Towards a Seamless Development Process for Automotive Engine Control System

download Towards a Seamless Development Process for Automotive Engine Control System

of 10

Transcript of Towards a Seamless Development Process for Automotive Engine Control System

  • 7/27/2019 Towards a Seamless Development Process for Automotive Engine Control System

    1/10

    Control Engineering Practice 12 (2004) 977986

    Towards a seamless development process for automotive

    engine-control system

    Wootaik Lee, Seungbum Park, Myoungho Sunwoo*

    Automotive Control & Electronics Laboratory, Department of Automotive Engineering, Hanyang University, 17 Haengdang-Dong, Seongdong-Gu,

    Seoul 133-791, South Korea

    Received 24 March 2003; accepted 26 November 2003

    Abstract

    This paper describes a new development platform for an automotive engine-control system and introduces a seamless

    development process with a practical example of model-based engine control. The development platform consists of a target-

    identical rapid control prototyping (RCP) system and a PC-based hardware-in-the loop simulation equipment. This RCP system is

    designed very similarly to the real production controller with the help of a customized target package and powerful microcontrollers.

    This RCP system insures rapidity of production as well as prototyping, by adopting a target-identical microprocessor. The resulting

    target identity is important from the viewpoint of practice and application. An organized development environment is provided by

    matching hardware-in-the-loop simulation (HILS) equipment with the RCP system. A control system can be easily tested and

    validated using PC-based HILS that uses commercial-off-the-shelf I/O boards. The development platform supports enhanced

    concurrent engineering, and results in a reduction of development time and cost. To examine the feasibilities of the proposed

    development environment, a model-based air-to-fuel ratio controller based on a sliding mode control scheme is implemented as a

    practical example.

    r 2004 Elsevier Ltd. All rights reserved.

    Keywords: Engine-control systems; Hardware-in-the-loop simulation; Rapid control prototyping

    1. Introduction

    Automotive electronics have been changed dramati-

    cally recently, and new features have been introduced to

    satisfy customer expectations. Moreover, the require-

    ments for existing functions are becoming more

    demanding because of the stricter environmental man-

    datory regulations and safety standards. Engine-control

    tasks, that were classically solved mechanically, are now

    being replaced by electronic control systems, and the

    design and implementation of control algorithms is a

    crucial element in the development of automotive

    engine-control systems. The pressure on the develop-

    ment process of electronic engine-control systems has

    increased rapidly because engineers are expected to

    implement more features in a less time. Ad hoc heuristic

    design and implementation methods are being replaced

    by systematic requirements-driven processes.

    In the last decade, researchers have enhanced the

    development process of the control system efficiently in

    both academia and industry. The requirements of the

    modern development process have been the subject of

    many studies. Isermann (1996) examined the importance

    of a systematic development process and software tools

    for the design. Hanselmann (1998) suggested that the

    modern development process is characterized by com-

    puter-aided support in all stages from specification to

    product. Smith (1999) also proposed that a more

    efficient development process is not intended to change

    the basic steps. Rather, improved software and hard-

    ware tools can make the process more efficient.

    Browne, Bass, Croll, and Fleming (1994) and Hajji

    et al. (1996) proposed a framework of tools which allow

    the design of distributed, potentially fault-tolerant, real-

    time control software. Kimura and Maeda (1996) also

    introduced two development tools for an engine-control

    system. One is the engine and vehicle simulator and the

    other is the control logic simulator substituting for a

    ARTICLE IN PRESS

    *Corresponding author. Tel: +82-2-2290-0453; fax: +82-2-2297-

    5495.

    E-mail addresses: [email protected] (W. Lee),

    [email protected] (S. Park), [email protected]

    (M. Sunwoo).

    0967-0661/$- see front matterr 2004 Elsevier Ltd. All rights reserved.

    doi:10.1016/j.conengprac.2003.11.016

  • 7/27/2019 Towards a Seamless Development Process for Automotive Engine Control System

    2/10

    part of the engine-control logic in production CPUs.

    Butts (1996) benchmarked many computer-aided con-

    trol system design (CACSD) tools and computer-aided

    software/systems engineering (CASE) tools to be

    adopted in the modern development process. Tradi-

    tional CASE and CACSD tools were integrated to

    improve hybrid systems development support for theautomotive power train control community, and CASE/

    CACSD products were applied to a simplified power

    train system model to investigate the feasibilities.

    Toeppe, Bostic, Ranville, and Rzemien (1999) per-

    formed a trial evaluation of commercially available

    CACSD tools, and Howold and Jupfer (2000) compared

    classical CASE-tools approach and tools for automated

    code generation.

