Vehicle Yaw Rate Control For Lane Change

2020 IEEE 2nd International Conference on Electronics, Control, Optimization and Computer Science (ICECOCS) | 978-1-7281-6921-7/20/$31.00 ©2020 IEEE | DOI: 10.1109/ICECOCS50124.2020.9314541

Maneuver Using Fuzzy PID Controller And Neural

Network Controller
Abdussalam Ali Ahmed Omer.S. M. Jomah
Mechanical Engineering Engineering and Information
Department Technology Research Center
Bani Waleed University (E.I.T.R.C)
Bani Waleed, Libya Bani Waleed, Libya
[email protected] [email protected]

Abstract— Vehicle safety and control is attracting attention between the steering tire and road should be replaced by
increasingly in an attempt to improve the stability and suitable electronic actuators to motivates the easier
manoeuvrability of vehicles. Three degrees of freedom vehicle implementation of active front-tire steering control based on
dynamic model ( called planar vehicle model) is established. the feedback (2).
Based on theories of fuzzy PID control and neural network
based-Controller, controller of vehicle stability is designed by
using the method of direct yaw rate control and the two different Intelligent controllers were used in this paper such as a fuzzy
control strategies. The controllers were compared under one PID controller and a neural network controller to improve
road condition which is a lane change of manoeuvre. By control of the vehicle yaw rate and manoeuvrability (3). Fuzzy
comparing and analyzing the control effect of fuzzy PID control PID controller has the benefit of not requiring a certain
and neural network based-Controller, the result shows as mathematical model for the plant, while still being highly
follows: the two controllers improved the yaw rate to follow the robust. Also, certain fuzzy controller designs improve
reference yaw rate but, using the fuzzy PID controller gave a performance by adapting and learning features. Because of
better and closer path for the desired path of yaw rate compared these characteristics, the fuzzy PID controller has been
to using of the neural network controller. successfully implemented in the automotive engineering field
for vehicle dynamic control.
Keywords— Vehicle control, Yaw rate, Fuzzy PID control,
neural network control. The neural network controller also used in this work to
improve the yaw rate performance of the vehicles, then to be
I. INTRODUCTION compared to the fuzzy PID controller. Neural network-based
In recent years by the development of electronic control control is a method of control that specifically used to
devices, vehicle handling and passenger safety systems have overcome the need for vehicle modelling and control of any
been enhanced efficiently. Active Front and Rear Steering kind form and its related to a complex set of vehicle parameters
(4). Neural networks are broadly considered an effective tool
(AFS and ARS) and Direct Yaw moment Control (DYC) are
for system modelling and simulation, as they are appropriate to
some of the systems which are developed through the fast
model the complex systems using their ability to describe and
extension of electric and electronic technologies. In line with identify the relationships from input-output data pairs.
this progress, electronic systems have been employed
successfully for vehicle control. The most prevalent and the The main requirements for completing this paper are 1) to
main features that the vehicle handling and stability involved implement a vehicle control system using the fuzzy PID
are the tracking of yaw rate and side-slip of the vehicle body controller and neural network controller, 2) comparing the yaw
(1). rate behaviour of the vehicle by using two controllers. Matlab
The yaw rate is one of the important aspects that indicates Software was used to implement all Simulink models.
vehicle safety, and the control in yaw rate can avoid The main challenge in this work is the use of two known
dangerous, unwanted behaviour of the vehicles in extreme intelligent controllers and not the use of traditional controllers
road conditions and maneuvers. The active front steering is such as PID controller, and the important contribution to this
one of the practical approaches for the vehicle yaw stability paper is to compare the performance of the vehicle using the
and control in which the correction of the front steering angle two controllers, which has not been studied by researchers
can be added to the driver steering input when the desired yaw previously.
rate varies due to an external disturbance. To provide
maximum benefits of the steering system, steering actuated
yaw stability and control, such as mechanical linkages

978-1-7281-6921-7/20/$31.00 ©2020 IEEE.

Authorized licensed use limited to: International Institute of Information Technology Bangalore. Downloaded on September 19,2023 at 15:08:31 UTC from IEEE Xplore. Restrictions apply.
 II. VEHICLE MODELS coordinates; a and b are the displacement of the C.G of the
vehicle to of front and rear axles; Bf and Br are the distances
A. Planar vehicle model between left and right tires; vx, vy are the velocities of
The vehicle planar model is formulated from the following longitudinal and lateral orientation, wz is the actual yaw rate, δ
three equations of motion of a four-wheel vehicle with front is the front wheel steering angle ( equal to 2o), whereas m is the
steering. Figure 1 describes the sketch of the vehicle model mass of the purposed vehicle, and I denotes to the moment
with the physical parameters concerned. The positive x-axis inertia due to yaw motion.
starts at the C.G and points in the front end direction of the
vehicle, this direction is also indicated to as a longitudinal The input signal of the vehicle dynamic system is the steering
direction, while the y-axis is corresponded to as the lateral δ, this angle will set as a lane change manoeuvre with an
direction and starts from the model centerline. The assumptions amplitude of front steering angle of 0.03492 radians (2π/180)
that were taken into account in this model are that the front as obtained in figure 2. The lane-change manoeuvre describes
wheels have the same steering angle and the roll, pitch, and the vehicle’s handling performance in the case of a series of
bob motions are neglected. rapid steering and counter-steering actions.

