Professional Documents
Culture Documents
Abstract: In this paper the control of the pressure inside the rear brake hydraulic circuit of
a sport motorbike is considered. A newly-designed electro-mechanical actuator, controlled by a
dedicated ECU, is capable of properly moving the piston inside the brake master cylinder and
is used to achieve the control purpose. In particular, this paper proposes a third-order linear
model for the hydraulic braking system and formalizes its theoretical derivation. Furthermore,
the work points out the relevance of friction inside the master cylinder chamber and discusses a
technique to compensate its negative effects. The pressure control system is designed referring
to the braking system linear model properly identified. The performances of such a controller
are assessed on the instrumented motorbike.
1. INTRODUCTION AND MOTIVATION vehicle. Even the majority of automobiles braking systems
are based on this working principle, but the devices used
A hard-braking maneuver on a sport motorbike might be in this field are quite different (tandem master cylinders,
a very critical task to achieve by a non-professional driver: brake boosters, ...). In order to integrate an active braking
in fact the driver has to be aware of the vehicle dynamics control on motorbikes, this classical structure has to be
so that he can control correctly and rapidly the behavior modified. One of the most interesting solution in this
of the vehicle; furthermore a braking maneuver strongly direction is the so-called ”Brake By Wire” (BBW): this
depends on several critical phenomena (such as road technology is based on the concept of physically decoupling
conditions) which yields to vehicle instability problems the braking command of the driver and the braking
(i.e. wheel locking). Even a professional driver might not action itself. In a BBW system, the entire braking process
be able to obtain ”optimal” braking performances in terms is governed by the ECU: the driver braking request is
of stability and maximum deceleration for every external acquired by a proper transducer and processed by the
condition, especially in cornering. These considerations control unit which generates the command for the braking
highlight the potential benefits of an active braking control actuator according to the available measurements and a
strategy in order to improve both the safety of the system specific control law. The power produced by the actuator
and the maneuver performances. In the framework of is then transferred to the brake pads, generating the brake
active braking control systems, the technological and the force.
control design aspects, which are strictly related to each From the control design view-point there is a huge gap
other, must be pointed out. between bikes and automobiles. Active braking control
From a technological point of view, the classical structure systems, such as Antilock Braking System, come standard
of motorcycle braking systems is based upon a hydraulic on most four-wheeled vehicles produced today; the inter-
circuit and passive mechanical devices. It basically consists action between braking system and other devices on the
of a master cylinder, a pipeline, brake calipers (and their automobiles (active suspensions, active differentials and
relative brake pads) and a brake disk (mounted on the active steering) have been analyzed and some appealing
wheel); the driver gets the braking force through the brake solutions have been presented (see e.g. Zanten (2002)).
lever which pushes the piston of the master cylinder: this In the framework of motorcycles, instead, active braking
action causes the oil pressure inside the master cylinder control systems are still little studied: from a cultural
chamber to increase, inducing the movement of the brake point of view, the motorcycle world and, especially, mo-
pads against the disk. The friction between the surfaces of torcyclists are very resistive to the introduction of active
these two devices produces the desired braking force on the control systems on their bikes, since automation and elec-
tronics usually mean ”unsafe” for them; from a practical
? This work has been partially supported by FIRB project ”Highly
point of view, the mechanical design represents a critical
Innovative motorbikes with ultra-low emission engines, active sus- task, in particular for what regards the choice of weights
pensions, electronic brakes and new materials” , MIUR project and dimensions of the devices introduced to control the
”New methods for Identification and Adaptive Control for Industrial
system. Hence, nowadays, motorcycles are not produced
Systems”, MV Agusta Motor S.p.A. and Brembo S.p.A.
525
Mechatronics'10
Cambridge, MA, USA, Sept 13-15, 2010
Current [A]
0.5
0.4
0.3
0.2 Motor Current Setpoint
(a) Physical layout of the actuator 0.1 Measured Motor Current
Simulated Motor Current
0
Actuator Rear brake Front brake 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Time [s]
Pressure Pressure
sensor sensor Fig. 3. Current Closed-Loop step response (Iset = 0.73 A)
526
Mechatronics'10
Cambridge, MA, USA, Sept 13-15, 2010
527
Mechatronics'10
Cambridge, MA, USA, Sept 13-15, 2010
By substituting the values of table 1 into (15), the dynamic force is proportional to the pressure in the master cylin-
behavior of the system can be analyzed. The transfer der chamber (pmc ), is adopted. The pressure dependence
function (15) is characterized by the three poles T1 , T2 describes the changes in the contact surfaces due to the
and TH , the zero τH and the gain K gathered in table 2 deformations of the piston seals as pressure in the master
(in (15) the pressure is expressed in bar and not in Pascal). cylinder increases.
