Hybrid Wired-Wireless Networks for Real-Time Communications

Embed Size (px)

Citation preview

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    1/13

     Incorporating the Best of Both Worlds for

     Improved Functionality in Industrial Environments

    Thanks to their pecu-

    liar features, wireless

    networks are currently

    becoming more and

    more attractive for use

    in industrial automa-

    tion and manufactur-

    ing scenarios. However, they cannotbe thought of as a total replacement

    for more traditional wired networks

    in such environments, at least in the

    short/mid term. This means that for

    the time being it is worth focusing on

    their integration with the existing in-

    dustrial wired solutions, in order to

    achieve enhanced flexibility, efficien-

    cy, and performance for the overall

    networked system.

    In this article, some considerations

    are presented about the way sev-eral well-known industrial networks

    (based on both fieldbus and industrial

    Ethernet solutions) can be practically

    extended with wireless subnetworks

    that rely on popular technologies,

    such as IEEE 802.11 and 802.15.4. This

    results in hybrid networks, which are

    able to combine the advantages of

    both wired and wireless solutions. Inparticular, advantages and drawbacks

    of several interconnection techniques

    are highlighted and, depending on the

    wired networks specifically taken into

    account, some hybrid configurations

    that are able to cope in a satisfactory

    way with the tight timing require-

    ments often imposed by industrial

    control systems are suggested.

    Overview 

    Recent years have witnessed the everincreasing adoption of digital com-

    munication networks in manufactur-

    ing environments. Field networks

    (also known as fieldbuses) [1] are

    certainly the most popular solution in

    those scenarios. For a long time they

    have been used at the lowest levels

    of factory automation systems (shop

    floor), which are often characterized

    by severe timing requirements. On

    the other hand, Ethernet networks,

    which were traditionally employed as

    factory backbones in order to handleinformation flows between different

    8  IEEE INDUSTRIAL ELECTRONICS MAGAZINE n  MARCH 2008  Digital O bject Identifie r 10.1109/MIE.2008.917155 1932-4529/08/$25.00©2008IEEE

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    2/13

    MARCH 2008  n  IEEE INDUSTRIAL ELECTRONICS MAGAZINE 9

    GIANLUCA CENA,ADRIANO VALENZANO,AND STEFANO VITTURI

    plant areas, are now deemed suitable

    also for control applications down to

    the field level. This is mostly due to

    enhancements of the network perfor-

    mance, which led to the development

    of industrial Ethernet protocols pur-

    posely designed to provide real-time

    communications [2]. In many cases,

    variants of such protocols have been

    defined that are able to support iso-

    chronous communications as well.

    However, the scenario mentionedabove is quickly evolving, and other

    types of communication technolo-

    gies, namely wireless networks, are

    currently widely available at reason-

    able prices. Wireless communica-

    tions technologies are becoming more

    and more pervasive everyday, and

    they are changing deeply several as-

    pects of human life by enabling new

    perspectives and opportunities that

    could hardly even be thought of only

    a few years ago. This is likely going to

    become true for industrial and factoryenvironments, too, where highly auto-

    mated production systems can get sig-

    nificant benefits from the introduction

    of the most advanced wireless com-

    munication techniques. Indeed, sev-eral studies have recently proved the

    suitability of some of the most popular

    wireless solutions for employment in

    industrial scenarios as well, including

    real-time communications [3].

    Adopting wireless communications

    in industrial environments is particu-

    larly attractive as it allows, in princi-

    ple at least, the avoidance of cabling,

    which in many applications turns out

    to be cumbersome and/or expensive.

    However, while several standard solu-

    tions and components are commonly

    available for wired industrial com-

    munications, wireless systems are far

    from being considered well settled for

    such kinds of applications. In fact, it is

    worth pointing out that the straightfor-

    ward introduction of highly innovative

    solutions in industrial and manufac-

    turing systems has never been either

    simple or fast. This is due to several

    reasons, such as reliability, efficien-

    cy, safety, cost, and security, just to

    mention a few. As a consequence, the

    growth of wireless technologies in in-

    dustrial applications is expected to be

    noticeably slower than in other areas,

    even though they are envisaged to play

    a crucial role in many next-generation

    industrial and production systems.

    A point most experts agree on is

    that wireless communications are un-

    likely to be able to replace completely

    the current wired solutions adoptedin many industrial scenarios, at least

     © PHOTODISC & ARTVILLE

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    3/13

    10  IEEE INDUSTRIAL ELECTRONICS MAGAZINE n  MARCH 2008

    in the mid term. Indeed, it is worth

    noting that in many practical appli-

    cations, this is not considered a real

    need. Actually, in factory automation

    systems there is often the need to

    connect a few components to an al-

    ready deployed wired communication

    system that cannot be reached (eas-

    ily and/or reliably) with a cable. The

    most common example is a sensor

    mounted on a moving/rotating device

    that has to be connected to a control-ler located on a wired segment.

    This kind of problem may be eas-

    ily solved by providing wireless ex-

    tensions to existing wired systems.

    Resulting configurations are hybrid

    networks in which, typically, wireless

    segments have limited geographic

    extension (some tens of meters) and

    connect only a few stations. Moreover,

    controllers are usually located on the

    wired segment in such networks. This

    implies that wireless (sub)networks

    will have to coexist with more tradi-

    tional wired communication systems

    for quite a long time. Thus, the inte-

    gration of wireless and wired com-

    munications in industrial scenarios

    is becoming a crucial issue that has

    to be dealt with carefully in order to

    achieve efficiency and performance.

    This article focuses on the most

    relevant aspects concerning the

    aforementioned hybrid networks. In

    particular, after a general analysis

    of hybrid wired/wireless systems,

    the ways wireless extensions of both

    fieldbuses and real-time Ethernet

    (RTE) networks can be effectively

    implemented are taken into account.

    Then, examples will be provided of

    extensions based on some popular

    networks. Specifically, we refer to the

    well-known Profibus DP [4] and Devi-

    ceNet [5] fieldbus networks, as well

    as the Ethernet/IP [6] and Profinet IO[7] RTE networks. (In the latest ver-

    sion of the Profinet specification, the

    acronym IO has been removed. None-

    theless, for practical reasons, in this

    article we maintain the original nota-

    tion.) It is worth noting that many oth-

    er interesting solutions actually exist

    that could be provided with a wireless

    extension. However, as it would have

    been practically unfeasible to deal

    with all of them, we decided to focus

    on a reduced set of networks that are

    nevertheless representative of a widerrange of industrial solutions.

    On the other hand, we consider

    both the IEEE 802.11 wireless lo-

    cal area network (WLAN) [8] and

    the IEEE 802.15.4 low-rate wireless

    personal area network (LR-WPAN)

    [9] as possible wireless extensions.

