Control Engineering Practice 22 (2014) 57 68

Contents lists available at ScienceDirect

Control Engineering Practice

journal homepage: www.elsevier.com/locate/conengprac

Fast and smooth clutch engagement control for dual-clutch transmissions

Koos van Berkel a,n, Theo Hofman a, Alex Serrarens b, Maarten Steinbuch a
a b

Department of Mechanical Engineering, Eindhoven University of Technology, P.O. Box 513, 5600 MB, Eindhoven, The Netherlands Punch Powertrain, Croy 46, 5653 LD Eindhoven, The Netherlands

art ic l e i nf o
Article history: Received 23 November 2012 Accepted 24 September 2013 Available online 24 October 2013 Keywords: Clutch engagement control Automotive control Dual-clutch transmission

a b s t r a c t
Automotive dual-clutch transmissions use two gear shafts and two clutches to perform automated gear shifts at a high comfort level. The two objectives of the clutch engagement controller are to realize a fast clutch engagement to reduce the gear shifting time, and a smooth clutch engagement to accurately track the demanded torque without a noticeable torque dip. This research work presents a new controller design that explicitly separates the control laws for each objective by introducing clutch engagement phases. Simulations and experiments in a test vehicle show that the control objectives are realized with a robust and relatively simple controller. & 2013 Elsevier Ltd. All rights reserved.

1. Introduction Automated transmissions have become increasingly popular in passenger vehicles for automatic gear selection and gear shifting (Wheals et al., 2007). There are several types of automated transmissions, such as the classic automatic transmission, the low-cost automated manual transmission, the continuously variable transmission, and the Dual Clutch Transmission (DCT). The DCT, sometimes referred to as a direct-shift gearbox, uses two gear shafts and clutches, as schematically depicted in Fig. 1. The inactive shaft can be prepared for the next gear shift in advance, so the two clutches can quickly switch between gear shafts when the actual gear shifting is demanded (Zhang, Chen, Zhang, Jiang, & Tobler, 2005). This gives the DCT some advantages over other automated transmission technologies: it overcomes the discomfort issue of torque interruption during gear shifting, without sacricing much of the transmission efciency. In addition, the fuel efciency can be improved by using a rapid upshifting strategy without sacricing much of the performance (Ngo, Hofman, Steinbuch, & Serrarens, 2012). The DCT is able to perform fast and comfortable shifts to a lower gear when the driver requests more engine power.

1.1. Control problem formulation The performance of DCTs heavily relies on the quality of the clutch engagements: the driver expects a quick response of the
n

Corresponding author. Tel.: 31 40 2472811; fax: 31 40 2461418. E-mail address: k.v.berkel@tue.nl (K. van Berkel).

powertrain without feeling a torque interruption and oscillations caused by the engine inertia. The task of the clutch engagement controller is to realize a fast and smooth clutch engagement, subject to the uncertainties in the actuator dynamics and sensor measurements. Furthermore, for its relevance in the competitive automotive industry, the controller is restricted to the use of only standard speed sensors to keep the cost low, and to the use of limited computation and memory resources for its implementation in real-time hardware. In order to enhance in-vehicle calibration, the design must be transparent to understand the impact of each calibration parameter, whereas the number calibration parameters needs to be as small as possible (Christen & Busch, 2012). The design of clutch engagement controllers for automated transmissions is widely studied in the literature, resulting in several solutions. Decoupling (linear) controllers show promising simulation and experimental results with easily implementable control laws in Slicker and Loh (1996), Pettersson and Nielsen (2000), Fredriksson and Egardt (2000), Garofalo, Glielmo, Ianelli, and Vasca (2001), Serrarens, Dassen, and Steinbuch (2004) and Glielmo, Iannelli, Vacca, and Vasca (2006). Many of these designs, however, neglect relevant actuator dynamics or consider only their frequency domain characteristics, thereby neglecting relevant transient effects caused by initial conditions of the states. Heuristic and fuzzy logic controllers as presented in Tanaka and Wada (1995) and Kulkarni, Shim, and Zhang (2007) are based on multi-dimensional control maps and require ad hoc calibration of many parameters. Finally, optimal model predictive controllers with mixed cost functions, as described in Bemporad, Borrelli, Glielmo, and Vasca (2001), van der Heijden, Serrarens, Camlibel, and Nijmeijer (2007) and Dolcini and Bchartde Wit (2010), can be

0967-0661/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.conengprac.2013.09.010

58

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768

implemented in a vehicle after ofine computation of the explicit control laws. In-vehicle calibration remains cumbersome as each adjustment in one of the calibration parameters requires a new, ofine computation. 1.2. Main contribution and outline Most of the controllers that aim at both fast and smooth clutch engagement can be captured in the same framework: rst, the clutch slip speed is reduced at a high rate to achieve a fast clutch engagement, followed by a controlled reduction of the slip acceleration to achieve a smooth clutch engagement. The controlled reduction has to be fast, but cannot be faster than the dynamics of the actuator. This research work presents a new controller design that explicitly separates the control laws for each of these phases. The switches between the phases are chosen such that the desired slip acceleration is achieved at the time of clutch engagement. The latter can be used as the single calibration parameter to determine the trade-off between fast and smooth clutch engagement. Robustness analysis shows that the desired slip acceleration can be achieved for a range of time constants of the actuator dynamics, yet at the cost of a conservative clutch engagement time. The control framework uses only the slip speed and slip acceleration, so it can be used for the different driving modes of the DCT-based powertrain. It can also be used for a wider class of systems that require fast and smooth clutch engagement, such as described for a mechanical-hybrid powertrain equipped with a ywheel and a continuously variable transmission in van Berkel, Veldpaus, Hofman, Vroemen, and Steinbuch (in press). This paper elaborates the clutch engagement controller for three relevant driving modes, which are gear upshifting and downshifting during propulsion, and vehicle launching from standstill. Other driving modes, such as gear shifting during braking, are not elaborated, yet the controller can be designed using the same framework. This paper will not consider the design issues related to the setpoint tracking of the clutch torque on component level, as in Horn, Bamberger, Michau, and Pindl (2003) and Montanari, Ronchi, Rossi,

Tilli, and Tonielli (2004), yet will focus on the high-level setpoint generation for the powertrain components that inuence clutch engagement, i.e., the engine and the clutch. In summary, the main contributions of this paper are:

 the design of a stable and robust controller for fast and smooth 
clutch engagement that is tunable with a single calibration parameter; and a generic control framework elaborated for three relevant driving modes.