This model is considered as a 3-DOF model because it takes

into account three motions which are the lateral motion the
longitudinal motion and the yaw motion.

Fig. 2. The steering input of vehicle with a lane-change maneuver.

The longitudinal Fx and lateral Fy tire forces in equations 1,2,

Fig. 1. Vehicle planar model. and 3 are defined using Dugoff's tire model. The Dugoff's tire
model gives the calculation of the force under the combined
The mathematical expression of vehicle motions can be generation of the longitudinal and lateral tire forces. These two
formulated as follows: forces can be calculated as follows (5):
• For the yaw motion:
(4)

(5)
(1) where α is The sideslip angle for each tire, C1 is the
longitudinal stiffness of the tire, C2 is the tire cornering
stiffness and λ can be expressed as:
• For the longitudinal motion:
(6)
cos sin
(2) And
• For the lateral motion: 2 1
cos sin 1 1
(3) μ denotes the friction coefficient at the tire and road contact, Fz
represents the force acting on the tire in the vertical direction,
whereas σx is the tire longitudinal slip ratio, this ratio can be
Where , , , , , , , are the derived in the accelerating case as follow:
components of forces for the front left tire, front right tire, rear 1 (7)
left tire, and the rear right tire along with x-axis and y-axis

Authorized licensed use limited to: International Institute of Information Technology Bangalore. Downloaded on September 19,2023 at 15:08:31 UTC from IEEE Xplore. Restrictions apply.
 The geometry of the tire and the velocity vector can be used to sideslip angle βd. The most important assumptions known in
calculate the value of the sideslip angle α at each tire, this angle this model are:
is considered as one of the parameters of equation 5. If the
velocity value ( vx and vy ) at the points of contact between the 1) the vehicle moves on a flat surface or a flat road.
tire and the road is known, it is easy to calculate the angle 2) the tire forces operate in a linear behaviour.
according to the following equations: 3) merging the front tires in one tire, and merging the two rear
tires in one tire also.
.
tan (8)
.

.
tan (9)
.

. .
tan (10) Fig.4. Vehicle reference model.
.
The equations of motion for the reference model can be
. expressed as: (6):
tan (11)
.
(13)

. . (13)
Where αij denotes the slip angle at each tire.
. (14)
B. Tire dynamic model
The rotational dynamics of the four tires are given by (Figure . (15)
3): .
(16)
(i=1,…4) (12) .
(17)
Where reff is the radius of each tire, Jw is the moment of inertia
of each tire about the rotational axis, wi represents the angular Where:
velocity, and T is the applied torque at the tire centre. variables wzd and βd denotes the desired yaw rate and sideslip
angle of the vehicle, these two variables represent the outputs
of the reference model. Caf and Car denote the stiffness
coefficients of front and rear tires, whereas a and b are the
distance between the COG of the vehicle to the front and rear
axle. The simulation result for the reference model of the
vehicle which represents the desired yaw rate with a lane-
change manoeuvre is obtained in figure 5.

Fig. 3. The rotational dynamics of each tire.

C. Vehicle reference model

The reference vehicle model shown in figure 4 which is called
the bicycle model, is a 2-DOF model, this model is considered
one of the most important parts in the field of vehicle design
and manufacturing because it considered for design and
analysis of controllers. What distinguishes this model is that it
is produced the desired yaw rate wzd of the vehicle and the Fig. 5. Path of the desired vehicle yaw rate.