    Some companies offer sophisticated development

    environments for automotive applications (Hansel-

    mann, 1998; Leharth, Baum, Beck, Werther, & Zur-

    awka, 1998). They support the specification of

    embedded systems, and verify and validate it on

    different abstraction levels. They also generate a code

    automatically for rapid control prototyping (RCP)

    controllers, and provide a hardware-in-the-loop simula-

    tion (HILS) experimental environment.

    This paper presents a new integrated platform-based

    development environment for automotive engine-con-

    trol system. The control algorithms can be developed by

    using the MATLABs/SIMULINKs off-line simulation

    environment. This control design can be easily imple-

    mented in the proposed RCP platform through the

    REAL-TIME WORKSHOPs and the customized

    target package. The generated execution codes areeffectively tested and verified by using PC-based HILS.

    The proposed platform-based environment enables the

    developers to design control laws, to generate executable

    code, and to test the control system in a unified way, and

    make the development process more seamless.

    The modern development process is introduced and

    the way to enhance the development process is proposed

    in the following section. The features of a newly

    proposed development platform are explained in detail

    in the next section. The results from the pilot project of

    an air-to-fuel ratio (AFR) control using the developed

    platform are presented to illustrate the feasibilities of the

    proposed development environment.

    2. Development process

    The major characteristics of the traditional develop-

    ment process are described as follows: (i) conventional

    textual specifications, (ii) a lack of comprehensive tools,

    (iii) a sequential approach. The inefficient traditional

    development process is being replaced with a modern

    development process, which is characterized by an

    integrated computer tool chain in all stages from the

    specification to the final product.

    2.1. Conventional development process V-model

    The modern development process can replace the

    expensive prototypes with the appropriate alternativesand provide an economical virtual test environment to

    minimize the expensive and time-consuming experi-

    ments on the test bench or in-vehicle. Furthermore,

    the modern development process eliminates an error-

    prone hand-coding process.

    Sivashankar and Butts (1999) and Smith (1999)

    addressed the modern development process in the form

    of a simplified V-model (see Fig. 1), and summarized the

    requirements of each development phase, respectively.

    On the left downward path, development becomes more

    detailed and concrete, eventually leading to the compo-

    nent. The right upward path leads to the final

    production electronic control unit (ECU), which is then

    compared to the original ideas, objectives and specifica-

    tions.

    At the beginning of a new project, the overall system

    functionalities are specified. After the initial concept

    phase, the requirements of each subsystem are defined

    according to the project characteristics. These require-

    ments are analyzed and the core functions and features

    of each subsystem are specified.

    Control algorithms are primarily designed in response

    to the pre-defined specifications, and they are also made

    robust with respect to extreme and abnormal operating

    conditions. A number of architectures and softwaredesign requirements are incorporated. A modeling work

    of both the physical plant and the control system occurs

    in this stage. This is the most mature stage in the

    development process. Advanced modeling tools are well

    established and offer powerful graphical user interface

    and modeling capabilities. These tools can be smoothly

    integrated into the development process.

    The control algorithm is implemented in the form of

    source codes through a traditional hand-coding or an

    auto code-generation method, and is validated with its

    ARTICLE IN PRESS

    Fig. 1. Modern development process V-model.

    W. Lee et al. / Control Engineering Practice 12 (2004) 977986978

  • 7/27/2019 Towards a Seamless Development Process for Automotive Engine Control System

    3/10

    requirements, using a software-in-the-loop simulation

    (SILS) method or an RCP method. As in the case of the

    modeling and simulation phase, both the SILS and RCP

    environment are used to represent the actual system as

    much as possible. RCP including an automatic code

    generator is the key step for the modern development

    process. With the help of the RCP environment, thecontrol engineers can easily validate and verify their

    own controller design in the vehicle or in the special test-

    benches, without the intervention of software or hard-

    ware specialists.

    The next step is the modification of the source codes

    to an appropriate form for the production target

    processor. Converting the control algorithms to the

    production level codes is a very time-consuming and

    tedious phase.

    After the functions are integrated in the production

    target, they should be tested against the above-

    mentioned functional requirements. Integrated control

    systems can be tested and verified using the specially

    designed test benches or real vehicles. As an alternative,

    a real-time model-based testing procedure, which is

    called HILS, may be used. HILS is characterized by the

    operation of real components in connection with real-

    time simulated components. Usually, the control system

    hardware and software is the real system, and the

    controlled process can be either fully or partially

    simulated.