The outline is as follows: Section 2 presents the simulation model of the powertrain. Section 3 formalizes the control problem and describes the controller design. The performance and robustness of the controller are evaluated by simulations in Section 4 and by experiments in Section 5. Section 6 summarizes the results and conclusions.

2. Dynamic powertrain model The simulation model describes the most important longitudinal dynamics of the DCT-based powertrain. The purpose is to simulate the engagement of either clutch in Fig. 1. Since the clutch engagement is not performed simultaneously with the cross shift switching between the two gear shafts, as will be discussed in Section 3.1.4, it is sufcient to model only one clutch and one gear shaft for the clutch engagement simulations. The remaining relevant components are the engine, clutch, gearbox, drive shaft, and the vehicle, as shown in Fig. 2. The powertrain parameters used throughout this paper are listed in Table 1, where the speed ratios are dened as the rotational speed on the vehicle side

Table 1 Powertrain model parameters. Symbol Value 0.15 0.01 0.01 90 400 6000 0.19 f0:25; 0:39; 0:62; 0:97; 1:42g 0.92 [ 20, 140] [0, 140] 0.08 0.04 Unit kg m kg m2 kg m2 kg m2 Nms/rad Nm/rad Nm Nm s s
2

Description Engine inertia Primary inertia Secondary inertia Vehicle inertia Drive shaft damping Drive shaft stiffness Final drive ratio Gearbox ratios Average gear efciency Engine torque Clutch torque Engine time constant Clutch time constant

clutches gear shafts

gearbox wheels

engine

dual clutch transmission

Fig. 1. Schematic representation of a DCT-based powertrain. The DCT enable fast and smooth gear shifting using two gear shafts and two clutches.

Je Jp Js Jv dd kd rd rg ^t e c e c

Fig. 2. Detailed dynamic powertrain model for simulation purposes.

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768

59

divided by the rotational speed on the engine side. Each component model is described in more detail in the sequel. 2.1. Engine The gasoline internal combustion engine generates mechanical power through complex chemical and mechanical processes. The Engine Control Unit (ECU) steers and monitors these processes to control the engine torque e , which can take any value in e ; e . The torque is indirectly controlled as a function of the throttle position, engine speed, temperature, and air/fuel mix, which may result in an offset due to modeling uncertainties. The actuation is delayed by at least two crankshaft revolutions for a four stroke combustion cycle (Serrarens, 2001), whereas the torque rate is limited to reduce pollutant emissions (Nesch, Wang, Voser, & Guzzella, 2012). The mean value dynamics can be approximated by a rst order linear model with a xed time constant e (Guzzella & Onder, 2010; van der Heijden et al., 2007), given by _ e t 1

temperature, slip speed, and oil pressure. The dominant dynamics can be approximated by a rst order linear model with a xed time constant c (Gao, Chen, Sanada, & Hu, 2011; van der Heijden et al., 2007), given by _ c t 1

c t c t ;

where c denotes the requested clutch torque. The higher order dynamics, as well as the dependency on temperature, slip speed, and oil pressure, are accounted for in an uncertainty in c . In addition to these dynamics, a substantial delay (O0:1 s) can be expected when the pressure chambers are not lled in advance, which is relevant for the torque response at vehicle standstill, but not during the critical clutch engagement phase when the clutch is already transmitting torque. In the simulations, no modeling uncertainties are assumed in the torque offset and time constant, yet the robustness of the controller against these modeling uncertainties will be discussed with its design in Section 3.8. 2.3. Gearbox The gearbox consists of a ve-speed gear set, a Drive-NeutralReverse (DNR) set, and a differential with a nal drive gear. For the transmission model, the DNR is neglected as only the Drive mode is of interest. The speed ratio rg of the gear set depends on the selected gear g A f1; 2; 3; 4; 5g and remains constant during clutch engagement. The xed speed ratio of the nal drive gear (subscript d) is denoted by r d , so d r d r g p . The speed dynamics are modeled as a function of the lumped transmission inertias on the primary shaft J p and the lumped transmission inertias on the secondary shaft J s , by _ p t 1 c t r d r g t d t t t ; J p r2 g t J s 5

e t e t ;

where e denotes the requested clutch torque, indicated by the accent. The higher order dynamics and the dependencies on the many variables are accounted for in an uncertainty in e . In the simulations, no modeling uncertainties are assumed in the torque offset and time constant, yet the robustness of the controller against these modeling uncertainties will be discussed with its design in Section 3.8. The requested engine torque usually equals the nominal engine torque determined by the ECU e , denoted by the bar, which is based on the accelerator pedal position and a manufacturerspecic look-up table. The DCT may request the ECU to temporarily reduce the torque request from its nominal setpoint, e.g., to decelerate the engine during a upshift, but may not request a torque increase for safety reasons, so

e r e:

Despite this constraint, all driving modes can be performed as will be explained in Section 3.1. The engine speed dynamics are modeled by 1 _ e t e t c t ; Je 3

where d is the drive shaft torque. The transmission losses due to the hydraulic pump and friction in the bearings are lumped in the loss torque t , which is described by an empirical model as a function of the rotational speed, torque, and gear ratio. 2.4. Drive shaft The drive shafts that connect the nal drive with the driven wheels are relatively long and slender. Assuming driving in a straight line they can be modeled by a linear spring-damper system with stiffness kd and damping dd resulting in Slicker and Loh (1996) and Dolcini and Bchartde Wit (2010) _ d t _ w t _ d t kd d t w t dd where w is the rotational speed of the wheel shaft. 2.5. Vehicle Since this work mainly focuses on the clutch engagement, wheel slip is neglected, so the vehicle mass and the wheel inertias can be lumped into an equivalent inertia J v at the wheel shaft. The dynamics is modeled by _ w t 1 t v t ; Jv d 7 6

where e Z 0 is the engine speed, J e is the engine inertia, and c is the torque transmitted through the clutch. 2.2. Clutch The considered clutch is a multi-plate system immersed in transmission oil and integrated in the DCT. This wet friction clutch type is often used for its compact design and the favorable controllability characteristics during clutch engagement. When slipping, the pressure force on the clutch plates can be manipulated by a hydraulic actuator to control the transmitted clutch torque (c ), which can take any value in c ; c . The torque is transmitted from the high-speed side to the low-speed side of the clutch, so signc signe p , where p denotes the primary speed of the gear box. The transmitted clutch torque is controlled by a relatively straightforward feedforward controller, which uses the friction characteristic of the friction surfaces as a (static) function of the transmission oil temperature and the slip speed. Due to wear in the friction surfaces and the transmission oil, the actual friction characteristic will (slowly) change in time, which may result in an offset in the clutch torque. The hydrodynamics in the uid lm and the valve give a small actuation delay, which depends on the

where v represents the external load, due to rolling resistance, air drag, and variations in altitude. 3. Controller design The powertrain controller consists of two layers: a supervisory controller selects the most suitable driving mode, whereas specic

