This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 1

Fully Electrified Regenerative Braking Control

for Deep Energy Recovery and Safety
Maintaining of Electric Vehicles
Guoqing Xu, Senior Member, IEEE, Kun Xu, Student Member, IEEE,

Chunhua Zheng, Member, IEEE, Xinye Zhang, and Taimoor Zahid


Abstract—Recycling the braking energy is significant to I. INTRODUCTION
improve the total energy efficiency of electric vehicles (EVs).
Moreover, the braking safety must be maintained in complex
conditions. As the actuator, the electric traction motor has much E LECTRIC vehicles (EVs) have drawn much attention from
academia, industry, and governments recently due to the
increasing concern over the serious problems worldwide, e.g.,
better features than the internal combustion engines, e.g., the
faster torque response, and the capability for energy feedback. shortages of non-renewable fossil fuels, environmental
From the perspective of energy efficiency and safety, highlighting pollution, and global warming. EVs have some great
the regenerative braking is a meaningful but challenging problem. advantages compared to conventional internal combustion
In this research, a braking system using only electric
engine (ICE) vehicles, e.g., the utilization of the renewable
motors/generators as the actuators is investigated, in which the
energy may be potentially fed back to the onboard energy storage energy sources, the high efficiency powertrains, and the
system (ESS) as much as possible. The potential energy that may advanced motion performances actuated by electric motors.
be recovered to the ESS is analyzed. In order to maintain the However, how to improve the energy efficiency and maintain
stability and to improve the performance of the regenerative the braking safety are two urgent problems to be solved [1], [2].
braking in unknown tire-road conditions, a knowledge-based For ICE vehicles, the kinetic energy during braking turns
methodology in a hierarchical control structure is proposed,
into the heat energy being dissipated via the mechanical friction,
where the maximum adhesion force and the motor reference
torque are determined online. The proposed methodology avoids which sometimes accounts for about 50% on average of all the
the complex determination of the optimum slip ratio, while effective traction energy. According to the national standard
acquiring nearly the optimum anti-skid braking performance. GB/T 18488.1-2006 of China, the stop and speed reduction
Simulation and experiment were carried out to validate the during basic city cycles occupy 30.77% and 17.44%,
effectiveness. respectively [3]. For EVs, the kinetic energy may be converted
to the electrical energy via the regenerative braking system
Index Terms—antilock braking, electric vehicle, regenerative
braking, slip control.
(RBS) in which the motor functions as a generator. And the
electrical energy may be stored in an onboard energy storage
system (ESS), which will be further used to propel the vehicle.
It has been pointed out that the RBS may increase the driving
Copyright (c) 2015 IEEE. Personal use of this material is permitted.
range as much as 8%-25% [4]. Therefore, regenerative braking
However, permission to use this material for any other purposes must be is one of the most effective methods to improve EV’s total
obtained from the IEEE by sending a request to [email protected]. energy efficiency, especially in the operation conditions with
Manuscript received May 11, 2014; revised October 13, 2014 and December
22, 2014; accepted February 04, 2015. This work was supported in part by
frequent start-stop drive patterns.
National Natural Science Foundation of China (Grant No. 51305437), and There are mainly two classes of braking systems according to
Shenzhen Science and Technology Innovation Commission (Grant No. the actuators in EVs. One is the conventional friction braking
JCYJ20130401170306801). system (see Fig. 1), in which the vehicle decelerates or stops by
Guoqing Xu is with School of Electronics and Information, Tongji
University, 4800 Cao’an Highway, Shanghai 201804, China. He is also with the mechanical friction forces exerted on wheels. Additionally,
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, the hydraulic antilock braking system (ABS) functions to
and The Chinese University of Hong Kong, Shenzhen 518055, China (e-mail: prevent wheels from locking. This approach is simple and
[email protected]).
Kun Xu is with Shenzhen Institutes of Advanced Technology, Chinese shows robustness and reliability especially for the emergency
Academy of Sciences, and The Chinese University of Hong Kong, Shenzhen braking process. However, the braking operations use the ICE
518055, China. He is also with Shenzhen College of Advanced Technology, and hydraulic actuators, which hold obvious time delay that is
University of Chinese Academy of Sciences, Shenzhen 518055, China (e-mail:
[email protected]). prone to exhibit the limit cycles. Additionally, it is essentially
Chunhua Zheng, Xinye Zhang and Taimoor Zahid are with Shenzhen an energy dissipation pattern with low energy efficiency.
Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen The other is the hybrid braking system (HBS) as shown in
518055, China.
Corresponding author: [email protected]
Fig. 2, in which the friction braking and the electric braking (i.e.,

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 2

Wheel speed signals Wheel speed signals

Battery
Master Vacuum
Cylinder Booster
SOC

Command
Vehicle
Inverter Control

Brake
Unit
ABS
Controller

Gear
On-Off On-Off On-Off
Switch On-Off Motor
Switch Switch Switch

Brake System
P

Front axel

Front axel
Brake System
Rear axel

Rear axel
Electronic
Hydraulic
Pressure
Regulating

Electronic
Hydraulic
M P
Valve
Accumulator Pressure
Regulating
Valve M
Accumulator
Fig. 1. The pure friction braking system.

Fig. 2. The hybrid braking system (HBS). The braking forces are generated
ESS from both the electric motor/generator and the electric hydraulic system by an
dc/dc

+ optimized distribution mechanism. The electric hydraulic system generates a

_
SOC controllable friction braking force, which is the basis of the advanced antilock
Vehicle braking system.
Inverter Control
Unit
perspective of the control effect, directly controlling the slip
Gear

