2007-01-0866

Modeling and Control of a Single Motor

Electronic Wedge Brake
J. Fox, R. Roberts, C. Baier-Welt, L. M. Ho, L. Lacraru, B. Gombert
Siemens AG Siemens VDO Automotive

Copyright © 2006 SAE International

ABSTRACT parameters within the new design, particularly to the

coefficient of friction between pad and disk, but also to
Siemens VDO is currently developing a brake-by-wire brake pad wear and thermal effects. Two different
solution called the Electronic Wedge Brake (EWB). control approaches are highlighted here: one is based
Earlier prototypes used a two-motor concept inside the on a cascaded controller which was originally developed
brake actuator to drive the wedge. In this paper, a new for the tandem-motor prototype. The second concept
prototype generation is presented where only one motor uses full or partial state-feedback and is able to generate
is needed. This is more efficient in terms of cost, and almost identical step responses for a broad range of
actuator weight, and also reduces the complexity of the parameter variations.
control strategy. A state-space model is derived for the
new actuator and two controller structures are explained. The performance of the second controller is demonstra-
Simulations and experimental results from a roller test ted in simulation results. For the cascaded controller,
bench are shown. experimental results are included, which show the poten-
tial of the new prototype to fulfill the requirements of a
INTRODUCTION modern brake system.

In the past few years, much effort has been spent at HISTORY OF THE ELECTRONIC WEDGE
Siemens VDO to develop a brake-by-wire solution called BRAKE
the Electronic Wedge Brake. Because the brake uses
the self-reinforcing wedge principle, the actuation forces The electronic wedge brake (EWB) is a self-reinforcing,
and therefore electrical power requirements are much electro-mechanical brake system. The self-reinforcing
lower than in conventional electromechanical brake operating principle allows reduction in the actuation
systems. Tests on dynamometers and in prototype force, such that system is able to realize the required
vehicles have already proven that the EWB can braking forces using existing 12V vehicle electrical
outperform hydraulic systems with regard to braking systems.
distance and driver comfort, particularly on low friction
surfaces. The initial investigation of the wedge brake concept is
described in [1]. The concept was then further developed
Previous publications have described a realization where by eStop, which introduced the alpha-prototype, with the
two motors are used to drive the wedge. By this means, ability to realize the full braking force expected from a
it is possible to create a preload in the drive system such modern disc brake system [2,3]. At the center of the
that no backlash can occur. In this paper, a new single alpha prototype design is the brake heart, which realized
motor concept is presented, which satisfies the low the wedge braking principle, and the dual motor concept,
backlash requirement mechanically. The actuator differs which removed backlash in the transmission (Figure 1).
from the tandem motor implementation in both its drive-
train concept and geometry, and therefore a new state- In 2003, this was followed by the beta prototype (Figure
space physical model is developed. The nonlinearities 2), which was a more robust development while retaining
within this system, most notably the effect of braking the same basic mechanical concept. The beta prototype
direction, and their effects on the accuracy of the model also contained other advanced functionalities, including
are discussed. brake pad wear adjustment, and a fail-safe mechanism
which reduces braking force to zero when electric power
The wedge principle which provides the self-reinforce- is lost. The fail safe mechanism is required to prevent
ment can also lead to an unstable open-loop system. uncontrolled braking due to the self reinforcment of the
Therefore, a control algorithm has to be carefully EWB. The beta prototype was subsequently installed
designed to guarantee the closed-loop stability under all onto dynamometers, test trailers and test vehicles,
circumstances. It must be robust to the variation of where it was validated in real-world conditions, including
demanding winter tests. Its development and the testing complexity over a single motor solution. As a result, the
results can be found in papers by Siemens AG [4,5]. next generation, more production-oriented prototype,
denoted in this paper as the prototype 1 (PT1, Figure 3),
was designed with the constraint that only one motor will
be used to control the wedge mechanism .