    Indeed, 802.11 WLANs have been al -

    ready considered in several studies

    and tested in practical applications

    that demonstrated their suitability

    for industrial use [10]–[12]. For such

    a reason, most of the examples pro-

    vided in this article are based on

    that solution. However, it has to be

    noted that 802.15.4 LR-WPAN is an

    emerging standard for short-range

    connectivity, characterized by long

    battery lifetime and low cost. Such

    features allow one to implement, for

    example, sensor networks distribut-

    ed over wide geographical areas and

    the connection of sensors/actuators

    located on mobile equipment.

    As a further possible network

    candidate for implementing wireless

    extensions, it is worth mentioning

    Bluetooth [13]. This network, which

    was originally designed as a mere

    cable replacement system, has pro-

    gressively gained widespread use

    thanks to its features, and it is now

    employed in several areas, including

    industrial automation. For example,

    in [14] a real -time wireless sensor/ac-tuator system based on Bluetooth is

    described. That system, which uses a

    modified version of the original me-

    dia access control (MAC) protocol,

    has been implemented and its com-

    ponents are currently available as

    commercial products. However, de-

    spite its interesting features, due to

    the same practical reasons outlined

    above for the wired networks (it is ac-tually impossible to consider all the

    available solutions here), we are not

    discussing Bluetooth in this article.

    Relevant Industrial NetworksThe basic operating principles of the

    networks we are going to consider in

    this article are briefly discussed in the

    following sections.

    Profibus DP and Profinet IO

    Profibus and Profinet are the maincommunication solutions foreseen by

    the Profibus and Profinet International

    (PI) community for industrial environ-

    ments, each of which is, in turn, made

    up of several profiles.

    Profibus DP is a widespread field-

    bus based on a master-slave ap-

    proach, implementing data exchange

    between controllers (masters) and

    field devices such as sensors and

    actuators (slaves) at the maximum

    speed of 12 Mb/s. (Actually, the Pro-

    fibus DP standard specifies a mas-

    ter-master communication as well.

    Nonetheless, this is typically meant

    for the nonreal-time exchange of di-

    agnostic and/or parameterization

    data. As such, this type of communi-

    cation will not be further considered

    in this article.) At the data-link layer,

    the medium access is granted exclu-

    sively to master stations via a token

    passing scheme. It is worth mention-

    ing, however, that most practical con-

    figurations make use of a single mas-

    ter (monomaster networks). Slave

    devices can only respond when que-

    ried by a master. The operation of the

    network is based on a polling cycle of

    slaves that is repeated continuously

    by the master. At each query of a de-

    vice, the master provides the slave

    with output data and, at the same

    time, fetches the input data. Suitable

    techniques are also available for thetransmission of acyclic data.

    Another issue that should be taken into account

     when interconnecting industrial Ethernet networks

    and WLANs concerns the mechanisms for broadcast/ 

    multicast traffic confinement.

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    4/13

    MARCH 2008  n  IEEE INDUSTRIAL ELECTRONICS MAGAZINE 11

    Profinet IO, instead, is an RTE net-

    work included in Part 2 of the IEC 61784

    International Standard [15]. In par-

    ticular, the communication profile CP

    3/6 of such a standard refers to both

    Profinet IO RT_Class 2 and 3 specifi-

    cations, whereas the communication

    profiles CP 3/4 and CP 3/5 both are

    concerned with Profinet IO RT_Class1. As specified by the standard, some

    different types of stations may be con-

    nected to the network, with the most

    significant of them being I/O control-

    lers (e.g., PLCs or industrial PCs) and

    I/O devices (sensors and actuators).

    Basically, Profinet IO makes use

    of a time division multiple access

    (TDMA) technique to guarantee fast

    and precise access to the transmis-

    sion medium. In practice, the traffic

    scheduling takes place according tothe cycle shown in Figure 1. The cycle

    is divided into four phases. In the first

    one (RED) only RT_CLASS 3 traffic

    (the most critical) can take place and

    over predefined physical paths en-

    forced by a special kind of switches

    purposely developed for Profinet IO.

    In the ORANGE phase, only RT_CLASS

    2 frames are exchanged. In this case,

    however, physical paths are not de-

    fined. The GREEN phase is reserved

    to both RT_CLASS 2 and RT_CLASS 1

    frames, either cyclic or acyclic. In this

    phase (which is the only mandatory

    one) access to the physical medium is

    regulated by the priority assigned to

    the frames, according to IEEE 802.1Q

    [16]. Finally, the YELLOW phase is

    used for nonreal-time traffic.

    Similarly to other RTE networks de-

    fined by the IEC 61784 standard, the ap-

    plication layer protocol of Profinet IO

    is based on the definition of communi-

    cation objects that can be exchanged

    over the network via communication

    relationships established between I/O

    controllers and I/O devices.

    DeviceNet and Ethernet/IP 

    DeviceNet and Ethernet/IP (together

    with ControlNet and CompoNet) rely

    on the well-assessed common indus-

    trial protocol (CIP) [17]—formerly

    known as the control and informa-

    tion protocol—which enables a highdegree of interoperability among

    these kinds of networks. Figure 2 de-

    picts the protocol stack foreseen by

    CIP. Seamless bridging and routing

    is supported natively by this proto-

    col, which enables parameterization

    information and process data to be

    easily exchanged in heterogeneous

    distributed systems that employ

    both fieldbus and industrial Ether-net transmission technologies.

    DeviceNet adopts the control-

    ler area network (CAN) protocol at

    the lower layers of its protocol stack

    (physical and data link). Despite the

    bus topology and the relatively low

    speed (up to 500 kb/s), it is able to

    guarantee deterministic behavior

    thanks to the bitwise arbitration tech-

    nique featured by CAN. This protocol,

    in fact, implements a nonpreemp-

    tive priority-based communicationsystem, which does not suffer from

    destructive collisions and where the

    message with the highest priority is

    always given precedence. This also

    means that network congestion is im-

    plicitly prevented.

    Instead, Ethernet/IP is conceived

    to rely on conventional Ethernet

    transmissions. Even though strictly

    deterministic operations are not

    granted, the Open DeviceNet Vendor

    Association (ODVA) (the organization

    that is in charge of managing CIP) has

    established a set of design rules [18]

    that achieve quasi real-time behavior

    in Ethernet/IP. Basically, they require

    that switched high-speed Ethernet

    is adopted (i.e., full duplex 100-Mb/s

    equipment) and that unnecessary

    network traffic is kept as low as pos-sible (e.g., by means of virtual LANs

    (VLANs) to confine multicast traffic

    generated as a consequence of pro-

    cess data exchanged over I/O connec-

    tions according to the producer/con-

    sumer model). As long as the load

    on the network is (fairly) below the

    theoretical available bandwidth, non-

    blocking switches guarantee that any

    message is eventually delivered to the

    intended destination within a period

    of time that, because of the high bitrate, is short enough for most control

    applications (at least from a practical

    point of view).

    IEEE 802.11

    The IEEE 802.11 family includes sever-

    al standard specifications for wireless

    local area networking. All devices be-

    longing to such a family (also referred

    to as WiFi) work in specific bands of

    the radio spectrum, centered around

    FIGURE 2 – Protocol stack of CIP-based networks.

