IJST, Transactions of Mechanical Engineering, Vol. 37, No.

M2, pp 161-174
Printed in The Islamic Republic of Iran, 2013
Shiraz University

AN OPTIMIZED MULTI-STAGE SCHEME TO COORDINATE

*
STEERING AND BRAKING
A. TAVASOLI1 AND M. NARAGHI2**
1

Sama Technical and Vocational Training College, Islamic Azad University, Karaj Branch, Karaj, I. R. of Iran
2
Dept of Mechanical Engineering, Amirkabir University of Technology, Tehran, I. R. of Iran
Email: naraghi@aut.ac.ir
Abstract A novel Integrated Vehicle Dynamics Control (IVDC) scheme is presented to
coordinate active steering and braking subsystems. The multi-stage coordination scheme is based
on the phase-plane Method. The first stage includes the high-level controller which integrates three
individual controllers according to vehicle states in the phase plane. In the next stage, an optimized
scheme is established to allocate the control objectives to individual braking and steering forces.
To achieve this, an inequality-constrained optimization problem, including driver's brake demand,
is solved analytically. Coefficients of the cost function are adapted based on the vehicle phaseplane trajectory to realize a proper coordination of braking and steering subsystems. Simulation
results validate the effectiveness of the proposed method to enhance the vehicle dynamics control.

Keywords Coordinated, vehicle dynamics, active steering, wheel braking, optimal, tire force distribution, phaseplane method

1. INTRODUCTION
Through the last two decades, numerous active safety systems have been developed to enhance handling
and stability of ground vehicles, and considerable improvements have been achieved in this regard.
However, saturation of tire forces, as well as limitations in control of every aspect of vehicle dynamics in
critical maneuvers keeps this area challenging and attractive to the researchers. The latest approaches to
this problem include: coordinated (as well as integrated) vehicle dynamics control, and optimal
distribution of tire forces.
A functional integration of slip-ratio control and active suspension for lateral vehicle dynamics was
discussed in [1]. Integration of individual wheel braking and active front and rear wheel steerings was
presented [2]. In another work, [3], the feedback linearization technique was applied to the problem of
integrated vehicle control using steering and brakes. In 2006, Crolla suggested a rule-based integration
scheme to coordinate active steering, driveline, and braking [4].
In the context of optimal distribution of tire forces, force distribution to maximize
acceleration/deceleration of a four-wheel vehicle during cornering was studied in [5]. Mokhyamar and
Abe [6] presented a method of tire force distribution to minimize entire tires workload usage. The notion
of Adaptive-Optimal Distribution of tire Forces (AODF) was introduced by the authors in [7]. An
optimizing scheme is suggested to achieve maximum handling with guaranteed vehicle dynamics stability
[8]. To assure vehicle dynamics stability, the standard stability constraints of the phase-plane are
incorporated into the optimal distribution of tire forces unit. To distribute vehicle control among
individual tire forces constrained to nonlinear saturation condition, static and dynamic control allocation

Received by the editors January 10, 2012; Accepted March 9, 2013.