Authorized licensed use limited to: International Institute of Information Technology Bangalore. Downloaded on September 19,2023 at 15:08:31 UTC from IEEE Xplore. Restrictions apply.
 controller (NNC). Regardless of the design approach, the
III. CONTROL STRATEGIES OF THE VEHICLE YAW
control system stability needs to be verified systematically.
RATE
Moreover, the problem becomes more complicated when any
The importance of yaw rate control is to enhance good vehicle parameter of the control system is uncertain(7).
performance, stability and handling in the linear and non-linear Neural Network is a well-known learning model for its ability
regions of the tire. Two types of control techniques are being to cope with the demands of the environment changing (8). In
used, including neural network-based control and fuzzy PID this work, the application of the Neural Network (NN)
control. controller is analyzed in the vehicle dynamics and will be
compared to the fuzzy PID control approach to control the
yaw rate of a vehicle. This network work with supervised
A. Fuzzy PID Control
learning where the data set is presented to train the network
A fuzzy PID control system can be divided into three parts (6): before a simulation is run to get output results.
part 1: fuzzification, part 2: fuzzy rules and inference, part 3: Neural networks in the computer world work on the
defuzzification. Figure 6 shows the structure of a vehicle mechanism of human brains of problem solving-strategy. In a
control system including a fuzzy PID controller. human brain, many linkages or connections are provided
The block diagram of the control system is implemented as in through networks of axons and synapses to the computing
figure 6, which consists of the reference model, the actual elements which called neurons. They communicate with each
model of the vehicle (planar model), and the fuzzy PID other in a chemical environment where electrical impulses are
controller. In the control system shown in figure 6, the desired generated among them to pass information (9). Figure 7 shows
vehicle yaw rate wzd and the desired body sideslip angle βd are the implementation of the NN control system for a vehicle
calculated by the reference vehicle model. The reference yaw (planar vehicle model)) through the available neural network
rate wzd is compared to the actual yaw rate wz which exiting controller block in MATLAB software.
from the actual planar model and in the same way the reference
sideslip angle βd is compared to the actual sideslip angle β
which is one of the outputs of the planar model. The fuzzy PID
controller was implemented with two inputs (the error and
derivative of error for both the yaw rate and sideslip angle).
The fuzzification part produced one output which is the direct
moment applied at the tire centre ∆Mz. The fuzzification
process uses to transform the real variables to corresponding
fuzzy variables depending on the selected member of
functions.
an estimator for sideslip angle and yaw is developed. Finally,
the fuzzy PID controller with two inputs and one output is
Fig. 7. Control system block diagram with neural network controller.
presented to improve the preference of a vehicle dynamics with
the lane-change manoeuvre.
IV. SIMULATION
In order to analyze and evaluate the control system
performance of the vehicle, some simulations are performed.
The dynamic behaviours and performance of the developed
control system using a neural network controller and fuzzy
PID controller for the planar vehicle model are performed
using Matlab/Simulink. It is assumed that the vehicle moves at
a constant speed of 75 Km/hr and the road friction coefficient
μ equals 0.4. One driving condition is performed which is the
input of steering. The input steering set as a lane change
manoeuvre that has an amplitude of two degrees (2π/180=
0.03492 radians) as obtained in figure 2. A comparison
Fig. 6. Control system block diagram with fuzzy PID controller. between the performance of the vehicle using two controllers
is illustrated in figures 8, 9, 10, and 11.
The vehicle parameters values which are used in the work and
B. Neural Network Control Systems simulations are listed below in table 1.
Neural networks (NNs) have been widely used in a broad
range of several control applications. Nowadays, there are
many approaches that are used to design a neural network

Authorized licensed use limited to: International Institute of Information Technology Bangalore. Downloaded on September 19,2023 at 15:08:31 UTC from IEEE Xplore. Restrictions apply.
 Table 1. The vehicle parameters. computer simulations are based on the following conditions:
Symbol Unit Value the initial vehicle velocity is 75 km/h and the friction
m Kg 1900 coefficient of the road is 0.4 (wet road).
I Kgm2 2760 In the simulation, the vehicle makes a lane change manoeuvre
a m 1.4 after travelling at a constant vehicle velocity of 75 km/h for
b m 1.6 one second. The front steering angle input is a sinusoidal wave
Bf and Br m 1.62 with the maximum steering angle of 2 degrees (about0.03492
reff m 0.35 rad).
J Kgm2 4.00 In fact, as is clear from the figures, there is a clear
μ --- 1.0 convergence between the performance of the vehicle in the
C1 N/rad 52000 case of using the two controllers, as the yaw rate in both cases
C2 N/rad 29500 approached the desired yaw rate and the reason for this
success is the use of two types of intelligent controllers in this
paper.
Figures 8 and 9 show the comparison of vehicle responses
with two different control systems on a wet road. It is obvious
that, compared with uncontrolled vehicle, the performances of
vehicles with control systems have been improved clearly. As
shown in Figures, the yaw rate response and sideslip angle of
a vehicle with the proposed control system using fuzzy PID
control can track the desired yaw rate and the vehicle sideslip
angle more satisfactorily than the neural network control.

VI. CONCLUSION
From the simulation results of this paper, it can be concluded
that the vehicle handling and stability are improved using both
of fuzzy PID controller and a neural network control
approach, which verify the effective use of the control
Fig. 8. Vehicle yaw rate. systems. The results of this work indicated that both the
controllers that were used gave good results as the yaw rate of
the vehicle improved due to using of a lane change of
manoeuvre steering, but simulation results show that the
proposed control system with fuzzy PID controller can
improve both of vehicle yaw rate and the sideslip angle better
than using of the neural network controller.