    Final validation and testing will always be required in

    the completed system. At the end of the design process,

    the control system is finally calibrated for the specific

    application using the HILS and the in-vehicle testprocedures. This phase has shown the greatest advance

    in the development process.

    2.2. Proposed V-model for engine-control system

    This RCP equipment enables the control engineers to

    verify their own control algorithm in an ideal environ-

    ment using an automatic code-generation technique. In

    the target implementation phase, this control algorithm

    is coded and linked with other parts of the software by

    the software engineers. This final software is validated

    against the pre-defined requirements or specifications to

    check the feasibilities of the target implementation in the

    HILS phase. The described V-model can be enhanced

    more seamlessly in an application-specific case. A

    development platform, which is composed of a target-

    identical RCP (Lee, Shin, & Sunwoo, under submission)

    and a convenient HILS (Lee, Yoon, & Sunwoo, 2003),

    alleviates some technical difficulties in transition from

    one phase to another phase (see Fig. 2).

    As shown in Fig. 1, the conventional V-model

    separates the RCP phase from the target-implementa-

    tion phase due to the insufficient computing power of a

    target microprocessor and the difference of the abstrac-

    tion levels of CACSD tools. The computing power

    difference can be alleviated by the recent advent of

    powerful floating-point microcontrollers, which are

    currently used in high-end engine-control applications.

    Furthermore, the difference in the abstraction levels can

    also be overcome by some CACSD tools, which have an

    automatic code-generation function and provide a

    flexible environment for integrating source codes. The

    powerful microcontrollers and these CACSD tools

    make a target-identical RCP viable in the engine-control

    application in spite of its immaturity. The target-

    identical RCP also makes control validation easier. This

    conventional V-model has two different phases to

    validate the control algorithms. One is the control

    verification phase, in which only the control algorithm

    can be tested by use of an RCP equipment, and the other

    is the control validation phase, in which the whole

    control system may be tested, including control algo-

    rithm and software implementation, by use of a target

    ECU. The target-identical RCP can constitute thefinal control system as easily as the RCP in the context

    of Fig. 1. Therefore, it makes the former validation

    phase redundant, and renders the later validation phase

    as easy as the former one.

    The conventional V-model enables software engineers

    to code the application part after the control algorithm

    is verified in the RCP phase. Hand coding of the control

    algorithm makes it difficult to shorten the development

    time. The automatic code-generation technique, which is

    widely used in RCP and HILS, has been developed to

    the production-quality level and this technique can be

    applied to convert the control algorithm to the execu-

    tion code. In spite of the maturity of the automatic code-

    generation technique, it is very difficult to generalize the

    code-generation procedure of the low layer codes,

    because of the hardware dependencies and efficiency of

    the codes. The easiness of automatic code generation

    can be compromised with the efficiency of hand coding

    in the aspect of the layered architecture. The application

    part of the software is converted to the codes by the

    automatic code-generation technique, and the low layer

    codes are programmed by software experts on a trial

    and error basis. This compromise makes the control

    design and the software design perform concurrently.

    ARTICLE IN PRESS

    Fig. 2. Proposed V-model for engine-control system.

    W. Lee et al. / Control Engineering Practice 12 (2004) 977986 979

  • 7/27/2019 Towards a Seamless Development Process for Automotive Engine Control System

    4/10

  • 7/27/2019 Towards a Seamless Development Process for Automotive Engine Control System

    5/10

    executed in real-time. The Mororola MPC555s is

    selected as the target processor. The HILS equipment

    is composed of a host computer and a target computer,

    which have some data acquisition cards to interface the

    physical signals. Fig. 4 shows the block diagram of the

    proposed development platform. Although this figure

    depicts the RCP host, HILS host, and measurement and

    calibration hosts separately, all the programs of the host

    computers are executed concurrently in a single host

    computer for the developers convenience.

    Fig. 5 shows the layered architecture of the proposed

    development platform. This RCP platform is designed

    for target implementation as well as RCP. To achieve

    this goal, layered architecture is strictly maintained and

    the RCP platform is designed similarly to the produc-

    tion controller. In order to utilize all the features of the

    MPC555s, hardware abstraction layer (HAL) is de-

    signed as similarly to the target codes as possible. The

    middle layer is carefully organized and designed for the

    TARGET LANGUAGE COMPILERs realizations.