Wedge THE NEW PROTOTYPE 1 CONCEPT

Rollers
Caliper

Stator

Motor

Motor

EPB, Fail-Safe
Brake Heart Rollerscrew Wedge and Wear
Mechanism Motor Adjustment
Force Stiffener Motor
Sensor
Figure 1: Dual Motor Solution as Used in the Alpha
Prototype. The Brake-Heart Mechanism is shown. Figure 3: PT1 Prototype (90kN / 4500 Nm)

The PT1 wheel unit concept uses two electrical motors

In previous publications, it was shown that the stability of in total, each with its own position sensor. While the
the wedge brake varies with the coefficient of friction EWB based service brake is actuated by the first motor,
between brake pad and disk. At low coefficient of friction, parking brake, wear adjustment and fail-safe function are
the net force on the wedge acts to push it back out of the realized using the second motor, compare Figure 3.
caliper, while at high friction coefficient, it pulls it in. A
change in this parameter can therefore result in the Position sensors on each motor and additional clamping
wedge jumping across the backlash in the drive force and fail-safe sensors allow a precise control and
mechanism, resulting in a step change in braking force. enhanced self diagnostics including plausibility checks of
To solve this problem, the alpha and beta prototypes motors and sensors.
both used a tandem motor design, such that the two
motors can be used to preload the drive train. An integrated Wheel Control Unit (WCU) allows the
minimization of cable harness efforts and electro-
magnetic interferences.

SINGLE MOTOR ACTUATION FOR THE SERVICE

BRAKE

The service brake is based on wedge brake technology

which is combined with a floating caliper, compare
Figure 4. In summary, the changes of the service brake
wedge drive include:

1. angled drive shaft to increase the spindle durability

of braking in the "forward" direction. This also results
in a more complex geometry when the wedge is
actuated in "backward" direction.
2. new, low backlash power screw concept
3. increased torque class (4500 Nm)

Figure 2: Beta Prototype

Although the tandem motor solution has proven to be a

reliable and functional method to remove backlash, there
is the obvious disadvantage of increased cost and
 Figure 6: PT1 force sensor integration

Figure 4: Single motor actuation for the service The achieved signal quality of the new force sensor
brake concept is significantly better compared with the former
prototypes. While the noise of the new force sensor is
limited to 0.1% (Figure 7), the noise of the classical
INNOVATIVE FORCE SENSOR CONCEPT sensor is approx. 0.4% (Figure 8). The main reason of
the better signal to noise ratio of the new force sensor is
Classical force sensors for brake by wire systems are based on the low sensibility related to side forces
typically based on quite expensive pressure sensors parallel to the disc. While the classical sensor shows a
derivates and need a very sophisticated mechanical low robustness related to side forces, the PT1 sensor is
integration due to a high sensitivity to side forces. As a not being affected by side forces mainly for 2 reasons:
result, noise to signal ratio is quite poor related to the
efforts. 1. the high stiffness of the loaded arm in side direction
2. the hall sensor does not measure the side
The force sensor principle used for the EWB system movement of the loaded arm, but the movement in
measures the caliper deformation. The sample shown in clamping direction.
Figure 5 uses the principle to provide a non loaded
reference arm which is outside the force flow of the Another reason for the high signal quality is the
caliper. This arm is created designing a slot inside the absense of stick-slip-effects inside the sensor system.
caliper. The typical relative movement of the loaded part
to the reference surface is approx. 1 mm for the
maximum clamping force.

loaded
arm
load free
arm
∆x

Figure 5: FEA simulation for the deformation of a

slotted caliper loaded with a high clamping force

Figure 6 shows the sensor integration in the caliper. The Figure 7: Signal of a classical force sensor in
base sensor used is a linear Hall sensor (10 bit) which is Prototype 1 (max. 90 kN). The noise level is approx.
combined with a NdFeB magnet. 100 N (0.1% of max. range)
 θ M , ωM
iM
K Axial
FR
RM
uM FM
α
LM β K Cal calliper
spindle/gear D Axial disc FB

µ
FN
x vwheel

Fig 9: Simplified Model of the Electronic Wedge Brake

The coordinate xW is defined to be parallel to the brake

disc. In this direction, the balance of forces then reads:

cos α FM + µ FN − sin α FR = mW v&W

mW is the mass of the wedge, vW its velocity in xW-

direction. In the direction of the normal force, the force
balance reads:

sin β FM − FN + FR cos α = mW tan α v&W

Both equations can be combined to

⎛ 1 ⎞
FM + ( µ − tan α ) tan αK Cal x ⎟ .
1
v&W = ⎜
mW (1 + tan 2 α ) ⎝ cos α ⎠

Figure 8: Signal of a classical force sensor in a To calculate the motor force FM, the system is modeled
BETA prototype (max. 40 kN). The noise level is in axial direction by a stiffness KAxial and a viscous
approx. 150 N (~0.4% of max. range) damping DAxial. The roller screw has a lead L and thus
transforms the motor angle θM, the motor rate ωM, and
The following chapters will concentrate on the wedge the motor torque MM to the wedge position, velocity, and
drive train of the service brake. motor force. Assuming a simple model of the roller screw
where one assigns a constant torque/force efficiency
STATE-SPACE MODEL OF THE SINGLE MOTOR 0<η<1 to it, one can write
BRAKE
L ⎛ L L ⎞1
MM = − ⎜ K Axial ( xW − θ ) + DAxial (vW − ωM )⎟ .
BASIC LINEAR EQUATIONS 2π ⎝ 2π cos α 2π cos α ⎠ η

In order to derive a mathematical model for the PT1, However, this model is inaccurate for many conventional
consider the simple model of the brake actuator from roller screws. They usually show a nonlinear behavior
Figure 9. The wedge has an angle α. Four forces act on such that the above equation could also be replaced by
it: a motor force FM, a normal force FN between the brake a general nonlinear term:
disc and the brake pads that are mounted on the wedge,
a braking force FB that results from the relative M M = M M ( FM , vW )
movement of pads and disc and the coefficient of friction
µ between them, and a reaction force FR which is normal
In the EWB, the nonlinearity of the roller screw can be
to the wedge edge because the friction on this edge can
approximated by a lookup-table for the simulation
be neglected. The angle under which the motor force
models. However, the controllers that are developed in
acts on the wedge is called β. If β=α, the relation
the following section are robust with respect to the
between motor force and braking force is optimal, and
magnitude of the roller screw's nonlinearity.
only this case is considered throughout this paper.
 ⎛ 0 1 0 0 0 ⎞
⎜ ( µ − tan α ) tan α K Cal − K Axial (cos α + tan α sin α ) DAxial (cos α + tan α sin α ) LK Axial (cos α + tan α sin α ) LDAxial (cos α + tan α sin α ) ⎟
⎜ − 0 ⎟
⎜ mW (1 + tan 2 α ) mW (1 + tan 2 α ) 2π mW (1 + tan 2 α ) cos α 2π mW (1 + tan 2 α ) cos α ⎟
⎜ 0 0 0 1 0 ⎟
A=⎜ LK Axial LDAxial L2 K Axial L2 DAxial d kM ⎟
⎜ − 2 − 2 − M ⎟
⎜ 2πJ Aη 2πJ Aη 4π J A cos α η 4π J A cos α η J A JA ⎟
⎜ 0 0 0
k
− M −
RM ⎟
⎜ LM ⎟⎠
⎝ LM

⎛ K Cal tan α 0 0 0 0⎞
⎜ ⎟
T
⎛ 1 ⎞
B = ⎜⎜ 0,0,0,0, ⎟⎟ C =⎜ 0 0 1 0 0⎟
⎝ LM ⎠ ⎜ 0 0 0 1 ⎟⎠
⎝ 0
Fig 10: State-Space Matrices of the PT1 Model

The motor that is used to drive the wedge is a high- that its efficiency depends on whether the rotational part
performance brushless DC motor. For the purposes of is driving the translational part or vice versa. In the
controller design, it can be modeled as a single-phase simulation models of the EWB, these effects have been
DC motor with resistance RM, inductance LM , a friction handled by a lookup table which is based on direct
(modeled as a viscous damping) DM, and a torque measurements of the roller screw. Another effect to be
constant kM. The motor angle θM is measured, its angular modeled is stick friction in the motor and the roller screw.
velocity ωM is obtained from an observer. The motor is
driven by a voltage uM. Using these variables, an Another mechanical nonlinearity comes into play when
equation for the motor current iM can be written: one looks closer at the wedge motion if the direction of
the brake disc changes. Since the basic brake
R k 1 mechanism of Figure 9 shows self-reinforcement only for
i& = − M i − M ω M + uM the depicted direction of the wheel, the real brake has a
LM LM LM 'double-wedge' construction. Its principle can be seen in
Figure 11. If α=β, then the spindle is parallel to the
STATE-SPACE MODEL wedge flank that is used during forward braking (the
case described above). If however, the wheel moves
The equations described in the previous section backwards, the wedge has to be driven into negative xW
represent a fifth order system which can be written in direction. There is an angle of 2α between the spindle
state-space form: axis and the velocity vector of the wedge. Therefore, a
rod with two joints has to be attached between the
x& = Ax + Bu spindle and the wedge. The nonlinear kinematics of this
mechanism make the state-space equations more
y = Cx complicated, but result in a similar matrix structure. The
new kinematics are a major difference to the Beta
prototype where the spindle axis was parallel to the xW
Here, x = ( xW , vW ,θ M , ω M , i M ) T is the state vector axis.
and y = ( FN ,θ M , i M ) T the vector of outputs. The
matrices A, B, and C are then given in Figure 10.