Motor ratio is preferred due to the intrinsic relationship between the

slip ratio and the adhesion state. Conventional ABS uses a
Front axel
Rear axel

rule-based method to regulate the slip ratio. Advanced control

methodologies with high robustness and adaptability have been
studied in the literature recently, e.g., the cascaded nonlinear
Wheel speed signals
control [9], the sliding mode control [10-13], and the learning
Fig. 3. The fully electrified regenerative braking system (FE-RBS). control [14]. However, as vehicle dynamics involves nonlinear
characteristics, road uncertainty, disturbances and parameter
the regenerative braking) are both equipped. The cooperative variations, directly controlling the slip ratio meets great
control strategy between the friction braking system and the challenges for practical use.
RBS is crucial for the HBS, which has been widely investigated In this paper, a fully electrified regenerative braking system
in the literature [4-7]. Present research for HBS mainly focuses (FE-RBS) for EVs is investigated, which only uses electric
on improving the energy recovery during normal braking motors as the braking actuators to improve the energy
process through designing an advanced braking force efficiency and safety. The potential ability of energy recovery
distribution methodology considering various braking from wheels is analyzed. Then the paper proposes a novel
requirements and constraints. Generally, for the emergency braking control methodology for FE-RBS of EVs. The
braking process, it still uses the conventional ABS. The transition point is captured dynamically, and the maximum
adoption of both the controllable friction braking mechanism adhesion force in an unknown tire-road condition is determined
and the electric braking mechanism adds complexity and high at the transition point via a knowledge base system. An
cost to the system. Moreover, the time response features of the appropriate motor torque reference is further determined based
two braking mechanisms are inherently different, which affects on the proposed law, which considers both the driver’s desired
the dynamic control performance significantly. braking operation and the available maximum tire-road
In order to recover the kinetic energy during deceleration or adhesion utilization. The proposed methodology avoids the
braking as much as possible, fully electrified regenerative determination of the optimum slip ratio reference, which is very
braking system (FE-RBS) is potentially a perfect solution, e.g., difficult in practical applications. Both the braking stability and
an FE-RBS structure as shown in Fig. 3. If the ESS is developed the maximum tire-road adhesion utilization are achieved.
sophisticatedly to release the constraints in practical use (e.g., Moreover, keeping only the regenerative braking functions, the
the state of charge (SOC) and the battery power), the energy deep braking energy recovery is potentially feasible. The
may be recovered sufficiently. Additionally, the bandwidth of contribution of this research mainly includes the investigation
the electric motor is much higher than the actuators in of the potential deep energy recovery via the FE-RBS, and a
conventional ABS, which indicates that the time delay is no novel braking control methodology considering the braking
longer a serious bottleneck. Therefore, using only motors as the stability, the braking performance, the coordination with the
actuators in the braking system may lead to a very fast and driver’s breaking operation, and the deep energy recovery.
precise torque control performance, which profoundly The paper is organized as follows. Section II analyzes the
advances the vehicle motion control in terms of dynamic potential energy recovery of EVs via the FE-RBS; section III is
characteristics and the time delay feature [8]. about the vehicle dynamics model and the tire-road contact
As to maintain the braking safety on slippery road or in model; section IV proposes a novel braking control
emergency braking process, the braking forces need to be methodology of the FE-RBS; sections V and VI report the
implemented properly in various tire-road conditions. From the simulation results and the experimental results, respectively;

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 3

Vehicle speed (km/h)

60
Max torque 95 40
200 Min torque
Efficiency 20
90
0
100
Motor torque (Nm)

0 100 200 300 400 500 600 700 800 900 1000 1100
85

Energy (kWh)
0 80
0.4 Traction
Regenerative braking
0.2
75
-100 0
0 100 200 300 400 500 600 700 800 900 1000 1100
70

Battery power (kW)

-200 40
65 20
-6000 -4000 -2000 0 2000 4000 6000 0 100% regenerative braking
Motor speed (rpm) No regenerative braking
Fig. 4. Performance characteristics of a PMSM. -20
0 100 200 300 400 500 600 700 800 900 1000 1100

0.6

Battery SOC
Max torque
0.58
200 Min torque
Efficiency 0.56 100% regenerative braking
No regenerative braking
100% regenerative braking
100 0.54
Motor torque (Nm)

No regenerative braking 0 100 200 300 400 500 600 700 800 900 1000 1100
Time (s)
Fig. 5. Simulation results of an EV.
0
TABLE I
PARAMETERS IN THE SIMULATION
-100
Vehicle total mass (kg) 1130
Overall drive gear efficiency (%) 90
-200
Tire radius (m) 0.3537

-6000 -4000
-2000 0 2000 4000 6000 Aerodynamic drag coefficient 0.37
Motor speed (rpm) Vehicle frontal area (m2) 2.59
Fig. 6. Operating points of the PMSM. 3
Air density (kg/m ) 1.21
section VII discusses the FE-RBS in the distributed vehicle Rolling resistance coefficient 0.014
structure; finally Section VIII is the conclusion and the future
work.
recovery during braking are conducted here. In Fig. 5, the
II. POTENTIAL ENERGY RECOVERY FROM WHEELS regenerative braking energy is around 52% of the total traction
energy and the battery energy is saved by recovering this part of
A. Energy Recovery Analysis energy as shown in the battery SOC trajectories. The battery
A significant amount of energy is dissipated during braking SOC is increased from 0.5440 to 0.5614 via full regenerative
in ICE vehicles. In EVs, however, traction motors can function braking. TABLE I lists the vehicle parameters and the given
as generators to convert the kinetic energy of vehicles during data used in this simulation. The same motor shown in Fig. 4
braking to the electrical energy that may be then stored into the and a 28 Ah battery are used here. Fig. 6 shows the operating
ESS. Fig. 4 shows the performance characteristics of a points of the motor in the two cases.
permanent magnet synchronous motor (PMSM) used in the The improvement of the total energy efficiency may lead to a
analysis of this section. PMSM has good features of high significant improvement of the system economy. Specifically,
efficiency, high control accuracy, high torque density, and low the cost and the weight of the battery pack may be reduced from
noise, being particularly suitable for driving EVs. In Fig. 4, the following analysis. Considering an EV with a total energy
quadrant I is for the traction mode and quadrant IV is for the efficiency  and an onboard battery pack with a capacity Eo ,
regenerative braking mode. if there is an improvement  of the total efficiency via the
The energy recovery effect of EVs is more remarkable for FE-RBS, the battery capacity may be reduced to a smaller one
urban areas in which there are lots of start-stop drive patterns. E r while providing the same traction energy as described in (1).
Fig. 5 illustrates the simulation results of an EV on Manhattan The greater the improvement of the total efficiency is, the
driving cycle. The case where there is no energy recovery smaller the capacity of battery may be selected. Smaller
during braking and the case with an assumption of full energy capacity not only contributes to reduce the cost, but also helps

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 4

ESS power (kW) Vehicle speed (km/h)

60 S
40

20 M
0 Motor
0 5 10 15 20 25 30 Battery dc/dc Inverter
converter
50

0
Supercapacitor
-50
0 5 10 15 20 25 30 Fig. 8. A hybrid energy storage system for FE-RBS.
0.1
ESS energy (kWh)

power electronic converters/inverters with various topologies

0.05
[21-26]. Fig. 8 shows a topological structure of an HESS. The
battery provides all the operating energy, and supplements the
0 super-capacitor using a bidirectional dc/dc converter. The
0 5 10 15 20 25 30
Time (s) super-capacitor quickly feeds the energy during vehicle starting
and quickly stores the electrical energy converted from the
Fig. 7. Power and energy requirements for ESS.
kinetic energy during deceleration and braking.

