J Eswa 2013 07 085

Accepted Manuscript
Adaptive Cruise Control of a HEV Using Sliding Mode Control
Behnam Ganji, Abbas Z. Kouzani, Sui Yang Khoo, Mojtaba Shams Zahraei
PII: S0957-4174(13)00586-1
DOI: http://dx.doi.org/10.1016/j.eswa.2013.07.085
Reference: ESWA 8778
To appear in: Expert Systems with Applications
Please cite this article as: Ganji, B., Kouzani, A.Z., Khoo, S.Y., Zahraei, M.S., Adaptive Cruise Control of a HEV
Using Sliding Mode Control, Expert Systems with Applications (2013), doi: http://dx.doi.org/10.1016/j.eswa.
2013.07.085
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Adaptive Cruise Control of a HEV Using Sliding
Mode Control
Behnam Ganji, Abbas Z. Kouzani*, Sui Yang Khoo, Mojtaba Shams Zahraei
School of Engineering, Deakin University, Waurn Ponds, Victoria 3216, Australia
bganj, kouzani, sui.khoo, mshams@deakin.edu.au
Corresponding Author: Abbas Z. Kouzani
Address: School of Engineering, Deakin University, Waurn Ponds, Victoria 3216, Australia
Tel: +61 3 522 72818
Fax: +61 3 522 72167
Email: kouzani@deakin.edu.au
1
Abstract
This paper presents adaptive cruise control of a hybrid electric vehicle. First, the
mathematical model of the vehicle is formulated. Next, a classical controller is applied to the
vehicle model. Swarm optimisation is implemented for self parameter tuning of the
controller. The model is simulated and the result of the response to a variable speed is
analysed. The results reveal that the controller is not a powerful means to manage the rapid
transformation of the desire set point. Accordingly, a sliding mode controller is developed
next. The performance of this controller is compared with the classical controller.
Keywords: Sliding Mode Control, Adaptive Cruise Control, Stability and Robustness,
Hybrid Electric Vehicle, Automated Speed Control
2
1. Introduction
The control of longitudinal vehicle motion has been pursued at various levels by researchers
and car manufacturers. The developed systems involved in the control of the longitudinal
vehicle motion include cruise control, anti-lock brake system (ABS), and toehold control
systems. In addition, radar-based collision avoidance and adaptive cruise control (ACC)
systems can be listed. This classification is referred to as "driver support systems" in the
literature (Takae et al., 2005). The ACC is a speed control system used in recent models of
vehicles. It is an extension of the conventional cruise control system, and allows the host
vehicle (the ACC vehicle) follows another vehicle with an appropriate clearance or a
predefined time gap (see Fig. 1). The time gap is the ratio of the distance between the ACC
vehicle and the front vehicle to the speed of the ACC vehicle.
Clearance
Time gap = Clearance/Vehicle speed