60

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768

driving mode controllers perform dedicated control tasks such as clutch engagement. In order to simplify the overall controller design, it is desired to have a generic clutch engagement controller that can be used for multiple driving modes. In the sequel, three relevant driving modes are explained in more detail for which the clutch engagement controller is explicitly designed. First, the control objectives are determined and formalized, after which the controller design model is introduced and the control laws are derived. Accordingly, the effect of uncertain actuator dynamics on the control performance is analyzed. 3.1. Driving modes The considered driving modes

gear g

gear g -1

engine speed shaft L speed shaft H speed

drive shaft torque

are

engine torque clutch L torque clutch H torque

 Power upshift: The DCT shifts one gear upwards while power  
is continuously transmitted from the engine to the wheels ( 1). Power downshift: The DCT shifts one gear downwards while power is continuously transmitted from the engine to the wheels ( 2). Vehicle launch: The vehicle is accelerated from standstill until clutch engagement ( 3).

phase 1 phase 2 cross shift phase 3

Fig. 4. Schematic representation of the speeds and torques of the main powertrain component for the power downshift.

Other relevant driving modes are the upshift and downshift during braking situations, when the engine drag torque (partly) decelerates the vehicle. Figs. 3, 4 and 5 illustrate the speeds and torques of the most relevant powertrain components. For each driving mode, three similar clutch engagement phases are identied, whereas the power upshift and power downshifts have an additional cross shift phase. 3.1.1. Power upshift ( 1) The power upshift is requested to improve the fuel efciency by decreasing the engine speed. The drive shaft torque reduces due to the higher gear ratio, which may be anticipated by the driver, or compensated by further pressing the accelerator pedal. Before the power upshift starts, the engine is connected to the gear shift with the lower gear denoted by the subscript L, i.e., e p;L . The inactive gear shaft is prepared with the higher gear denoted by the subscript H, to obtain e p;H afterwards.

gear 1

engine speed shaft speed drive shaft torque

engine torque clutch torque

phase 1

phase 2

phase 3

Fig. 5. Schematic representation of the speeds and torques of the main powertrain components for the vehicle launch.

gear g

gear g + 1 engine speed shaft L speed shaft H speed

drive shaft torque

The power upshift starts with a cross shift to change the torque transmission from one clutch to the other, while engine power can be continuously transmitted to the drive shaft since p;L 4 p;H . After the cross shift, the clutch engagement starts by reducing the engine torque to reduce the slip speed s e p;H at a high rate (phase 1), followed by a controlled reduction in the slip _ s (phase 2), and clutch closure (phase 3). After acceleration clutch closure, the engine torque is returned to its nominal value. In the short time period when the engine torque is reduced, the drive shaft torque is (partly) delivered by the engine inertia. 3.1.2. Power downshift ( 2) The power downshift is requested to increase the engine power by increasing both its speed and torque. The drive shaft torque does not immediately increase, as a part of the engine power is required to accelerate its inertia rst, yet the driver perceives a very quick auditive response, i.e., increasing engine noise frequency, without experiencing any torque interruption. Before the power downshift starts, the engine is connected to the higher

engine torque clutch L torque clutch H torque

cross shift phase 1 phase 2

phase 3

Fig. 3. Schematic representation of the speeds and torques of the main powertrain component for the power upshift.

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768

61

gear e p;H , whereas the inactive gear shaft is connected to the lower gear. The power downshift is triggered by a sudden increase in the nominal engine torque (or, accelerator pedal position). Since the cross shift can only take place when e 4 p;L , rst the slip speed s e p;L is reduced at a high rate (phase 1), by keeping the clutch torque low with respect to the engine torque. Accordingly, _ s is reduced (phase 2) to obtain a controlled the slip acceleration overshoot, which is needed to perform the cross shift. After the cross shift, the clutch is immediately closed (phase 3). 3.1.3. Vehicle launch ( 3) The vehicle launch is requested to accelerate the vehicle from standstill (or, from very low vehicle velocities). Before the vehicle launch starts, the engine is running at a fuel-efcient idling speed e e;idle , whereas during the vehicle launch, a higher launch speed e e;launch is desired for more engine power, which may be a function of the accelerator pedal position. After clutch engagement, the drive shaft torque slightly reduces due to the change in the acceleration of the engine inertia. The vehicle launch is triggered by a sudden increase in the nominal engine torque, after which the clutch torque must be kept relatively low in order to accelerate the engine to its launch speed. Meanwhile, the clutch torque must gradually increase to accelerate the vehicle, such that the driver immediately notices the powertrain response. After reaching the launch speed, the clutch torque is increased to reduce the slip speed s e p at a high rate (phase 1), followed by a controlled reduction in the slip _ s (phase 2), and clutch closure (phase 3). acceleration 3.1.4. Cross shift During the cross shift, the torque transmission shifts from the leaving gear shaft to the upcoming gear shaft, by (linearly) decreasing the transmitted torque of the leaving clutch while simultaneously increasing that of the upcoming clutch. In order to avoid any actuation delays, the upcoming clutch needs to be prepared with a pre-lled clutch chamber, whereas the torque capacity of the leaving clutch needs to be reduced to the transmitted torque. Considering a feasible time frame for the actuator dynamics (i.e., 0.3 s), a synchronous cross shift is relatively straightforward, especially when identical actuators are used,

resulting in a gradually changing drive shaft torque. Since the cross shift does not coincide with any clutch engagement phase, it is not considered in the design of the clutch engagement controller below.

