz y_ þ Fx
X ð9Þ
€ ¼ mv h_ cz x_ þ
mv y Fy

3.2. Bond graph model

In the word bond graph of the bicycle vehicle model shown in Fig. 5b, CTF blocks represent necessary coordinate trans-
formations [39]. Maintaining the model structure defined in the word bond graph, the complete bond graph model as shown
2138 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150

Fig. 6. Bond graph model of bicycle vehicle model.

in Fig. 6 can be drawn using Eqs. (7)–(9). The terms mv h_ cz y_ and mv h_ cz x_ in Eq.(9) are conservative pseudo-forces which are
implemented by a gyrator (GY) element in the bond graph model. In Fig. 6, lines with half-arrow at their end are power
bonds and those with a full arrow at their end are information bonds. The vehicle inertias (I:Mv) are modeled in the moving
frame and the pseudo-forces transform them to the inertial frame. The rotary inertia (I:Jv) is modeled at a 1-junction. The
transformer elements (TF) are used to calculate the tangential and normal velocities at tyres according to Eqs. (7) and (8).
The flow detectors (Df) connected to velocity points in the inertial frame are not present in the actual system, i.e., the actual
system is not instrumented with an inertial sensor. These flow detectors are simply added to plot the x and y positions of the
vehicle centre of mass in the inertial frame and also to modulate the modulated transformer elements in the part of the junc-
tion structure.
Port number 7 shown within a circle is used to interface this model to the model of the brake system. The two modulated
3-port R-fields (MR-elements) implement Pacejka’s magic formulae. First of all, the velocity of the contact patch on the tyre,
the longitudinal velocity of the tyre and the lateral velocity are used to compute the longitudinal and lateral slip ratios
according to Eqs. (1) and (2). Thereafter, Eq. (5) is used to compute the coefficient of friction and the net frictional force.
The components of the net frictional force are used to compute the longitudinal and lateral tyre forces and the reaction mo-
ment on the drive. The self-aligning moment is not included in the bicycle model. Thus, the MR-element, as causalled in the
model, receives the information of three generalized flow variables and computes three generalized effort variables.

3.3. Braking system

A comparison between the bond graph model of mechanical equivalent braking system (Fig. 2a) and the conventional
hydraulic braking system is shown in Fig. 7. In hydraulic braking system, the pedal force Fpedal is represented by the Se ele-
ment (Fig. 7b). The fluid flow through supply or hold valve is modeled by a resistive element Rsv. The element C connected
with 0 junction, which determines pressure of brake cylinder, indicates the brake fluid compressibility Kb (a function of bulk
modulus and fluid volume). Rrv is the resistance in the pressure relief valve and atmospheric pressure is indicated by the
zero-valued Se element. Braking torque depends on the area of the brake cylinder and the radius of the brake drum.
In the bond graph model of the mechanical equivalent braking system [20] shown in Fig. 2a, the controlled voltage from
the ABS controller is fed to the motor whose generated torque (sm) is represented by a Se element. The resistive element
connected to 1 junction denotes mechanical losses (R1m). The cable stiffness is represented by C element connected to 0-
junction. The return spring having stiffness Kre is represented by C element. The other end of the return spring is anchored
to the ground by a zero-valued source of flow. The output braking torque is applied on the bicycle vehicle model (Fig. 6) and
wheel model for full vehicle through port 7.

3.4. Model validation

The model parameters used in the simulation of the bicycle vehicle model are given in Table 2. Implementations of ABS on
the planar bicycle model are first carried out to fine-tune and test the control algorithm. Once the controller implementation
(program) was found to be satisfactory, it was ported to the full vehicle model.
 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150 2139

Fig. 7. Bond graph model of (a) mechanical equivalent braking system and (b) hydraulic braking system.

Table 2
Parameter values of bicycle vehicle model.

Subsystem Parameter values

Vehicle body mv = 1600 kg jv = 100 kg m2 a = 1.0 m b = 1.0 m
Wheel Jw = 15 kg m2 rw = 0.3 m
C1 = 1.029 C2 = 17.16 C3 = 0.523 C4 = 0.03
Antilock brake rlow = 0.04 rhigh = 0.5 Kca = 104 N/m Kre = 106 N/m
rbd = 0.15 m Rlm = 0.04 N s/m la = 1 m lm = 0.4 N m/A
Raero = 0.1 kg/s