ACC Vehicle Vehicle Ahead
Fig. 1. ACC vehicle time gap to maintain clear from the front vehicle
The control strategy in the ACC is such that when the vehicle faces a slower moving vehicle
or a vehicle that is changing lane, the controller automatically controls the throttle or the
brake to slow down the ACC Vehicle. The objective is to maintain a predetermined distance
between the two vehicles.
3
Many components of the ACC systems are similar to those of the conventional cruise control
systems. The conventional cruise control systems adjust the vehicle’s speed so that it follows
the driver's command and maintains the speed. The desired speed and the feedback from a
speed sensor are compared, and the vehicle speed is adjusted by the throttle angle. The
control module is integrated with anti-lock brake, stability and traction control, and engine
and transmission module control systems. A number of studies relating this research area
have been carried out and published, of which the most noticeable studies are reviewed in the
following paragraphs.
Takae et al. (2009) proposed a distance control assist system together with an ACC to
maintain the constant distance between a host vehicle and a lead vehicle. This means giving a
push-back force to the accelerator pedal in order to help the driver to release the accelerator
and push the brake. The test results revealed that the system can reduce the workload on the
driver when following the forward vehicle especially in heavy traffic.
Moon et al. (2009) considered an ACC with a collision avoidance approach. They considered
various aspects such as driver comfort, safe-driving and avoiding rear-end collision. Van
Arem et al. (2006) introduced a cooperative ACC. In addition to measuring the distance
between the host vehicle and the front vehicle, they considered that the two vehicles can
exchange information via a wireless system. Their study focused on the impact of this
information on traffic flow in order to enhance the driving comfort on highways by helping
the driver to frequently regulate the speed and match it with the speed of the preceding
vehicle while also sustaining constant time gap. A cooperative ACC that communicates with
traffic signals was proposed by Malakorn and Byungkyu (2010). They showed that the
cooperative ACC in collaboration with intelligent traffic signals can improve the quality of
driving from the standpoint of mobility and environmental impacts. However, using a large
variety of data in the system increased its complexity and reduced its practical appeal.
4
Even though ACC is a valuable innovation in vehicle control systems, its controllability and
stability have not been considered as a major feature in the existing studies. The key
component in cruise control is controller. The traditional methods of linear control such as
using the PID controller cannot provide a reliable control particularly in the ACC vehicle that
has variable set-points.
Although fine-tuned PID controllers can provide a practical solution to many applications, in
presence of disturbance or variable set-points, they cannot be reliable and especially with the
large difference between the desired speed and the actual speed value, the system output
produces high overshoot and sometimes volatility. Consequently, this paper presents the
development of a sliding mode controller for the nonlinear model of the vehicle to enhance
the performance of the ACC system.
2. Adaptive Cruise Control (ACC)
In addition to the interface switches in the conventional cruise control system, which are used
for cruise control settings, there are two other inputs, an input for setting the time gap
between the ACC vehicle and the front vehicle, and another input for a radar sensor that
measures the distance from the front vehicle. A driver engages the ACC, based on the road
situation, and the brake parameter, and the safe clear distance from the front vehicle can be
calculated. A schematic layout of the ACC, its accessories and the communications are
shown in Fig. 2.
3. Vehicle Dynamics
Vehicle dynamics can be represented by the longitudinal motion of the vehicle. Newton's
second law is applied to show the moving vehicle. The longitudinal forces acting on the
vehicle are expressed as acceleration, rolling, gravitational, and drag. The forces are shown in
Fig. 3.
5
Fig. 2. Physical layout of the ACC vehicle
Fig. 3. Longitudinal forces acting on a vehicle
6
The longitudinal forces on a vehicle are as follows:
(1)
2
3
4
In the above equations, Faccel is the acceleration force, Froll is the difference between the
forward friction forces and the backward rolling resistances on the tyres with Crr being the
rolling coefficient, Fgrv represents the component of the gravitational force in the x-direction
which is resistive when traveling uphill and tractive when moving downhill, CD is the drag
coefficient, and vw is the wind velocity.
7
Therefore, considering the vehicle as a moving solid body, the net traction force (Fd) is equal
to the sum of the above longitudinal forces, i.e.:
5
The equations of the longitudinal forces describe the dynamics of the vehicle, and the net
traction force that is given by Eq. (5) pushes the vehicle forward.
For a given vehicle and a road, the speed and torque at the output of mechanical coupling can
be calculated from the forces acting on the vehicle as revealed in Eq. (5).
Simplifying a power-train, considering the constant efficiencies at final drive (ηf),
transmission, torque converter (η) and neglecting the slip in wheels (i.e. driving is such that
slip can be considered zero), the torque at the output of mechanical coupling can be
calculated from the following equation:
6
And the rotational speed at the output of mechanical coupling is given by:
8
7
where gr is the gear ratio which is a function of speed, fr is final drive ratio and rw is the wheel
effective radius.
4. Transfer Function and State-space Model of Cruise Control
The cruise control in a vehicle can be considered as a common feedback system in the