Corresponding author

162

A. Tavasoli and M. Naraghi

techniques are introduced into IVDC [9], [10], and the results of both methods are compared [10].
Optimized coordination of brakes and active steering for a 4WS passenger car was considered [11].
In the present work, a multi-stage IVDC based on adaptive-optimal distribution of tire forces is
considered. In a new approach, the concept of coordination based on phase-plane method has been utilized
to make both consistency between high-level control objectives and the other to integrate steering and
braking subsystems in an optimal way. While this method applies merits of ODF to the phase-plane
approach, it contributes to the related literature in the following ways.
- The phase plane method is utilized to organize several high-level control objectives for the ODF system.
In this regard, a new dynamics stability controller uses both steering and braking subsystems in an optimal
manner.
- An adaptive optimal approach is established to allocate the integrated high-level control objectives to the
individual steering and braking tire forces in the presence of braking demand of the driver. A new
optimization problem is defined and solved analytically.
- An adaptation mechanism is considered which adjusts different weighting coefficients of the cost
function in the optimization problem according to the phase-plane notion. Thus, a phase-plane based
adaptive-optimal approach is utilized to distribute tire forces.
The effectiveness of the proposed method is evaluated through simulation results. The operation of a
nonlinear vehicle model under the proposed method is examined, and by comparison with the previous
results, the conclusions of the work are drawn.
The rest of the paper is organized as follows: The next section presents the high-level control design
procedure. In section 3, the AODF method is demonstrated. Sections 4 and 5 are devoted to simulation
results and concluding remarks.
2. COORDINATED HIGH-LEVEL CONTROL DESIGN
In general, vehicle handling and stability are achieved through controlling the yaw rate and side-slip angle,
respectively. To this end, according to the vehicle lateral dynamics, there are two available virtual control
inputs of total body lateral force and yaw moment. To design virtual control inputs through a coordinated
high-level controller, individual high-level control laws for handling and stability are organized according
to the phase-plane methodology. A description of various regions in the phase-plane of the side-slip angle,
, has been shown in Fig. 1, for reference region of stability control definition [12]. In what follows, the
design procedures for different high-level control objectives are first presented. Then, a phase-plane based
adaptation mechanism is proposed to coordinate several high-level control objectives.
a) Individual controllers design
In this section, three individual control laws are demonstrated. The controller design is based on a
2DOF linear vehicle model, with constant speed. The basic equations of motion for this model are [13]:
(1)
(2)
where m and Iz denote the total mass and yaw moment of inertia of the vehicle, respectively, V is the
vehicle velocity, and r represents the yaw rate. M and Y are desired body yaw moment and lateral force.
To deal with the substantial nonlinear characteristics and uncertainties of the vehicle dynamics, individual
controllers based on sliding-mode methodology are designed to ensure robustness.

IJST, Transactions of Mechanical Engineering, Volume 37, Number M2

October 2013

163

An optimized multi-stage scheme to

Stability controller
activated

Stability controller
activated

Stable region
(Reference
region)

Stability
error

Control
Boundaries

Fig. 1. Different regions in

phase-plane

1. Controller design for the vehicle handling: To improve vehicle manoeuvrability, generalized body
lateral force and yaw moment are used to control side-slip angle and yaw rate, respectively. Design
procedure is based on Eqs. (1) and (2) which represent a 2DOF linear model. A reference model [13]
generates the desired yaw rate for the vehicle handling. To account for un-modelled dynamics and
uncertainties, the disturbance terms, and r, are embedded into each equation to get
(3)

(4)

The unknown, but bounded, disturbances could be time dependent functions of different vehicle states. To
overcome the substantial uncertainties in vehicle dynamic modelling, we adopt sliding mode control that
has been proven effective in achieveing a robust vehicle control design [14]. To create the total lateral
force, Y, for a zero desired side-slip angle, the sliding surface is defined as:
(5)
Differentiating this equation and considering (3)
(6)
where the term

is assumed to be bounded to a known value, , so that

(7)

To meet the sliding condition [15]

,

(8)

the desired body lateral force is considered as

(9)
in which

is to be designed. Insert (9) in (6) to get the left side of (8) as

(10)

whose combination with (7) is given by

Then, to reach the desired behaviour of sliding condition (8),

(11)

is considered to be
(12)

October2013

IJST, Transactions of Mechanical Engineering, Volume 37, Number M2

164

where

A. Tavasoli and M. Naraghi

. is the signum function, and

0 is a design parameter which is chosen as

(13)

Substituting (12) into (9), the desired body lateral force is attained. In practical implementation, to leave
out the chattering effects of the sign switching function, it is replaced by a saturation function with a
boundary layer thickness of >0. Thus, the final control law becomes
(14)
In order to compute the body yaw moment, Mr, to track the desired yaw rate, rd, an integral sliding surface
is adopted
,

(15)

where the integral term is used to mitigate the undesirable yaw angle, or vehicle heading offset, and to
ensure the desired vehicle heading. Differentiating (15), along with (4), leads to
,

(16)

The desired body yaw moment for handling can be derived in a similar way as presented before to get (14)
,

(17)

where

with

0 being an upper bound for |

|, and

(18)
0 the sliding surface parameter.