The validation of the ABS system has been performed by comparing the results with those reported in Ref. [20]. Consid-
ering the fact that the steering effect was not considered in Ref. [20], the steering angle in the bicycle model was kept zero.
The vehicle weight and the initial linear and angular velocities of the wheels are taken to be the same as those in Ref. [20]
and the rear wheel of the vehicle is considered to be freely rolling. With these adjustments to match the scenarios, the ABS
bicycle model developed here is nearly equivalent to that presented in Ref. [20].
All simulations presented in this article are performed in Symbols software [40] which is an integrated platform for port-
based modeling using bond graphs and block diagrams, simulation, animation, control system analysis and fault detection
and isolation system design [41,42]. The software runs in Windows and Linux operating systems. It follows an object ori-
ented modeling framework in which subsystem models can be developed, validated and integrated [43]. One of the major
advantages of Symbols software is its implementation of polymorphic objects which allows template matching for varying
boundary causalities. The modeling environment is shown in Fig. 8 where the bond graph model of the bicycle vehicle has
been displayed. The C-elements with signal bonds in Fig. 8 together represent the detector of flow (Df) elements and sinks.
The MR elements in Fig. 6 are modeled as a subsystem (represented by its icon) in Fig. 8. This user created subsystem appears
in the list of subsystem objects called capsules. The software has many in-built capsules for basic engineering components.
The equations of motion are derived by Symbols software from the bond graph model. These equations can be exported in
various ways such as in S-function form for Matlab simulink, C++ source code for user written solvers, etc. Symbols software
also has a powerful native solver (see Fig. 9) which offers a vast range of simulation and post processing tools.
The simulation of ABS system is a combination of continuous time and discrete time simulation. The measurements are
acquired from the model at intervals of 0.2 ms (sampling frequency of 5 kHz) and a zero-order-hold is used with each mea-
surement. The simulation time step is chosen to be 0.1 ms. The same simulation setting is used for all results presented in
this article.
Fig. 10 shows a comparison of the results reported in Ref. [20] with those obtained from the developed ABS bicycle model.
Both the results show almost linear variations in the forward linear wheel speed and a regular pattern of deceleration and
acceleration of the wheel angular speed. The slip ratios are initially maintained within a bound and as the vehicle slows
down, their maximum values approach unity. The variations in the results are mainly due to the differences in the consid-
ered dynamics. The four wheel model of Ref. [20] is influenced by vehicle’s pitch during braking and thus the load transfer to
the front wheels (where ABS brakes are applied) produces marginally higher braking force than the bicycle model. Therefore,
the results in Fig. 10a show slightly fast braking action as compared to those in Fig. 10b.

3.5. ABS performance while braking during maneuvering

The ABS is meant to provide better maneuverability while brakes are applied on the steered wheel. To evaluate the
maneuverability of the vehicle under braking situation, a particular scenario has been considered here. First of all, the vehicle
starts from rest and it is brought to a steady linear speed of 50 km/h over a straight path. Then it is steered at t = 12 s to
2140 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150

Fig. 8. Modeling environment in Symbols software: the left panel contains bond graph elements and subsystem models called capsules, the bottom panel
displays equations derived from bond graph model in the equations tab and the user can enter model parameter descriptions, model nonlinearity
expressions, etc. in the expressions tab.

Fig. 9. The simulation environment in Symbols software contains tabs for parameter, initial condition, plot variables and simulation setting entries. The
slider tab is for runtime parameter variation by the user. The runtime display is not scaled properly and a separate graphics display window appears after
the end of simulation.

follow a circular path with a constant steering angle of 0.1 radians. The steering action causes a decrease in the vehicle linear
speed. Once the vehicle linear speed is stabilized and it moves at constant speed over the circular path, front wheel brakes
are applied at t = 17 s. Note that to handle large side and forward slips during wheel lockup, a composite slip based formu-
lation [44] developed by US Department of Transportation (DOT) has been adopted.
Fig. 11a shows the predefined path (2.5 m wide lane) over which the vehicle is supposed to move and the actual paths
taken by the vehicle under two different braking conditions: the first when ABS is used and the second when a conventional
mechanical brake is used. In the ABS, a slip control strategy is applied to keep the slip ratios between 0.2 and 0.25. Initially, a
sustained brake force is applied till the slip ratio becomes 0.2 and then the slip control algorithm takes over. The result shows
that the vehicle veers off from the lane due to application of the conventional brake whereas it stays within the lane till it
stops when ABS is used.
 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150 2141

Fig. 10. Vehicle speed, wheel speed and wheel slip ratio (a) as reported in Ref. [20] and (b) as obtained from the bicycle model during full braking by ABS.

Fig. 11. (a) Vehicle centre of mass trajectory in inertial frame; (b) wheel slip ratio; (c) vehicle speed; and (d) wheel speed for ABS and conventional braking
system.

The results in Fig. 11b show the change in slip ratios due to braking. The slip ratios for ABS remain within 0.2–0.25 and
thus the vehicle is able to follow the steering. The slip ratio quickly approaches a value of 1.0 due to application of conven-
tional brake. This causes the wheels to lock up and thus the vehicle cannot steer.
2142 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150

Fig. 12. Word bond graph of four wheel vehicle model.

Fig. 11c shows the vehicle speed variations due to steering and braking and Fig. 11d shows the angular velocity of the
front wheel under the same condition. The wheel locks up due to application of conventional brake as soon as the wheel
speed becomes zero. However, on application of ABS there is an initial sudden drop in wheel speed (till the slip ratio becomes
0.2) and then the wheel speed decelerates and accelerates much like the results shown in Fig. 10b. The linear speed of the
vehicle reduces almost linearly to zero in both cases; the deceleration (see Fig. 11c) being slightly faster for ABS as compared
to the conventional brake. This deceleration almost doubles when brakes are applied both on the front and the rear wheels.
With the ABS based on slip control mechanism, the vehicle stops faster and at a shorter distance. The results suggest that an
automatic lane departure monitoring and control system [45] should also involve the brake system in the control loop.