The coefficient of friction, µ, is a parameter that will

change heavily during the brake's operation. While in a
normal operation point, µ would be around 0.4, the
control must be able to handle very high friction (it is spindle
designed to handle even µ = 1 situations) and zero
friction (during standstill, under no load, no braking force α
is acting and therefore µ = 0). When looking at the rod wedge
system matrix A, one notes that µ can be found in only
one matrix element. One can also see that if the axial xW
stiffness is large compared to the caliper stiffness, the
influence of variations in µ is reduced. Therefore, a high
axial stiffness is advantageous. Fig 11: Forward and backward braking

KNOWN NONLINEARITIES

It has already been mentioned that the roller screw can

be a major source of nonlinearity. This is due to the fact
FDEM Force ωDEM Rate iDEM Current VDEM Motor θMOT Brake
Controller Controller Controller

FMEAS ωMEA iMEAS

Figure 12: Cascaded Controller Structure

Force
Limited
Force Demand Voltage Voltage
TG PI Current Limiter Brake Actuator

State-Feedback
Observer
Regulator
Force, Motor Position
Motor Speed, Current

Figure 13: PI State Feedback Controller

CONTROLLER CONCEPT both directions to build up force. The direction chosen

depends on the direction of rotation of the wheel. This
CASCADED CONTROLLER problem is solved by allocating signs within the controller
based on the demanded direction.
For initial commissioning of the brake, a standard
cascaded controller has been used. A simplified repre- The individual controllers within this structure are built up
sentation of the concept is provided in Figure 12. This from P, I and D elements, with some additional filtering
structure has the advantage that each loop can be to shape the frequency response, and varying degrees
consecutively built up, tested and validated. The clear of feed-forward. The majority of the work involved in
relationship between the loops and the physical developing the controller is associated with producing
parameters within the brake also simplifies handling the good basic brake feel for comfort braking on the one
nonlinearities in the system. Additionally, the loops can hand and high dynamics for ABS on the other.
run at different sample times and therefore allow the
computational effort to be minimized. The cascaded controller concept for the single motor
solution is structurally simpler than the control for the
The inner current loop is in fact a field-orientated motor Beta prototype which had two motors. In the Beta
torque control. This generates holding and acceleration solution, which was used in the references [2] and [3], a
torques at the motor and is fundamental to the operation preload control had to be added to prevent backlash
of the brake. The rate control loop is tuned to guarantee effects disturbing the control quality. Since the new
the linear stability of the brake controller for all values of prototype is designed such that the freeplay can be
friction coefficient. The advantage of this approach is neglected over the lifetime of the brake, a preload
that complete flexibility is retained for the outer loop, control is no longer required and the controller has a
since there are no longer restrictions on its closed-loop pure cascaded structure, as depicted in Fig. 12.
bandwidth. The sensor from which the motor rate is
measured is also less affected by mechanical PI STATE FEEDBACK CONTROL
disturbances and noise than the force sensor.
A new controller concept which is currently under
The force loop is responsible for implementing the brake development for the PT1 is depicted in Figure 13. The
demand required by the driver. One notable difference core of the controller is a full state feedback which has
between control of a wedge brake and a conventional been designed as an output Riccati controller in discrete
electro-mechanical brake is that it requires motion in time. This approach minimizes a weighted integral of the
output signals. Since not all system states can be a moderate value. If higher dynamics are needed (for
measured, namely the wedge position and the wedge example in braking scenarios where ABS is applied),
velocity which are decoupled from the motor's position different controller parameters can be used to generate
and velocity because of the finite axial stiffness, these faster responses.
states must be observed or estimated. Steady-state
Kalman filter gains have been used for tuning the As can be seen in Fig. 14, if the parameters of the
observer. For this part of the controller, alternative system are not known, state-space controllers do not
designs can also be used, for example a partial state have zero steady-state error. Therefore, an outer PI loop
feedback regulators or output feedback. In the latter is added which controls the error between the measured
case, no observer would be needed anymore. Indeed, normal force and the force demand.
recent research on this topic has shown that output
feedback is sufficient to give a good control The tuning of the Riccati regulator is very intuitive
performance. because with the weighting coefficients, you can chose
whether you want to put more controlling effort to
If one uses a Riccati regulator, the controlled system is influencing the outputs, state, or inputs. The outer PI
guaranteed to be stable for the parameters used during loop is also very easy to tune because it doesn't regulate
the design. However, as stated above, at least the the whole system but only the errors of the inner state
coefficient of friction µ must be regarded as highly feedback controller.
varying. Also KCal will vary due to brake pad wear and
other environmental influences. Therefore, a robustness Two blocks in Fig. 13 need further explanation. Just
analysis of the designed controller has to be carried out. before the actuator, the controller's output can be
Here, the influence of µ will be examined more closely. modified by the current limiter block. This is done to
allow only for a limited motor current. In the cascaded
controller, this can be done by simply limiting the input to
4 the inner current loop to a certain value. Since however
normal force [N]