In particular, TH is an high-frequency pole and is related At first, it is useful to express the friction force Ff r in
to the master cylinder pressure/position dynamics; the the master cylinder in terms of friction torque Tf r at the
other two poles are located at frequencies f ≈ 1Hz and DC motor. The relation between these two variables is
f ≈ 10Hz: they are strictly connected to the elastic described by the kinematic equation (7). Let Tf r , Tcf ,
behavior of both the fluid and the pads and they determine Tpr and µΩ be the total friction, the Coulomb friction,
the dynamic response of the braking system. The system the preload and the viscous friction torques respectively.
step response in a nominal pressure range will be shown The friction model is represented by the following set of
further in this paper (see Fig. 6). equations:
Tcf 1 + (kbrk,p − 1)e−c|Ω| + µΩ
if Ω ≥ 0,
Tf r =
4. FRICTION IDENTIFICATION AND
−Tcf 1 + (kbrk,n − 1)e−c|Ω| + µΩ if Ω < 0.
COMPENSATION
Tpr + fcf r,p pmc if Ω ≥ 0,
Tcf =
Friction in the actuator is mainly caused by the relative Tpr + fcf r,n pmc if Ω < 0.
motion between the piston O-ring seals and the internal (16)
cylinder surface. Friction is responsible of undesired behav- The parameters introduced in (16) are:
iors of the braking system, such as stick-slip phenomenon,
hysteresis and non-repeatability, which involve problems - c: it expresses the transition between the static fric-
in both identification and control. These considerations tion torque Tsf and the Coulomb friction torque Tcf ;
lead to the necessity of compensating friction, in order - µ: represents the viscous friction coefficient;
to correctly estimate the system parameters and achieve - kbrk,i (i = p, n): it defines the relationship between
better control performances. The pursued approach can be the breakaway friction torque (Tbrkf = Tcf + Tsf )
divided into two steps: at first a suitable friction model is and the Coulomb friction torque;
formalized and identified; then a compensation technique - fcf r,i (i = p, n): it governs the Coulomb friction
is discussed and its performances are evaluated. torque increasing with pressure.
The reader observes that the model is able to describe
4.1 Friction modeling and identification friction asymmetric behavior, since kbrk and fcf r may
have different values for positive or negative motor angular
A ”modified” version of the Stribeck friction model (see speeds. The vector of parameters:
e.g. Olsson et al. (1998)), in which the Coulomb friction ξ = [c µ kbrk,p fcf r,p kbrk,d fcf r,d ]
Table 1. Physical parameter values is estimated via numerical simulation. The identification
approach (see Corno et al. (2009a)) is divided in the
Symbol Value following steps:
DOT4 Bulk Modulus β 1 × 109 [Pa]
DOT4 Cinematic viscosity ν 5 × 10−4 [m2 /s] (1) the slowly-varying current ramp profile:
DOT4 Oil damping kdamp 2 × 104 [N/(m/s)]
0.2 t if t ∈ [0 25]
DOT4 Density ρ 1 × 103 [kg/m3 ] iset = (17)
10 − 0.2 t if t ∈ (25 50]
Pipeline cross-section di-
dpl 8 [mm] is applied to the motor (the current control loop
ameter
Pipeline length Lpl 1 [m] ensures that i ≡ iset ) and the pressure response pmc
is measured. The same current profile is applied to
Master Cylinder piston the non linear model (defined by (13) and (16)); the
Amc 1.2 × 10−4 [m2 ]
area
pressure ps,mc is registered;
Master Cylinder nominal
xmc 7 × 10−3 [m] (2) the vector of estimated parameters ξ̂ is obtained
position
Pad area Apd 3 × 10−4 [m2 ]
by minimizing,
Pn with respect to ξ, the loss function
2
Pad spring constant Kpd 1 × 106 [N/m] J(ξ) = k=1 (pmc (k) − ps,mc (ξ, k)) through a non-
Gear ratio kg 3 × 10−4 [m/rad] linear least square algorithm.
Gear efficiency ηg 0.9
Rotor inertia Jmot 1 × 10−5 [kg m2 ] The estimated values are listed in table 3 and the related
Back EMF constant Ke 30.4 × 10−3 [Wb] Stribeck curves are depicted, for different pressure values,
Master cylinder piston in Fig. 4.
