ABSTRACT:

In this paper a neuro-fuzzy-sliding mode control (NFSMC) with extended state observer (ESO) technique is
designed to ensure the traction of an electric vehicle with two separate permanent magnet synchronous
motor (PMSM). Each PMSM systems ( source-converter-motor) are coupled to an electronic differential
(ED) in order to compensate the tendencies of direction of the vehicle and maintain a steady speed by
adjusting the difference in speed of each motor-wheel according to the direction in the case of a turn. To
ensure the control and performance of the vehicle a hybrid control scheme employs two types of controllers
, neuro-fuzzy sliding mode control on the direct current loop and disturbance rejection control laws (ESO)
on the Speed loop ,and quadrature current loop of the PMSM by taking into account the dynamic of the
vehicle. Simulation under Matlab/Simulink to evaluate the efficiency and robustness of the proposed control
method and ED on the closed loop system.
INTRODUCTION
The The number of vehicles in circulation continues to increase, and even though new engines are
becoming less polluting, pollution problems are increasingly critical, especially for the greenhouse gases
responsible for global warming of the planet.
Currently, the most effective way to combat these emissions is the implementation of increasingly
restrictive standards for new vehicles. Despite the technological advances, the limits imposed by the
standards will be too restrictive to be respected by the thermal engines, in more or less long term. These
restrictive standards are forcing automakers to introduce new technologies that will make vehicles less
polluting. Electric vehicles have emerged, by definition an electric vehicle is a vehicle whose propulsion is
provided by an engine operating exclusively with electrical energy. The electric car is advanced by all
players in the automotive field as one of the cleanest and greenest transportation solutions. It could indeed
be an alternative to this alarming pollution, especially since the road transport sector emits more pollutants
into the atmosphere than the industry sector. However, despite the extensive research on the power train and
batteries, the electric vehicle is still in the experimental stage and is subject to modification or improvement.
Like anything else, cars adopting this type of consumption have advantages as well as disadvantages. The
greatest strength of this class of cars lies in the electric nature of their engine. At the same time for technical
reasons because the electric motors are more flexible than the thermal engines (high torque at low speed).
Much research has been done on electric vehicles in particular the control of its traction chain, which in
turn translates into the control of these engines wheels. The performance-dependent operation of the engine-
wheel assembly system is characterized by smooth rotation throughout the engine speed range, total torque
control at zero speed, and rapid acceleration and deceleration.
On a classic vehicle the drive wheels are connected by half-shafts controlled by a bevel gear. On each of
the wheels there is a torque equal to half of that provided by the return angle. Because of the rigid
connection formed with the drive shaft, the two drive wheels rotate at the same angular speed, which has no
particular disadvantage in walking in a straight line [5] [6]. The path of the inner wheel is reduced compared
to that of the outer wheel; these two distances to be performed in the same time interval, it is necessary that
the angular velocity of the two wheels is different. If the wheels are not driving, no problem, otherwise it is
necessary to interpose a differential mechanism allowing the wheels to rotate at different speeds. there is a
system in conventional vehicles called a mechanical differential that ensures the difference in speeds while
ensuring torque for the two driving wheels [7].
In this work we will replace the mechanical differential with an electronic differential in order to reduce
the drive line components, thus improving the overall reliability and efficiency. This option will also reduce
the drive line weight since mechanical differential and gear reduction are not used. However, one of the
main issues in the design of these EVs (without mechanical differential) is to ensure vehicle stability in
particular while cornering or under slippery road conditions [8] . This calls for a specific traction control
system . The control of the traction effort transmitted by each wheel is at the base of the command strategies
aiming to improve the stability of the vehicle. Each wheel is controlled independently by using an electric
motorization.
in this context a Permanent magnet synchronous motor has been adopted as the electric vehicle (EV)
propulsion. Permanent magnet synchronous motors (PMSMs) have extensive industrial applications
including electric vehicles, robotics, and wind energy conversion systems[1],[2],[3].
This can be attributed to desirable performance characteristics including high power density, high torque
to weight ratio, high reliability and high efficiency [1]. Many control techniques for PMSMs have been
developed in literature studies. While the PI controller is still a popular choice due to its simplicity and ease
of implementation, it does not take disturbances and uncertainties into account, leading to poor performance
[4].
Due to the development of digital signal processor technology, power electronics, control theories, more
and more advanced control methods are introduced to the PMSM control problem, e.g., adaptive control [9],
robust control [10], sliding mode control [11], neural network control [12], and fuzzy control [13], etc.
These methods can improve the performance of the electric vehicle based on the PMSM from different
aspects.
Sliding mode control is a popular control due to its ability to reject internal parameter variations and
external disturbances. SMC has found extensive applications in the areas of power electronics and electric
machines [14]. Consequently, SMC has been widely and successfully applied into the position and velocity
control of PMSM. However, SMC has its own disadvantage, i.e., chattering phenomenon, which originated
from the interaction between parasitic dynamics and high frequency switching control [15]. In order to avoid
the phenomenon, a control method was proposed to reduce or completely eliminate the chattering, such as a
neuro-fuzzy-sliding mode control (RNFSMC) with extended state observer (ESO) to ensure the traction of
the electric vehicle. The hybrid neuro-fuzzy sliding mode control with extended state observer [16] is one of
the most common controls associated several techniques: fuzzy control [17] and sliding mode control [18],
neural network , and disturbances rejection laws to exploit the advantages of all the techniques to limit the
disadvantages of regulation by conventional adjustment algorithms and to improve the performance of the
controller system (stability, accuracy, speed, robustness, etc...).
The FIS forms are a useful computing framework based on the concepts of fuzzy set theory, fuzzy if–then
rules and fuzzy reasoning. The ANFIS [19],[20] is a FIS implemented in the framework of an adaptive fuzzy
neural network. It combines the explicit knowledge representation of a FIS with the learning power of
ANNs. Usually, the transformation of human knowledge into a fuzzy system (in the form of rules and
membership functions) does not give the target response accurately. So, the parameters of the FIS should be
determined optimally [21].
Extended state observer (ESO) [22],[23] regards the lumped disturbances of the system, which consists of
internal dynamics and external disturbances, as a new state of the system. This observer can estimate both
the states and the lumped disturbances. Based on the ESO, a feed forward compensation for the disturbances
can also be employed in the control design.
In this work, combining the extended state observer (ESO) with the adaptive neuro-fuzzy inference
system technique together, a robust control for the whole system is developed to improve the adaptation of
the closed-loop system.
The reminder of this paper is organized as follows: section 2 reviews the principle components of the
traction system and their model equations (vehicle dynamic, Electronic differential , electric motor model).
Section 3 shows the theory of control ( sliding mode control and Neuro-fuzzy-sliding mode control with
disturbance rejection law model ) to remedy the chattering phenomena. Section 4 a simulation results verify
the validity of the proposed technique of control. Finally we terminate by a the conclusion is in section 5
ELECTRIC VEHICLE TRACTION SYSTEM
This section is dedicated to the dynamics of the electric vehicle and the different components on board the
vehicle and their equations models.
Vehicle dynamics
In this work the proposed control strategy takes into account the dynamic vehicle, so the first step in
vehicle performance modeling is to write an electric force model. This is the force transmitted to the ground
through the drive wheels, and propelling the vehicle forward. This force must overcome the road load and
accelerate the vehicle [24].