2. Dynamic stability control design: The Dynamic Stability Control (DSC) system employs the body
yaw moment to control the side-slip angle. Usually, a counter-steering controller near the stability borders
plus a PD control action for the braking subsystem is used to control the side-slip angle [1]-[4]. However,
adopting the PD control action necessitates measurement of the side-slip angular acceleration, which is
undesirable from a practical point of view. Here, a control law is designed to compute the body yaw
moment directly for the side-slip angle control. Thus, a DSC law, which requires only side-sip angle and
side-slip angle velocity based on sliding mode is introduced. To get the body yaw moment, MDSC, for the
side-slip angle control, first differentiate (3) and replace (4) for to get
(19)
In Eq. (19), since
is the control input to be designed, the actual value of the body lateral force, Fy, is
represents total unknown uncertainties in achieving (19),
used instead of its desired value Y. Also,
and is bounded by the known value
|

(20)

The boundaries of the reference region in Fig. 1 are defined as [4]

1

(21)

Since the main task of DSC subsystem is to draw the vehicle phase-plane trajectory back into the stable
region, whenever it exits this region, the sliding surface outside the stable regime is defined as

IJST, Transactions of Mechanical Engineering, Volume 37, Number M2

October 2013

165

An optimized multi-stage scheme to

1,

1,

(22)

To realize the sliding surface dynamics, differentiate (22) and use (19) and (3)
4

(23)

Growth of the side-slip angle saturates individual lateral tire forces outside the stable region. In such
conditions, we can assume the total body lateral force is limited to the vehicle weight multiplied by the
tire/road friction coefficient , i.e. mg, and
is expected to be bounded as well. As a result, by
considering (7) and (20), the total disturbance-like term on the right of (23) can be assumed to fulfill
4

(24)

To achieve the sliding condition, the same procedure applied before can be implemented such that
4

(25)

where

with

(26)

0 being the sliding surface parameter.

b) Coordination of high-level control objectives

The key problems to be addressed in coordinating different high-level controllers are which
controllers to use in each regime, and how to make consistency between various control objectives to rule
out possible conflicts. Accordingly, the control objectives need to be put in order of priority. To this end,
based on the phase-plane concept, inside the stable region where tire forces are considered to have linear
characteristics, for handling improvement, the total body yaw moment and lateral force, given by (14) and
(17), are utilized to control yaw rate and side-slip angle, respectively. At higher side-slip angles, the highlevel control objective transits from handling to stability. The vehicle stability is related directly to the
side-slip motion, and this motion should be restricted to the stable region (Fig. 1). As the vehicle phaseplane trajectory approaches the reference boundaries, with growth of the side-slip angle, lateral tire forces
saturate, and using exclusively the body lateral force, Eq. (14) is no longer a proper function to reduce
side-slip motion. In such conditions, the notion of DSC is exploited to stabilize the vehicle, and the total
yaw moment is used for stability control through DSC system (Eq. (25)). Consequently, the final body
yaw moment is computed as:
1

(27)

is the high-level adaptation gain which is adjusted according to the vehicle state in the phasewhere
plane, as shown in Fig. 2. By this adaptation mechanism, hard switchings between several types of control
objectives is avoided, and the control task transits from one to another instead. As a result, abrupt system
responses that can be induced by sudden hard switching actions are excluded. The coordinated high-level
control scheme can be found in Fig. 3.
October2013