Mmc 1 × 10−2 [kg]
mass
4.2 Friction compensation
Table 2. Gain, poles and zeros for the braking
system transfer function (15) Since the identified friction model is affected by great
uncertainty, friction is compensated adopting the dither-
K τH T1 T2 TH ing technique. Formally, this strategy provides an open-
loop friction compensation by using an high-frequency
10.1 1/(2π160) 1/(2π0.866) 1/(2π9.75) 1/(2π4.7 104 )
disturbance, named dither signal, to ”smooth” the effect
528
Mechatronics'10
Cambridge, MA, USA, Sept 13-15, 2010
Identified Stribeck curves for different pressure values Pressure / Current: quasi-static characteristic
0.15 5.5
pressure p = 2 [bar] With Dither
mc 5
pressure pmc= 20 [bar] Without Dither
pe
3.5 S lo
0.05 ve
3 is ti
Po
p:
2.5 am
0 R
2
1.5
−0.05
1
e Slope
gativ
0.5 p: Ne
Ram
−0.1 0
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2 0 5 10 15 20 25 30 35
Motor velocity Ω [rad/sec] Measured master cylinder pressure [bar]
529
Mechatronics'10
Cambridge, MA, USA, Sept 13-15, 2010
27.5
- bandwidth: fp = 10 Hz;
27 - phase margin: ϕm ≥ 60◦ ;
- overshoot: OS% ≤ 15%;
26.5
- zero steady-state error;
26
Theorethical Non-Linear Model Hard−Braking Test
25.5 Measured
Identified Model (medium pressure) 40
25
0 0.5 1 1.5 2 2.5 3 3.5 4 35
Time [s]
30
Pressure [bar]
Fig. 6. Time comparison between the braking system,
25
the identified system and the theoretical model (Step
response of Iset = 0.35 A; medium pressure range) 20
15
10
5 Pressure Setpoint
duced in (21) in order to achieve a thorough class of models Measured Pressure
which accurately catches this low frequency dynamic too: 0
0 0.2 0.4 0.6 0.8 1 1.2
Time [s]
P (s) (1 + sτL )
G(s) = =K (22) (a) Hard-braking test (the step setpoint is generated via software)
I(s) (1 + sT1 )(1 + sT2 )(1 + sTL )
Tracking Test
Let ξ = [K τL T1 T2 TL ] be the vector of the
30
parameters in (22); for each pressure working point, the
strategy adopted for the estimation of ξ consists of two
steps: 25
20
system dynamics;
(2) identification of the high frequency (1 Hz ≤ f <
15 Hz) braking system dynamics. 15
530
Mechatronics'10
Cambridge, MA, USA, Sept 13-15, 2010
The pressure controller is tuned with respect to the lin- Cossalter, V., Doria, A., and Lot, R. (2000). Optimum sus-
ear model identified at medium pressures: this range of pension design for motorcycle braking. Vehicle System
pressure is indeed considered the nominal working region Dynamics, 34(3), 175–198.
of the braking system. The control strategy consists of a Hirsch, M., Alberer, D., and del Re, L. (2006). Stability
PID regulator whose action is based on the pressure error control by advanced full-braking systems of motorbike.
(pset − p); this controller generates the current reference In Proceedings of IMECE2006: 2006 ASME Interna-
i∗set : this signal, in addition to the dither current idit , repre- tional Mechanical Engineering Congress and Exposition.
sents the reference iset for the internal current control loop Ho, H.P., Day, A.J., Hussain, K., and Johnstone, A.
(see Fig. 1). In order to get a decoupling action between the (2009). Modelling and simulation of the characteristics
internal and the external control loop, a second order low- of a hydraulic brake master cylinder. In SAE Interna-
pass filter with cut-off frequency f = 80 Hz is applied to tional - SAE 2009 Brake Colloquium.
the signal iset . The PID has been tuned and implemented Isermann, R. (2005). Mechatronic systems. Springer
adopting an anti-wind up configuration. Berlin.
The experimental results depicted in Fig. 7 show the Junge, C., Senicar, F., Wegener, R., and Soter, S. (2009).
excellent performances in terms of overshoot and settling Pressure control of a nonlinear system with a linear-
time obtained in the whole ranges of pressure. In Fig. 7(a) pmsm and a standard inverter. In Electric Machines
an hard-braking maneuver is considered: it basically con- and Drives Conference, 2009. IEMDC ’09. IEEE Inter-
sists of a step response where the pressure setpoint pset national, 83–88.