3.2. Objectives and criteria The controller design objectives are to prescribe setpoints for the requested engine torque e t in (1) and the requested clutch torque c t in (4) to

 track the demanded drive shaft torque;  realize a clutch engagement that is fast to reduce frictional  
losses and smooth to ensure a high comfort level with a limited torque dip at the moment of clutch closure; make the trade-off between fast and smooth engagement tunable with just one calibration parameter; and warrant stability and robustness against various uncertainties such as errors in the actuator models and measurement noise.

These objectives are translated into mathematical criteria that are suitable for the controller design.

3.2.1. Demanded torque The demanded drive shaft torque d is determined by the nominal engine torque of the ECU ( e ), the selected gear ratio (rg), the torque to accelerate the relevant inertias in the powertrain, and the transmission loss torque t :

d t

 1  _ p t t t ; e t J e J p r g r d J s rg rd

It is not necessary to exactly track d , since the driver is quite capable to correct the accelerator pedal position to adjust the torque demand. For the driving comfort, however, it is important to limit sudden changes in the drive shaft torque, such as the torque dip caused by the clutch closure at time t2. So the rst control objective is to minimize the torque dip for t 4 t 2 : min maxj d t d t j:
e ;c
t

_ s) Fig. 6. The generic control framework is based on the typical slip speed prole during fast and smooth clutch engagement. (a) Slip speed (s ) and slip acceleration ( trajectories during a typical clutch engagement. The three clutch engagement phases are separated using the slip speed thresholds 1 t and 2 and (b) owchart of the clutch engagement controller.

62

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768

3.2.2. Fast and smooth clutch engagement The slip speed of the clutch is dened as

s e p :

10

3.2.3. Calibration parameter The value 2 can be used as the single calibration parameter to quantify the trade-off between fast and smooth clutch engagement. A small value for 2 gives a smooth clutch engagement, but requires a relatively long time period to reduce the slip acceleration, thereby affecting the engagement time performance given by (12). On the other hand, a large value for 2 gives a fast, but harsh clutch engagement, thereby affecting the maximum torque dip performance given by (9). The objective is not to nd an

_ s as a Fig. 6(a) shows the slip speed s and slip acceleration function of time, for a typical clutch engagement. Three phases are distinguished:

 Phase 1: Fast reduction of the slip speed s t for t A t 0 ; t 1 ;  Phase 2: fast reduction of the slip acceleration _ s t for 
t A t 1 ; t 2 ; Phase 3: clutch closure and driving for t Z t 2 .

optimal value for 2, as the optimal trade-off remains subjective and should be calibrated in the vehicle. Instead, for a given value of the calibration parameter 2, the maximum torque dip is already implicitly determined, so the only relevant control objective is then to minimize t2, subject to the requirements (11), (14), and (16). In order to minimize t2, the switch time t1 has to occur as late as possible to prolong the time period of fast slip reduction, but just in time to enable the actuator dynamics to change the slip acceleration to the least-restrictive value of the smoothness _ s t 2 criterion (16), so that 2 when s t 2 2 . 3.2.4. Uncertainties Only sensor information that is standardly available in automotive applications may be used. The number of sensors in a massproduced vehicle is very limited in order to keep production costs low. Here, the rotational speeds e , p , d and w are measured. Accelerations are estimated from the measured speed signals, using a differentiator in series with a rst order low-pass lter and a rate limiter, where a cut-off frequency of 20 Hz gives a good trade-off between noise and phase lag. The controller must be robust against measurement uncertainties, but also against modeling uncertainties. The time constants e and c are a function of many unknown and changing variables, but also crucial to determine the duration of phase 2. The controller must be stable and robust for a family of time constants e A e ; e and c A c ; c . 3.3. Framework

These phases are considered in more detail in the sequel. Ideally, the clutch will be closed and mechanically locked when the slip speed s approaches zero. However, the sign of the slip speed may not change, since it gives an uncomfortable change in the sign of the clutch torque. Therefore, closure is enforced by over-clamping of the clutch plates at time t2 when

s t 2 2 ;

11

for robustness. 2 can be considered as the slip speed threshold to start phase 3. A fast clutch engagement is desired to keep the gear shifting time and the thermal load in the clutch low, so the second control objective is to minimize the clutch engagement time: mint 2 t 0 :
e ;c

where j2 j is close to zero but larger than the measurement noise

12

A fast clutch engagement is realized when after a small starting period, for all t A t 0 ; t 1 with t 1 r t 2 , the slip speed is reduced to a certain threshold

s t 1 1 ;
at a high rate, given by ( _ s r 1 if s 4 0 ; _ s Z 1 if s o 0

13

14 The framework explicitly separates the three different phases of clutch engagement, as previously described. The controller ows through these subsequent phases in one direction as shown in the ow diagram in Fig. 6(b), so there can be no chattering between two phases. The switch between two phases takes place once the slip speed s t intersects one of the slip speed thresholds 1 t and 2, as indicated in Fig. 6(a). This framework uses only the measured slip _ s t , which provides speed s and the estimated slip acceleration robustness against disturbances in the powertrain dynamics (due to feedback control), and is generic for different driving modes as well as for different powertrain topologies. 3.4. Design model The design model captures only the most relevant powertrain dynamics in order to design a relatively simple yet accurate controller. The additional assumptions are:

where j 1 j a sufciently large parameter and sign 1 signs in order to reduce the slip speed. The fastest clutch engagement is obtained when t 1 t 2 , so the slip speed is reduced at a high rate in the entire time interval t 0 ; t 2 . However, the discontinuity in the drive shaft acceleration at closure time t2 is proportional to the slip acceleration just before closure, and is given by Garofalo et al. (2001) and Glielmo et al. (2006):
_ d t 2 _ d t 2 J e t 2 J v J e t 2 _ s t 2 r g t 2 r d :

15

2 2 where J e t 1=r d r g t J e denotes the equivalent engine inertia at the drive shaft. This discontinuity causes an uncomfortable dip in the drive shaft torque, thereby affecting the control performance given by (9). So, to achieve a smooth clutch engagement it is required that ( _ s t 2 Z 2 if s t 2 40 ; 16 _ s t 2 r 2 if s t 2 o0

 The inuences of the primary and secondary gearbox inertia  The clutch engages so smooth that no drive shaft oscillations are excited, so d w .  The variation in transmission efciency of the transmission has
a negligible inuence on the clutch engagement, so the torque ^ t c . ^t loss can be approximated using a constant efciency The controller design model describes the controlled outputs as a function of the actuator inputs and the external variable. The are small enough to neglect them, so J p 0 and J s 0.

where 2 is close to zero and sign 2 signs . Fig. 6(a) illustrates what the requirements (14) and (16) imply: the absolute _ s t j must be reduced from a large value like slip acceleration j j 1 j to a small value like j 2 j at the point of time t2 when s _ s has to be fast, but is equals 2. The controlled reduction of restricted by the time constant of the physical system, i.e., the time constants e and c of the actuator dynamics.