4. The bond graph for the four wheel model

The four wheel vehicle system is modeled by composing six subsystem models. These are: vehicle body, suspension,
wheel, steering, antilock braking system and differential. The word bond graph representation of the full vehicle model is
shown in Fig. 12 where bonds represented by two parallel lines are multi-bonds. The flow variables at the interface of dif-
ferent subsystems are marked in Fig. 12. The complimentary power variables or generalized effort variables (force for linear
velocity and torque for angular velocity) are not shown in Fig. 12 to maintain clarity of the figure. The four wheels are con-
nected with the vehicle body through suspensions. The steering and ABS are coupled with the axle by scalar bonds. Likewise,
scalar bonds connect the differential to the rear wheels and the vehicle body. The engine and the gear box are disengaged
during braking, so they are not shown here although they are included in the full model. The integrated model may be re-
duced further for specific operational conditions [39].

4.1. Vehicle body

The full vehicle schema is shown in Fig. 13. It is assumed that the vehicle is symmetrical with respect to its longitudinal
axis. The suspension is approximated by simple linear spring-damper suspension system which corresponds to energy stor-
age and energy dissipation at each corner of the vehicle body while allowing free rotation of the wheel about axle (y-axis).

Fig. 13. Four-wheel vehicle model.

 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150 2143

The vehicle body motion is described by three linear displacements along three body-fixed axes and the rotational motion of
the body is defined by the three Cardan angles. The Newton–Euler equations with attached body fixed axes aligned with the
principal axes of inertia are as used for modeling the vehicle body [24].
The vehicle body is modeled as a rigid body with six degrees of freedom, i.e., pitch, roll, yaw, heave, surge, and sway mo-
tions. The rigid body motion is described with respect to a coordinate system rotating and translating with it. This local coor-
dinate frame attached at the centre of mass of the body is assumed to be aligned with the inertial principal axes. The motion
of the vehicle body is described by three linear displacements along body-fixed x, y and z axes (see Fig. 13) and rotations
about those axes. The vehicle orientation is defined by the three Cardan angles. The Newton–Euler equations of the vehicle
body [24] with attached body fixed axes aligned with the principal axes of inertia are as follows:
X
F x ¼ mc €xc þ mc ðz_ c h_ cy  y_ c h_ cz Þ ð10Þ
X
€cb þ mc ðx_ c h_ cz  z_ c h_ cx Þ
F y ¼ mc y ð11Þ
X
F z ¼ mc€zc þ mc ðy_ c h_ cx  x_ c h_ cy Þ ð12Þ
X
M x ¼ J cx €hcx  h_ cy h_ cz ðJ cy  J cz Þ ð13Þ
X
M y ¼ J cy €hcy  h_ cz h_ cx ðJ cz  J cx Þ ð14Þ
X
M z ¼ J cz €hcz  h_ cx h_ cy ðJ cx  J cy Þ ð15Þ

The first three equations (Eqs. (10)–(12)) are Newton’s equations where pseudo-forces appear due to the use of body-
fixed or non-inertial frame. The later three equations (Eqs. (13)–(15)) are Euler’s equations which account for gyroscopic mo-
ments. Note that the linear and angular velocities used in Newton–Euler equations are velocities as seen from a frame that is
momentarily aligned with the body-fixed principal axes. The whole set of six equations is referred to as Newton–Euler equa-
tions. Euler’s equations are represented in a bond graph form called Euler Junction Structure (EJS) [1–3,5,24].
Consider a rigid body of mass m, moments of inertia about principal axes (x, y and z) with origin at the mass centre as Ixx,
Iyy and Izz, angular velocities as seen from an inertial frame momentarily aligned along body fixed principal axes as xx, xy
and xz, external moment components as Mx, My and Mz, external force components as Fx, Fy and Fz. The general form of EJS
for the rigid body is shown in Fig. 14a. Another such gyrator ring structure is used to model the Newton’s equations. The
Newton–Euler equations are represented in bond graph form by combining both to form a star-shaped junction structure
given in Fig. 14b.
Fig. 14b is the basic building block to construct the bond graph model of a multibody system. The next modeling step
involves computation of velocities of certain points on the rigid body, transformation of those into inertial frame velocities
and implementation of the physical constraints at the chosen points. Note that in a bond graph model, implementation of
kinematic constraints at a set of points also accounts for the dynamic constraints, i.e., the external forces and moments
resulting from those constraints. This happens due to the power conserving junction structure. Therefore, bond graph mod-
eling is very useful in developing complex system models.
The equations for three linear velocities of left–front suspension reference point (point 1 in Fig. 13 whose body fixed coor-
dinates are x1, y1, z1) in the moving system of axes are as follows:

Fig. 14. Bond graph model of (a) Euler junction structure and (b) Newton–Euler equations.
2144 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150

x_ 1 ¼ x_ c þ z1 h_ cy  y1 h_ cz ð16Þ
y_ 1 ¼ y_ c þ x1 h_ cz  z1 h_ cx ð17Þ
z_ 1 ¼ z_ c þ y1 h_ cx  x1 h_ cy ð18Þ
The equations for the three angular velocities of left front suspension reference point are similarly expressed. Likewise,
the equations for three linear velocities and three angular velocities of other suspension reference points are derived.
The velocity transformation from the moving frame to the inertial frame can be written using successive multiplication of
rotation matrices as follows:
8 9 8 9 8 9 8 9
_
>
<X> = < x_ >
> = < xX >
> = < xx >
> =
Y_ ¼ Tw;h;u y_ and xY ¼ Tw;h;u xy ð19Þ
>
:_> ; >
:_> ; >
: > > >
Z z xZ ; :
xz ;
where
2 32 32 3
cos w  sin w 0 cos h 0 sin h 1 0 0
6 76 76 7
Tw;h;u ¼ 4 sin w cos w 05 4 0 1 0 5 4 0 cos /  sin / 5
ð20Þ
0 0 1  sin h 0 cos h 0 sin / cos /
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Tw Th Tu

and w, h and / are the Z–Y–X Cardan angles (a form of Euler angles). This power conserving Coordinate Transformation can
be represented by a bond graph junction structure (CTF or coordinate transformation block) [39]. Inverse transformation of
velocities in inertial frame to body-fixed frame can be represented by a similar junction structure [39].
The angles used in Eq. (20) are the Cardan angles (Euler angles). Transformation from body-fixed angular velocities to
Euler angle rates is given by
8 9 2 31 8 9 2 38 9
_
>
</> = 1 0  sin h < xx >
> = 1 tan h sin / tan h cos / > < xx >
=
6 7
h_ ¼ 4 0 cos / cos h sin / 5 xy ¼ 6
40 cos /
7
 sin / 5 xy : ð21Þ
>
:_> ; >
: > > >
w 0  sin / cos h cos / xz ; 0 sin /= cos h cos /= cos h
:
xz ;
The transformations given in Eq. (21) are represented in bond graph form as a transformer junction structure (EATF or
Euler angle transformation block). The Euler angle rates are integrated to obtain the Euler angles, which are used in all coor-
dinate transformations, i.e., in CTF and EATF bond graph sub-models. Note that for the vehicle body,
xx ¼ h_ cx ; xy ¼ h_ cy ; xz ¼ h_ cz , Ixx ¼ Jcx ; Iyy ¼ Jcy and Izz = Jcz. For the geometric parameters defined in the nomenclatures, the
suspension reference point 1 coordinates are x1 = a, y1 = c and z1 = h. The other reference point coordinates are similarly
expressed and used in kinematic relations.
The bond graph in Fig. 15 models the vehicle body inertia and transformations of the three linear and three angular veloc-
ities into velocities at relevant suspension reference points. The inertias are coupled by a pair of gyrator rings (Euler junction
structure [1,2,13,16]), one for translational and the other for rotational velocities.
Three sets of forces and moments act on the body. First of all, weight of the body and the aerodynamic forces (modeled by
Raero element) in the inertial frame act the vehicle body model in non-inertial frame through coordinate transformation
(CTF). Secondly, engine torque which is in a different body fixed frame (wheel frame) is transformed twice (first from wheel
body fixed frame to inertial frame and then to vehicle body fixed frame) to act on the body. Lastly, the suspension forces and
moments (constraint forces) acting in the inertial frame are transformed to get forces in the body fixed direction. Then these
forces are multiplied with the moment arms. The forces at all the four corners of the body are added together to get the sus-
pension force acting at the centre of gravity of the body.

4.2. Wheels

The bond graph model of wheel is shown in Fig. 16. The wheel is modeled as a rigid body with six degrees of freedom.
Similar to vehicle body, the inertias are coupled by a pair of gyrator rings, one for the translational and the other for the rota-
tional velocities.
There are some basic differences between the vehicle body model and the tyre model. The suspension reference points in
the vehicle body are fixed points. On the other hand, the tyre-road contact point in the wheel is not a point fixed on the
wheel; it changes as the wheel rotates about the axle. The second difference is that the wheel rotation about its axle axis
does not change its orientation in the inertia frame. Here, the wheel and its axle are assumed to be rotationally symmetric
about an axis.
The tyre-road normal contact force and the gravity force always act along inertial Z-axis. Thus, the wheel vertical dynam-
ics is decoupled from longitudinal and cornering dynamics. The ground reaction force (Fz) and wheel radius (rw) are used to
modulate longitudinal and cornering dynamics which are given in axle body-fixed frame. The vertical velocity of the wheel is
computed through a set of transformations at a 0-junction which is connected to a 1-junction where the vertical stiffness and
T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150 2145

Fig. 15. Bond graph model of the vehicle body.