TABLE II III. DYNAMICS OF VEHICLE BRAKING

COMPARISON OF THE BATTERY COST AND THE WEIGHT

Without With FE-RBS A. One-quarter Vehicle Braking Dynamics

FE-RBS   50%   50% +30% There are mainly two kinds of forces during vehicle braking
Battery capacity 20kWh 12.5kWh as shown in Fig. 9. One is the wheel braking force generated by
Battery cost the braking actuators directly applied on wheels to make them
6200 USD 3875 USD
(e.g., with 310 USD /kWh) slow down. Another is the vehicle braking force generated by
Battery weight the tire-road contact to decelerate vehicle. From the wheel
117.65kg 73kg
(e.g., with 170Wh/kg)
dynamics point of view, the vehicle braking force makes the
wheel accelerate around the axel. Assuming that the forces on
to reduce the weight. For example, Table II compares the cost the left and the right tires are the same, the dynamic equations
and the weight of the battery with respect to the price and the of the wheel and the vehicle take the forms as,
gravimetric energy density. M b  rFd  J , (2)
E 
Er  o (1)  F  mV ,  (3)
(   ) d

where M b is the braking force generated by the motor that

B. The Energy Storage System (ESS)
functions as a generator; Fd , J ,  are the tire-road friction
It is observed from Fig. 5 that the battery power trajectory
force, the wheel inertia, and the wheel rotational speed,
shows many transients which may be divided into a
respectively; m is the vehicle mass; and V is the vehicle
near-constant power signal and a frequent power signal. In
longitudinal speed.
actual practice, however, the battery power density may not
The tire-road friction force has a relationship with the wheel
satisfy the frequent power requirement. Here, it is assumed that
normal force Fz , which is described with the friction
an ESS meets the requirement. Fig. 7 illustrates the power and
energy requirements for the ESS on a start-stop route for the coefficient  as shown in (4). And the tire-road contact
same EV used in Fig. 5. condition can be depicted using the relation between  and the
The electrical energy stored in the ESS is converted to the slip ratio  .
vehicle mechanical energy during starting and acceleration  ( )  Fd / Fz (4)
process and vise versa during braking and deceleration process.
The energy provided by the ESS may be divided into two parts: B. The Tire-Road Interaction
one is the recycled part used for fast acceleration and fast The slip ratio describes the difference between the wheel
braking; the other is used to maintain a near-constant speed of speed and the vehicle speed, which is defined in (5), where R
the vehicle. The former corresponds to the ESS power capacity, is the effective wheel radius.
and the latter corresponds to the ESS energy capacity. Such   (V  R) / V , V  0 (5)
information is useful to design a hybrid ESS (HESS). Here, the
Fig. 10 shows the typical relationship between the friction
battery size should satisfy the latter requirement and the
coefficient  and the slip ratio  , which can also be referred
super-capacitor size should satisfy the former requirement. The
HESS may consist of a battery pack, a super-capacitor, and/or in the literature focused on the vehicle traction control, e.g., [6],

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 5

Wheel rotational
speed ω ω 1.0

Adhesion coefficient
V 0.8 Dry asphalt road
Wheel brake
Vehicle
torque Mb 0.6
speed

Tire Force Fd 0.4

(Vehicle brake force) Wet asphalt road
Fig. 9. Dynamics of vehicle braking. 0.2
[10], [20]. The well-known model of this relationship is the Icy road
Magic Formula tire model, which is a semi-empirical formula 0.2 0.4 0.6 0.8 1.0
and has a good correlation with the experimental data. Fig. 10 Slip ratio
indicates that the increase of the slip ratio may lead to an
increase of the adhesion coefficient until arriving at the Transition (or optimal) points
maximum value with the operation point d / d  0 . Stable region
However, further increase of  will lead to an obvious Unstable region
decrease of  , thereby, inducing an unstable motion of wheels.
Fig. 10. The   curves of various tire-road contact conditions under
Here, the operation point d / d  0 is the optimum operation driving/braking.
point corresponding to the optimum slip ratio opt and the
IV. CONTROL DESIGN
maximum friction coefficient  max . Additionally, in Fig. 10,
the region with   opt is described as the stable region, and A. Overall Control Scheme
the unstable region is the one with   opt . The optimum The objectives of the FE-RBS control, including the braking
stability, the braking performance, and the energy recovery,
operation point is also the transition point from the stable need to be considered in combination. Specifically, it is desired
region to the unstable region. that the operation point is maintained near the optimum one
The main safety concern during braking is that the wheels during braking. Consequently the maximum adhesion
may be more prone to skid or even lock up in emergency utilization in the current tire-road condition can be achieved to
braking process or on slippery road, which prolongs the meet the braking strength while keeping braking stability in
stopping distance and severely impairs the stability and emergency braking process or on slippery road, and thereby the
steerability due to the drastic loss of longitudinal and lateral tire shortest stopping distance is achieved. Additionally, the
forces. Therefore, maximizing the vehicle adhesion force while maximum electric braking power is achieved as well, which
preventing the wheel from severely skidding or locking up is potentially helps to recover the kinetic energy as much as
crucial to maintain the stability and achieve the shortest possible.
stopping distance during braking. According to (3) without As there is still difficulty in determining the optimum slip
considering the influences of the vehicle resistance, the ratio and designing the adaptive and robust controller for the
maximum deceleration and the consequent shortest stopping conventional optimum slip ratio control during braking, we are
distance may be acquired if the vehicle adhesion force keeps interested in developing a novel methodology that does not
the maximum value associated with the specific tire-road need the optimum slip ratio and complex controller design. The
condition. The maximum vehicle adhesion force Fdmax proposed methodology aims to achieve the optimal braking
corresponds to the maximum friction coefficient  max and the effect via determining the maximum friction force. The overall
optimum slip ratio opt with a constant wheel vertical load as hierarchical structure of the control scheme is shown in Fig. 11.
The lower-level subsystem is the closed-loop torque controller
shown in (6). Therefore, controlling the slip ratio at the
of the motor, which is independent of the high-level controller.
optimum value can acquire the maximum adhesion braking
The high-level controller is designed as a knowledge-based
force while maintaining the stability. However, as both Fdmax controller, which determines the appropriate reference input
and opt highly depend on the tire-road interaction that is (i.e., the motor torque reference) to the low-level subsystem.
usually unknown, nonlinear, and time-varying, there is still This structure allows no constraints on the lower-level of the
great challenge in the optimum braking control. motor control. Therefore, the commercially designed motor
Fdmax   max  Fz   (opt )  Fz (6) controller may be capable of performing more sophisticated
braking control.