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768

63

For the power upshift ( 1), the control laws for the engine and clutch are for t A t 0 ; t 1 given by ( e t ut 20 if 1; c t c t _ p t can be interpreted as the nominal clutch where c e t J e torque. The fastest slip reduction is achieved when the engine torque is reduced to its drag torque level, i.e., ut e t , but the change in torque sign might give problems with play and pretension in the driveline. For a sufciently high e , a small positive torque can be chosen instead e.g., ut 0:05 e , but for a relatively low e it may be necessary to reduce the engine torque to its minimum to avoid a long clutch engagement time. For the power downshift ( 2) and vehicle launch ( 3), the control laws for t A t 0 ; t 1 are given by ( e t e t 21 if A f2; 3g: c t ut For the power downshift ( 2), the fastest slip reduction is achieved when the clutch torque is reduced to its minimum, i.e., ut 0, but that would give an unacceptable interruption in the drive shaft torque. A better strategy is to prolong the clutch torque value that corresponds to the engine torque right before the sudden increase that triggered the power downshift, i.e., ut c t 0 , so the driver will only experience a small delay in the torque increase, yet directly hears the acceleration of the engine speed. For a relatively small toque increase, it may be necessary to reduce the clutch torque further to avoid a long clutch engagement time. For the vehicle launch ( 3), the fastest clutch engagement is achieved when the clutch torque is increased to its maximum, i.e., ut c , but that would give a drive shaft torque that is much higher than demanded, and prevents the engine speed to accelerate to e;launch , or even cause engine stalling when e o e;idle . Instead, a proportional feedback controller is proposed to track e;launch while gradually increasing the clutch torque, e.g., with ut c t pe;launch e t where p 4 0. Then, an additional safety mechanism is required to guarantee a minimum drive shaft torque and to prevent engine stalling in extreme situations, e.g., when the vehicle stands on a slope. The design of such a mechanism is beyond the scope of this paper.

Fig. 7. Block diagram of the control framework: the controller creates the actuator inputs e T and c based on the demanded drive shaft torque d and the selected control phase . The powertrain dynamics are excited by the actuator inputs and _ p , resulting in the controlled drive shaft torque d and clutch the external variable _ s . The state estimator uses the measured speed signals to slip acceleration ^ for the phase selector and the controller. ^ and accelerations _ estimate speeds

controlled outputs are the drive shaft torque d and the clutch slip _ s , whereas the actuator inputs are the setpoints acceleration for the engine e and the clutch c . The external variable is the _ p . The formulation with the acceleraprimary shaft acceleration _ p instead of the drive shaft torque d and the load torque v tion follows from the assumption that sufciently accurate acceleration estimates can be determined from the available sensor signals. The structure of the control signals and the framework are schematically represented in the block diagram in Fig. 7. For the slipping clutch in phases 1 and 2, the model can be described using the speed ratio denitions and the time-derivatives of (10) and (3), by

d t
_ s t

1 ^ t t ; c t rd rg 1 1 _ p t ; e t c t Je Je

17

18

plus the actuator dynamics as described by (1) and (4). When the clutch is mechanically locked in phase 3, the clutch torque is no longer an actuator input and the slip acceleration is no longer controlled. The model is then given by

d t

1 _ p t ^ t t : e t J e rd rg

19

plus the engine torque dynamics as described by (1). From (17), it follows that for a slipping clutch, the drive shaft torque is completely determined by the clutch torque, whereas from (19) it follows that for a closed clutch, the drive shaft torque can only be inuenced by the engine.

3.6. Controller phase 2 and switch thresholds In phase 2, the main control objectives are to track the demanded drive shaft torque and to realize a change from a fast to a smooth slip acceleration before the clutch engages. The drive shaft torque and the slip acceleration are controlled by the same actuators as in phase 1. The main difference is that now the trajectory of the slip acceleration is important to warrant that the slip acceleration is reduced to 2 when the slip speed equals 2. So, the actuator dynamics should be considered in the control law. The control law is composed of a nominal feedforward term (denoted by the bar), which keeps the slip speed constant, and a feedback term that aims at smooth clutch engagement using a new control variable u(t) to control the slip acceleration. Using a generic description of the slip speed dynamics next, a single expression for u(t) can be derived that is suitable for each driving mode. The control laws for t A t 0 ; t 1 are given by ( e t e t J e e ut 22 if 1; c t c t

3.5. Controller phase 1 In phase 1, the main control objectives are to track the demanded drive shaft torque and to realize a fast reduction of the slip speed. In principle there are two actuators to achieve this. The clutch determines the drive shaft torque, so the engine can be used to reduce the slip speed. However, as the engine torque can only be reduced cf. (2), the engine can only be used for the power upshift ( 1), whereas the clutch must be used for the power downshift ( 2) and the vehicle launch ( 3). Since the exact slip speed trajectory is of minor interest in this phase, the actuator dynamics may be neglected and the physical limitations of the system may be chosen to achieve the fastest slip speed reduction. In the sequel, suitable values for a new control variable u(t) are proposed to control the slip acceleration for each driving mode.