IJST, Transactions of Mechanical Engineering, Volume 37, Number M2

166

A. Tavasoli and M. Naraghi

1
1

24
24
1.
1
Fig. 2. Adaptation mechanism for high level control

Sliding Mode
Control

, r

Y
Mr

Prioritizing
high-level
control
objectives

DSC

System
Phase plane

nr
Gain adaptation
mechanism

Fig. 3. Coordinated high-level control scheme

3. ADAPTIVE-OPTIMAL DISTRIBUTION OF TIRE FORCES

In this section, the notion of adaptive-optimal distribution of tire forces is used to distribute the total body
force and moment of the first stage among individual tire forces. An optimization problem subjected to
equality and inequality constraints is defined, and solved analytically to utilize the maximum capacity of
tire forces for control objectives. The analytical solution of the inequality constrained optimization
problem makes the proposed IVDC scheme more suitable for realistic applications. In order to coordinate
the steering and braking subsystems, another adaptation mechanism is considered to adjust the coefficients
of the cost function in this phase. Therefore, the part of the braking action that is undesirable by driver,
and is used only for DSC system, brake-based DSC, is disabled inside the stable region of the phase-plane.
Accordingly, the negative effect of the brake-based DSC on the longitudinal motion is averted in this
region. Once the side-slip angle increases and vehicle goes beyond the stable regime the brake-based DSC
is activated to limit the side-slip motion.
a) Formulation and solution of the constrained optimization problem
With the proposed AODF, we seek to fulfill two main goals. One is to utilize the maximum capability
of tire forces and the other is to manage the active steering and braking subsystems. These are
accomplished by considering and optimizing a convenient two part cost function. The first part, which is a
summation of squared normalized resultant tire forces (tires workload), is to minimize tires workload and
the second one, being the squared summation of normalized longitudinal tire forces, enables adjusting the
steering and braking contributions in IVDC. The cost function is written according to:

(28)

where i denotes wheel number, Fxi and Fyi are the total longitudinal and lateral tire forces, Fxdi and Fxbi are
the longitudinal braking forces due to DSC operation and drivers demand, respectively, Fzi is the vertical
IJST, Transactions of Mechanical Engineering, Volume 37, Number M2

October 2013

167

An optimized multi-stage scheme to

load, and i is the tire friction coefficient, all defined in the vehicle body fixed coordinate system, as
shown in Fig. 4. In addition, Ai and Bi are weighting coefficients to be adjusted and X is a 121 vector that
contains tire forces, as dependent variables for the cost function. It is assumed that each wheel can be
steered independently and only braking torques at wheels are possible.
The weighting coefficient Bi is updated by means of another phase-plane based adaptation
mechanism. This value determines the contribution of the braking due to the DSC system in the objective
function. Thus, when the vehicle phase-plane trajectory is within the stable region, in order to prevent
negative effects of braking on longitudinal dynamics and disturbances to the driver, the coefficient Bi is set
to a large value (1000 in simulations). Accordingly, in this region, the contribution of braking due to DSC
in the cost function is increased and this subsystem is disabled. As the side-slip dynamics moves toward
unstable region of the phase-plane the coefficient Bi transits to 0, thereby decreasing the contribution of
brake-based DSC in the cost function. Therefore, in this regime brake-based DSC is activated to take
advantage of this system for vehicle stabilization. Additionally, to further reduce the influence of brakebased DSC on the longitudinal motion, the operation of this system is postponed by delaying the
adaptation gain of the AODF relative to that of high-level control, as seen in Fig. 5. By this delay, when
the vehicle stability is in question, the DSC system will first operate through steering subsystem, by
applying active counter-steering, to stabilize the vehicle. However, as the lateral acceleration grows and
the phase-plane curve goes further out of the stable region, the effectiveness of the steering subsystem for
vehicle stabilization fades away and, hence, braking is activated. In this condition, the task of the stability
control through DSC is shared between steering and braking actuation concepts by AODF system.
1

FY 2

FX 1

FX 2

Lf

Adaptation gain

FY1

4
Lr

FY 3

FY 4

Integrated high-level control

AODF System
1

FX 3

1.1 1.2

1
24

4
24

FX 4

Fig. 4. Vehicle dynamics model schematic in top view

Fig. 5. Adaptation gains in different phases of IVDC

The sum of the lateral forces and yaw moments acting on the vehicle by each tire force should equal
to the required total lateral force, Y, and yaw moment, M, computed based on the presented integrated
high-level control. Therefore, referring to Fig. 4, all of the variables in objective function must satisfy the
two equality constraints as follows:

(29)

(30)

In addition, the longitudinal braking forces Fxbi should assure the demanded total braking force, ax<0, by
driver

October2013

(31)

IJST, Transactions of Mechanical Engineering, Volume 37, Number M2

168

A. Tavasoli and M. Naraghi

On the other hand, we assume only braking torque is possible at each wheel and the driveline torque is
excluded. These words are written as
0,

1, ,4

(32)

Further, the statement that the positive braking torque has no physical meaning results in:
0,

5, ,8

(33)