    TCP/UDP

    IP

    Ethernet

    IEEE 802.3

    WiFi

    IEEE 802.11

    DeviceNetConnectionManager

    Encapsulation Protocol

    CAN

    ISO 11898

    CIP (Explicit Versus I/O Messaging)

    Device Profiles and Application ObjectsUser Layer

    Application

    Layer

    Transport

    Layer

    Data-Link

    and PhysicalLayers

    FIGURE 1 – Profinet IO cycle.

    Red

    RT_Class 3

    Orange

    RT_Class 2

    Green

    Class 2/Class 1

    Yellow

    Nonreal Time

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    5/13

    12  IEEE INDUSTRIAL ELECTRONICS MAGAZINE n  MARCH 2008

    either 2.54 GHz (legacy 802.11, 802.11b,802.11g) or 5 GHz (802.11a). Very

    high data rates are foreseen in the

    802.11 standard, ranging from 11 Mb/s

    (802.11b) to 54 Mb/s (802.11g/a). More-

    over, they are expected to increase up

    to (about) one order of magnitude with

    the next-generation devices (802.11n).

    The IEEE 802.11 standard specifies

    a mandatory distributed coordination

    function (DCF) to access the physical

    medium, based on the use of a carrier

    sense multiple access with collisionavoidance (CSMA/CA) technique. As a

    further, nonmandatory, option, 802.11

    encompasses also a centralized MAC

    protocol [point coordination function

    (PCF)]. In this case, the access to the

    network is regulated by a specific sta-

    tion, namely, the point coordinator

    (PC), which grants exclusive access to

    any wireless station, thus preventing

    collisions (the main cause of nonde-

    terminism in WLANs). Although PCF

    is not currently supported by the vast

    majority of commercial WiFi boards, it

    nevertheless represents an interesting

    option for industrial applications, at

    least from a theoretical point of view.

    Finally, it is worth mentioning that

    devices complying with the 802.11e

    standard also support the concept of

    quality of service (QoS) in the form of

    traffic prioritization/parameterization.

    Such a property is particularly appeal-

    ing for industrial applications since it

    allows (at least in principle) for faster

    delivery of critical messages. In partic-

    ular, the 802.11e specification defines

    an additional hybrid coordination

    function (HCF), which in turn consists

    of two distinct mechanisms; namely,

    the enhanced distributed channel ac-

    cess (EDCA) and the HCF controlled

    channel access (HCCA). Despite HCCA

    being able to provide a fairly more

    deterministic behavior than EDCA,

    thanks to the presence of a hybrid co-ordinator (HC) that effectively ensures

    contention-free access to the wirelessmedium by allocating stations trans-

    mission opportunities (TXOPs), com-

    pliant devices are, at the time of writ-

    ing, not available yet. EDCA-compliant

    equipment, instead, is currently avail-

    able off the shelf at (relatively) low

    cost. EDCA is noticeably simpler and

    cheaper than HCCA, as it basically re-

    lies on a distributed approach. Hence,

    it will likely be adopted more and more

    in several real-world wireless indus-

    trial installations.

    IEEE 802.15.4

    The IEEE 802.15.4 specification deals

    with ad-hoc wireless interconnections

    of electronic devices within a limited

    transmission range (some tens of me-

    ters) at low data rates (up to 250 kb/s).

    It has been designed in order to target

    several application fields ranging, for

    example, from home and building au-

    tomation to simple cable replacement

    and from smart tags to automotive

    sensing. Devices of the IEEE 802.15.4

    family typically operate in the indus-

    trial, scientific, and medical (ISM)

    band, centered around 2.4 GHz (actu-

    ally, a second band is also available,

    depending on the country considered,

    but it is not widely adopted since it is

    limited to data rates of 20 kb/s).

    Two types of MAC protocols are en-

    compassed by the standard: a beacon-

    enabled MAC protocol, characterized

    by the presence of a network coordina-

    tor that periodically sends beacons, and

    a nonbeacon MAC protocol. Most of the

    commercially available devices operate

    in beacon-enabled mode. In this mode,

    the network coordinator periodically

    issues superframes that are divided in

    two parts. The first part, called conten-

    tion access period (CAP), takes place

    immediately after the beacon and is

    based on a distributed CSMA/CA mech-

    anism for controlling the access to thechannel. After the CAP, there might be a

    (optional) contention-free period (CFP),

    in which guaranteed time slots (GTSs)

    are exclusively allocated to the nodes.

    Conversely, in the nonbeacon mode a

    pure CSMA/CA protocol is adopted for

    channel access control. In this way, a

    totally decentralized and asynchronous

    MAC is obtained.

    Implementationof Hybrid NetworksFrom a general point of view, the in-

    terconnection of different communi-

    cation networks is achieved through

    specific devices called intermediate

    systems (ISs) [19]. These devices have

    different structure, functionality, and

    complexity, depending on the layer

    of the open systems connection (OSI)

    reference model they operate.

    Interconnection at the Physical Layer 

    The simplest form of ISs are repeaters.

    They operate at the physical layer and

    are used to interconnect subnetworks

    that share the same MAC mechanism.

    They are usually adopted for regenerat-

    ing electrical signals flowing across dif-

    ferent network segments. In Ethernet

    networks, for example, they operate

    according to the so called 3-R princi-

    ple: reshaping, retiming, and retrans-

    mitting. In this way, it is possible to face

    the problem of signal attenuation over

    the cable when the network extension

    grows longer. Moreover, they enable an

    increase in the number of nodes that

    can be networked, which is practically

    limited by the current absorbed by de-

    vices attached to the bus.

    Besides simpler two-ported de-

    vices, a special kind of repeater exists

    (called a hub) that is provided with

    several ports and can be adopted to

    achieve star network topologies. Be-

    cause of the increased reliability (a

    defective cable no longer affects the

    whole network operation), hubs are

    particularly suitable for deploying

    network infrastructures and for the

    cabling of buildings. More advanced

    versions of repeaters may be some-

    times employed to interconnect sub-

    networks that rely on different media

    (e.g., copper wires and fiber optics)

    or signaling techniques (e.g., voltage/current levels).

    Highly automated production systems can get

    significant benefits from the introduction of the most

    advanced wireless communication techniques.

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    6/13

    MARCH 2008  n  IEEE INDUSTRIAL ELECTRONICS MAGAZINE 13

    As the MAC mechanism that regu-

    lates the access to the shared transmis-

    sion medium is the same for the whole

    network, care has to be taken about the

    increased propagation delays when net-

    work segments are connected through

    repeaters. In many cases, such as Eth-

    ernet, CAN, or Profibus networks, this

    leads to limitations on the maximumnumber of repeaters that can be insert-

    ed between any two nodes.