REFERENCES
[1] Ali Roshanbin and Mahyar Naraghi, “Adjustable Robustness Method
for Fuzzy Logic Integrated Control of Active Steer Angle and Direct
Yaw Moment,” International Journal of Control and Automation, Vol.
6, No. 4, August, 2013.
[2] S. Krishna, S. Narayanan and S. Denis Ashok,“ Fuzzy logic based yaw
stability control for active front steering of a vehicle, ” Journal of
Mechanical Science and Technology, Vol. 28, Issue 12, 2014.
[3] K. R. Buckholtz, “ Use of Fuzzy Logic in Wheel Slip Assignment Part
I: Yaw Rate Control”, SAE 2002 World Congress, (2002),Paper No.
2002- 01- 1221.
[4] Chindamo, D., Gadola, M., Benini, C., Romano, M, “Estimation of
Fig. 9. Vehicle sideslip angle . Vehicle Side-Slip Angle Using an Artificial Neural Network”, In
Proceedings of the ICMAA 2018 Conference, Singapore, 24–28
February 2018.
[5] Li Feiqiang, Wang Jun and Liu Zhaodu," On the Vehicle Stability
V. RESULTS AND DISCUSSION Control for Electric Vehicle Based on Control Allocation", IEEE
The two presented control systems are evaluated through Vehicle Power and Propulsion Conference (VPPC), September 3-5,
2008, Harbin, China.
several computer simulations using Matlab. One typical [6] Zimmermann, H. J, “Fuzzy set theory and its applications”, (Kluwer,
simulation test is presented in this paper, i.e. a lane-change Boston, Massachusetts), 1991.
maneuver. In the process of simulation, the response of a [7] Pornchai Khlaeo-om and Suwat Kuntanapreeda, “A Stability Condition
vehicle with the two control systems used are compared with for Neural Network Control of Uncertain Systems”, european
symposium on artificial neural networks, Bruges , 2005.
the response of the uncontrolled vehicle. All results of the

Authorized licensed use limited to: International Institute of Information Technology Bangalore. Downloaded on September 19,2023 at 15:08:31 UTC from IEEE Xplore. Restrictions apply.
 [8] Liu, Y., Cukic, B., & Gururajan, S, “Validating neural network-based Annual Conference of the IEEE Industrial Electronics Society, Lisbon,
online adaptive systems: A case study. Software Quality Journal”, Portugal, 2019, pp. 473-478, doi: 10.1109/IECON.2019.8927733.
15(3), 309-326. http://dx.doi.org/10.1007/s11219-007-9017-4. [12] B. Lenzo, M. Zanchetta, A. Sorniotti, P. Gruber and W. De Nijs, "Yaw
[9] Kadam, D. B., Patil, A. B., & Paradeshi, K. P, “ Neural network based Rate and Sideslip Angle Control Through Single Input Single Output
intelligent process control system”, IEEE International Conference on Direct Yaw Moment Control," in IEEE Transactions on Control
Recent Trends in Information, Telecommunication and Computing Systems Technology, doi: 10.1109/TCST.2019.2949539.
(ITC), pp. 356-358, 2010. [13] S. Wang, X. Zhao, Q. Yu, S. Zhang, P. Shi and M. Yu, "Research on
[10] Wei Huang, Pak Kin Wong, Ka In Wong, Chi Man Vong & Jing Zhao, Strategy of the Stability Control System of Dual-Motor Drive Electric
“Adaptive neural control of vehicle yaw stability with active front Vehicle," 2019 IEEE International Symposium on Circuits and
steering using an improved random projection neural network”, Systems (ISCAS), Sapporo, Japan, 2019, pp. 1-5, doi:
International Journal of Vehicle Mechanics and Mobility, 2019, DOI: 10.1109/ISCAS.2019.8702238.
10.1080/00423114.2019.1690152. [14] X. Jin, G. Yin, J. Chen and N. Chen, "Analysis of Lateral Stability
[11] K. Miyahara, H. Fujimoto, Y. Hori and V. Ivanov, "Performance Region for Lightweight Electric Vehicle Using Phase Plane
Benchmark of Yaw Rate Controllers by Active Front Steering: Approach," 2019 Chinese Control And Decision Conference (CCDC),
Comparative Analysis of Model Predictive Control, Linear Quadratic Nanchang, China, 2019, pp. 5461-5466, doi:
Integral Control and Yaw Moment Observer," IECON 2019 - 45th 10.1109/CCDC.2019.8832747.

Authorized licensed use limited to: International Institute of Information Technology Bangalore. Downloaded on September 19,2023 at 15:08:31 UTC from IEEE Xplore. Restrictions apply.