    By use of the implemented scheduler, any task of the

    application program can be easily activated in accor-

    dance with the pre-defined condition, such as a certain

    event or time period. A customized toolbox, called

    engine-control toolbox, is developed to interface an

    application program, e.g. the control algorithm, with the

    lower layer. The frequently used functions are more

    abstracted as in a block library form (see Fig. 6). All theblocks are categorized as initialization, I/O functions,

    and scheduling functions. The I/O signals of the ECU,

    which generally interact with the engine sensors and

    actuators at the off-line simulation phase, are replaced

    with the appropriate sink and source blocks, such as

    rpm read, and update fuel duration. The execution of the

    control algorithms can be configured by use of the

    trigger blocks and the task subsystem.

    To develop a new engine-control system, highly

    sophisticated software, calibration, measurement and

    diagnosis equipment has to be used. It would be

    beneficial to use the same tool for the calibration and

    rapid prototyping. CAN calibration protocol (CCP) is

    widely used in the automotive industry, and is becoming

    a standard tool. Thus, the CCP is selected as a

    measurement and calibration tool in this study.

    The HILS environment requires signals from the

    computational platform to be interfaced with the

    ARTICLE IN PRESS

    Fig. 4. Hardware block diagram of the development platform.

    Fig. 5. Layered architecture of the development platform.

    Fig. 6. SIMULINKs toolbox of rapid prototyping platform.

    W. Lee et al. / Control Engineering Practice 12 (2004) 977986 981

  • 7/27/2019 Towards a Seamless Development Process for Automotive Engine Control System

    6/10

    hardware. In the layered architecture perspective, the

    hardware layer can be divided into two categories. One

    is the logic-level I/O layer in which the microcontroller

    directly interfaces, and the other is the real I/O layer

    containing signal conditioning subsystems, power am-

    plifying subsystems, and other subsystems. During the

    design phase, one must determine the proper portioningof signal processing into available software and hard-

    ware. This results in a trade-off among system complex-

    ity, computational burden, and maintainability. Though

    the lowest hardware interface layer is desired to emulate

    more realistic environment, this layer can be by-passed

    for the purpose of economy and convenience. If the

    logic-level signals, such as the ignition-triggering signal,

    can be used instead of real I/O signals, these complicated

    devices can be eliminated. In this study, the logic-level I/

    O is used and major parts of the by-passed layer are

    modeled. For seamless integration of the development

    process, xPC targets of REAL-TIME WORKSHOPs

    is used in the HILS platform.

    4. Engine-control experiments

    An engine-control experiment is performed as a pilot

    project, to prove the feasibility of the proposed

    development process. Two air-to-fuel ratio (AFR)

    control algorithms are evaluated in this experiment.

    One employs a relatively simple PI control law with a

    feed-forward compensation function, and the other

    employs a sliding mode fuel injection control law with

    a Smith predictor (Yoon, Park, Lee, & Sunwoo, 2001).

    4.1. Off-line simulation model

    The off-line simulation model mainly consists of anengine model and a controller, as shown in Fig. 7. In this

    stage, all SIMULINKs library blocks can be used free

    of implementation problems, and a variable-time step

    ordinary differential equation solver can be also used for

    simulation speed and accuracy. However, it is difficult to

    model event-based or crank angle-based controller

    behaviors because SIMULINKs basically provides a

    time-based simulation method. To model event-based

    behavior in a time-based simulation environment,

    special simulation mechanisms must be designed, but

    this may cause significant overhead in designing and

    simulating the model. Therefore, the event-based

    behaviors are generally simplified in an off-line simula-

    tion phase by minimizing a sampling period to an

    affordable value. To increase the accuracy of the

    simulation, multi-rate simulation methods are adopted.

    Mean value engine models are widely used in

    developing model-based engine controllers. Addition-

    ally, these engine models are generally appropriate for

    the HILS platform. The nonlinear dynamic engine

    model, which was introduced in the previous study

    (Yoon & Sunwoo, 2001), is used for designing a fuel-

    injection controller and HILS.

    ARTICLE IN PRESS

    Fig. 7. Off-line simulation model using SIMULINKs.

    W. Lee et al. / Control Engineering Practice 12 (2004) 977986982

  • 7/27/2019 Towards a Seamless Development Process for Automotive Engine Control System

    7/10

    The controllers are expressed in relatively complicated

    differential and algebraic equations. Some equations are

    related with engine models, and they are already

    expressed in block diagram representation of the engine

    model. These are imported from the previous engine

    model blocks and other equations are expressed in

    combination of SIMULINKs

    blocks and user-definedS-functions.