µ=0
the new approach does not have an inner current loop,
the current limiter is slightly more complicated. Here, a
2 model based approach is used which also has the
flexibility to limit other values like the motor power
µ=1
consumption.
0
0 2 4 6 8 10
-3
time [s]
x 10
normal force [N]

5 1.5
current [A]

µ=0 1
0
0.5
µ=1 0
-5 0 0.05 0.1 0.15
0 2 4 6 8 10
time [s] time [s]
-3
x 10
5
current [A]

Figure 14: Step Responses of the PI State-Feedback µ=0

Controller for Different µ (no outer PI Loop) 0

As can be seen in Fig. 14, the step response in force of µ=1

the controlled system shows for µ=1 a damped system -5
0 0.05 0.1 0.15
with no overshoot. For lower µ, the steady-state value is
time [s]
higher and the system shows increasingly less
dynamics, but for µ ranging from 0 to 1, the system is
always stable. This is due to the fact that the controller Figure 15: Step Responses of the PI State-Feedback
has been designed for µ=1, i.e. for the highest admitted Controller for Different µ (including outer PI Loop)
friction. The higher µ is, the more the wedge is pulled in
and the more control effort is required – the system can The block called TG is a trajectory generator. It is
be stabilized more easily for lower µ. If however, the inserted to shape the force demand that comes from
controller had been designed for the same step outside the actuator. The aim of the trajectory generator
response at lower µ, it would have been unstable for µ = is to ensure that only demands that can be handled
1. physically by the actuator are passed to the control
algorithm. The trajectory generator can also modify the
The step response for µ = 1 shows a rise time of 0.1 s. demand in different braking scenarios and thereby slow
For this performance, the motor power consumption has down the controller, for example during comfort braking,
or speed it up when ESP, ABS, or emergency braking the force loop would be replaced by a PI state feedback.
are in action. This has the following advantages: By doing this, one combines the flexibility and
robustness of the new approach with the simple
• the regulator reactions can be designed without implementation of the current limits. Additionally, one
touching the control loop itself – therefore, stability is can process the state feedback control loop at a lower
not influenced by the trajectory generator sample rate than in the full state feedback approach.
• energy saving algorithms can be designed if high Only the inner loop has to be processed at the high rate.
brake performance is not needed Thus, one reduces the computational effort. These
• the generated trajectory can be used for diagnosis: issues are under investigation. The new controller
since it is designed such that it outputs only concept has only been tested in simulation so far. But
physically possible trajectories. Its outputs can be since the design step is partially automated, it can be
compared to the measured force and, if the easily adjusted to mechanical design variants of the
difference between them is too large, a malfunction actuator. An efficient implementation of the controller is
of the brake can be detected ongoing while the tests described below are carried out
with the original cascaded controller.