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 6

Vehicle speed
Fd / t (Non-driven wheel speed)
Knowledge d Feature

Maximum Adhesion Force

Base dt Extraction
 / t Wheel speed
(Drive wheel speed)

Forward


Single gear
Controller

reduction
Inverter
Motor
Knowledge- T*
based + e
M
Driver’s
controller Motor Reference
Torque (MRT)
- Mb
Desired
T
Braking Force
Vehicle brake force
Feedback Signal
High-level Subsystem Calculation

Low-level Subsystem Vehicle Model

Fig. 11. The overall control scheme.

In Fig. 11, the driver’s desired braking force is given analyzing the physical process during braking and the general
corresponding to the current position of the brake pedal features of various    curves, some knowledge can be
measured by a position sensor. The maximum adhesion force is concluded to determine Fdmax . From Fig. 10, the slope of the
dynamically determined via an inference engine in the
   curve may be used to identify the current region of the
knowledge base. The knowledge-based controller determines
the reference input of the lower-subsystem. The low-level operation point. In actual practice, the operation point moves
subsystem is a closed-loop torque controller, which tracks the with respect to time, while the slope of    curve only
torque reference and generates the actual electric braking force indicates the relationship without the time information.
implemented on the wheels through a single gear reduction. Therefore, the following relationship is derived introducing the
The tire-road friction force generates the decelerating action to information of the time domain,
the vehicle as shown in equation (3). The feature extraction unit   / t
 . (7)
generates the feature variables that represent the system   / t
dynamic characteristics. The feature variables here refer to the Consequently, the current operation point may be identified
adhesion force and the slip ratio, where the slip ratio is whether being in the stable region or not using the differential
calculated according to the definition in (5), and the adhesion
signals  / t and  / t . If the transition point in Fig. 10 is
force is estimated via a disturbance estimator that will be
discussed at the later stage. The differential signals of the captured dynamically in the process of the wheel skidding from
feature variables with respect to time are then sent to the the stable region to the unstable region, the maximum adhesion
knowledge base. force Fdmax under the current tire-road contact is approximately
As the tire-road condition is generally unknown, the kernel determined, which is then used to generate the motor braking
of the high-level controller is to determine the maximum torque reference to avoid the unsafe motion towards more
adhesion force and the torque reference input of the low-level serious skidding or locking up in future. Here, the knowledge
subsystem, which are conducted by the knowledge base and the base in Fig. 11 uses the following IF-THEN production rule to
knowledge-based controller in Fig. 11, respectively. imitate the decision process of the braking dynamics and
determine the maximum adhesion force. It is worth noting that
B. Determination of the Maximum Adhesion Force
the knowledge base uses the input signal Fd / t instead of
For a certain kind of tire-road contact condition, there is an
available maximum adhesion force Fdmax as shown in (6).  / t which is based on a reasonable assumption that the
However, the maximum adhesion forces are different in various wheel normal force is constant or changes slowly during
braking.
tire-road conditions. Here a methodology to determine Fdmax in
an unknown tire-road condition is proposed via capturing the
transition point. As there are strong nonlinear and uncertain
characteristics in vehicle braking dynamics, it is difficult to
give an analytical solution to capture the transition point from
the stable region to the unstable region. However, based on

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 7

IF (Fd / t | t  k 1  0) AND (Fd / t ) | t  k  0) AND V J

D   0. (12)
( / t | t  k 1  0) AND ( / t | t  k  0) V mR 2
THEN (i ) The transition point occurs in the time interval (8) It shows that at the steady operation point, the adhesion force
t  [k  1, k ]; is lower than the available maximum adhesion force in the
current tire-road contact condition. Therefore the adhesion loss
(ii) The available maximum adhesion force is
is introduced using the MRT Generation Law I for the anti-skid
Fdmax  Fd | t  k . braking control. In this case, although the stability is ensured,
C. Generation of the Motor Reference Torque (MRT) both the maximum vehicle deceleration and the maximum
regenerative braking power are not achieved.
When the transition point from the stable region to the
In order to compensate the adhesion loss and to keep the
unstable region is captured and the available maximum
consequent steady operation point near the optimum point, an
adhesion force is thus determined using the IF-THEN
additional amount of the wheel braking torque is exerted by the
production rule in the knowledge base at the time t=k, the
electric motor as shown in (13),
knowledge-based controller in Fig. 11 determines the motor
reference torque (MRT) that aims to maintain the braking
Tadmax  Tdmax  TWdec , (13)
safety and to acquire the maximum braking force in various where Tadmax is the admissible maximum wheel braking torque,
unknown tire-road contact conditions. The following MRT TWdec is the additional compensation torque to provide the
generation law I gives a rule to determine the MRT.
appropriate wheel deceleration. TWdec may be further derived
MRT Generation Law I. If the transition point is captured at
the time t  k , the MRT is set as the smaller one of two values using the necessary condition in (14) for the slip ratio that will
by the high-level controller: the maximum adhesion torque not change after the implementation of the electric braking
Tdmax and the driver’s desired braking torque Tdes , i.e., torque. As we have V  0 and V  0 during braking,
equation (15) is further determined from (14). On the contrary,
T *  min{Tdmax , Tdes }  min{Fdmax  R, Fdes  R} . (9)
the condition in (15) may also derive the expression   0 .
Assuming that Tdes is greater than Tdmax and the low-level Thus the sufficient condition is also proved. Hence, we have the
controller will track the MRT ideally (i.e., M b  Fdmax  R ), we following Lemma II and the improved MRT Generation Law II
can investigate the dynamic behavior of the operation point. accordingly.
The differential equation of the transient slip ratio is derived in   0  V  V  V  V  0
  (14)
(10) from (2)-(5). Then the transient differentiation of the slip
V V
ratio at t=k is negative (i.e., (k )  0 ), which means the slip  (15)
V V
ratio will be decreased towards the stable region. Such dynamic Lemma II. If V  0, V  0 , the sufficient and necessary
behavior of the operation point with the proposed MRT
condition to maintain the slip ratio at a steady state is that the
Generation Law I is summarized in Lemma I.
parameters of V , V , V , V meet the relationship in (15).
VV  V V V F F  R  Mb
   2    2 d  R d (10) MRT Generation Law II. If the transition point is captured at
V V m JV
the time t  k , the MRT is set as the smaller one of two values
Lemma I. At the transition point from the stable region to
the unstable region, if the wheel braking force (or torque) is by the high-level controller: the admissible maximum adhesion
implemented equally to the available maximum adhesion force torque Tadmax , and the driver’s desired braking torque Tdes , i.e.,
(or torque), the slip ratio will be decreased and the operation T *  min{Tadmax, Tdes } , (16)
point will move to the stable region. where,
Such dynamic behavior of the operation point is also easy to
Tadmax  Tdmax  TWdec
understand from the physical process. At the transition point, 
T J V  . (17)
the wheel braking force is limited to the available maximum  Wdec   V R
adhesion force, which causes an approximate zero deceleration  R2 V
of the wheel dynamics immediately. As the vehicle speed Overall, the proposed control methodology is triggered and
continues to decelerate, the difference of the wheel speed and implemented with the following two critical steps: 1) the
the vehicle speed becomes smaller, thereby the operation point operation points traverse from the stable region to the unstable
slides to the stable region until arriving at a steady state. The region; 2) the control mechanism captures the transition point
and determines the motor reference torque according to the
necessary condition for the steady state is   0 . Then
MRT Generation Law II. The first step is easily met. For
considering   0 and M b  Fdmax , the steady adhesion force example, for drivers who want to apply vehicle's brake in an
is derived as, emergency situation or on a slippery road, they will exert more
F force than normal causing hard braking due to the emergent
Fd  dmax  Fdmax , (11) condition and their lack of experience. Hence the operation
1 D
where point may easily slide to the unstable region. If the transition
point is captured, the control system limits the MRT based on