    4.2. Engine model and HILS platform

    As shown in Fig. 8, the engine model should be

    supplemented with analog I/O and timing I/O signal

    modules to interface actual signals in real-time. Engine

    I/O signals can be categorized into two groups. One

    group is composed of analog signals, and the other is

    composed of digital signals, including timing signals.

    Synchronization of timing I/O signals, such as crank

    shaft signal, cam shaft signal, and spark and injectionsignals, with the engine event make the interface more

    difficult. Sometimes, these interfaces are achieved

    through sophisticated I/O boards with dedicated pro-

    cessors. With the help of the powerful computing power

    of a PC and efficient kernel of xPC targets, these

    synchronized timing I/O signals are interfaced through

    the digital I/O and the timing I/O subsystems.

    4.3. Control algorithm and RCP platform

    Since the engine model can be easily modified for the

    HILS platform, the controller can also be altered into an

    RCP form with the developed engine-control toolbox

    (see Fig. 9).

    The major changes are substitution of I/O blocks forthe I/O signals in the off-line simulation. Each I/O block

    calls appropriate functions, which are already designed

    and implemented in the lower layer. Some blocks and

    modules should be replaced with more simple ones

    because the generated code should be compact for an

    embedded microcontroller and it should be executed in a

    timely manner for real-time control. In this case, the

    controller is configured to generate a stroke-based event

    trigger signal in every engine stroke, to execute the

    control algorithm at the same rate.

    4.4. Closed-loop experiments

    An off-line simulation model is tested under the

    proposed development platform (see Fig. 10). The

    engine model is executed on the xPC target desktop

    PC, and the engine-control algorithm is run on the

    target processor. As described previously, all I/O signals

    are connected at the logic level.

    ARTICLE IN PRESS

    Fig. 8. Engine HILS model.

    W. Lee et al. / Control Engineering Practice 12 (2004) 977986 983

  • 7/27/2019 Towards a Seamless Development Process for Automotive Engine Control System

    8/10

    The results of the off-line simulation and a virtual

    experiment using the proposed platform are depicted

    and compared in Figs. 11 and 12. In Fig. 11, the PI AFR

    control algorithm is tested in the time-based manner,

    and the synchronization of the engine is achieved by a

    60-2 type crankshaft sensor. On the other hand, in Fig.

    12, the sliding mode AFR control law is executed in the

    event-based manner, and it is assumed that a 36-1 type

    crankshaft sensor is mounted on the engine.

    For performance evaluation of the proposed con-

    troller, the throttle angle is changed as shown in Figs.

    11(a) and 12(a), to simulate a fast tip-in and tip-out

    situation that allows the engine to be operated abruptly

    between 2000 and 4000 rpm. In addition, the engine

    is assumed to be operated under a constant load

    condition. The AFR sensor is assumed to have a

    measurement delay of two engine cycles because of the

    event-based nature of the engine, and it is assumed to

    have band-limited white noise.

    Figs. 11(c) and 12(c) represent the off-line simulation

    results of the models. The off-line simulation model is

    also tested under the proposed development platform

    and the results are shown in Fig. 11(d) and 12(d).

    Compared with the off-line simulation, the performance

    of the experiment using the development platform shows

    some degradation. A non-zero execution time of the

    plant model and the control algorithm, measurement

    noise, quantization, and other factors in the experiment

    cause this degradation. The different execution period of

    the control task also degrades the control performance.

    These problems, which may occur in the target

    implementation stage and the validation stage, can be

    efficiently handled with the help of this virtual develop-

    ment environment.

    ARTICLE IN PRESS

    Fig. 9. Engine controller (using an engine-control toolbox).

    Fig. 10. Photograph of the proposed development platform.

    W. Lee et al. / Control Engineering Practice 12 (2004) 977986984

  • 7/27/2019 Towards a Seamless Development Process for Automotive Engine Control System

    9/10

    5. Conclusions

    A new development platform for automotive engine-

    control systems is introduced.

    MATLABs/SIMULINKs/REAL-TIME WORK-

    SHOPs tool chain is used as a base environment for

    implementation and evaluation of the engine controller

    as well as the development of the control algorithm.

    This platform provides a target-identical RCP platform

    and PC-based HILS environment. With the help of a

    customized target package and the advent of powerful

    microcontollers, the RCP system is organized very

    similarly to the real production ECU. This feature

    alleviates many implementation problems, which may

    occur between the RCP system and the production

    system. The control system is easily investigated and

    validated using the PC-based HILS system. This system

    uses xPC targets with commercially available off-the-

    shelf I/O boards and logic-level signals for connection

    with the controller. This platform-based development

    process enables the developers to design control laws, to

    generate executable codes, and to evaluate the control

    system in a unified manner.