    An interesting example of hybrid

    wired/wireless interconnection car-

    ried out at the lowest layer of the pro-

    tocol stack is provided in [20]. In that

    case, the authors refer to a hybrid ar-

    chitecture between Profibus and Ra-

    dio Fieldbus [10]. Both networks use

    the same data-link layer protocol—

    specifically, the fieldbus data link (FDL)

    defined by Profibus—but have differentphysical layer protocols. Indeed, Profi-

    bus is based on the well-known RS-

    485 standard whereas Radio Fieldbus

    uses a direct spread spectrum (DSS)

    technique similar to the one adopted

    in 802.11 WLANs. However, it is worth

    noting that pure repeaters are unusual

    in real-life hybrid wired/wireless net-

    works since the resulting physical com-

    munication support in most practical

    cases is not seen as a (sufficiently) uni-

    form medium by the MAC protocol (the

    access techniques adopted by wireless

    networks are, typically, rather different

    from wired solutions).

    Interconnection at the

    Data-Link Layer 

    An IS operating at the data-link layer

    is called a bridge. Its function is to

    transfer frames between systems

    that are not (or cannot be) treated

    as a uniform communication support

    by protocols at the MAC level (e.g.,

    switched LANs that are made up ofseveral separate collision domains).

    A bridge that is equipped with more

    than two ports is commonly referred

    to as a switch (this kind of devices

    is currently present in almost every

    real-life Ethernet installation).

    Even though the MAC mechanism

    does not need to be the same for all the

    subnetworks interconnected through

    bridges/switches, the set of communi-

    cation services they provide at the data-

    link layer should be similar. Limitationson the kind of networks that can be

    interconnected concern, for example,

    their addressing scheme (which should

    be uniform on the whole network) and

    the maximum transfer unit (MTU),

    which affects the allowed payload size

    as fragmentation is not permitted at the

    data-link level. When subnetworks with

    different MTUs are interconnected, care

    has to be taken so that the payload of

    the exchanged frames never exceeds a

    threshold equal to the minimum among

    the supported MTUs.

    Figure 3 shows an example of inter-

    connection through a bridge for the

    hybrid networks considered in this

    article. From a practical point of view,

    the bridge to be used in hybrid wired/

    wireless networks is a device equipped

    with two transmitting/receiving inter-

    faces (one for each subnetwork) and,

    optionally, a protocol converter. A

    frame originated from whatever seg-

    ment (either the wired network or thewireless extension) is propagated to

    the other by the bridge, which receives

    the frame, converts it in the suitable

    (target) format, and then retransmits

    it. Access points (APs) are a very popu-

    lar example of a bridge for hybrid net-

    works. APs are used both to enable

    communication in an infrastructure

    WiFi network [also known as a basic

    service set (BSS)] and to interconnect

    it to an existing Ethernet backbone

    (portal function).

    Interconnection at Higher Layers

    For ISs working at the network layer or

    above, the generic term of gateway is

    often adopted. A gateway is responsi-

    ble for transferring protocol data units

    (PDUs), as well as generic streams of

    information and application services

    between the application processes

    of the interconnected systems, by

    performing the required protocol

    translations. From a general point of

    view, there is no practical restriction

    on the kinds of networks that can be

    interconnected through gateways.

    FIGURE 3 – Interconnection via a bridge.

    Data-Link Protocol Converter

    Wired

    Network

    WirelessExtension

    Bridge

    (Access Point)

    Physical

    Data Link

    Network

    Transport

    Application

    Application

    Program

    Physical

    Data Link FrameFrame

    Transceivers Physical

    Data Link

    Network

    Transport

    Application

    Application

    Program

    Physical

    Data Link

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    7/13

    On the other hand, gateways are usu-

    ally quite complex, expensive devices,

    and the performance they achieve is

    somehow lower than bridges (higher

    latencies have to be expected).

    ISs that operate specifically at the

    network layer are known as routers.

    At this level, a uniform addressing

    scheme and protocol are used [i.e., theinternet protocol (IP)], which ensure

    worldwide connectivity regardless of

    the physical medium, MAC mecha-

    nism, and data-link services of the in-

    terconnected subnetworks. However,

    we are not directly interested in them

    since most industrial networks are de-

    ployed as local networks.

    When the interconnection takes

    place at the application layer (conver-

    sion of high-level services), the term

    proxy is sometimes adopted. Indeed,a proxy is more than an application-

    layer gateway. It usually represents

    parts of the network as if they were a

    single node, possibly hiding the under-

    lying structure for security or perfor-

    mance reasons. For instance, proxies

    are used in fieldbus/IP connections

    to represent legacy fieldbus segments

    (as in Profinet).

    An example of a wireless extension

    implemented at the application layer is

    shown in Figure 4. As can be seen, the

    gateway incorporates two interface

    boards fully compliant with the over-

    all protocol stacks of the subnetworks

    they are connected to, and a protocol

    converter, usually implemented as a

    software application.

    General Issues Concerning

    Wireless Extensions of

    Wired Industrial Networks

    Although technically feasible, wire-less extensions of industrial networks

    to be used at the shop floor (either

    fieldbuses or RTE networks) are not

    so straightforward. This is basically

    due to three main reasons.

    First, the transmission support (wire-

    less medium) is shared among all the

    nodes. Even operating at the maximum

    allowable speed (e.g., 54 Mb/s for cur-

    rently available 802.11a/g/e networks),

    this may result in a low per-station net

    throughput compared to that of wirednetworks (especially switched Ethernet)

    when the number of connected devices

    grows higher. Furthermore, the non-

    negligible overheads introduced by the

    communication protocol (larger proto-

    col control fields, acknowledgement and

    reservation mechanisms, no provision

    for full-duplex operations, etc.) have

    to be considered as well. For example,

    the minimum time taken to exchange

    reliably 8 B of user data over a WLAN (54

    Mb/s, no RTS/CTS, no TCP/IP encapsu-

    lation, ad-hoc network mode) including

    interframe spaces (as they effectively

    waste network bandwidth) is about 100

    µs. Such a value is only slightly shorter

    than the transmission time over a CAN

    network operating at 1 Mb/s.

    Second, random access techniques

    (e.g., CSMA/CA) are often employed

    by wireless networks. This means

    that, on the one hand, unpredictable

    (and unbounded) transmission delaysmight occur (because of the occur-

    rence of repeated collisions) while,

    on the other hand, network conges-

    tions could be experienced, and this

    is surely the worst aspect. In other

    words, when the offered load exceeds

    a given threshold (even for a limited

    time), a condition may likely arise be-

    cause of positive feedback so that the

    effective net throughput decreases as

    the load increases. This means that

    the network may become temporarilyunavailable to carry out timely data

    exchanges, which is not compatible

    with proper operation of distributed

    control systems. Moreover, even in the

    case of lightly loaded networks, non-

    negligible jitters may affect the trans-

    mission unless some form of prioritiza-

    tion scheme is suitably adopted.

    Third, and last, wireless chan-

    nels are much more error prone than

    wired cabling [21], and this is a seri-

    ous drawback when they are used

    in industrial environments, which

    are usually affected by nonnegligible

    FIGURE 4 – Interconnection via a gateway.