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 8

Start
Input filtered signals Fd
Calculating the differential signals r s is the Laplace operator


Yes No
Is transition point? Mb  1
Yes
flag  1?
No 1/ J 
flag  1 s
Determine Fdmax as (8)
  
 noise
Fdmax (k )  Fdmax (k  1)
LPF Js LPF 
Fdmax (k )  Tmmax / R
1/r F̂d
Disturbance Observer
Determine MRT based on (16)
Fig. 13. Disturbance observer to estimate the friction force.
Return
Fig. 12. Flowchart of the control algorithm of the high-level controller.
E. Acquisition of Slip Ratio and Friction Force
the proposed MRT Generation Law II, with which the brake The slip ratio is calculated based on the definition in (5)
safety and the brake performance are both ensured. A flowchart using the measurements of the wheel speed and the vehicle
is presented as shown in Fig. 12 to clarify the control algorithm, speed. Here the vehicle speed can be obtained via measuring
wherein Tmmax is the maximum drive torque from motors. the rotational speed of the non-driven wheel (i.e., the free
spinning wheel). The friction force cannot be measured directly.
D. Stability From the wheel dynamics, the friction force can be regarded as
The braking stability requires that the operation points stay in the load disturbance to the motor driving system. A disturbance
the stable region during braking with the proposed controller, observer (DOB) of the friction force is designed, which uses the
i.e.,    c  [0, * ] . The following Lemma III states that the measurements of the driving torque and the wheel rotational
braking stability is ensured with the proposed control speed. The wheel rotational speed is obtained using a speed
methodology. measuring device, e.g. a rotary encoder. The inverse model of
the dynamic equation is consequently derived as shown in (23).
Lemma III. Considering *  [0,1] , and the set  c  [0, * ] ,
The block diagram of DOB is shown in Fig. 13, which uses a
 (0)   c . If the electric braking torque is implemented as low pass filter (LPF) and an inverse model of the wheel
described in (18) or (19), then we have  (t )   c , t  0 , i.e., dynamics.
 c is a positive invariant set. Fˆd  (M b  J ) / r (23)
M b  Tdmax (18)
V. SIMULATIONS AND ANALYSIS
M b  Tadmax  Tdmax  TWdec (19)
The proposed control methodology is validated using the
Proof: Defining the Lyapunov function V ( )  0.52 , then numerical simulation, which uses the parameters of a vehicle
we have, prototype presented in the following Section VI.
M  Fd  R V Fd
V ( )        ( R b  2 ). (20) A. Simulation on High Adhesion Road
JV V m
The simulation assumes that the vehicle brakes on a high
At the boarders of  c , i.e., {0, * } , we have, adhesion road surface. The initial wheel speed and vehicle
M  Fdmax  R V Fdmax speed are the same, i.e., 20 m/s. The driver’s desired braking
V (* )  *  ( R b  2 ) force is set as a ramp function with respect to the time to
JV V m
simulate an increasing braking demand. Fig.14 shows the
 * V F (21)
  (0  2 dmax )  0, for (18) simulation results. In Fig.14 (a)，the solid grey line represents
 V m
 the    curve of the tire-road condition, and the dotted red
0, for (19)
line indicates the trajectory of the operation point during the
V (0)  0 (22) braking process. It shows that the operation point remains in the
As V ( )  0 at the boarders of  c , we can state that  c is a stable region during braking. No wheel locking up or serious
skidding occurs. Fig.14 (b) presents the wheel speed and the
positive invariant set. Therefore, the slip ratio will be
vehicle speed. Fig.14 (c) presents the adhesion coefficient and
maintained in the stable region, and the brake stability is
the slip ratio. Fig.14 (d) is the stopping distance.
ensured.

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 9

1 20
Friction coefficient Vehicle speed
0.8 Wheel speed
15

Speed (m/s)
0.6
10
0.4

0.2 5
   curve (icy road)
Actual working point
0 0
0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 6 7
Slip ratio Time (s)
(a) (b)
1
Friction coefficient 250
Normalized Slip, Mu

0.8 Slip ratio

Stop distance (m)

200
0.6
150
0.4
100
0.2 50
0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Time (s) Time (s)
(c) (d)
Fig. 14. Simulation results on the high adhesion road. (a) The    curve and the operation point; (b) Vehicle speed vs. wheel speed; (c) Slip ratio and
friction coefficient. (d) Braking distance.