Thus, the cost function contains twelve variables, which must satisfy three equality constraints (29)(31), and eight inequality constraints in (32) and (33).
It can be readily shown that the objective and constraint functions are convex. Therefore, the KarushKuhn-Tucker (KKT) conditions are necessary and sufficient for a global minimum [16]. In this paper, the
solution of the optimization problem is simplified by developing an intuition to assess which inequality
constraints may be active at each time step, according to the vehicle dynamics. In a typical optimization
problem, all inequality constraints that are treated as equalities at the optimum point constitute the active
set. The recognition of the active set is a key factor to solving any optimization problem. The inequality
constraints of the active set are regarded as equalities and the remaining inequality constraints are
disregarded [17]. Hence, as the active set is realized, the inequality optimization problem turns into an
equality one. To find out the feasible active sets for our problem, off-line computations were performed in
[18]. At each stage, a combination of gj-s was treated as an active set and, based on this specific active set,
KKT conditions were solved to get the optimization variables, i.e. individual tire forces. Then, the
inequality constraints in KKT conditions were checked and only the solutions satisfying these inequalities
were chosen as a feasible solution. All possible combinations of gj-s were examined and it was concluded
that, according to the sign of the body yaw moment M, only five cases might yield a true active set. The
possible solutions are presented in the Appendix so that the desired tire forces can be obtained through
these equations.
Then, using the inverse of simple tire model the active steering angle of each wheel, i, can be
determined, as depicted in [7]. Further, by ignoring the rotational dynamics of wheels, it is assumed that
the applied torque at each wheel is equal to the wheel radius times the desired longitudinal force.
Nonetheless, to avoid wheel lock, specifically in critical maneuvers wherein the demanded longitudinal
forces by AODF might be high, an ABS for a brake by wire system is considered to compute the final
braking torque, Ti, at each wheel. The overall scheme of the proposed IVDC can be seen in Fig. 6.
4. SIMULATION RESULTS
In this section, the operation of the vehicle under suggested IVDC is examined through simulations. A
9DOF nonlinear vehicle model is used for this purpose [7]. Also, a driver model, validated experimentally
[19], is used to simulate the drivers behavior in different maneuvers. To compare the proposed method
with previous works, we present the results of AODF with a combined high-level control, combined ODF,
for which no coordination is made among high-level control objectives [6], [7], [20], and [21]. In all
maneuvers, the vehicle is assumed to move with an initial velocity of 130 km/h on a slippery road where
the coefficient of friction is 0.3.
a) Single-lane changing without drivers braking command
A Single-lane change (SLC) without any braking demand by driver is considered as the first
maneuver. Simulation results for this scenario are shown in Figs. 7-13, wherein the results of the present
work have been denoted by Coordinated AODF. As depicted in Figs. 7-9, the vehicle with no active
control and only guided by driver has an oscillatory and completely unstable response in this critical
IJST, Transactions of Mechanical Engineering, Volume 37, Number M2

October 2013

169

An optimizedd multi-stage sccheme to

maneuver. Nonethelesss, using active safety coontrol throu

ugh AODF system,
s
withh either com
mbined or
integrated hhigh-level coontroller, maakes the vehiicle converg
ge to the inteended path bby driver. Fu
urther, the
vehicle withh integrated high-level control has the best perrformance in
n convergingg to the desired path,
following thhe desired yaaw rate, and reducing thee side-lip ang
gle. In Fig. 10,
1 the phasee-plan trajecttory under
Coordinatedd AODF hass converged to
t the stablee region moree effectively due to efficcient function
n of DSC.
This approaach has also kept
k the tiress away from saturation more
m
efficienttly, as seen bby Fig. 11. The
T plot of
longitudinall forces, Figg. 12, and acttive steeringg, Fig. 13, arre shown to assess how tthe suggesteed method
makes use of the availlable actuato
ors. Outside the referencce region, Coordinated A
AODF has demanded
d
braking forcces at wheells on the veh
hicle right oonly in a while to stabilize the vehiccle. At the saame time,
Combined A
AODF has applied
a
moree braking forrces, thus inccreasing the negative effe
fects of brakiing on the
vehicle longgitudinal dynnamics. Furth
her, active stteering by bo
oth AODF sy
ystems is idenntical until th
he vehicle
phase-planee trajectory leaves
l
the staable region iin the phase-plane. Then
n, Coordinateed AODF haas applied
counter-steeering to furrther stabilizze the vehiccle. Also, th
he active steeering by C
Coordinated AODF is
smoother annd more propper.
- Esstimation
Drivers
Commands