process control class. It attempts to maintain a constant speed in the presence of the
disturbances such as wind and slope of the road. Different from a conventional cruise
controller in which the speed set-point is a constant parameter, in an adaptive cruise
controller the speed can be a variable factor. This variation of the speed is an additional
burden to the cruise controller system from the standpoint of reliability. In practice, a cruise
control system with a PID controller cannot be robust against the swift speed changes, and
usually falls to instability. This is the main reason for switching to the manual mode in the
ACC vehicle due to the rapid change in the speed. Before evaluating the ACC with the PID
controller and the sliding mode controller, the structure of the model is explained.
A conceptual presentation of a forward model of the cruise control is shown in Fig. 4. In the
forward model, the desired speed starts from accelerator or brake, and then the control system
where different components act such that the desired speed is transferred to the wheels. In
contrast, in a backward model, the demand power starts from the wheels and then propagates
toward the engine in conventional vehicle, and the engine and the electric motor in hybrid
electric vehicles.
9
Fig. 4. Conceptual model of the vehicle and the cruise control
It can be seen from Fig. 4 that the command for acceleration and deceleration is initiated
from the accelerator, brake or ACC. This command is interpreted as the control signal, U in
the block diagram. The control signal determines the required traction force Fd provided by
the electric motor or/and the internal combustion engine in the vehicle. Removing the
resistive forces such as gravitational, aerodynamic and friction from the traction force, the
momentum force can be obtained. Accordingly, the speed can be calculated and sent back to
the ACC as shown in the figure.
In order to obtain the transfer function of the drive force system, the actuator and the force
sources are considered to be a first order system and delay. In other words, the actuator and
the vehicle propulsion system can be modelled as a time delay in cascade with a first order
and a lag (Sheikholeslam & Desoer, 1989) as follow:
8
10
From (Sheikholeslam & Desoer, 1989), C1=743, T=1s, τ=0.2s. For the modelled vehicle, the
equivalent mass, Meq≈1500 kg, g=9.8 m/s2 and the total coefficient of drag force is 0.72
N/(m/s)2.
In order to simplify the model and to extract a straightforward mathematical model, the initial
conditions and fraction are considered to be equal to zero and the road is considered to be as a
level road. The velocity and traction force are determined as the states of the problem.
Accordingly, the sate equations are written as follow:
9
10
11
11
From the presented state equations, the only nonlinear term is speed in Equation (9). With the
perturbation around the operation points via differentiating the state equations, the new state
equations can be presented as follows:
12
13
14
In these equations, δv is the output, and physically means the amount of perturbation around
the speed operation point. Also, δFd is the amount of perturbation around the demand force
operation point, and v is the desired speed value. If the linear state equations solved for the
ratio of ∆V(s)/∆U(s) in the frequency domain, the transfer function can be written in the
following format:
12
15
If the time delay expanded as the power series expansion, the time delay transfer function can
be approximated as:
16
Substituting Eq. (16) in Eq. (15), the vehicle transfer function can be obtained as the
following third order system:
17
Then, for the modelled vehicle after substituting the values in the transfer function and with
the desired speed of 20 m/s, the system will be presented as follows:
13
18
As described in the first paragraph of this Section, the cruise control acts to maintain the
output speed V of the vehicle at a desired speed in the presence of disturbances. The
Proportional-Integral-Derivative (PID) controller has been used successfully for regulating
different processes for many years, and nowadays for the cruise control applications. In order
to have a fair comparison between the results from the application of the PID controller and
the sliding mode controller, the gain values Kp, Ki and Kd need to be tuned to provide enough
robustness during the variable set-point and other disturbances. There are different methods
of tuning which can be classified into online methods and methods relying on a model of the
vehicle. Some traditional approaches with the controller in loop use an optimisation
algorithm such as the steepest descent or Newton’s method (Hjalmarsson & Birkeland, 1998;
Johnson & Moradi, 2005) to minimize some costs, for example, the error between the input
and the output. In these methods, usually the calculations are time intense; for instance, the
gradient method requires the calculation of Hessian which is computationally expensive.
To avoid the burden of the calculation for the parameter tuning and in order to obtain fairly
optimum PID gain values, a heuristic approach of optimisation (Yoon and Shin, 1997;
Aggarwal et al., 2006) is employed for the transfer function in Eq. (18).
5. Controller Tuning Using Hybrid Particle Swarm Optimisation
14
Particle swarm optimisation (PSO) is implemented to tune the PID parameters as shown in
Fig. 5. In the framework shown in the figure, H(s) is the obtained transfer function in Eq.
(18), V(t) is the actual speed and Kp, Ki and Kd are the PID controller gain values.
Fig. 5. Framework of the PSO tuning
The PSO has been proven to be very effective for many optimisation problems. It was first
developed by Eberhart & Kennedy (1995), then improved by introducing a linearly varying
inertia weight to the original PSO (Shi & Eberhart, 1998; Shi & Eberhart, 1999) which
increased the overall performance of the PSO. The PSO commences with a swarm of
particles. The position of the particles is defined as a potential solution to a problem in a D-