Although the stopping distance is big, the vehicle may not lose
B. Simulation for Braking on Slippery Road
the lateral force and the vehicle is in a controllable state.
A low adhesion tire-road condition (i.e., the solid grey line as Moreover, as the utilization of the maximum adhesion force
shown in Fig. 15 (a)) is used in the simulation, with the from the tire-road contact, the braking process is close to the
optimum slip ratio opt  0.15 and the maximum friction optimum.
coefficient  max  0.2 . The initial vehicle and wheel speed are C. Comparison
20m/s. The driver’s desired braking force is set as a ramp Additionally, the traction control based on the maximum
function with respect to the time to simulate an increasing transmissible torque estimation (MTTE) in [18] is also
braking demand, which is an aggressive braking behavior on conducted on the low adhesion road in the simulation for
the low-adhesion road. comparison with the proposed methodology in the paper.
In Fig.15 (a), the dotted red line represents the trajectory of The existing control approaches in the literature mainly fall
the operation point during braking. It shows there is no wheel into two classes. One is directly controlling the wheel slip ratio,
locking up or skidding. The vehicle is very safe during braking e.g., the sliding mode control [10]-[13], learning control [14].
on the low adhesion road. Fig.15 (b) shows the curves of wheel The other approach does not need to control slip ratio directly,
speed versus vehicle speed, where the turning point of the which has been studied recently as it is more practical and is
wheel speed results from the switching operation at the independent from the driving conditions, e.g., the model
transition point, i.e., the motor reference torque is set as the following control [8], and the most recent MTTE-based control
admissible maximum adhesion torque according to the MRT proposed in [18], [19]. The proposed methodology in this paper
generation law II. Fig.15 (c) shows the slip and friction falls into the second class, as it does not control the slip ratio
coefficient curve. The slip ratio is maintained near the optimum directly. Therefore, we conduct the MTTE-based control in the
one, i.e.,   opt  0.15 . And the friction coefficient is near simulation for the comparison. Fig. 15 shows the simulation
the maximum, i.e.,    max  0.2 . Fig.15 (d) shows the results, where the dotted lines in Fig. 15 (b) - Fig. 15 (d) are the
stopping distance. As the road in the simulation is very slippery, results using the MTTE-based control. It indicates that the
the stopping distance reaches more than 200 meters due to the proposed methodology in the paper has a better adhesion
small deceleration generated from the tire-road contact. utilization and a shorter stopping distance.

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 10

0.2 20
Vehicle speed (Proposed)
Wheel speed (Proposed)
Friction coefficient
0.15 15 Vehicle speed (MTTE)

Speed (m/s)
Wheel speed (MTTE)

0.1 10

0.05    curve (icy road) 5

Actual operation point (Proposed)
Actual operation point (MTTE)
0 0
0 0.2 0.4 0.6 0.8 1 0 5 10 15 20 25
Slip ratio Time (s)
(a) (b)
0.4
Friction coefficient (Proposed) F2
250
Normalized Slip, Mu

Slip ratio (Proposed)

Stop distance (m)

0.3 Friction coefficient (MTTE) F1
Slip ratio (MTTE) 200

0.2 150
100
Distance (Proposed)
0.1 F1 : (20.7 s,207.9m)
50
Distance (MTTE)
F2 : (24.2s,244.5m)
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time (s) Time (s)
(c) (d)
Fig. 15. Simulation results on the icy road. (a) Road curve (blue) and the operation points (pink); (b) Vehicle speed vs. wheel speed; (c) Braking distance; (d)
Slip ratio and friction coefficient.
TABLE III
PARAMETERS OF THE VEHICLE
As to the comparison with the approach that controls the slip Vehicle Total Mass (103 g) 1130
ratio directly, a wheel slip control of EVs based on the sliding Type Front steering, Rear driving
mode technique is proposed in [10], which aims to track the Voltage (V) 72
desired slip ratio reference. The slip ratio reference is usually Lead-Acid Battery
3
Capacity (10 Wh) 5
the optimum slip ratio, however, which is difficult to determine,
because the specific tire-road conditions are very complex and Rated Speed (rpm) 3000
PMSM Motor
uncertain in real scenarios. Herein, it considers a constant value Rated Power (103 W) 5
(i.e., *  0.2 ) instead, and does not focus on the determination Tire radius (m) 0.3537
of the optimum slip ratio reference. Therefore, the optimum
control performance (e.g., maximizing the adhesion utilization
in an unknown tire-road condition) cannot be achieved. The EV prototype
proposed methodology in this paper does not employ the slip
ratio control that avoids determining the optimum slip ratio
reference and designing the complex nonlinear controller.
Additionally, the peak adhesion force is estimated via capturing
the transition point to determine the motor reference torque, Sandy road Wet iron planes
which consequently maintains the operation point around the
optimum one that corresponds to the optimum slip ratio.
Fig. 16. A low-cost micro electric vehicle prototype.
Therefore, both the safety and the maximum adhesion
utilization are achieved. kg (50% front, 50% rear, and the wheelbase of 1.73 m), and is
powered by lead-acid batteries (72 V, 5 kWh). All the control
functionalities are implemented in low-cost micro control units
VI. EXPERIMENTAL RESULTS (MCUs). The parameters in details are shown in Table III.
A. EV Prototype B. Experimental Results
An EV prototype was built (see Fig. 16), featuring one In the experiment, the mechanical braking system is released.
low-voltage, 5.0-kW PMSM motor (72 V, 3000 rpm), coupled Only the electric braking force generated from the motor is
to the rear wheels through a single-gear transmission with the implemented on the rear wheels. The vehicle runs on a slippery
fixed ratio 10:1. Without passengers, the vehicle weighs 980 sandy road. In the beginning, the vehicle accelerates to a certain

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 11

10 10
Wheel speed Wheel speed
9 Vehicle speed 9 Vehicle speed
Slip ratio × 10 8 Slip ratio × 10
8
7 7

speed (m/s)
speed (m/s)

6 6
5 5
4 4
3 3
2 2
1 1
0 0
0 500 1000 1500 2000 2500 3000 3500 4000 0 500 1000 1500 2000 2500 3000 3500
Time (ms) Time (ms)
(a). Result of the FE-RBS test without control. (b). Result of the FE-RBS test with control.
Fig. 17. Experimental results of the FE-RBS with an EV Prototype.

speed. Then the driver begins to apply vehicle’s brake without

Wheel speed signals
and with the braking control, respectively. Fig. 17 (a) shows the
Motor ESS Motor
test result without control, where the braking force generated + SOC

dc/dc
from the motor follows the driver’s braking command without _

any control effect. The curves of the wheel speed and the slip Vehicle

Front axel
Rear axel Inverter Control Inverter
ratio show that there is a large wheel locking up. Fig. 17 (b) Unit
shows the test result with the braking control, which shows that Inverter Inverter
the braking process is smooth. No severe wheel skidding or
wheel locking up occurs. Motor Wheel speed signals Motor