64

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768

e t e t c t c t J e c ut

if A f2; 3g;

23

where W(z) is the Lambert W-function (Lambert, 1758), i.e., the inverse function of z W zeW z , and z _1 x _ 2 =x _ 2 x2 ex : _ 2 x2 x 31

where time constants e and c are constant and assumed known for now. In the sequel, it turns out to be convenient to introduce a modied slip speed xs o , where o is a yet to be chosen constant offset. After a lengthy yet straightforward calculation it follows that, for each driving mode, the slip speed dynamics of (18) can be written as a second order linear differential equation t x 1

The offset o of the slip speed state can now be used to _ 1 with drastically simplify the relation between x1 and x

o 2
_1 x

2 ;

32

so relation (30) simplies to x1

_ t ut wt : x

24

The associated time constant and external variable w(t) are given by ( e  _ : if 1; 25 e c t w t 1 c J p t
e

33

 _ : c 1 e t w t J e p t e

if A f2; 3g:

26

Using this result, the controller ows to the next phase at time t1 once xt x1 t and at time t2 once xt x2 . It follows that for a _ t 0 in phase 1, the initial slip speed given initial slip acceleration x xt 0 needs to be sufciently large to allow the actuator to change the slip acceleration in time, i.e., jxt 0 j Z jx1 t 0 j. On the other hand, phase 2 can be skipped, when the slip acceleration in phase 1 is sufciently, i.e., when jx1 t 0 j r jx2 j. The corresponding slip speed thresholds, as illustrated in Fig. 6(a) and (b), are then given by the predened

The external variable w includes the variables that are not available for the controller design, such as the time-derivatives of _ c , and the primary shaft _ e , clutch torque the engine torque p . The value of w, however, is usually small (O100) acceleration s (O1000) when keeping the compared to the desired slip jerk demanded drive shaft torque d constant during this phase,1 so the external variable can be neglected in the controller design, i.e., wt 0. The main objective for u is to quickly reduce the slip acceleration before the slip speed reaches 2. The slip acceleration trajectory should be a function of the slip speed rather than a function of time to be robust against disturbances in the powertrain. If the feedback-controlled closed loop system is critically damped, the slip speed and slip acceleration exponentially decay to zero without overshoot in the minimum amount of time (Franklin & Emami-Naeini, 2001), as desired. The time constant of this closed loop system should be small for a quick reduction of the slip acceleration, but obviously larger than the time constant of the physical system . A control law that realizes these criteria is given by ut 1 4
2

2 and
34

_ s t : 1 t 2 2v 2 3.7. Controller phase 3

In phase 3, the clutch is immediately closed by applying the maximum pressure on the clutch plates and mechanically locked. As a consequence, the clutch torque is not a control variable anymore. The control law for t Z t 2 is simply the normal situation where the engine tracks its nominal setpoint to provide the demanded drive shaft torque:

e t e t :

35

The end of this phase is decided by the supervisory controller, e.g., when the next gear shift is desired, which is not considered in this work, but for example in Ngo et al. (2012). 3.8. Robustness The modeling uncertainties of the engine and clutch are reected in an offset in the drive shaft torque and uncertain time constants of the actuator dynamics. An offset in the drive shaft torque gives in general no comfort issues, since the driver is able to correct the accelerator pedal position to change the demanded torque. When the engine and clutch have different torque offsets, a change in the drive shaft torque may become noticeable at the moment of clutch engagement (t t2), yet the torque change is usually dominated by the torque dip caused by the discontinuous slip acceleration at the end of the critical phase 2. This torque dip is mainly sensitive to the uncertain time constant of the actuator (), which follows from (27). Assuming that the time constant is unknown, yet xed and bounded A ; , the slip acceleration criterion given by (16) can then be ^ . The reasonguaranteed by estimating the time constant as ^ ing is as follows: if o , the controller assumes a faster actuator dynamics than realizable, so the control action to change the slip acceleration is always too late and the smoothness criterion ^ 4 , the controller assumes cannot be satised. Vice versa, if a slower actuator dynamics than realizable, so the smoothness criterion is easily satised, yet with a longer clutch engagement ^ is necessary to meet the smoothtime. As a result, choosing ness criterion at the cost of a conservative clutch engagement time.

x t :

27

Since 4 0, the closed loop system described by (24), (27), and wt 0 is asymptotically stable and critically damped with a double pole 1=2. The state trajectories are for t A t 1 ; t 2 given by _ 1 x1 t t 1 x1 et t 1 ; xt x _ t x _ 1 x1 t t 1 x _ 1 et t1 ; x 28 29

_ 1 x _ t 1 are the initial states of this phase. where x1 xt 1 and x Given the desired slip speed and slip acceleration at the engage_ 2 2 , the state trajectories ment time t2, i.e., x2 2 o and x (28) and (29) can be used to determine the combination of initial states that realizes these nal states. This combination is found to be   _1 1 x ; 30 x1 1 W z
1 This is not the case when the load torque suddenly changes, e.g., when driving into a curb, yet accurate control for smooth clutch engagement is then of less importance.

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768

65

4. Simulations Simulations are performed to evaluate the performance and robustness of the presented controller design. A set of three simulations is selected to answer three research questions: (i) How does the controller perform for the three driving modes? (ii) What is the effect of the calibration parameter 2 on the clutch engagement performance? (iii) How robust is the controller against variations in the time constant of the actuator?

4.1. Control performance The control performance is evaluated for the three driving modes using the simulation parameters listed in Table 2. Figs. 810 show for each driving mode from top to bottom, respectively, the engine and primary speeds in the powertrain , the slip speed s

Table 2 Simulation and experiment parameters. Parameter 2 2 e t r t 0 e t 4 t 0

a

Power upshifta 2 100 60 60

Power downshift 10 150 60 120

Vehicle launch 2 100 0 80

Unit rad/s rad/s2 Nm Nm

For simulation and experiment.

Fig. 9. Simulation results for the power downshift. Phase 3 is not shown because it occurs after the cross shift, which is not considered in the clutch engagement controller design.

Fig. 8. Simulation results for the power upshift. (Top to bottom) Three powertrain speeds (), slip speed s and thresholds 1 (shown in phase 1) and 2 (shown _ s , engine torque e , clutch torque c , and drive in phase 2), slip acceleration shaft torque d . The dotted vertical lines indicate the start times t0, t1 and t2 of each phase.

together with the slip speed thresholds 1 and 2, the slip _ s , the engine torque e , the clutch torque c , and acceleration the drive shaft torque d . The three dotted vertical lines indicate the time instants t0, t1 and t2 for the start of the each control phase. Recall that cross shifts as shown in Figs. 3 and 4 are not shown in the simulations, so the power upshift simulation starts after the cross shift, whereas the power downshift simulation stops before the cross shift, i.e., when the clutch is ready to be closed in phase 3 afterwards. As can be seen in Fig. 8, the clutch engagement for the power upshift initiates at time t t 0 by reducing the engine torque e in order to reduce the slip speed s (phase 1). At time t t 1 , when the slip speed s equals the threshold 1, the engine torque e is _ s (phase 2). increased in order to reduce the slip acceleration Meanwhile, the clutch torque c is kept constant, so the drive shaft torque d is not interrupted. At time t t 2 , when the slip speed s equals the threshold 2, clutch closure is enforced and the engine torque request e is returned to its initial value (phase 3). _ s s o (only in The setpoints for the slip acceleration phase 2) and the drive shaft torque d are tracked very well, even though the control laws are simplied by neglecting (i) the modeling details as listed in Section 3.4, (ii) the actuator dynamics in phases 1 and 3, and (iii) the term w in phase 2. These simplications cause a small tracking delay (neglected dynamics) and offset (neglected efciency variations and transmission iner_ s at the end of phase tias) of d , and a small deviation from 2 when w is not completely negligible anymore compared to the s . The small torque dip after clutch engagedecreasing slip jerk ment in phase 3 is caused by the discontinuity in slip acceleration, of which the size can be reduced with the calibration parameter 2, at the cost of a longer engagement time.