dimensional space. Each particle has two state vectors, the velocity vector
, and the position vector . The velocity of the
particle is the added value to the position of the particle to find its position in the next
generation. These particles are randomly initialized and freely then fly across the multi-
dimensional search space. During the flight, the particles evaluate their positions based on an
iterative schedule and a fitness function. Each particle maintains a memory of its previous
best position, , and particles in a local neighbourhood share memories of
15
their “best” positions, designated as , then those memories are utilized
to update the velocity and position of every particle, as follows:
19
20
Here: and is the size of the swarm. t=1, 2,…, n determines the iteration number.
, are acceleration coefficients. are random numbers uniformly distributed in [0 1].
Velocity is restricted to a maximum value
In the proposed approach, are considered as particles. In other words, the
optimisation of the PID parameters can be considered as the evolution of the particles in a 3-
dimentional space. The possible particles are shown in the following matrix:
21
16
The other necessary consideration is the fitness function. An appropriate cost function can be
chosen based on the required criteria. As in the ACC, the demand traces the desired speed;
the minimum error and settling time are considered as the criteria of the fitness function:
21
where e(t) is the system error, the difference between the desired and the actual speed, ts is
the settling time, w1 and w2 are the coefficients to smooth the response.
Each variable represents a tune parameter at each instance of simulation, and then the cost of
the selected particles will be found from Eq. (21).
Similar to other evolutionary algorithms, the PSO algorithm has the drawback of premature
convergence, especially in the final stage where similar particles with a loss in diversity are
created which increases the possibility of being trapped in local optima. To cope with the
problem, the Chaos sequence (Alatas et al., 2009; Song et al., 2007) is adopted to avoid local
optima. The chaotic sequence is usually produced by the following well-known logistic
mapping:
22
17
where µ is a control parameter. The variable is considered to be in a chaotic situation
when µ = 4. By making use of the characteristic of chaos of being sensitive to the initial
value, m different variables can be obtained randomly by setting m different initial values
between 0 and 1 except the fixed point (0, 0.25, 0.5, 0.75, 1).
To make sure that the search is around the neighbourhood of the optimum point, the process
of the chaotic local search can be defined as follows:
23
where is the current optimal controller parameters in Eq. (21), , mapped
to [0, 1], is the next optimal value after chaotic disturbance. is the disturbance factor
between (0, 1), which can be defined as follows:
24
where n is an integer, and in this work (Alatas et al., 2009).
The improved PSO formulas can be expressed as,
18
25
26
27
is the reinitialised particle velocity, is a threshold value and . Velocity is
restricted to a maximum value , meanwhile,
The applied hybrid PSO with the chaos algorithm can be summarized as follows:
Step (1): Initialize the positions vector X and associated velocity V of Kp, Ki and Kd
randomly, 30 particles are chosen for each parameter in the population.
19
Step (2): Evaluate the fitness function (Eq. (21)) for each particle. Compare particles' fitness
evaluation with particle's best solution . If the current value is better than , then set the
value equal to the current value.
Step (3): Compare the fitness evaluation with the population's overall previous best If the
current value is better than , then reset , to the current particle's value and position.
Step (4): Change velocities and position via using Eq. (25), (26), (27).
Step (5): Repeat Step (2)-Step (5) until a stop criterion is satisfied or a predefined number of
iterations is completed.
The PID tuning is simulated through the hybrid PSO with chaos. The result of movement of
the particles with the hybrid method converges into the circle as shown in Fig. 6. The
obtained optimum values for this process are Ki=0.08, Kp=2.14 and Kd=1.81.
Fig. 6. The convergence of swarms to optimal point
The model of the cruise controller of hybrid electric vehicle (shown in Fig. 5) with the
achieved gain values and the linear transfer function (Eq. (18)) is simulated in
20
Matlab/Simulink. The output of the radar sensor is presented as a variant speed including the
step and the ramp signals. This profile of speed is applied as the desired value in the model.
Since in practice the variation of the desired speed cannot be the same as a step function, the
speed profile is cascaded to a first order filter. The result of simulation is revealed in Fig. 7.
Fig. 7. The input and the response signal of the linear model of the cruise controller
The reference speed is shown by a red solid line and the output by the dashed line. The
response function overshoots and undershoots based on the domain of the step change which
varies from 3% to 50%. For the ramp variation, the output follows the input signal. The
response shows that the system is stable and the output reasonably follow the input.
However, the response is valid for the simplified linear model. To investigate the reliability
and the validity of the response, the nonlinear model of the cruise controller, shown in Fig. 4,
is developed and simulated. The same input for the linear model is applied as the desired
speed. In order to take into account the genuine road states such as downhills and uphills, the
real road data is presented in the model. The (x,y,z) coordinates of the Nowra-Bateman Bay
highway (Ganji et al., 2011) in Australia over a 10 km distance is considered. A sample of the
Cartesian coordinates of the data is presented in Table 1.
21
Table 1. Real Cartesian coordinates for a small section of the Nowra-Bateman Bay Highway
used in the present study
Road location DIST EAST NORTH ALT