F zi- Estimaation
F yi

ax

Inversse i
Tire Mo
odel

AODF

Integrrated

High-L
Level Y & M

System

Ti

F xi
AB
BS

Conttrol
AODF
Gain Adaptation

g
High-Level
Gain Adaptation

Vehicle Displacement & Velocity

a x,a y

Nonliinear Vehiclee
Model

V, r

Tires Slip Ratio

, - Estimattion

F yi

T
Tire Model

Fig
g. 6. Structure of the overalll IVDC system
m

Fig. 7. Vehhicle path in SLC without drrivers brakinng

October2013

Fig. 9.
9 Side-slip an
ngle in SLC w
without drivers braking

IJST
T, Transaction
ns of Mechanica
al Engineeringg, Volume 37, Number
N
M2

170

A. Tavaasoli and M. Na
araghi

Fig. 8.
8 Yaw rate inn SLC withoutt drivers brak
king

y in SLC
Figg. 10. Phase-pllane trajectory
without drivers braking

Fig. 11. Tirres work-load in SLC witho

out
drivers brakking

manded by DS
SC in SLC wiithout driverss braking
Fig. 12. Tiire longitudinaal forces comm

Fig. 13. Active

A
steering
ng in SLC with
hout drivers braking
b
IJST, Transacctions of Mechanical Engineeering, Volume 337, Number M2
M

Occtober 2013

An optimizedd multi-stage sccheme to

171

b) Single-laane change maneuver

m
wiith drivers bbraking com
mmand
To exaamine the fuunction of th
he proposedd IVDC in a more criticcal maneuveer, SLC with
h drivers
braking dem
mand of -0.3g, between
n t=1s and tt=4s, is con
nsidered. Thee response oof the vehiccle to this
maneuver iis in shownn in Figs. 14-18. Againn it is seen that the veehicle underr exclusively
y drivers
managemennt is compeleetely unstable. On the oother hand, the
t vehicle under
u
Coordi
dinated AOD
DF has the
best responsse in terms of
o vehicle paath in Fig. 144, yaw rate in
n Fig. 15, and side-slip an
angle in Fig. 16. These
are due to the fact thaat Coordinateed AODF uutilizes the DSC
D
system to keep thee vehicle ph
hase-plane
trajectory inside the sttable region
n so effectivvely, as shown by Fig. 17. Figure 18 reveals how the
Coordinatedd AODF disstributes the yaw momennt demanded
d by DSC am
mong tire loongitudinal forces.
f
As
vehicle phaase plan trajeectory leavess the stable region, the DSC
D
system
m applies braaking torques on right
wheels to sstabilize the vehicle.
v
Thiss braking torrque, which is
i undesirable for passenggers, has beeen applied
only for a shhort period of
o time due to
t proper funnctioning of the
t proposed
d adaptation m
mechanisms. Then, as
the vehicle phase plan trrajectory retu
urns to the sstable area, Coordinated
C
AODF
A
system
em command
ds positive
longitudinall forces for DSC
D operatio
on to release the brakes, thereby
t
decreasing the tootal tires worrk-load.