Fig. 16. Bond graph model of a wheel.

2146 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150

Fig. 17. Bond graph model of (a) steering system and (b) modified steering system required for comparing vehicle’s yaw rate with the test and simulation
result given by Drozdz and Pacejka [24].

damping of tyre (represented by Kt and Rt, respectively) are modeled. The angle / used in the transformations (to obtain
wheel vertical motion at contact patch and to model self weight of the wheel and axle) is the wheel tilt with respect to
the ground, i.e., the angle of rotation of the wheel about X-axis of an momentarily conceived inertial frame whose Z-axis
is along inertial vertical direction and y-axis is aligned along the axle. The title angle may include additional camber. The
coordinate transformations are performed through the transformer structures shown in the model, i.e., a separate CTF block
is not required here.
The MR-element which is used for modeling road-tyre interactions is modulated by the normal reaction in the inertial
frame. Kinetic phenomenon and weight of the wheel are considered for the un-sprung mass vertical dynamics. The tyre-road
interactions and braking action are considered for wheel longitudinal and cornering dynamics. The cornering force and self-
aligning moment are dependent on vertical load and lateral slip angle, while the longitudinal force is dependent on vertical
load and longitudinal slip rate. The characteristic relations for the MR-element are given according to Burckhardt’s formulae
[21] or the composite slip based formulation [44].
Ports 1–6 in Fig. 16 are connected to the corresponding velocities of the suspension reference points. The braking torque
is applied on the front wheel through port 7 and the engine torque is transmitted to the rear wheel through the port 7. Port 8
is (shown as by dotted bond) is present only for front wheels. This port is used to interface the front wheel model to the anti-
roll bar model, which constrains the relative roll between two front wheel axles. The steering torque is applied at port 4
through the steering column model given in Fig. 17.

4.3. Steering system

In this article, we are directly modeling the input from the human operator. The moment applied on steering wheel about
the z-axis is supplied as shown in Fig. 17a. The actual the rate of rotation of steering wheel is measured by a yaw rate sensor.
It is assumed that the human vision system prescribes the desired rate of rotation. A feedback yaw rate controller may be
developed where the controller action can mimic the driver’s response to the consequences of steering. The steering model
output is used to rotate the front wheels by applying torques on the front axle (and reactive torques on the vehicle body). The
rotary inertia Jst of the steering wheel is represented by I-element. The remainder of the system comprises mechanical trans-
mission to the front axle. The rate of rotations of left and right wheels about the z-axis are represented at 1h_ l and 1h_ r junc-
tions. These are interfaced to the front axle model through ports 9 and 10. The modulated transformer moduli m1 and m2 are
determined from Ackermann’s formulae given as


ða þ bÞ cos2 hl þ c tan hl cos2 hl _
d_ i ¼ d ð16Þ
ða þ bÞ cos2 hst  c tan hst cos2 hst


ða þ bÞ cos2 hr  c tan hr cos2 hr _
d_ o ¼ d ð17Þ
ða þ bÞ cos2 hst þ c tan hst cos2 hst
In experimental investigations, steering wheel rotations (driver inputs) are measured. The model in Fig. 17a may be mod-
ified to the form shown in Fig. 17b to take the rate of steering wheel rotation as the input. The transformer with modulus lstw
is the gear ratio between steering wheel rotation and steering rotation.

5. Simulation results for four wheel model

The normal loads or load transfer on the wheels vary due to pitch, yaw and roll motions. At high speeds, aerodynamic
forces can as well affect the load transfer. Therefore, the load on each wheel, which is the deciding factor of maximum brak-
ing and driving torques, is influenced by various factors; broadly the operational conditions (speed, steering angle, etc.) and
the environmental conditions (wind speed, road condition, etc.). The model parameters in the simulations for the full vehicle
model are given in Table 3.
In the next step, the developed vehicle model is further validated by comparing the results with experimental maneuver
data available in literature. Once the initial simulation is satisfactory, simulation of ABS was carried out using square pulse
braking input. The pulse signal width is modulated for above instances to get the output. Then a comparative study between
 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150 2147

Table 3
Parameter values of full vehicle.

Subsystem Parameter values

Vehicle body mc = 1600 kg Jcx = 260 kg m2 Jcy = 1110 kg m2 Jcz = 1370 kg m2
a = 0.9 m b = 1.5 m c = 0.7 m h = 0.1 m
Suspension Ksx = 107 N/m Rsx = 2000 N s/m Ksy = 107 N/m Rsy = 2000 N s/m
Ksz = 80 kN/m Rsz = 500 N s/m Kstx = 107 N m/rad Rstx = 2000 N m s/rad
Ksty = 0 N m/rad Rsty = 0 N m s/rad Kstz = 106 N m/rad Rstz = 360 N m s/rad
Wheel mw = 15 kg Jwx = 0.1 kg m2 Jwy = 0.2 kg m2 Jwz = 0.1 kg m2
rw = 0.3 m
Tyre Kt = 305 kN/m Rt = 200 N s/m
C1 = 1.029 C2 = 17.16 C3 = 0.523 C4 = 0.03
Brake rlow = 0.2 rhigh = 0.25 sg = 0.01 kg = 250 N m
rbd = 0.15 m R1m = 0.04 N s/m Kca = 104 N/m Kre = 106 N/m
la = 1 m
Steering wheel Jsw = 1 kg m2 d = 0.1 rad

conventional braking and ABS is carried out. Finally, various simulations are performed to check the effectiveness of the con-
troller from vehicle performance perspective. The simulations were conducted in Symbols software with the specifications
given in Section 3.4.