Y (km)
D (km) X (km) Z (km)
R[0000001,0010,C1,0.009] 0.0085 9687.603 4421.904 0.0155
R[0000001,0010,C1,0.019] 0.0188 9687.599 4421.895 0.0158
R[0000001,0010,C1,0.029] 0.0291 9687.596 4421.885 0.0162
R[0000001,0010,C1,0.039] 0.0386 9687.592 4421.876 0.0164
R[0000001,0010,C1,0.049] 0.0488 9687.588 4421.867 0.0167
The simulation result, presented in Fig. 8, shows higher overshoot of 225% and fluctuation in
change points. This reveals the fact that using a PID controller in the adaptive cruise control
system is not a powerful means to manage the rapid transforms of the variable desire set
point.
Fig. 8. The input and the response signal of the non-linear model of the cruise controller
22
In other words, in the PID controller with the variable set-point, the high overshoot and
settling time or even oscillation is unavoidable. Alternatively, the sliding mode control
which is explained in the next section is a powerful methodology to address a reliable
solution for such problems.
6. Sliding Mode Control
A robust and reliable approach to control a nonlinear and uncertain system is the sliding
control methodology. Intuitively, in a control problem, it is more straightforward to describe
a system based on a 1st-order system, i.e. 1st-order differential equation. Accordingly, in the
sliding control methodology, a notational simplification is introduced, which allows nth-order
problems to be substituted by corresponding 1st-order problems. Then, for the altered
problems perfect performances can be achieved. The sliding mode control has been used to
control various systems including robot manipulator (Capisani & Ferrara, 2012), automotive
transmission (Kim & Wang, 2009; Song & Sun, 2012), under water vehicle (Bian et al.,
2011; Javadi-Moghaddam & Bagheri, 2011), electric motor and uninterrupted-power-supply
(UPS) (Corradini et al. 2012; Komurcugil, 2012). However, in an adaptive cruise control, it is
a novel approach which is considered and investigated in this study.
The dynamic of a single input system is described as follows:
x n φ x β x u 29
23
where x is scalar and the variable of interest, for instance the position of a mechanical system
T
or in the our case the speed of vehicle, and is the state vector. In
Eq. (29), the function φ(x) is usually nonlinear and not exactly known. However, it is upper
bounded by a known continuous function of x. β(x) is the control gain and similar to φ(x) is
not exactly known, but is bounded by known continuous functions of x. For example, the
inertia of a mechanical system is only known to a certain precision or friction in a model that
describes only part of the actual friction forces. The finite control variable u gets the state x to
track a specific time varying state in the presence of model
imprecision on φ(x) and β(x).
In order to achieve the tracking task by the finite control u, the initial desired state xd(0) must
satisfy the following condition:
xd 0 x 0 30
For instance in a second-order system, the position or the velocity cannot jump, therefore the
desired trajectory from time t = 0 unavoidably starts with the same position and velocity of
the vehicle. Otherwise, the tracking can only be attained after a transient.
A time-varying nonlinear second order system can be presented as follow:
31
24
32
where x(t) = [x1(t) x2(t)]T is the state vector, ∆φ and d are functions representing the system
uncertainty and external disturbances, respectively. It is assumed that there exists a strictly
positive constant ρ which is the lower bound of the absolute value of β(x, t), i.e. 0 < ρ =
inf{|β(x, t)|}. Additionally, the functions ∆φ and d are unknown and bounded.
Hence, for every pair (x, t), there exists a constant ε such that it satisfies the following
inequality:
33
The initial conditions of the system are designated as x1(0) and x2(0) where x1(0) = x1(t0) and
x2(0) = x2(t0). The system that is presented by the state Eq. (31) and Eq.(32) is supposed to
track the desired path given by a function of time x(d) (t) = [x1(d)(t) x2(d)(t)]T, where x2(d)(t) =
d(t) and x2(d) (t) is a differentiable function of time. The path tracking error vector is
defined as follows:
34
25
Therefore, the errors can be shown as e1(t) = x1(t) - x1(d) (t) and e2 (t) = x2(t) - x2(d) (t).
If the initial time considered as t = t0, the tracking error and the error and its derivative can be
stated as:
e1 t0 e0 0, e2 t0 0 35
where e0 is a non zero random real number. In practice, this condition is satisfied in many