Vehicle path inn SLC with drrivers brakingg

Fig. 14. V

Fig. 16. Side-slip angle in SLC

C with driverss braking

Fig. 15. Yaw rate in SLC with drivers

d
brakin
ng

hase-plane traj ectory in SLC

C with driverss braking
Fig. 17. Ph

October2013

IJST
T, Transaction
ns of Mechanica
al Engineeringg, Volume 37, Number
N
M2

172

A. Tavaasoli and M. Na
araghi

F 18. Longiitudinal forcess by DSC in SLC

Fig.
S
with drivers braking

5. CO
ONCLUSIO
ON
a
steerin
ng and brakiing subsystem
ms was preseented in thiss paper. A co
oordinated
Optimal inttegration of active
high-level ccontroller avverts possible conflicts between vaarious contro
ol objectivess. A new DSC was
designed annd by the prroposed meth
hod the DSC
C exploits bo
oth active stteering and bbraking actu
uations for
vehicle conttrol in an opttimal manner. Then, the the high-level control is distributed oover the indiv
vidual tire
forces throuugh an adapttive-optimal approach in the presencee of drivers braking accceleration dem
mand. So,
the maximuum capacity of tire forces is used to fulfill the co
ontrol objecttives. At the same time, the active
steering annd braking subsystems are coorddinated through another phase-plan
ane based adaptation
a
mechanism.. To this endd, the weightting coeffici ents of the cost
c function
n are adjustedd based on the
t phaseplane methood. An anallytical solution was deriived for the proposed op
ptimization pproblem. Th
he vehicle
behavior unnder the propposed method
d was investiigated during
g some criticcal maneuverrs and the ressults were
compared w
with those off earlier work
k. It is conclluded that un
nder the prop
posed schem
me the vehicle stability
is enhancedd significanlyy compared to previous w
works of the literature.
l
RE
EFERENCES
1.

Smakmaan, H. T. (20000). Function

nal integratioon of slip con
ntrol with acttive suspensioon for improv
ved lateral
vehicle ddynamics. PhD
D Dissertation
n. s.l. : Delft U
University of Technology, The
T Netherlannds.

2.

Selby, M
M. (2003). Inteelligent vehiclle motion conttrol. Leeds : PhD
P Dissertatiion, Universityy of Leeds, UK.
U

3.

Burgio, G. & Zegelaaar, P. (2006). Integrated

I
vehhicle control using
u
steering
g and brakes. IInternational Journal of
Control,, Vol. 79, No. 5, pp. 53454
41.

4.

He, J. (22006). Coordiination of acttive steering, driveline, and

d braking for integrated veehicle dynamiics control.
Proc. IM
MechE, Vol. 2220, Part D: J. Automobile E
Engineering, pp.
p 1401-1421
1.

5.

Peng, H
H. & Hu, J. S. (1999). Tracction/braking force distribu
ution for optim
mal longitudinnal motion du
uring curve
followinng. Vehicle Sysstem Dynamiccs, Vol. 26, Noo.4, pp. 301-3
320.

6.
7.

8.
9.

Mokhyaamar, O & Abe,

A
M. (2004). Simultaneeous optimal distribution of lateral tiree forces for the model
followinng control. Jouurnal of Dynamic system, M
Measurement, and Control, Vol. 126, pp. 753-763.
an adaptiv
Naraghi,, M., Roshannbin, A. & Tavasoli,
T
A. ((2010). Vehiccle stability enhancement
e
ve optimal
approachh to the distribbution of tyre forces. Proc. IMechE, Parrt D: J. Automobile Engineeering, Vol. 224, pp. 443453.
Tavasolii, A. & Naraaghi, M. (2011). An optim
mizing schemee to achieve maximum haandling with guaranteed
vehicle ddynamics stabbility. Australiian Journal off Basic and Ap
pplied Sciencees, Vol. 5, Noo. 11, pp. 1989
9-2001.
Tavasolii, A. & Naragghi, M. (2013)). Interior poinnt method to optimize
o
tire force
f
allocatioon in 4-wheeleed vehicles
using high-level slidinng mode conttrol with adapptive gain. Asian Journal of
o Control, Vool. 15, No. 4, pp. 1188
1200.