66

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768

_ c is clearly nonzero in this value of w given (26), where the term transient phase. For the power downshift, this implies a slightly smaller overshoot during the cross shift, whereas for the vehicle launch this implies a smaller torque dip. 4.2. Calibration parameter The effect of the calibration parameter 2 on the clutch engagement performance is evaluated for the power upshift using three different values, i.e., 2 A f 50; 150; 300g rad/s2. Fig. 11 shows from top to bottom, respectively, the slip speed s , the slip _ s , the drive shaft torque d , and the engine torque acceleration

e . As expected, a low absolute value of 2 ( 50 rad/s2) results in a slow, yet very smooth clutch engagement with a limited torque dip. Conversely, a high absolute value ( 300 rad/s2) results in a fast, yet rather harsh clutch engagement, after which the engine needs some time to recover its torque. For a better understanding of this calibration parameter, two performance indices are introduced that quantify, respectively, fast and smooth clutch engagement. The index d maxjd t d j for t 4 t 2 gives a measure of the torque dip that is caused by the smoothness of the clutch engagement. The index t 2 t 2 t n 2 gives a measure of the engagement delay, where t n 2 is dened as the shortest possible clutch engagement time for a given drive shaft torque, i.e., by omitting phase 2. Fig. 12 shows the values of these performance indices on each axis for 2 A f 25; 50; 100; 150; 250; 350g rad=s2 . The curve through these points can be interpreted as a Pareto-optimal front of the two performance
indices: for a given value of 2, or implicitly, for a given torque dip d , the engagement time t 2 is optimal, because the switch from phase 1 to 2 is performed as late as possible, determined by the time constant of the actuator. 4.3. Robustness The time constants of the actuators are crucial control parameters that determine the duration of phase 2. Fig. 13 shows the effect of the uncertainty in the engine time constant for the power upshift. The relative performance indices d = d and t 2 =t n 2 are evaluated for various values of the estimated ^ = A 0:5; 2 and calibration parameter A time constant e e 2 340; 20 rad=s2 . As expected, an underestimated time con^ = o 0) has a large inuence on the smoothness stant (i.e., ln e e of the clutch engagement d , but only a small inuence on the engagement time t 2 . An overestimated time constant on ^ = 4 0) has a large inuence on the the other hand (i.e., ln e e

Fig. 10. Simulation results for the vehicle launch. The slip speed is rst increased to change the operation speed of the engine, which is necessary to deliver sufcient torque.

Fig. 11. Simulation results for the power upshift: the effect of calibration parameter 2 A f 50; 150; 300g rad=s2 on the clutch engagement performance. Fig. 12. Simulation results for the power upshift: the effect of calibration parameter 2 A f 25; 50; 100; 150; 250; 350g rad/s2 on the relative torque dip (d = d ) and the relative engagement delay (t 2 =t n 2 ). Lower values correspond to better performance.

The power downshift in Fig. 9 and vehicle launch in Fig. 10 show qualitatively similar results, except for the smaller slip _ s at the end of phase 2. This is due to the larger acceleration

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768

67

^ e =e on the relative torque dip (left) and the relative Fig. 13. Simulation results for the power upshift: the impact of the (wrongly) estimated engine time constant engagement delay (right), as a function of the calibration parameter 2.

engagement time, but only a small inuence on the smoothness of the clutch engagement. This result conrms the reasoning in ^ to Section 3.8, that it is necessary to choose a large estimate of e meet the smoothness criterion, but at the cost of a conservative clutch engagement time.

5. Experiments Experiments are performed on a test vehicle to validate the control performance and its robustness against realistic sensor noise and modeling errors. The test vehicle is a 2005 Smart Forfour, equipped with a 1.5 l gasoline spark ignition internal combustion engine and a prototype DCT with a customized transmission control unit that is able to communicate with the ECU. For practical reasons, no torque measuring shafts are implemented, so the engine torque estimate (e ) is provided by the ECU and only the setpoints are available for the clutch torques. The demanded drive shaft torque ( d ) is estimated using (8), where the nominal engine torque ( e ) is provided by the ECU. The drive shaft torque (d ) is estimated using (7), where the (estimated) external load (v ) only has a minor impact on the estimated drive shaft torque, since the vehicle is driven on a at road. The overall transmission control architecture for the DCT is relatively complex with interconnected system controllers such as the gear shift strategy and the clutch engagement controller, sub-system controllers for the shift drum and clutches, and safety layers. For the validation of the presented controller design, the clutch engagement controller for the power upshift is implemented and evaluated under on-road driving conditions. A relatively large time ^ 0:16 s) for robustness constant is estimated for the engine ( e against modeling uncertainties. Fig. 14 shows the cross shift and the clutch engagement for a power upshift. As expected, the slip speed trajectory s behaves qualitatively similar as with the simulation. The engine torque e , however, shows a much faster reduction in phase 1, and a delayed increase in phase 2, which can be explained by the relatively simple rst-order modeling of the engine torque dynamics given by (1). The faster torque reduction, however, has no consequences for the clutch engagement controller in phase 1, since the slip acceleration trajectory is of minor interest in this phase. The delayed engine torque increase in phase 2 has only minor consequences on the clutch engagement performance, since the slip acceleration trajectory is determined by the measured slip speed, instead of a prescribed trajectory as a function of time. Despite the large difference in engine torque, the relatively simple engine torque model has proven to be sufciently accurate in

Fig. 14. Experimental results for the cross shift and clutch engagement with a power upshift. Despite the different engine dynamics, the slip acceleration at clutch engagement matches the desired value 2  100 rad=s2 quite accurately, resulting in a fast and smooth clutch engagement that hardly disturbs the drive shaft torque.