practical applications such as position control or set point change of the second order
systems.
In order for effectively control the system presented by Eq. (31) and Eq. (32), i.e. to achieve a
control strategy that is insensitive to both external disturbance d(t) and model uncertainty
∆φ(x,t), a time varying switching line needs to be defined. The line slope should not change
during the control process, which implies that the line moves on the phase plane without
rotating, and it is shifted in the state space with a stable angle of inclination. Initially, the line
moves with a constant deceleration in the state space and then it stops at a time instant tf > t0.
In the model of the cruise control system shown in Fig. 4, the engine and/or the electric motor
are considered as a first order system with the constant time of T. Therefore, the power
demand Fd based on the control signal u can be presented as follows:
36
26
Hence, the state equations of the system from the Eq. (5) and Eq. (36) can be written as:
37
In practice, in order to be able to design a suitable sliding surface, the demand power Fd must
be a measurable and an observable parameter. In reality, from the engine torque meter and
the stated relation in Eq. (6), Fd can be computed.
Accordingly the sliding mode is defined as follows:
38
where vd and x2d are the variation of the measurable parameters. By substituting Eq. (37) in
Eq. (38), we get:
39
27
In Eq. (39), if Fd is substituted by its equivalent s2 + x2d from Eq. (38) and Eq. (39), it can be
rewritten as follows:
40
In order to find an effectual solution to the system, x2d in the fowling format is a
satisfactory selection:
41
where k1 is a selectable constant. In order to satisfy the stability of the system, the constant k1
must be positive, i.e. k1 > 0. If the relation in Eq. (41) is substituted in the Eq. (40) then:
42
If we assume that s2 is a differentiable parameter then from Eq. (38):
28
43
and by considering the system presented by Eq. (37), therefore:
44
Accordingly, the selection of the optimal control parameter leads to the following relation:
45
where k2 is a positive definite selectable constant. By substitution Eq. (45) in Eq. (44), we
conclude:
46
29
It is noticeable that the solution of the differential Eq. (46) converges to zero exponentially.
Taking into account the revealed result in Eq. (46) and from Eq. (42), it can be concluded
that:
47
Also, the solution of the differential Eq. (47) converges to zero exponentially, which proves
the stability of the sliding motion and the convergence of the error to zero.
The nonlinear model of the cruise control system with using the real data of the road is
modelled in MATLAB/SIMULINK. The described method of control is applied instead of
the PID controller. In the simulation, the same desired speed input for the PID controller is
used. The output speed and reference speed are presented in Fig. 9. As can be seen from the
figure, the output response reveals the 0% overshoot.
7. Conclusion
Adaptive cruise control systems are used in the new generation of vehicles. However, with
the variable set-point of ACC, the PID controller is not a reliable controller in plant. The
results of the simulation in this study showed the high overshoot and settling time or even
oscillation is unavoidable. Alternatively, the sliding mode control is a powerful methodology
to address a consistent solution for such problems.
The task of the sliding mode control method is to design a controller, in order for the sliding
variable "s" to converge to zero. In the sliding mode control, the desired system dynamics is
30
called "sliding mode or sliding mode surface", and the controller designed to conduct the
sliding variable "s" to converge to zero is called "sliding mode controller". Accordingly, the
sliding variables based on the measurable and differentiable parameters of vehicle are defined
and it has been proved that they converge to zero.
The cruise control of hybrid electric vehicle with using the real data of the road was
modelled. The simulation result shows the stable output speed with 0% overshoot in
comparison the oscillation and overshoot of 225% for the PID controller in plant with the
same situation.
Fig. 9. The input and the response signal of the non-linear model of the cruise controller
based on the sliding mode control
References
Aggarwal, V., Mao, M. & O’Reilly, U.-M. (2006). A self-tuning analog proportional-
integral-derivative (PID) controller. In Proceedings of the First NASA/ESA Conference