The elementary equation describing the longitudinal dynamics of a vehicle in the road is in the
following form:
The speed gear ensures the transmission of the motor torque to the driving wheels. The gear is modeled
by the gear ratio i , the transmission efficiency and its inertia.
Several types of batteries can be distinguished, but for the current electric vehicles, lithium-ion, lead-acid
and nickel-cadmium batteries are frequently used .
 Modeling of The Electronic Differential
The proposed archictecture propulsion system allows one to develop an electronic differential to ensure that
the over astraight trajectory of the two wheel drives roll exactly at the same velocity; and in a curve
trajectory the difference between the two wheel velocities ensure the vehicle trajectory over the curve. See
Figure 3. Since the two rear wheels are directly driven by two separate motors, If the vehicle is turning right,
the left wheel speed is increased and the right wheel speed remains equal to the common reference speed
 r _ ref . If the vehicle is turning left, the right wheel speed is increased and the left wheel speed remains equal
to the common reference speed r _ ref [8],[30],[31]. this helps the tyres from losing traction in turns. Figure
3(b) shows the vehicle structure describing a curve, where L represents the wheelbase,  the steering
angle, d the distance between the wheels of the same axle and r _ L and r _ R the angular speeds of the left
and right wheel drives, respectively.

When the vehicle begins a curve, the driver imposes a steering angle to the wheels. The electronic
differential however acts immediately on the two motors reducing the speed of the inner wheel and increases
the speed of the outer wheel.

Design of sliding mode controller

Sliding modes is phenomenon may appear in a dynamic system governed by ordinary differential
equations with discontinuous right-hand sides. It may happen that the control as a function of the system state
switches at high frequency, this motion is called sliding mode.
SMC has its own disadvantage, i.e., chattering phenomenon, which originated from the interaction between
parasitic dynamics and high frequency switching control.
In order to avoid the phenomenon, several control methods were proposed in the literature.
Three surface control strategy
We take the general equation Proposed by J.J.Slotine to determine the sliding surface :
n : Relative degree, equal to the number of times it derives the output for the command to appear.