In Fig. 15 the step response of the complete controller EXPERIMENTAL RESULTS

structure is shown. Because of the outer PI loop, the
differences between µ = 0 and µ = 1 have been reduced ROLLER TEST BENCH RESULTS
to a very small value. The controller is stable for all µ
and shows a step response without overshoot. The step Initial testing of the new brake was conducted on a roller
response in current however is different for different µ. test bench. This consists of a driven drum upon which
This is due to the varying amount of self-reinforcement. the test wheel is mounted. The test bench is capable of
producing a torque of 5000 Nm and a longitudinal
Another way to realize a controller for the EWB would be velocity of 250 Km/h. The speed of the drum is normally
to combine the cascaded controller with the PI state controlled but it is also possible to brake against the
feedback. The inner current loop of the cascade is then rotating inertia of the test rig. The vertical force on the
the same as in the original approach, but the rate and

4
x 10
5
4

Force 3

[N] 2
Demand
1
Actual
0
99 99.5 100 100.5

30

20
Test
Velocity [m/s] 10 Drum
Wheel
0
99 99.5 100 100.5

3000

2000
Motor
Position [deg] 1000

0
99 99.5 100 100.5

2000
1000
Motor 0
Speed [RPM] -1000 Desired
-2000 Actual

99 99.5 100 100.5

50

Motor 0
Current [A]
-50
99 99.5 100 100.5

30

20
Battery 10
Current [A]
0

-10
99 99.5 100 100.5
Time [s]

Figure 16: 40kN Step Response

tire can be varied to limit the maximum braking moment. For these tests, the actuator was driven directly from a
12V vehicle battery, so the measurements of battery
The cascaded controller described in the previous current taken correspond exactly to what would be seen
section is embedded on a microcontroller, and the data on the road. Battery current is positive when power is
communication flow between the EWB electronics and being taken out of the battery, and negative when
the GUI is done via CAN. regeneration occurs.

The results presented in this paper were generated The first example is a 40kN step demand in brake
using an initial version of the controller, prior to normal force, illustrated in Figure 16. The battery current
optimization of the mechatronic system. Consequently peaks during the initial acceleration of the motor at about
they represent relatively slow dynamics compared both 25 A and afterwards oscillates somewhat in response to
to earlier results from the tandem motor concept [3] and the demands of the rate controller. Once the brake
to the current state of development of this actuator. demand is reached however, it almost settles back to its
quiescent level and the power draw is negligible (of the
order of 15 W). This is typical of the results seen with all
versions of the EWB to date.

The motor rate control is a little uneven due to the motor

running close to the voltage limit, resulting in some
oscillation in both the motor and battery currents. This is
an aspect of the control which is currently being worked
on.

For comparison, a smaller step response is shown from

1kN normal force to 5kN and back (Figure 18). A similar
pattern is evident in the battery current in this case too,
with both steps leading to sharp peaks in the power as
the motor is accelerated. The much smaller demands
are both followed accurately by the controller, so comfort
braking with this actuator will be satisfactory.
Figure 17: Roller Test Bench
As a final example, a slow ramp in force is shown up to

6000

4000
Force
[N] 2000 Demand
Actual
0
20 21 22 23 24 25 26

15

10
Test
Velocity [m/s] 5 Drum
Wheel
0
20 21 22 23 24 25 26

1000

Motor 500
Position [deg]

0
20 21 22 23 24 25 26

2000
1000
Motor 0
Speed [RPM] -1000 Desired
-2000 Actual

20 21 22 23 24 25 26

50

Motor 0
Current [A]
-50
20 21 22 23 24 25 26

30

20
Battery
Current [A] 10

20 21 22 23 24 25 26
Time [s]

Figure 18: 5kN Step Response

 4
x 10
2

1.5
Force 1
[N] Demand
0.5
Actual
0
27 27.5 28 28.5 29 29.5 30 30.5 31 31.5 32

15

10
Test
Velocity [m/s] 5 Drum
Wheel
0
27 27.5 28 28.5 29 29.5 30 30.5 31 31.5 32

1500

1000
Motor
Position [deg] 500

0
27 27.5 28 28.5 29 29.5 30 30.5 31 31.5 32

500
Desired
Actual
Motor 0
Speed [RPM]

-500
27 27.5 28 28.5 29 29.5 30 30.5 31 31.5 32

10

5
Motor 0
Current [A]
-5

-10
27 27.5 28 28.5 29 29.5 30 30.5 31 31.5 32

10

5
Battery 0
Current [A]
-5

-10
27 27.5 28 28.5 29 29.5 30 30.5 31 31.5 32
Time [s]