    Application Protocol Converter

    WiredNetwork

    WirelessExtension

    Gateway

    Physical

    Data Link

    Network

    Transport

    Application

    ApplicationProgram

    Transceivers Physical

    Data Link

    Network

    Transport

    Application

    ApplicationProgram

    Physical

    Data Link

    Network

    Transport

    Application Services

    Physical

    Data Link

    Network

    Transport

    Application

    Controllers

    TCP/UDPIP

    14  IEEE INDUSTRIAL ELECTRONICS MAGAZINE n  MARCH 2008

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    8/13

     electromagnetic interferences. Besides

    causing higher communication laten-

    cies and jitters, transmission errors

    also directly affect the network reliabil-

    ity and, consequently, the robustness

    of the overall system. Unless acknowl-

    edged transmission schemes are ad-

    opted (which, incidentally, are not pos-

    sible when broadcast and producer/consumer-like multicast traffic are

    taken into account), there is a nonneg-

    ligible chance that messages sent over

    the air never reach the intended des-

    tination. This may also cause consis-

    tency problems; i.e., when a multicast

    message is received only by a part of

    the addressed devices.

    Each one of the three drawbacks

    mentioned above can be tackled (at

    least partially) by means of already

    existing or soon to be available tech-nologies. In particular, the throughput

    problem could be somehow lessened

    by the upcoming 802.11n specification,

    which is based on the multiple-input

    multiple-output (MIMO) technology

    and is theoretically able to provide

    a big leap (about one order of mag-

    nitude) in the network throughput. A

    more reasonable (and viable) choice,

    however, could be the use of several

    (smaller) separate wireless subnet-

    works (e.g., WLANs), interconnected

    by means of a wired backbone in or-

    der to limit the packet rate on each

    one of them. Keeping the size of each

    wireless subnetwork small also helps

    in operating it at higher speeds (be-

    cause of the automatic rate adaptation

    mechanism featured by most APs and

    WLAN adapters).

    The second problem (determin-

    ism) can be dealt with through the

    adoption of specific medium access

    techniques; e.g., PCF in the case of

    IEEE 802.11. Unfortunately, PCF is sel-

    dom implemented in commercially

    available APs. Variants of PCF, such

    as the iPCF mechanism introduced by

    Siemens [22], have the big drawback of

    being proprietary solutions, and hence

    compatibility and interoperability can

    hardly be ensured. Alternatively, the

    new prioritization features offered

    by the 802.11e standard (already sup-

    ported by several WLAN adapters cur-rently available off the shelf) could be

    exploited [23]. While not guaranteeing

    strict determinism, such a technology

    (originally developed for multimedia

    traffic) is able to provide significant

    improvements that might often be

    enough for many industrial systems.

    For example, it is possible to ensure

    quasi-real-time behavior by assign-

    ing higher priorities to the messagescharacterized by tight timing/safety

    constraints. Also TDMA techniques

    may be employed to solve the deter-

    minism problem. In such a direction,

    for example, the guaranteed time slots

    provided by IEEE 802.15.4 represent

    an interesting opportunity.

    The third problem (robustness) can

    be faced in several ways. Whenever

    possible, a proper placement of anten-

    nas is envisaged, optionally by exploit-

    ing antenna diversity. In the case thatthis is not enough, leaky wave anten-

    nas [24] can be used to ensure reliable

    communication over radio links. The

    drawback in this case is the need to

    deploy the leaky cable infrastructure,

    which makes it a nonuniversal solu-

    tion. As an alternative, if true mobility

    is required, devices operating in the

    5-GHz band (802.11a) can be adopted

    as a satisfactory choice [21]. In this

    way, in fact, interferences with other

    devices (mobile phones, Bluetooth-

    enabled equipment, notebooks with

    wireless connections, etc.) operating

    in the standard (and already jammed)

    2.4-GHz ISM band are avoided.

    IEEE 802.11-BasedExtensions: FieldbusesTypically, the transmission protocols

    adopted by fieldbuses are noticeably

    different from those employed bywireless networks at all levels of the

    communication stack. For example,

    random access techniques typical of

    WLANs are not employed by any field-

    bus network to the best of our knowl-

    edge. Other problems may arise that

    are related to either the frame pay-

    load, which in fieldbuses is typically

    very small, or the addressing scheme.

    As a consequence, wireless exten-

    sions of fieldbuses have to be imple-

    mented mainly at the applicationlayer; i.e., through gateways. However,

    some remarkable exceptions exist.

    DeviceNet, for instance, provides in-

    teroperability with all the networks

    based on the CIP protocol thanks to

    some interesting routing techniques

    that allow, in principle, implementing

    extensions at a lower level (i.e., through

    direct routing of CIP messages).

    Extension of Profibus DP 

    Figure 5 shows that the gateway

    needed to extend Profibus DP can be

    FIGURE 5 – Gateway with proxy functionality.

    DPProto-

    col

    Profibus DP

    IEEE

    802.11

    Gateway (Profibus DP Master Station)

    GatewayCode

    InputBuffers

    WirelessProtocol

    OutputBuffers

    MARCH 2008  n  IEEE INDUSTRIAL ELECTRONICS MAGAZINE 15

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    9/13

    implemented directly on a Profibus DP

    master station. An 802.11 board has to

    be installed on the master as well, and

    suitable code has to be developed for

    this particular implementation in or-

    der to enable the exchange of data be-

    tween the two segments. The wireless

    nodes implementing the extension are

    organized in a basic service set (BSS)

    which, as specified in the IEEE 802.11

    standard, is the basic building block

    of a WLAN. A similar solution hasbeen proposed in [11] for the former

    version of Profibus; namely, Profibus

    FMS. In that paper, some useful results

    derived from a set of practical experi-

    ments are provided. In particular, the

    stations of the BSS, which periodically

    exchange 40 B, are handled via a vir-

    tual polling algorithm with a resulting

    cycle time of 5 ms.

    In order to limit the effects of the

    delays that are unavoidably introduced

    by the gateway, the use of a proxy is

    recommended. The proxy hosts a pro-

    cess image (decoupled from the wire-

    less segments) that can be accessed

    from the wired side at any time, with-

    out timing restrictions introduced by

    the wireless communication protocol.

    This is accomplished, as shown in

    Figure 5, by means of a set of I/O buf-

    fers. Stations on the wireless segment

    access these buffers in order to store

    the input data acquired from the con-

    trolled system. In the same way, output

    data written by the proxy into the out-

    put buffers are subsequently retrieved

    by the wireless stations.

    The proxy code is responsible for

    interfacing the Profibus DP protocol

    to the input/output buffers (it is worth

    noting that buffers and data caching,though reasonable for a proxy, from a

    general viewpoint are not sufficient). In

    detail, when a data exchange requests

    that carries output data is issued over

    the DP protocol, the proxy code ex-

    tracts data from the DP protocol data

    unit, writes them into the output buf-

    fers, and then generates the acknowl-

    edgment for the DP protocol. At the

    same time, input data are taken from

    the input buffers and returned to the

    requester through the DP protocol.