5.1. Validation of four wheel model

The test data of Ref. [24] are used to validate the developed four wheel model. In Ref. [24], slalom maneuver test was
carried out on the vehicle at fixed vehicle speed of 50 km/h (13.88 m/s) and the steering wheel angle and the yaw rate were
measured. The steering wheel angle was maintained between ±100 degrees and the gear ratio between the steering wheel
and axle rotation was 25:1. With this small change in axle orientations, the under-steered vehicle’s linear speed remains al-
most constant at 50 km/h.
The experimental steering angle values (dstw in Fig. 18) at fixed time intervals were obtained from the graphs in Ref. [24]
and were used as input (h_ ref ¼ ddstw =dt) in the developed four wheel bond graph model. The simulated yaw rates are com-
pared with the experimentally obtained yaw rates of Ref. [24] in Fig. 18. Moreover, the results obtained in Ref. [24] from a
reduced order vehicle model are also plotted in Fig. 18 for comparison. It is found that the two models accurately predict the
actual handling behavior of the vehicle.

5.2. Antilock braking system in four wheel vehicle

Implementation of ABS has been carried out on the planar bicycle model to fine tune and test the control algorithm. Once
the controller implementation (program) was found to be satisfactory, it was ported to the full vehicle model.
The vehicle was brought to a steady longitudinal velocity of 33.3 m/s (120 km/h) along a straight path and then the ABS
was applied on the front wheels. The brake input signal as a square pulse was simulated to get the ABS characteristic. Fig. 19a
shows how the variation in the rotational speed of the wheel and vehicle speed take place when ABS is operational in an

Fig. 18. Comparison of vehicle’s yaw rate with the test and simulation results of Ref. [24] during slalom maneuver at vehicle speed of 50 km/h.
2148 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150

Fig. 19. (a) Vehicle and wheel speed; (b) wheel slip ratio; and (c) braking force during braking from 120 km/h.

emergency braking condition. Then the simple control logic, whose flow diagram is shown in Fig. 2b, was implemented to
maintain the slip ratio of the vehicle in the so called sweet spot of the friction-slip curve (Figs. 4 and 5) and the result in
Fig. 19b shows that the slip ratio lies between the desired limits, i.e., 0.2–0.25. The maximum braking force required for
ABS is 4988.28 N (Fig. 19c). The time taken for stopping the car is approximately 4.16 s.
The brake system performance at different initial vehicle speeds was also tested through simulations. The results given in
Fig. 20 are for dry asphalt road condition. It is seen that the time taken to bring the vehicle to a complete halt is 4.16 s, 2.65 s
and 1.35 s for initial vehicle speeds of 120 km/h, 80 km/h and 40 km/h, respectively. Although the curves in Fig. 20 look
apparently parallel, they are not exactly parallel. This discrepancy is due to the non-linear nature of the slip–friction curve.
ABS is more useful in adverse driving conditions. A comparison of ABS performance for different road conditions is given
in Fig. 21. It is found that the braking time increases as the friction coefficient reduces due to the road condition. The ABS
actuation is clearly seen from the variations in wheel speed during braking (Fig. 21b). Specific parts of the wheel speed re-
sponse are zoomed in Fig. 21b for closer inspection. There are insignificant variations in results for dry and wet asphalt road
conditions. However, in snow driving condition, it is difficult to control the wheel slip due to lack of grip between the road
and the tyre. While braking on snow, not only is the stopping time large as expected but also the switching frequency is
large. This means that the ABS actuation has to be very fast in adverse braking conditions.
The brake system design needs an integration of control system with the physical system model. Moreover, the brake sys-
tem actuation rate is a factor to be taken into consideration during its design. The brake system is repeatedly used in city
driving conditions where traffic is congested and often chaotic. Increase in longevity and reliability of brake system compo-
nents may contribute towards improvement of the overall road safety standard and fewer accidents.

6. Conclusions

A detailed bond graph model of a vehicle with sub-models for its components, especially the braking system, has been
developed in this paper. ABS controllers have been designed for the vehicle under consideration. The full vehicle model
 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150 2149

Fig. 20. (a) Vehicle and (b) wheel speed during ABS application at various initial vehicle speeds.