VII. DISCUSSIONS Fig. 18. The structure of distributed wheel driving/braking.

A. Centralized Versus Distributed FE-RBS
In the research above, we used a single wheel model to braking mechanism to meet such standard.
investigate the FE-RBS, which may be regarded as a Fd   Fz _ i    m  g , i  front, rear (26)
centralized drive structure. However, the distributed drive i

structure has more advantages. Fig. 18 shows a distributed V    g

(27)
vehicle structure using independently driven wheels. From the There are also technical challenges for the distributed
perspective of energy efficiency, the distributed wheel electric braking mechanism due to more complex structure and
regenerative braking may recover more energy than the dynamics. In the following discussions, two critical issues, i.e.,
centralized one. Additionally, the distributed wheel the slip estimation without non-driven wheel and the load
regenerative braking may take full advantage of the adhesion movement during braking will be presented and the solutions
forces from four tire-road contacts. Thus the braking will be discussed.
performance, especially for the emergency braking process, is
improved significantly. B. Estimating Friction Force and Slip Ratio Without Vehicle
Assuming that the load distribution between the front axle Speed
and the rear is 1:1, the available braking force of the rear axle In the distributed wheel braking system, each tire-road
braking is given by (24), where Fz _ rear is the normal force of contact has the relationship as shown in the    curve of Fig.
the rear axle. The deceleration of the vehicle is consequently 10. Therefore, the maximum adhesion force may be determined
calculated by (25). using the proposed method. However, the slip ratio cannot be
Fd  Fz _ rear    (m / 2)  g (24) calculated directly due to the lack of non-driven wheel. A slip
ratio observer/estimator needs to be designed without the
V    g / 2 (25) vehicle speed.
However, for the distributed braking structure with full One solution is to estimate the vehicle speed using the wheel
wheel braking, the available total braking force is calculated by speed measurement. Some nonlinear observers may be used to
(26), and the deceleration during braking is (27). Considering obtain the vehicle speed, such as the sliding mode observer with
that commercial vehicles must pass the strict safety standard the LuGre transient tire model as presented in [15]. Another
with a required braking strength, it is easier for the distributed approach is to estimate the slip ratio directly as presented in

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 12

[16], in which a model-based slip ratio estimator is proposed, effectiveness. Results show that the proposed methodology not
and an additional vehicle longitudinal acceleration sensor is only prevents the wheels from locking but also recovers the
used to improve the convergence. kinetic energy as much as possible via the motor, which
functions as a generator during braking.
C. Distribution of the Electrical Braking Forces
From the energy savings and safety point of view, the
The load movement occurs from the rear axle to the front one FE-RBS is an ideal solution. In this paper, the technical
due to the vehicle deceleration during braking. Hence, such feasibility is investigated. Future research will focus on
factor should be considered in the braking distribution between investigating and validating the onboard hybrid ESS, which is
the rear and the front axles. free of the battery constraint and the power constraint so as to
The braking strength z is defined as, store the recovered energy as much as possible.
z  V / g . (28)
The braking forces of the front and the rear axles are then ACKNOWLEDGMENT
presented as,
The authors would like to thank the reviewers for their
 mg
 F f  uFzf   L (b  z  h g )
valuable suggestions to improve the quality of the paper.
 , (29)
mg
 Fr  uFzr   (a  z  hg ) REFERENCES
 L
[1] F. L. Mapelli, D. Tarsitano and M. Mauri, “Plug-In hybrid electric vehicle:
where, F f and Fr are the braking forces of the front and rear modeling, prototype realization, and inverter losses reduction analysis,”
axles, respectively. h g is the height of the center of gravity. IEEE Trans. Ind. Electron., vol. 57, no. 2, pp. 598–607, Feb. 2010.
[2] G. Xu, K. Xu, Q. Zhang, C. Zheng, J. Liang and Y. Zhou, “The prospect
L is the distance between the front and the rear axles. a and of electric vehicles: pure electrification, intelligence and human care,” J.
Integr. Tech., vol. 3, no. 1, pp. 1–17, Jan. 2014.
b are the distances from the center of gravity to the front and [3] The Electrical Machines and Controllers for Electric Vehicles - Part 1:
the rear axles in the horizontal direction, respectively. General Specification, National Standard of People’s Republic of China
The total braking force is, GB/T 18488.1-2006.
[4] G. Xu, W. Li, K. Xu and Z. Song, “An intelligent regenerative braking
F f  Fr  u  m  g . (30) strategy for electric vehicles,” Energies, vol. 4, no. 9, pp. 1461–1477, Apr.
Then the ideal distribution curve is derived in (31) from 2011.
[5] G. Sovran and D. Blaser, “Quantifying the potential impacts of
(29)-(30) [4]. regenerative braking on a vehicle’s tractive fuel consumption for the U.S.,
1  mg 
4h g L  mgb European, and Japanese driving schedules,” SAE Paper, 2006-01-0664,
Fr   b2  Ff    2F f  (31) 2006.
2  hg mg  hg  [6] G. Yin and X. Jin, “Cooperative control of regenerative braking and
  
antilock braking for a hybrid electric vehicle,” Math. Probl. Eng., vol.
The distribution ratio is defined as, 2013, Article ID 890427, pp. 1-9, 2013.
 i  Fi /(F f  Fr ), i  f , r . (32) [7] K. Muta, M. Yamazaki and J. Tokieda, “Development of new generation
hybrid system THS II - drastic improvement of power performance and
In order to meet the ideal distribution curve,  i needs to be fuel economy,” SAE Paper, 2004-01-0064, 2004.
[8] Y. Hori, Y. Toyoda and Y. Tsuruoka, “Traction control of electric vehicle:
determined in real time. The method as shown in (33) can be basic experimental results using the test EV ‘UOT electric march’,” IEEE
used to determine  i [2], [17], where only the deceleration of Trans. Ind. Appl., vol. 34, no. 5, pp. 1131–1138, Sep./Oct. 1998.
[9] W. Pasillas-Lepine, A. Loria and M. Gerard, “Design and experimental
vehicle is used. It does not depend on the tire-road condition. validation of a nonlinear wheel slip control algorithm,” Automatica, vol.
48, no. 8, pp. 1852–1859, 2012.
 b  V  hg / g
 f 
 F f /( F f  Fr )  (33)
[10] R. De Castro, R. E. Araujo and D. Freitas, “Wheel slip control of EVs
 L based on sliding mode technique with conditional integrators,” IEEE
 r  1  f Trans. Ind. Electron., vol. 60, no. 8, pp. 3256–3271, Aug. 2013.