    Note that the above technique

    does not impose any restriction on

    the protocol(s) actually used on the

    wireless extension (e.g., as already

    mentioned, a virtual polling algorithm

    is used in [11]), as data exchanges take

    place via I/O buffers.

    Extension of DeviceNet 

    DeviceNet was designed bearing in

    mind the option of having several sub-

    networks interconnected by means of

    routers. Such a feature, besides making

    the interconnection with other ODVA

    networks (e.g., ControlNet and Ether-

    net/IP) simpler, turns out to be very

    helpful when implementing wirelessextensions. Routing in heterogeneous

    CIP networks is achieved by means of

    so-called port segments, which are

    added to messages and explicitly de-

    scribe the route they should follow

    (this is often referred to as source

    routing). Such a route is specified by

    means of a list of items. Each item, in

    turn, describes a hop by means of the

    output port of the router to be used in

    forwarding the message and the ad-

    dress of the intended destination inthe next subnetwork (either the target

    node or another router). As the frame

    travels along its route, the port seg-

    ment is shortened by routers, as each

    one of them removes the information

    concerning the hop it has processed.

    Obviously, the routing process takes

    some time, and, hence, response times

    over the interconnected network grow

    longer. This aspect has to be taken

    into account carefully when demand-

    ing real-time applications have to be

    designed (e.g., it might be necessary

    to relax timing constraints).

    Using WLANs together with con-

    ventional DeviceNet devices can be

    accomplished in several ways. The

    first, and simpler, solution is to rely

    on a DeviceNet to Ethernet/IP router

    that, in turn, is connected to a conven-

    tional AP (see Figure 6). In this case,

    no new hardware component has to

    be developed. However, this solution

    causes each message to traverse (at

    least) two intermediate devices before

    reaching its target node (e.g., the rout-

    er and the AP), and this introduces ad-

    ditional delays.

    A second and more sophisticated

    solution requires the modification of

    existing DeviceNet routers so as to

    include an 802.11 port, as shown in

    Figure 7 (this should not be a dif ficult

    task for manufacturers of DeviceNet

    products). In this case, the routeris completely aware of the exactFIGURE 6 – Extension of DeviceNet via a CIP router and an AP.

    IEEE802.11

    RequestingDevice

    DeviceNet toEthernet/IP

    RouterAP

    TargetDevice

    Ethernet/IPDeviceNet

    16  IEEE INDUSTRIAL ELECTRONICS MAGAZINE n  MARCH 2008

    The integration of wireless and wired

    communications in industrial scenarios is becoming

    a crucial issue that has to be dealt with carefully in

    order to achieve efficiency and performance.

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    10/13

     requirements of each CIP connection,

    and, hence, it can select the most suit-

    able QoS when forwarding messages

    over the radio channel (provided

    that a QoS-enabled network, such as

    802.11e, is adopted), thus providing

    better real-time behavior.

    It is worth noting that currently

    some devices are actually availableoff the shelf that can be used to intro-

    duce wireless extensions to DeviceNet

    networks. For instance, [25] describes

    a solution that relies on the routing of

    messages for carrying out information

    exchanges over explicit connections.

    Such a solution foresees a wireless

    master (implemented as a DeviceNet

    slave) connected to the primary

    DeviceNet bus, and one or more wire-

    less slaves (which are seen as virtual

    DeviceNet masters), each one con-necting a different DeviceNet remote

    bus. To improve communication ef-

    ficiency, process data collected by

    wireless slaves from each remote bus

    can be provided to the master as a sin-

    gle process image. Unfortunately, solu-

    tions like this are mostly based on pro-

    prietary communication technologies,

    and, hence, they cannot be considered

    as universal proposals.

    IEEE 802.11-Based Extensions:Real-Time Ethernet NetworksDespite the availability of both hard-

    ware components and protocols, the

    implementation of wireless extensions

    to RTE networks is not immediate. In-

    deed, interoperability between Eth-

    ernet and WLANs is always ensured

    by using inexpensive off-the-shelf

    devices (i.e., APs), along with some

    commonly available protocols. As an

    example, the IEEE 802.1D bridging

    protocol [26] allows for interconnec-

    tions at the data-link layer; however,

    such a solution, which has good effi-

    ciency, is sometimes difficult to imple-

    ment since it relies on the availability

    of IEEE 802.2 logical link control (LLC)

    services, which are not always explic-

    itly accessible on Ethernet and 802.11

    off-the-shelf components.

    A more straightforward intercon-

    nection technique can be obtained via

    the TCP/IP protocol suite but, unfor-tunately, it is unsuitable for achieving

    real-time behavior. For this reason, in

    some cases the User Datagram Proto-

    col (UDP) protocol [27] is preferred.

    UDP, in fact, has lower overheads than

    TCP and offers simpler access to com-

    munication thanks to its unacknowl-

    edged transmission scheme (which

    lacks the sliding window mechanism).

    In this case, however, the weakpoint is found in the access protocols

    adopted by RTE networks, which often

    are not completely compliant with the

    original Ethernet specifications. For

    example, Ethernet Powerlink (EPL)

    [28], which actually relies on standard

    Ethernet controllers, adopts a protocol

    based on a fast and precise periodic

    polling carried out on (wired) nodes

    that ensure rapid operation (cycle time

    under 200 µs) and ultra-low jitter (be-

    low 1 µs). Unfortunately, this does notallow for the integration of wireless

    devices, due to the unavoidable uncer-

    tainty of communication times over

    radio channels (because of interfer-

    ences and electromagnetic noise) that

    impairs determinism unacceptably.

    Extension of Profinet IO

    The extension of Profinet IO has to

    face all the problems discussed in the

    previous sections. Indeed, it is not

    possible to implement an extension at

    the data-link layer, since the CSMA/CA

    technique employed by IEEE 802.11

    cannot cope with the very tight time

    schedule imposed by the TDMA medi-

    um access technique used by Profinet

    IO. Consequently, the extension may

    only be implemented at the applica-tion layer, via a gateway, in a way simi-

    lar to that described for Profibus DP.

    There are, however, two alterna-

    tive options for carrying out such

    a task. When real-time behavior is

    not required, data exchange with

    the wireless nodes may simply take

    place in the nonreal-time period of

    the Profinet IO cycle (the YELLOW

    phase shown in Figure 1). The second

    option is more appealing, since it en-

    ables preserving real-time properties,and is based on the use of the PCF. As

    shown in Figure 8, an AP implement-

    ing the PCF technique can be used as a

    bridge. With such a solution, wireless

    stations can be included in the Profi-

    net IO cycle, too, since the AP is able

    to assign them specific time slots that

    provide contention-free (and, hence,

    deterministic) access to the wireless

    medium. Specific actions should be

    FIGURE 7 – Extension of DeviceNet via a WLAN-enabled CIP router.