on daptive Hardware and Systems (pp. 12-19).
Alatas, B., Akin, E., & Bedri Ozer, A. (2009). Chaos embedded particle swarm optimization
algorithms. Chaos, Solitons and Fractals, 40(4), 1715–1734.
31
Bian, X. Q., Cheng, X. Q., Jia, H. M., Yan, Z. P., & Zhang, L. J. (2011). A bottom-following
controller for underactuated AUV based on iterative sliding and increment feedback.
Kongzhi yu Juece/Control and Decision, 26, 289-292.
Capisani, L. M., & Ferrara, A. (2012). Trajectory planning and second-order sliding mode
motion/interaction control for robot manipulators in unknown environments. IEEE
Transactions on Industrial Electronics, 59(8), 3189-3198.
Corradini, M. L., Ippoliti, G., Longhi, S., & Orlando, G. (2012). A quasi-sliding mode
approach for robust control and speed estimation of PM synchronous motors. IEEE
Transactions on Industrial Electronics, 59 (2), 1096-1104.
Eberhart, R. C., & Kennedy, J. (1995). A new optimizer using particle swarm theory," In
Proceedings of the Sixth International Symposium on Micro Machine and Human
Science (pp. 39-43), Nagoya, Japan.
Ganji, B., Kouzani, A. Z., & Hesami, M. A. (2011). Backward modeling and look-ahead
fuzzy energy management controller for a parallel hybrid electric vehicle. Control and
Intelligent Systems, vol. 39 (3), 179-189.
Hjalmarsson, H., & Birkeland, T. (1998). Iterative feedback tuning of linear time-invariant
mimo systems. In Proceedings of the 37th IEEE Conference on Decision and Control
(pp. 3893-3898), Tampa, Florida, USA.
Javadi-Moghaddam, J., & Bagheri, A. (2010). An adaptive neuro-fuzzy sliding mode based
genetic algorithm control system for under water remotely operated vehicle. Expert
Systems with Applications, 37 (1), 647-660.
Johnson, M. A., & Moradi, M. H. (2005). PID control: New identification and design
methods. Science+Business Media, Springer.
32
Kim, G. W., & Wang, K. W. (2009). Switching sliding mode force tracking control of
piezoelectric-hydraulic pump-based friction element actuation systems for automotive
transmissions. Smart Materials and Structures, 18, 1-15.
Komurcugil, H. (2012). Rotating-sliding-line-based sliding-mode control for single-phase
UPS inverters. IEEE Transaction on Industrial Electronics, 59, 3719-3726, OCTOBER
2012 2012.
Malakorn, K. J. & Park, B. (2010). Assessment of mobility, energy, and environment
impacts of intellidrive-based cooperative adaptive cruise control and intelligent traffic
signal control. In Proceedings of the 2010 IEEE International Symposium on Sustainable
Systems and Technology (ISSST), Charlottesville, VA, USA.
Moon, S., Moon, I., & Yi, K. (2009). Design, tuning, and evaluation of a full-range adaptive
cruise control system with collision avoidance. Control Engineering Practice, 17(5),
442-455.
Sheikholeslam, S., & Desoer, C. (1989). Longitudinal control of a platoon of vehicles. PATH
Research Report (UCB-ITS-PRR-89-6), Institute of Transportation Studies University of
California at Berkeley.
Shi, Y., & Eberhart, R. (1999). Empirical study of particle swarm optimization. In
Proceedings of the IEEE Congress on Evolutionary Computation (pp. 1945-1950).
Shi, Y., & Eberhart, R. (1998). A modified particle swarm optimizer. In Proceedings of the
IEEE International Conference on Evolutionary Computation (pp. 69-73).
Song, X. & Sun, Z. (2012). Pressure-based clutch control for automotive transmissions using
a sliding-mode controller. IEEE/ASME Transactions on Mechatronics, 17(3), 534-546.
Song, Y., Chen, Z., & Yuan, Z. (2007). New chaotic PSO-based neural network predictive
control for nonlinear process. IEEE Transactions on Neural Networks, vol. 18(2), 595-
601.
33
Takae, Y., Iwai, M., Kubota, M., & Watanabe, T. (2005). A study of drivers' trust in a low-
speed following system. SAE Technical Paper (2005-01-0430).
Takae, Y., Seto, Y., Yamamura, T., Sugano, T., Kobayashi, M., & Sato, K. (2009).
Development and evaluation of a distance control assist system with an active accelerator
pedal. SAE International Journal of Passenger Cars - Electronic and Electrical Systems,
2(1), 46-55.
Van Arem, B., Van Driel, C. J. G., & Visser, R. (2006). The impact of cooperative adaptive
cruise control on traffic-flow characteristics. IEEE Transactions on Intelligent
Transportation Systems, 7(4), 429-436.
Yoon, M.-H. & Shin, C.-H. (1997). Design of online auto-tuning PID controller for power
plant process control. In Proceedings of the 36th SICE Annual Conference (pp. 1221-
1224).
34
Highlights
Adaptive cruise control of a hybrid electric vehicle is presented.