Fig. 21. (a) Vehicle and (b) wheel speed during braking from 120 km/h on various road surfaces.

includes dynamics of the load transfer on the wheels while the vehicle makes a maneuver. The basic scheme for ABS con-
troller modeling, simulation and analysis to maintain the wheel slip within a desired range are presented. ABS was modeled
using an equivalent mechanical braking system of the actual hydraulic system. The vehicle models and control algorithms
have been comprehensively validated by comparing the simulation results with experimental and numerical results avail-
able in the literature.
The model and the algorithm developed here for controller design are good enough to take care of different road and
operating conditions. Moreover, the bond graph approach being a multi-disciplinary modeling tool makes it convenient
to integrate models in other energy domains such as electrical drives and fuel cells into the developed bond graph model.
It found from simulation studies that the ABS actuation rate needs to be extremely fast in adverse driving conditions. The
actuation rate and peak brake force, etc. are to be considered during mechanical design of the brake system actuator so that
desired maximum forces, actuator response time, etc. are achieved while proper care is taken to increase the fatigue life of
brake system components and reduce the wear in the brake pads. The reconfigurable simulation model developed in this
article in which vehicle and brake system parameters can be changed serves as a design and controller prototyping platform.
This integrated physical and control system model avoids expensive field trials during early stages of brake system control
law tuning and actuator sizing.

References

[1] D.C. Karnopp, D.L. Margolis, R.C. Rosenberg, System Dynamics, Modeling and Simulation of Mechatronic Systems, John Wiley & Sons, NY, 2000.
[2] A. Mukherjee, R. Karmakar, A.K. Samantaray, Bond Graph in Modeling, Simulation and fault Identification, CRC Press, FL, 2006.
2150 T.K. Bera et al. / Simulation Modelling Practice and Theory 19 (2011) 2131–2150