Figure 19: 15kN Ramp Response

15 kN and back. This is illustrated in Figure 19. It can be

seen that subjectively the ramp is followed quite well,
although there is some unevenness in the rate demand. CONCLUSIONS AND OUTLOOK
Because of the low wedge angle, the wedge is being In this paper, an overview has been given of the current
pulled in throughout all the experiments described here. status of the controller development of the EWB project.
At the start of the response, a positive motor current can While the design and the performance of the new
be seen which indicates that the motor has to overcome prototypes are being optimized, other aspects of the
losses in the drive mechanism and push the wedge in. brake development are focused. The vehicle dynamics
As the force increases however, it has to prevent the group is developing advanced control strategies for
wedge being drawn further into the brake. Because of safety and driver comfort, while reliability, robustness
the different driving and back-driving efficiencies of the with respect to environment, functional safety, and
roller-screw, the motor current required to 'allow' the system development are some of the major topics that
wedge to move into the brake is considerably smaller are being investigated. Vehicle testing is also being
than that required to pull it back out again. The battery performed on high and low friction surfaces to examine
current, however, is hardly different from its 'no load' the advantages of the new brake system.
value, demonstrating the low power requirements.
Fading situations with a very low friction coefficient have The control performance on the brake actuator itself is
also been tested and it has been shown that the the backbone for all systems on top of it. The high
controller handles these with the same performance. dynamics of the electrically driven system allow for faster
control reactions during, for example, ABS braking. This
One final aspect which can be seen on this plot is that results in shorter stopping distances and this benefit has
the brake caliper has a nonlinear stiffness characteristic. already been shown in test vehicles. Thus, driver safety
This is clear when one compares the triangular form of will be improved dramatically in cars equipped with the
the force response with the distorted triangle of the EWB.
motor position. At low forces, the motor has to move
relatively much further to generate the same change in
force, indicating a much lower value of the caliper
stiffness.
REFERENCES Dr. Christian Baier-Welt leads the wheel unit develop-
ment which includes mechanics, electronics and func-
[1] Hartmann, Schautt, Pascucci, & Gombert. “eBrake® tional development. After the study of mechanical
– the mechatronic wedge brake”. SAE Paper 2002- engineering, he received his PhD at the Technical
01-2582. University of Darmstadt.
[2] Roberts, Schautt, Hartmann, & Gombert. “Modeling
and Validation of the Mechatronic Wedge Brake”. Lok Man Ho is a control engineer from the Wedge
SAE Paper 2003-01-3331. Algorithm group. His current responsibilities include
[3] Roberts, Gombert, Hartmann, Lange, & Schautt. controller design and system architecture analysis. He
graduated in mechatronic engineering / computer
"Testing the Mechatronic Wedge Brake". SAE Paper
science at University of New South Wales, Australia.
2004-01-2766.
[4] Ákos Semsey, Richard Roberts. "Simulation in the
Dr. Lucian Lacraru is a Function Developer within the
Development of the Electronic Wedge Brake". SAE
Wedge Algorithm group. He is interested in control of
Paper 2006-01-0298 safety critical systems, an area in which he also obtained
[5] Lok Man Ho, Richard Roberts, Henry Hartmann, his PhD, at Loughborough University in the UK.
Bernd Gombert. "The Electronic Wedge Brake –
EWB. SAE Paper 2006-01-3196 Bernd Gombert is CTO of the Body & Chassis division
and of the Advanced X-by-Wire department of Siemens
CONTACT VDO. He founded the eStop company where the EWB
was developed originally.
Joachim Fox leads the group Wedge Algorithm at the
EWB team of Siemens VDO. He graduated in electrical Siemens AG Siemens VDO Automotive,
engineering/control science at Karlsruhe, Germany, and SV C BC AX WA, Postfach 10 09 43,D-93009
did research on stochastical state estimation, system Regensburg, Germany
identification, and inertial measurement systems. joachim.fox@siemens.com
http://www.siemensvdo.com/products_solutions/chassis-
Dr. Richard Roberts has a background in aerospace carbody/body_chassis_electronics/braking-technology
dynamics and control. He has developed the brake
actuator control for the EWB, from October 2002 as part
of eStop and since 2005 with Siemens VDO.