Sliding mode control of the electric traction system using the principle of the cascade control method, the
structure comprises a speed control loop which generates the current reference iqref which imposes the control
vqref , the control vdref is imposed by the current regulation idref
Direct axis control design
The resulting error will be corrected by a regulator operating in the sliding mode and the surface of this
control is given by:
s1  idref  id (18)

Considering the expression of the current id deducted in the equation system Eq. (6), the derivative of the
surface becomes:
A typical architecture of an ANFIS (Adaptive Neuro Fuzzy Inference System) is shown in Fig.6 , in which a
circle indicates a fixed node, whereas a square indicates an adaptive node. For simplicity, we consider two
inputs x, y and one output z. Among many FIS models, the Sugeno fuzzy model is the most widely applied
one for its high interpretability and computational efficiency, and built-in optimal and adaptive techniques
[34].
The principe of extended state observer method
Different kinds of disturbances appears in real applications of electric vehicle based on PMSM motors,
including internal disturbance of the motor, like parameters variation, back electromotive forces, torque
ripples, friction forces and unmodeled dynamics, and external disturbance, like load disturbances. These
disturbances may degrade the performance of the electric vehicle system if the controller does not have
enough ability to reject them. Since the control structure employed here is of cascade control loops for the
speed regulation problem, the control scheme includes a speed loop and two current loops. For the d-axis
current loop, Adaptif Neuro Fuzzy Sliding control algorithm is employed. For the speed loop and q-axis
current loop, two ESOs are introduced to estimate the different disturbances in each loop respectively.
ESO regards the lumped disturbances of system, which consists of internal dynamics and external
disturbances, as a new state of system. This observer is one order more than the usual state observer. It can
estimate both the states and the lumped disturbances. Based on the ESO, a feed-forward compensation for
the disturbances can also be employed in the control design.
Current controller design
It is well known that the current loops have parameters variations, e.g., variations of stator resistance,
inaccurate back-electromotive force (EMF) model. These parameter variations may primarily degrade the
control performance [36].
In order to eliminate unfavourable effects caused by these factors, an ESO-based controller is designed
for the q-axis current loop. The equation of q-axis current loop is written as follows:

It should be noted that the simulation run under a Japanese urban driving cycle 11 mode 4 that shall be
requested by the driver which shows wheel speed during steering for 0 < t < 481 sec. The 11 mode 4
Japanese cycle is being used for emissions and fuel consumption certification in Japan. It simulates both
urban and motorway cycle, including idling, accelerations, cruising and decelerations. The measurements
are performed while engine is hot, after a standard warming procedure.
During this cycle the vehicle approaches a right turn at 62 (positive steering angle) s and a left turn at 180
s (negative steering angle) .It is obvious that the electric vehicle operates satisfactorily according to the
complicated series of accelerations, decelerations, and frequent stops.
Referring to the figure 8 and figure 9, the vehicle turn right at t= 62 s, in compensation the torque of the
right wheel decreases, briefly becoming negative. Negative torque is, in essence, braking mode. This
working phase can be exploited for energy recuperation. Once the speed of the right wheel is stabilized , the
torque returns to its initial value, as it is shown in Figure 9. At t= 67 s the vehicle exits the curved section of
road; thus the driver applies an inverse steering angle to the front wheels. The electronic differential
subsequently acts to equalize the speed of the drive motors; The speed difference of the driving wheels is
shown in figure 8. At t= 180 s the vehicle approaches a left turn, it can be seen that the electronic
differential acts immediately on both electric motors by lowering the speed of the left wheel and, unlike that
on the right. This speed is illustrated in Figure 8. The variation of phase currents are shown in Figure 10.

CONCLUSION
This paper has dealt with an ED based EV. The ED system will reduce the drive line components, thus
improving the overall reliability and efficiency since mechanical differential and gear reduction are not used.
In this case, The EV traction drive system uses two separate PMSM back drive based wheels.
The proposed ED has been developed to handle the EV stability while cornering or under slippery road
conditions. For that purpose, a neuro-fuzzy-sliding mode control (NFSMC) with extended state observer
(ESO) technology is used to realize asymptotic tracking of speed trajectory and stator currents and
guarantee stability of whole the field-oriented PMSM drive system. The results obtained by simulation
show that this structure allows the design and implementation of an electronic differential and ensures
good dynamic and static performances. The paper shows that the electronic differential controls the
driving wheels speeds with high accuracy either in flat roads or curved ones. The disturbances do
not affect the performances of the driving motors.