(generated via software) changes from 6 bar to approx- Limebeer, D.J.N., Sharp, R.S., and Evangelou, S. (2001).
imately 37 bar. Fig. 7(b), instead, presents the tracking The stability of motorcycles under acceleration and
results achieved when the pressure reference comes from braking. Journal of Computational and Nonlinear Dy-
the instrumented brake pedal: its position is translated, namics, 9(215), 1095–1110.
through a static map, into the the setpoint pset . Michaux, M.A., Ferri, A.A., and Cunefare, K.A. (2007).
Effect of tangential dither signal on friction induced
7. CONCLUSION oscillations in an sdof model. Journal of Computational
and Nonlinear Dynamics, 2, 201.
In this paper the control of an electro-mechanical brake- Olsson, H., Astrom, K.J., Wit, C.C.D., Gafvert, M., and
by-wire actuator has been presented. The identification Lischinsky, P. (1998). Friction models and friction
of the system have exploited a model-based approach, compensation. European Journal of Control, 4, 176–195.
in order to achieve a better understanding of the brake Savaresi, S.M. and Tanelli, M. (2010). Active braking
behavior: the third-order model developed in this paper control system design for vehicles. Springer-Verlag. To
turns out to be a suitable model for the pressure con- appear.
trol, since it is capable of taking into account the most Schwarz., R., Isermann, R., Bohm, J., Nell, J., and Rieth,
relevant dynamics of the braking system. The friction P. (1999). Clamping force estimation for a brake-by-wire
compensation has been obtained via dithering technique: actuator.
the effectiveness of such a method for this specific problem Sharp, J.J. (1981). Hydraulic modelling. Butterworths
has been proven by the experimental results. The pressure London. 1981. 242.
control loop, based upon a PID controller plus the dither Sharp, R.S. (2009). Limit braking of a high-performance
signal, guarantees the 10 Hz bandwidth and the bounded motorcycle. Vehicle System Dynamics, 47(5), 613–625.
overshoots. The capabilities of such an actuator represent Tanelli, M., Astolfi, A., and Savaresi, S.M. (2007). Mixed
the basis from which more sophisticated high-level braking slip-deceleration control in automotive braking sys-
control strategies can be further implemented. tems. ASME Transactions: Journal of Dynamic Sys-
tems, 129(1), 20–31. Awarded with the 2007 ASME
ACKNOWLEDGEMENTS Dynamic Systems and Control Rudolf Kalman Best
Paper Award.
Thanks are due to ing. Nicola Iovine for having supported Tanelli, M., Astolfi, A., and Savaresi, S.M. (2008a). Robust
the test analysis, to ing. Carlo Cantoni and ing. Riccardo Nonlinear Output Feedback Control for Brake by Wire
Arrigoni from Brembo S.p.A. and to ing. Luciano Spag- Control Systems. Automatica, 44(4), 1078–1087.
giari from MV Agusta Motor S.p.A. Tanelli, M., Osorio, G., Bernardo, M.D., Savaresi, S.M.,
and Astolfi, A. (2009). Existence, stability and robust-
REFERENCES ness analysis of limit cycles in hybrid Anti-lock Braking
Systems. International Journal of Control, 82(4), 659–
Corno, M., Panzani, G., Maggio, G., Mazzocchi, P., Goggi, 678.
A., and Savaresi, S.M. (2009a). Nonlinear modeling and Tanelli, M., Prandini, M., Codeca, F., Moia, A., and
control of a dual-stage hybrid ride-by-wire throttle body Savaresi, S. (2008b). Analysing the interaction between
for a sport motorbike. In 2nd ASME Annual Dynamic braking control and speed estimation: the case of two-
Systems and Control Conference (DSCC09). Hollywood. wheeled vehicles. In 47th IEEE Conference on Decision
Corno, M., Savaresi, S.M., and Balas, G.J. (2009b). On and Control, 2008. CDC 2008, 5372–5377.
Linear Parameter Varying (LPV) Slip-Controller design Zanten, A.T.V. (2002). Evolution of electronic control
for two-wheeled vehicles. International Journal on systems for improving the vehicle dynamic behavior.
Robust and Nonlinear Control, 19, 1313–1336. In 6 th International Symposium on Advanced Vehicle
Corno, M., Savaresi, S.M., Tanelli, M., and Fabbri, L. Control, AVEC, Hiroshima, Japan.
(2008). On optimal motorcycle braking. Control En-
gineering Practice, 16(6), 644–657.
531