    IEEE802.11

    DeviceNet

    to WLANRouter

    DeviceNet

    TargetDevice

    Requesting

    Device

    MARCH 2008  n  IEEE INDUSTRIAL ELECTRONICS MAGAZINE 17

    FIGURE 8 – Extension of Profinet IO via a bridge.

    ProfinetIO

    Interface

    IEEE802.11

    APwith

    PCFProfinet IO

    Bridge

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    11/13

    taken in this case so as to improve net-

    work reliability as well (e.g., by means

    of leaky cables).

    An example of an application based

    on this technique is described in [29]

    (actually, the iPCF technique men-

    tioned earlier in this article is used).

    In [29], 21 Profinet IO devices are lo-

    cated on a wireless segment, whereasthe Profinet IO controller relies on

    wired connections. The bridge is a

    commercial product [22] implement-

    ing iPCF.

    Extension of Ethernet/IP

    As long as Ethernet/IP networks are

    taken into account, there is no theo-

    retical need for any change to adapt

    the CIP protocol to 802.11 networks.

    In fact, APs can be used that trans-

    parently forward each message tothe intended destination by means of

    its Ethernet address (which, in turn,

    is derived from the IP address used

    at the application level), as shown in

    Figure 9. While this is certainly the

    most appealing (and inexpensive)

    solution, it might lead to suboptimal

    results when real-time process data

    have to be exchanged. This is because

    in the existing conventional 802.11b/g

    WLANs, priorities cannot be enforced

    for process data higher than those

    for background traffic, unless special

    techniques, such as those provided by

    802.11e, are adopted.

    In the latter case, however, existing

    implementations of CIP are likely to

    be modified in order to map the tim-

    ing constraints of each message (I/O

    versus explicit connection, expected

    packet rate, and so on), which are

    known at the application level, onto

    the most suitable QoS as provided by

    the underlying communication sys-

    tem. From this point of view, Ethernet

    is not the same as 802.11e WLANs. In

    fact, IEEE 802.1Q [16] enables specify-

    ing up to eight traffic classes to encode

    frame priority in switched Ethernet,

    whereas 802.11e foresees four differ-

    ent access categories (ACs) to provide

    traffic prioritization.

    It is worth noting that developingWLAN-enabled Ethernet/IP devices,

    such as the target node on the right

    side in Figure 9, is not really a difficult

    task. The advantage of adopting this

    kind of solution is that the device can

    exploit the knowledge about the timing

    requirements of the CIP connection to

    select the most suitable QoS. A viable

    choice could be, for example, to map

    I/O connections with tight timing re-

    quirements (i.e., high expected packet

    rate) on voice-grade (AC_VO) traffic,whereas those having less severe con-

    straints on the transmission times can

    be mapped on video-grade (AC_VI)

    transmissions. Finally, explicit con-

    nections devoted to parameterization

    can rely on either best-effort (AC_BE)

    or background (AC_BK) traffic.

    In the case where conventional

    APs are used to interconnect wired

    Ethernet/IP segments with wireless

    extensions, exploiting 802.11e QoS

    becomes more difficult. Indeed, it is

    impossible for a conventional AP to

    determine the priority of each frame,

    since it is defined at the application

    layer, whereas the AP operates at the

    data-link layer. A possible solution is

    to rely on APs that are able to map

    802.1Q traffic classes directly onto

    IEEE 802.11e ACs (and vice versa).

    Besides traffic prioritization, an-

    other issue that should be taken into

    account properly when interconnect-

    ing industrial Ethernet networks and

    WLANs concerns the mechanisms

    for broadcast/multicast traffic con-

    finement (process data are usually

    exchanged according to the producer/

    consumer paradigm). In wired net-

    works such as Ethernet/IP, this is dealt

    with by means of VLANs. Although

    this mechanism is not available in

    802.11 wireless networks to reducethe amount of multicast traffic, some

    devices are currently available off the

    shelf that can map VLANs to different

    service set identifiers (SSIDs) [30].

    Such a feature is quite interesting,

    as it can be used to ease the integra-

    tion of WLAN extensions to existing

    Ethernet/IP networks.

    IEEE 802.15.4-Based ExtensionsIn analogy with the networks defined

    by the IEEE 802 committee(s), the IEEE802.15.4 specification only refers to the

    lower two layers of the communication

    stack and does not specify any type of

    upper-layer protocol. Thus, in order to

    grant access to 802.15.4 networks by

    the higher layers, two options are rec-

    ommended by the standard:

     using LLC type 1 services [31]

     accessing MAC services directly.

    Both techniques can be employed

    profitably, even if it is worth men-

    tioning that LLC services are seldom

    made available by 802.15.4 board

    manufacturers.

    Due to the limited transmission

    speed of IEEE 802.15.4, the resulting

    hybrid networks are not generally

    able to provide (very) fast communica-

    tions. On the other hand, the 802.15.4

    protocol has good efficiency, especial-

    ly when compared to 802.11. Finally,

    the guaranteed time slots made avail-

    able in the beacon-enabled mode al-

    low a co-ordered (e.g., deterministic)

    access scheme to the shared medium

    by wireless nodes, with a consequent

    reduction of the jitters that typically

    affect pure CSMA/CA protocols.

    Extension of Fieldbuses

    The use of IEEE 802.15.4 for implement-

    ing wireless extensions of fieldbuses

    basically presents the same problems

    already discussed for IEEE 802.11-

    based extensions. In particular, theremarkable differences between the

    FIGURE 9 – Extension of Ethernet/IP via an AP.

    IEEE802.11eAP

    Ethernet/IP

    Target

    DeviceRequesting

    Device

    18  IEEE INDUSTRIAL ELECTRONICS MAGAZINE n  MARCH 2008

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    12/13

    MAC protocols of the two kinds of

    networks do not practically allow for

    the interconnection at the lower lay-

    ers. Thus, from the above consider-

    ations it follows that fieldbus exten-

    sions based on 802.15.4 networks can

    only be achieved through a gateway

    that translates the services foreseen

    by the application layer of the field-bus to calls to 802.15.4 services. The

    gateway is typically implemented as

    a station that embeds the interface to

    the wired fieldbus and, at the same

    time, acts as network coordinator for

    the 802.15.4 system.

    Similarly to the case of the extension

    of Profibus DP through IEEE 802.11, also

    in this case there are no restrictions on

    the protocol(s) actually employed over

    the wireless extension. Indeed, the gate-

    way may either use the 802.15.4 data-linklayer services (LLC or MAC) directly or,

    alternatively, it may rely on the func-

    tions provided by a proprietary proto-

    col explicitly defined for managing data

    exchanges over the wireless extension.

    Once again, note that significant ben-

    efits may be obtained by the use of

    the guaranteed time slots. As an ex-

    ample, the virtual polling algorithm

    described in [11] for 802.11-based ex-

    tensions does not need to be imple-

    mented from scratch for IEEE 802.15.4

    since, in practice, such a mecha-

    nism is natively provided by the

    beacon-enabled mode of the access

    technique.