2 100 rad=s2 at time t t2, resulting in a continuous drive

shaft torque during the clutch engagement. The demanded drive shaft torque is tracked quite well, where a decrease can be observed during the cross shift (t A 0:4; 0s) due to an increasing gear ratio (average effect of the cross shift) and a decreasing engine torque demand. Overall, the experiment shows a good control performance during both the cross shift and clutch engagement, as well as robustness against realistic modeling uncertainties and measurement noise.

phase 2 to achieve the desired slip acceleration value of

68

K. van Berkel et al. / Control Engineering Practice 22 (2014) 5768 Franklin, G. F., & Emami-Naeini, A. (2001). Feedback control of dynamic systems (4th Edition). Prentice-Hall. Fredriksson, J., & Egardt, B. (2000). Nonlinear control applied to gearshifting in automated manual transmissions. In Proceedings of the 39th IEEE conference on in decision and control (Vol. 1, pp. 444449). Gao, B., Chen, H., Sanada, K., & Hu, Y. (2011). Design of clutch-slip controller for automatic transmission using backstepping. IEEE Transactions on Mechatronics, 16(3), 498508. Garofalo, F., Glielmo, L., Ianelli, L., & Vasca, F. (2001). Smooth engagement of automotive dry clutch. In Proceedings of the 40th IEEE CDC (pp. 529534). Glielmo, L., Iannelli, L., Vacca, V., & Vasca, F. (2006). Gearshift control for automated manual transmissions. IEEE/ASME Transactions on Mechatronics, 11(1), 1726. Guzzella, L., & Onder, C. (2010). Introduction to modeling and control of internal combustion engine systems (2nd Edition). Springer-Verlag. van der Heijden, A. C., Serrarens, A. F. A., Camlibel, M. K., & Nijmeijer, H. (2007). Hybrid optimal control of dry clutch engagement. International Journal of Control, 80, 17171728. Horn, J., Bamberger, J., Michau, P., & Pindl, S. (2003). Flatness-based clutch control for automated manual transmissions. Control Engineering Practice, 11(December (12)), 13531359. Kulkarni, M., Shim, T., & Zhang, Y. (2007). Shift dynamics and control of dual-clutch transmissions. Mechanism and Machine Theory, 42(February (2)), 168182. Lambert, J. H. (1758). Observations variae in mathesin puram. Acta Helvitica, Physico-Mathematico-Anatomico-Botanico-Medica, 3, 128168. Montanari, M., Ronchi, F., Rossi, C., Tilli, A., & Tonielli, A. (2004). Control and performance evaluation of a clutch servo system with hydraulic actuation. Control Engineering Practice, 12(November (11)), 13691379. Ngo, D., Hofman, T., Steinbuch, M., & Serrarens, A. (2012). Optimal control of the gearshift command for hybrid electric vehicles. IEEE Transactions on Vehicular Technology, 61(8), 35313543. Nesch, T., Wang, M., Voser, C., & Guzzella, L. (2012). Optimal energy management and sizing for hybrid electric vehicles considering transient emissions. In Proceedings of the IFAC ECOSM 12, Rueil-Malmaison, France (pp. 211218). Pettersson, M., & Nielsen, L. (2000). Gear shifting by engine control. IEEE Transactions on Control Systems Technology, 8(3), 495507. Serrarens, A. (2001). Coordinated control of the zero inertia powertrain (Ph.D. thesis). Netherlands: Technische Universiteit Eindhoven. Serrarens, A., Dassen, M., & Steinbuch, M. (2004). Simulation and control of an automotive dry clutch. In Proceedings of the American control conference (Vol. 5, pp. 40784083). Boston, Massachusetts, U.S.A. Slicker, J., & Loh, R. (1996). Design of robust vehicle launch control systems. IEEE Transactions on Control System Technology, 4(4), 326335. Tanaka, H., & Wada, H. (1995). Fuzzy control of clutch engagement for automated manual transmission. Vehicle System Dynamics, 24(45), 365376. Wheals, J., Turner, A., Ramsay, K., O'Neil, A., Bennett, J., & Fang, H. (2007). Double clutch transmission (DCT) using multiplexed linear actuation technology and dry clutches for high efciency and low cost. SAE Technical Paper 2007-011096, 18 pp. Zhang, Y., Chen, X., Zhang, X., Jiang, H., & Tobler, W. (2005). Dynamic modeling and simulation of a dual-clutch automated lay-shaft transmission. Journal of Mechanical Design, 127(2), 302307.

6. Conclusions The control problem of fast and smooth clutch engagement is considered for a powertrain equipped with a dual-clutch transmission. This paper contributes with a new, simple and robust controller design with a consistent performance under different initial conditions. The controller is based on the relevant actuator dynamics and the clutch slip dynamics. The clutch slip dynamics is controlled using the standard speed sensors such that the desired slip acceleration is achieved at the time of clutch engagement. The desired slip acceleration is useful as the calibration parameter to choose the trade-off between fast and smooth clutch engagement. Simulations show that the generic controller is effective for three relevant driving modes. A parameter variation study is performed to illustrate how the calibration parameter and the uncertainty in the actuator time constant inuence the clutch engagement performance. For uncertain time constants of the actuator, smooth clutch engagement can still be achieved, but at the cost of a conservative clutch engagement time. Experiments in a test vehicle validate the effectiveness of this relatively simple controller and its robustness against realistic modeling errors and measurement noise. Acknowledgments The authors acknowledge Luc Rmers, Wesley van de Ven, and others at Drivetrain Innovations for their contributions to this paper. References
Bemporad, A., Borrelli, F., Glielmo, L., & Vasca, F. (2001). Hybrid control of dry clutch engagement. In Proceedings of the European control conference (pp. 635639). Porto, Portugal. van Berkel, K., Veldpaus, F., Hofman, T., Vroemen, B., & Steinbuch, M. Fast and smooth clutch engagement control for a mechanical hybrid powertrain, IEEE Transactions on Control Systems Technology. http://dx.doi.org/10.1109/TCST.2013. 2279935, in press. Christen, U. & Busch, R. (2012). The art of control engineering: Science meets industrial reality. In Proceedings of the IFAC ECOSM 12. Rueil-Malmaison, France. Dolcini, J., & Bchart de Wit, H. (2010). Dry clutch control for automotive applications. London, United Kingdom: Springer-Verlag.