[11] M. Amodeo, A. Ferrara, R. Terzaghi and C. Vecchio, “Wheel slip control
via second-order sliding-mode generation,” IEEE Trans. Intell. Transp.
VIII. CONCLUSION AND FUTURE WORKS Syst., vol. 11, no. 1, pp. 122–131, Mar. 2010.
[12] E. Kayacan, Y. Oniz and O. Kaynak, “A grey system modeling approach
In this paper, a new methodology to control the braking for sliding-mode control of antilock braking system,” IEEE Trans. Ind.
process of the fully electrified regenerative braking system Electron., vol. 56, no. 8, pp. 3244–3252, Aug. 2009.
[13] K. Xu, G. Xu, W. Li, L. Jian and Z. Song, “Anti-skid for electric vehicles
(FE-RBS) that only uses electric actuators for EVs is proposed. based on sliding mode control with novel structure,” in Proc. IEEE ICIA,
It maintains the wheel in the stable adhesion region without 2011, pp. 650-655.
directly controlling the slip ratio, so that the complex [14] C. Mi, H. Lin and Y. Zhang, “Iterative learning control of antilock
braking of electric and hybrid vehicles,” IEEE Trans. Veh. Technol., vol.
determination of the optimum slip ratio reference and the 54, no. 2, pp.486–494, 2005.
nonlinear controller are avoided. The kernel of this method is to [15] G. A. Magallan, C. H. De Angelo and G. O. Garcia, “Maximization of the
capture the transition operation point from the stable region to traction forces in a 2WD electric vehicle,” IEEE Trans. Veh. Technol., vol.
60, no. 2, pp. 369–380, Feb. 2011.
the unstable region during the braking process, and to identify [16] K. Fujii and H. Fujimoto, “Traction control based on slip ratio estimation
the maximum adhesion force in unknown tire-road contact without detecting vehicle speed for electric vehicle,” in Proc. PCC '07,
condition, which is further used to determine the motor 2007, pp. 688–693.
[17] N. Mutoh, Y. Hayano, H. Yahagi and K. Takita, “Electric braking control
reference torque (MRT) based on the proposed determination methods for electric vehicles with independently driven front and rear
law. The simulation and experimental results validate the

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TVT.2015.2410694, IEEE Transactions on Vehicular Technology
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL., NO. , MARCH 2015 13

wheels,” IEEE Trans. Ind. Electron., vol. 54, no. 2, pp. 1168–1176, Apr. Chunhua Zheng received the B.Sc.
2007.
degree in mechanical engineering in 2005
[18] D. Yin, S. Oh and Y. Hori, “A novel traction control for EV based on
maximum transmissible torque estimation,” IEEE Trans. Ind. Electron., from Harbin University of Science and
vol. 56, no. 6, pp. 2086–2094, Jun. 2009. Technology, Harbin, China, and received
[19] J. S. Hu, D. Yin, Y. Hori and F. R. Hu, “Electric vehicle traction control a the M.Sc. degree in mechanical
new MTTE methodology,” IEEE Ind. Appl. Mag., vol.18, no. 2, pp.
engineering in 2007 from Harbin Institute
23–31, Dec. 2012.
[20] B. Subudhi and S. S. Ge, “Sliding-mode-observer-based adaptive slip of Technology, Harbin, China, and
ratio control for electric and hybrid vehicles,” IEEE Trans. Intell. Transp. received Ph.D. degree in mechanical
Syst., Vol. 13, No. 4, pp. 1617–1626, Dec. 2012. engineering in 2012 from Seoul National
[21] L. Gao, R. A. Dougal, and S. Liu, “Power enhancement of an actively
University, Seoul, Korea. She is currently an Associate
controlled battery/ultracapacitor hybrid,” IEEE Trans. Power Electron.,
vol. 20, no. 1, pp. 236–243, Jan. 2005. Professor of Shenzhen Institutes of Advanced Technology,
[22] M. Ortuzar, J. Moreno, and J. Dixon, “Ultracapacitor-based auxiliary Chinese Academy of Sciences. Her working field is powertrain
energy system for an electric vehicle: implementation and evaluation,” design and power management strategies of hybrid vehicles.
IEEE Trans. Ind. Electron., vol. 54, no. 4, pp. 2147–2156, Aug. 2007.
[23] J. Cao and A. Emadi, “A new battery/ultracapacitor hybrid energy storage
system for electric, hybrid, and plug-in hybrid electric vehicles,” IEEE Xinye Zhang received the B.Sc. degree in
Trans. Power Electron., vol. 27, no. 2, pp. 122–132, Jan. 2012. electrical engineering from Shandong
[24] Z. Amjadi and S. S. Williamson, “Prototype design and controller University of Science and Technology,
implementation for a battery-ultracapacitor hybrid electric vehicle energy
Qingdao, China, in 1998. He is currently a
storage system,” IEEE Trans. Smart Grid, vol. 3, No. 1, pp. 332–340,
Mar. 2012. master student with Shenzhen College of
[25] A. C. Baisden and A. Emadi, “An ADVISOR based model of a battery Advanced Technology, University of
and an ultra-capacitor energy source for hybrid electric vehicles,” IEEE Chinese Academy of Sciences, Shenzhen,
Trans. Veh. Technol., vol. 53, no. 1, pp. 199–205, Jan. 2004.
China.
[26] S. M. Lukic, J. Cao, R. C. Bansal, F. Rodriguez and A. Emadi, “Energy
storage systems for automotive applications,” IEEE Trans. Ind. Electron., From 2009 to 2012, he was a Research Assist with the Center
vol. 55, no. 6, pp. 2258–2267, Jun. 2008. for Automotive Electronics, Shenzhen Institutes of Advanced
Technology, Chinese Academy of Sciences. His research
interests include advanced motor control, motion control and
the applications in embedded systems.
Guoqing Xu received the B.Sc., M.Sc.,
and Ph.D degrees in electrical engineering Taimoor Zahid received the B.Sc. degree
from Zhejiang University, Hangzhou, in electrical engineering from Center for
China, in 1988, 1991, and 1994 Advanced Studies in Engineering (CASE),
respectively. Pakistan, in 2011. He is currently working
In 1997, he joined Tongji University, towards his Ph.D. degree at Shenzhen
Shanghai, China, where he has been a College of Advanced Technology,
Professor in the Department of Electrical University of Chinese Academy of
Engineering since 2000. Since 2007, he has Sciences, Shenzhen, China.
been an Associate Director of the CUHK-CAS Shenzhen His research interests include state estimation and intelligent
Institute of Advanced Integration Technology, The Chinese control for electric vehicles and energy management systems.
University of Hong Kong (CUHK), where he was promoted as
the Director in 2010. His current research interests are electric
vehicle control, energy processing, and automotive electronics.

Kun Xu received the B.Sc. degree in

automation, and the M.Sc. degree in
control theory and engineering from
Shandong University, Jinan, China, in
2006 and 2009, respectively. He is
currently working towards his Ph.D.
degree at Shenzhen College of Advanced
Technology, University of Chinese
Academy of Sciences, Shenzhen, China.
From 2009 to 2010, he was a Research Assist with the R&D
Center for Electric Vehicles, Chinese Academy of Sciences.
Since 2011, he has been a Research Assist with the
CUHK-CAS Shenzhen Institute of Advanced Integration
Technology. His research interests include state estimation and
intelligent control for electric vehicles and robots.

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.