[3] W. Borutzky, Bond Graphs-A Methodology for Modelling Multidisciplinary Dynamic Systems, SCS Publishing House, Erlangen, San Diego, 2004.
[4] A.K. Samantaray, B. Ould-Bouamama, Model-based Process Supervision, Springer Verlag, London, 2008.
[5] A.M. Bos, Modeling Multibody Systems in terms of Multibond Graphs, with Application to a Motorcycle, PhD thesis at Twente Univ., 1986.
[6] W. Cho, A bond graph approach to the modeling of general multibody dynamics systems, KEME International Journal 12 (5) (1998) 888–898.
[7] P.J. Gawthrop, Physical model-based control: a bond graph approach, Journal of the Franklin Institute 332 (3) (1995) 285–305.
[8] G. Dauphin-Tanguy, A. Rahmani, C. Sueur, Bond graph aided design of controlled systems, Simulation Practice and Theory 7 (5–6) (1999) 493–513.
[9] P.M. Pathak, R.P. Kumar, A. Mukherjee, A. Dasgupta, A scheme for robust trajectory control of space robots, Simulation Modelling Practice and Theory
16 (9) (2008) 1337–1349.
[10] R.F. Ngwompo, P.J. Gawthrop, Bond graph-based simulation of non-linear inverse systems using physical performance specifications, Journal of the
Franklin Institute 336 (8) (1999) 1225–1247.
[11] P.J. Gawthrop, Bond graphs: A representation for mechatronic systems, Mechatronics 1 (2) (1991) 127–156.
[12] J.J. Granda, The role of bond graph modeling and simulation in mechatronics systems: an integrated software tool: CAMP-G, MATLAB–SIMULINK,
Mechatronics 12 (9–10) (2002) 1271–1295.
[13] D. Hrovat, J. Asgari, M. Fodor, Automotive Mechatronic Systems, in: Mechatronic Systems Techniques and Applications, vol. 2, Gordon and Breach
Science Publishers, Amsterdam, 2000, pp. 1–98.
[14] G. Filippini, N. Nigro, S. Junco, Vehicle dynamics simulation using bond graphs, in: International Conference on Integrated Modelling and Analysis in
Applied Control and Automation (IMAACA 2007), Buenos Aires, Argentina.
[15] D. Margolis, T. Shim, A bond graph model incorporating sensors, actuators, and vehicle dynamics for developing controllers for vehicle safety, Journal
of the Franklin Institute 338 (2001) 21–34.
[16] H.B. Pacejka, Modelling complex vehicle systems using bond graphs, Journal of the Franklin Institute 319 (1/2) (1985) 67–81.
[17] L.S. Louca, J.L. Stein, D.G. Rideout, Generating proper integrated dynamic models for vehicle mobility using a bond graph formulation, in: Proceedings
of the 2001 International conference on Bond Graph Modeling (ICBGM’01) , Phoenix, vol. AZ 33(1), 2001, pp. 339–345.
[18] H.B. Pacejka, Tyre and Vehicle Dynamics, Butterworth-Heinemann, Elsevier, UK, 2006.
[19] Y. Oniz, E. Kayacan, O. Kaynak, A dynamic method to forecast the wheel slip for antilock braking system and its experimental evaluation, IEEE
Transactions on Systems, Man, and Cybernetics Part B Cybernetics 39 (2) (2009) 551–560.
[20] R.G. Longoria, A. Al-Sharif, C.B. Patil, Scaled vehicle system dynamics and control: a case study in anti-lock braking, International Journal of Vehicle
Autonomous Systems 2 (1/2) (2004) 18–39.
[21] M. Oudghiri, M. Chadli, A.E. Hajjaji, Robust fuzzy sliding mode control for antilock braking system, International Journal on Sciences and Techniques of
Automatic Control 1 (1) (2007) 13–28.
[22] C.B. Patil, R.G. Longoria, J. Limroth, Control prototyping for an Anti-lock braking control system on a scaled vehicle, in: Proceedings of 42nd IEEE
Conference on Decision and Control, USA, 2003, pp. 4962–4967.
[23] G.F. Mauer, A fuzzy logic controller for an ABS braking system, IEEE Transactions on Fuzzy Systems 3 (4) (2005) 381–388.
[24] W. Drozdz, H.B. Pacejka, Development and validation of a bond graph handling model of an automobile, Journal of the Franklin Institute 328 (5/6)
(1991) 941–957.
[25] R. Merzouki, B. Ould-Bouamama, M.A. Djeziri, M. Bouteldja, Modelling and estimation of tire–road longitudinal impact efforts using bond graph
approach, Mechatronics 17 (2–3) (2007) 93–108.
[26] S.L. Koo, H.S. Tan, M. Tomizuka, An improved tire model for vehicle lateral dynamics and control, in: Proceedings of the American Control Conference,
Minneapolis, Minnesota, USA, 2006, pp. 5879–5884.
[27] R. Rajamani, Vehicle Dynamics and Control, Springer, US, 2006.
[28] B. Ozdalyan, Development of a slip control anti-lock braking system model, International Journal of Automotive Technology 9 (1) (2008) 71–80.
[29] K. Li, J.A. Misener, K. Hedrick, On-board road condition monitoring system using slip-based tyre–road friction estimation and wheel speed signal
analysis, Journal of Multi-Body Dynamics, IMechE 221 (2007) 129–146.
[30] J. Svendenius, Tire Modeling and Friction Estimation, Lund University, Lund, Sweden, 2007.
[31] M. Denny, The dynamics of antilock brake systems, European Journal of Physics 26 (2005) 1007–1016.
[32] T.D. Gillespie, Fundamentals of Vehicle Dynamics, SAE International, 1992.
[33] H.K. Fathy, Z.S. Filipi, J. Hagena, J.L. Stein, Review of hardware-in-the-loop simulation and its prospects in the automotive area, Proceedings of SPIE –
The international Society for Optical Engineering 6228 (2006). 62280E.1–62280E.20.
[34] M.L. Kuang, M. Fodor, D. Hrovat, M. Tran, Hydraulic brake system modeling and control for active control of vehicle dynamics, in: Proceedings of the
American Control Conference, San Diego, California, 1999, pp. 4538–4542.
[35] S. Drakunov, U. Ozguner, P. Dix, B. Ashrafi, ABS control using optimum search via sliding modes, IEEE Transactions on Control Systems Technology 3 (1)
(1995) 79–85.
[36] Matteo Corno, S.M. Savaresi, G.J. Balas, Linear, parameter-varying wheel slip control for two-wheeled vehicles, in: Proceedings of the 47th IEEE
Conference on Decision and Control Cancun, Mexico, December 9–11, 2008, pp. 5030–5035.
[37] C. Song, J. Wang, L. Jin, Study on the composite ABS control of vehicles with four electric wheels, Journal of Computers 6 (3) (2011) 618–626.
[38] A. Sanyal, R. Karmakar, Directional stability of Truck-Dolly-Trailer system, Vehicle System Dynamics 24 (8) (1995) 617–637.
[39] T. Ersal, H.K. Fathy, J.L. Stein, Structural simplification of modular bond-graph models based on junction inactivity, Simulation Modelling Practice and
Theory 17 (2009) 175–196.
[40] Symbols User’s Manual, HighTech Consultants. <http://www.htcinfo.com>.
[41] B. Ould Bouamama, K. Medjaher, A.K. Samantaray, M. Staroswiecki, Supervision of an industrial steam generator. Part I: Bond graph modeling, Control
Engineering Practice 14 (2006) 71–83.
[42] A.K. Samantaray, K. Medjaher, B. Ould Bouamama, M. Staroswiecki, G. Dauphin-Tanguy, Diagnostic bond graphs for online fault detection and
isolation, Simulation Modelling Practice and Theory 14 (2006) 237–262.
[43] B. Ould Bouamama, Bond graph approach as analysis tool in thermofluid model library conception, Journal of the Franklin Institute 340 (2003) 1–25.
[44] H.T. Szostak, W.R. Allen, T.J. Rosenthal, Analytical Modeling of Driver Response in Crash Avoidance Maneuvering, An Interactive Model for Driver/
Vehicle Simulation, vol. II, US Department of Transportation Report NHTSA DOT HS-807-271, 1988.
[45] K. Huh, D. Hong, J.L. Stein, Development of a lane departure monitoring and control system, Journal of Mechanical Science and Technology 19 (11)
(2005) 1998–2006.