    In order to prove the feasibilities of the proposed

    environment, a pilot project for the development of an

    air-to-fuel control system is performed, and the simula-

    tion results are presented. The simulation results show

    that the proposed development process and the virtual

    experiment environment can efficiently handle various

    ECU design problems caused by transitions among

    separate development steps. The proposed environment

    can be a basis for the model-based approach in engine-

    control application.

    References

    Browne, A. R., Bass, J. M., Croll, P. R., & Fleming, P. J. (1994). A

    prototype framework of design tools for computer-aided control

    engineering. IEEE/IFAC joint symposium on computer-aided control

    system design, Tucson, AZ (pp. 369374).

    Butts, K. R. (1996). An application of integrated CASE/CACSD

    to automotive powertrain systems. Proceedings of the 1996

    ARTICLE IN PRESS

    Fig. 11. Comparison of the off-line simulation and the experiment

    using development platform (PI control system, 60-2 type crank signal,

    time-based execution).

    Fig. 12. Comparison of the off-line simulation and the experiment

    using development platform (sliding mode control system, 36-1 typecrank signal, event-based execution).

    W. Lee et al. / Control Engineering Practice 12 (2004) 977986 985

  • 7/27/2019 Towards a Seamless Development Process for Automotive Engine Control System

    10/10

    IEEE international symposium on CACSD (pp. 339345),

    Dearborn, MI.

    Hajji, M. S., Bass, J. M., Browne, A. R., Schroder, P., Croll, P. R., &

    Fleming, P. J. (1996). The development framework: Work in

    progress towards a real-time control system design environment.

    IEE Colloquium on Advances in Computer-Aided Control System

    Design, London, UK (pp. 4/14/3).

    Hanselmann, H. (1998). Development speed-up for electronic controlsystems. Convergence 98, Dearborn, USA.

    Howold, C., & Jupfer, R. (2000). Integration of new methods and tools

    for automotive control unit developmenta project management

    task. SAE2000-01-0392.

    Isermann, R. (1996). On the design and control of mechatronic

    systemsa survey. IEEE Transactions on Industrial Electronics,

    43(1), 415.

    Kimura, A., & Maeda, I. (1996). Development of engine control system

    using real time simulator. Proceedings of the 1996 IEEE Interna-

    tional Symposium on CACSD, Michigan, USA (pp. 157163).

    Lee, W., Shin, M., & Sunwoo, M. (under submission). Target-identical

    rapid control prototyping platform for model-based engine

    control. Proceedings of the Institution of Mechanical Engineers,

    Part D. Journal of Automobile Engineering.

    Lee, W., Yoon, M., & Sunwoo, M. (2003). A cost- and time-effectivehardware-in-the-loop simulation platform for automotive engine

    control systems. Proceedings of the Institution of Mechanical

    Engineers, Part D. Journal of Automobile Engineering, 217, 4152.

    Leharth, U., Baum, U., Beck, T., Werther, K., & Zurawka, T. (1998).

    An integrated approach to rapid product development for

    embedded automotive control systems. Control Engineering Prac-

    tice, 6, 529540.

    Sivashankar, N., & Butts, K. (1999). A modeling environment for

    production powertrain controller development. Proceedings of the1999 IEEE international symposium on CACSD (pp. 563568),

    Hawaii, USA.

    Smith, M.H. (1999). Towards a more efficient approach to automotive

    embedded control system development. Proceedings of the 1999

    IEEE international symposium on CACSD (pp. 219224), Hawaii,

    USA.

    Toeppe, S., Bostic, D., Ranville, S., & Rzemien, K. (1999). Automatic

    code generation requirements for production automotive power-

    train applications. Proceedings of the 1999 IEEE International

    Symposium on CACSD (pp. 200206), Hawaii, USA.

    Yoon, P., Park, S., Lee, W., & Sunwoo, M. (2001). Robust nonlinear

    control of air-to-fuel ratio in spark ignition engines. KSME

    International Journal, 15(6), 699708.

    Yoon, P., & Sunwoo, M. (2001). A nonlinear dynamic modeling of SI

    engines for controller design. International Journal of VehicleDesign, 26(2/3), 277297.

    ARTICLE IN PRESS

    W. Lee et al. / Control Engineering Practice 12 (2004) 977986986