A mathematical model of the vehicle is formulated.
A classical controller optimised by swarm optimisation is applied to the vehicle model.
A sliding mode controller is developed is applied to the vehicle model.
Cruise control of the vehicle by using the real road data is modelled.
35

J Eswa 2013 07 085

Uploaded by

Copyright:

Available Formats

You might also like

J Eswa 2013 07 085

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

J Eswa 2013 07 085

Uploaded by

Copyright:

Available Formats

Accepted Manuscript

Adaptive Cruise Control of a HEV Using Sliding Mode Control

To appear in: Expert Systems with Applications

Adaptive Cruise Control of a HEV Using Sliding

School of Engineering, Deakin University, Waurn Ponds, Victoria 3216, Australia

bganj, kouzani, sui.khoo, mshams@deakin.edu.au

Corresponding Author: Abbas Z. Kouzani

Tel: +61 3 522 72818

Fax: +61 3 522 72167

Hybrid Electric Vehicle, Automated Speed Control

Time gap = Clearance/Vehicle speed

between the two vehicles.

driver when following the forward vehicle especially in heavy traffic.

has variable set-points.

the performance of the ACC system.

2. Adaptive Cruise Control (ACC)

Fig. 2. Physical layout of the ACC vehicle

Fig. 3. Longitudinal forces acting on a vehicle

The longitudinal forces on a vehicle are as follows:

coefficient, and vw is the wind velocity.

to the sum of the above longitudinal forces, i.e.:

Simplifying a power-train, considering the constant efficiencies at final drive (ηf),

calculated from the following equation:

4. Transfer Function and State-space Model of Cruise Control

controller in which the speed set-point is a constant parameter, in an adaptive cruise

Fig. 4. Conceptual model of the vehicle and the cruise control

the ACC as shown in the figure.

and a lag (Sheikholeslam & Desoer, 1989) as follow:

Accordingly, the sate equations are written as follow:

equations can be presented as follows:

following third order system:

the desired speed of 20 m/s, the system will be presented as follows:

Proportional-Integral-Derivative (PID) controller has been used successfully for regulating

gradient method requires the calculation of Hessian which is computationally expensive.

5. Controller Tuning Using Hybrid Particle Swarm Optimisation

Fig. 5. Framework of the PSO tuning

particles. The position of the particles is defined as a potential solution to a problem in a D-

, and the position vector . The velocity of the

best position, , and particles in a local neighbourhood share memories of

their “best” positions, designated as , then those memories are utilized

to update the velocity and position of every particle, as follows:

, are acceleration coefficients. are random numbers uniformly distributed in [0 1].

Velocity is restricted to a maximum value

In the proposed approach, are considered as particles. In other words, the

the selected particles will be found from Eq. (21).

where µ is a control parameter. The variable is considered to be in a chaotic situation

of the chaotic local search can be defined as follows:

where is the current optimal controller parameters in Eq. (21), , mapped

between (0, 1), which can be defined as follows:

where n is an integer, and in this work (Alatas et al., 2009).

The improved PSO formulas can be expressed as,

is the reinitialised particle velocity, is a threshold value and . Velocity is

restricted to a maximum value , meanwhile,

randomly, 30 particles are chosen for each parameter in the population.

value equal to the current value.

Fig. 6. The convergence of swarms to optimal point

Cartesian coordinates of the data is presented in Table 1.