IJST, Transacctions of Mechanical Engineeering, Volume 337, Number M2

M

Occtober 2013

173

An optimized multi-stage scheme to

10. Tavasoli, A. & Naraghi, M. (2012). Vehicle sliding mode control with adaptive upper bounds: static versus
dynamic allocation to saturated tire forces. Mathematical Problems in Engineering, Vol. 2012, pp. 1-31.
11. Tavasoli, A. & Naraghi, M. (2012). Optimized coordination of brakes and active steering for a 4WS passenger
car. ISA Transactions, Vol. 51, pp. 573583.
12. Inagaki, S., Kshiro, I. & Yamamoto, M. (1994). Analysis on vehicle stability in critical cornering using phaseplane method. AVEC '94. pp. 287-292.
13. Rajamani, R. (2006). Vehicle dynamics and control. New York : Springer Science.
14. Guclu, R. & Yagiz, N. (2004). Comparison of different control strategies on a vehicle using sliding mode
control. Iranian Journal of Science & Technology, Transaction B, Engineering, Vol. 28, No. B4, pp. 413-422.
15. Khalil, H. K. (2002). Nonlinear systems. 3rd ed: Prentice Hall.
16. Boyd, S. & Vandenberghe, L. (2009). Convex optimization. 7th Ed, Cambridge University Press.
17. Nocedal, J. & Wright, S. J. (2006). Numerical optimization. 2nd Ed, Springer Science, USA.
18. Roshanbin, A. (2008). Vehicle stability enhancement- an adaptive-optimal approach to distribution of tire
forces. Iran- Tehran, Amirkabir University of Technology.
19. Abe, M. (2009). Vehicle handling dynamics: theory and application. Elsevier Ltd, UK.
20. Wang, J. & Longoria, R. G. (2009). Coordinated and reconfigurable vehicle dynamics control. IEEE
Transactions on Control Systems Technology, Vol. 17, No. 3, pp. 723-732.
21. Tjnns, J. & Johansen, T. A. (2010). Stabilization of automotive vehicles using active steering and adaptive
brake control allocation. IEEE Transactions on Control Systems Technology, Vol. 18, No 3, pp. 545-558.
22. Allen, R. W., Rosenthal, T. J. & Szostak, H. T. (1988). Steady state and transient analysis of ground vehicle
handling. SAE paper, No 870495, pp. 482-512, 1988.

APPENDIX: POSSIBLE SOLUTIONS OF KKT CONDITIONS

The optimization problem in Section 3 is summarized as

0
1, ,8

0,

The vector of desired individual tire forces is defined as

,

To realize u, KKT conditions [16] are applied to the optimization problem and the following equations are derived

0,
0,
0
0

1, ,3
1, ,8

in which
and are the Lagrange multipliers corresponding to equality and inequality constraints respectively.
Then, KKT conditions result in these five sets of linear equations as follows [18]:
1- If

1,3,5, 7 0, 2 , 4, 6,8 0,

x 1 [ Fxd 1
October2013

Fxd 2

F xd 3

F xb1

g1,3,5, 7 0,
F xb 2

Fxb 3

g 2, 4, 6,8 0,
F y1

Fy 2

then

[A 1 ] x 1 b 1
1313

Fy 3

131

where

131

8 ]T

IJST, Transactions of Mechanical Engineering, Volume 37, Number M2

174

A. Tavasoli and M. Naraghi

2- If

1,3,5, 7 0 , 2, 4, 6,8 0,

x 2 [ Fxd 1
3- If

Fxb1

Fxb 2

Fxd 2

Fxd 3

F xd 2

j 0,

x 5 [ F xd 1

F xd 2

The elements of the matrices,

F xd 3

Fxb1

Fxb 2

Fxb 3

F xb 2

g j 0,

and

F y1

Fy 2

F y1

g 2 , 4,5, 6, 7 ,8 0,

F xb1

F xd 3

Fxb 3

F xb 3

Fy 2

j 1,...,8

Fy 2
then

then

F xb 2
,

F xb 3

F y1

Fy 2

then

where

131

7 ]T

[ A 3 ] x 3 b 3
111

where

111

4 ]T

[A 4 ] x 4 b 4
1111

Fy3

131

1111

111

where

111

3 ]T

[ A 5 ] x 5 b 5
99

F xb1

[A 2 ] x 2 b 2
1313

Fy 3

g 1,3 0,

F y1

then

Fy3

g1,3,5, 6, 7 ,8 0, g 2 , 4 0,

2, 4,5,6 ,7 ,8 0, 1,3 0,

x 4 [ F xd 1
5- If

Fxd 3

1,3,5,6, 7 ,8 0, 2 , 4 0,

x 3 [ Fxd 1
4- If

Fxd 2

g 1,3,5, 7 0, g 2, 4, 6,8 0,

91

where

91

F y 3 ]T

1, ,5, are the functions of Ai, Bi, i, Fzi, d, Lf, Lr, Fy, and M.

IJST, Transactions of Mechanical Engineering, Volume 37, Number M2

October 2013