    Extension of Real-Time Ethernet

    Concerning RTE networks, the avail-

    ability of the 802.15.4 LLC protocol

    could enable, in theory, the imple-

    mentation of extensions right at the

    data-link layer, since most RTE net-

    works provide access to such a layer

    as well. However, it has to be con-

    sidered that LLC type 1 services are

    nonconfirmed, connectionless ser-

    vices. Such a feature may represent

    a serious drawback, which could lead

    to a decrease of the overall system

    performance due to the nonnegligible

    probability of errors affecting com-

    munications in the wireless segment.

    In fact, since 802.15.4 does not imple-

    ment frame retransmission at theMAC level, a (possible) corruption of

    a frame cannot be recovered by the

    LLC type 1 services (because they are

    unconfirmed) with the consequent

    definitive loss of the related protocol

    data unit. This means that also in this

    case an effective extension can only

    take place via a gateway.

    Finally, it is worth mentioning that

    ZigBee [32] offers a complete proto-

    col stack based on IEEE 802.15.4 and,

    obviously, represents an interesting

    opportunity for implementing hybridnetworks when the wired segment

    consists of either a fieldbus or an

    RTE network. In particular, it is inter-

    esting to notice that ZigBee enables

    connectivity with whatever type of

    network based on the IP protocol

    by means of the purposely defined

    ZigBee expansion devices (ZEDs).

    As a consequence, it is possible (at

    least in principle) to interface easily

    with some RTE networks (basically,

    all those that allow for TCP/IP com-

    munication, such as, for example,

    Ethernet/IP and Profinet IO) via a

    conventional IP router.

    ConclusionIt is envisaged that network configu-

    rations resulting from wireless exten-

    sions of conventional (wired) indus-

    trial networks will be adopted more

    and more in the near future, thanks to

    the availability of wireless technolo-

    gies and devices able to cope with in-

    dustrial communication requirements

    properly. Concerning wired networks,

    the analysis presented above focused

    on two popular fieldbuses (Profibus

    DP and DeviceNet) and two RTE net-

    works (Profinet IO and Ethernet/IP),

    since they are based on different

    protocols. In the same way, both the

    IEEE 802.11 family of WLAN standards

    and IEEE 802.15.4 wireless sensor net-

    works were taken into account as wire-less extensions, as they are two of the

    most promising technologies for in-

    dustrial and control applications, and

    many devices are already available off

    the shelf at (relatively) low cost.

    As pointed out in the article, the

    extension of Profibus DP and, in gen-

    eral, of almost all fieldbuses, can

    typically take place at the applica-

    tion layer. While providing greater

    flexibility, this may worsen the over-

    all performance, due to the delays

    gateways unavoidably introduce. Asan interesting exception, DeviceNet

    can be interconnected at a lower

    layer via CIP routers. This means

    that in hybrid networks based on

    DeviceNet all devices can be seen

    (and accessed) as conventional CIP

    nodes, whereas in the case where a

    gateway/proxy is adopted (as hap-

    pens in Profibus DP) wireless nodes

    are no longer effectively part of the

    Profibus network.

    When real-time (industrial) Eth-

    ernet networks have to be provided

    with a wireless extension, more

    alternatives are possible, in prin-

    ciple, including interconnection

    at the data-link level. However, as

    these networks often rely on pecu-

    liar protocols to ensure predictable

    communications over Ethernet,

    the integration with WLANs is not

    straightforward. As a consequence,

    a loss of performance/determinism

    in hybrid networks may occur also

    in this case. Moreover, when 802.11

    is selected for the wireless exten-

    sion, the native randomness of the

    DCF medium access technique has

    to be taken into account carefully

    in the design of the hybrid network,

    as it can give rise to additional (and

    unwanted) delays. Significant im-

    provements in this direction could

    be achieved by using deterministic

    techniques such as, for example, PCFin WLANs (which, however, is not so

    The IEEE 802.15.4 specification deals with ad-hoc

     wireless interconnections of electronic devices within

    a limited transmission range (some tens of meters)

    at low data rates (up to 250 kb/s).

    MARCH 2008  n  IEEE INDUSTRIAL ELECTRONICS MAGAZINE 19

  • 8/19/2019 Hybrid Wired-Wireless Networks for Real-Time Communications

    13/13

    frequently available in real devices)

    or guaranteed time slots in 802.15.4

    networks (which, however, suffer

    from a noticeably lower bit rate).

    Finally, an interesting option for

    achieving quasi-real-time communica-

    tions is also offered by the IEEE 802.11e

    standard, which supports the concept

    of QoS. Indeed, it has been shownthat, as an example, traffic prioritiza-

    tion could be used to deal with differ-

    ent types of messages (and, hence,

    of timing requirements) defined and

    handled by the CIP protocol, so as to

    transparently enable control applica-

    tions to meet real-time constraints

    even when hybrid interconnections

    are adopted.

    Biographies

    Gianluca Cena is director of researchwith the Istituto di Elettronica e di

    Ingegneria dell’Informazione e delle

    Telecomunicazioni of the Italian Na-

    tional Research Council (IEIIT-CNR),

    where he is engaged in activities con-

    cerning communications in manufac-

    turing and automotive environments.

    He is the author of many technical

    papers in the area of computer com-

    munications. His current research

    interests include industrial commu-

    nication systems and protocols and,

    in particular, real-time networks. He

    is also active in the international sci-

    entific community operating on these

    subjects, and he served as program

    cochairman for the sixth and seventh

    editions of the IEEE Workshop on Fac-

    tory Communication Systems.

     Adriano Valenzano  is director of

    research with the Istituto di Elettron-

    ica e di Ingegneria dell’Informazione

    e delle Telecomunicazioni of the Ital-

    ian National Research Council (IEIIT-

    CNR), where he is responsible for re-

    search on computer communications

    and formal methods for the analysis

    of security-critical systems. He is

    the author/coauthor of about 200 pa-

    pers published in international jour-

    nals, books, and conferences in the

    area of computer engineering. He is

    an associate editor of the IEEE Trans- 

    actions on Industrial Informatics. He

    has also served as a general cochair-man for the Sixth IEEE International

    Workshop on Factory Communica-

    tion Systems. His research interests

    concern computer communications

    (in particular, industrial communi-

    cations) and the formal analysis of

    security-critical systems.

     Stefano Vitturi  received the Lau-

    rea degree (summa cum laude) in

    electronics engineering from Univer-sity of Padova, Padova, Italy, in 1984.

    He has been a senior researcher with

    the Istituto di Elettronica e di Ingeg-

    neria dell’Informazione e delle Tele-

    comunicazioni of the Italian National

    Research Council (IEIIT-CNR) since

    January 2002. His research interests

    include industrial real-time communi-

    cation networks (wired and wireless)

    and implementation and performance

    analysis of devices conforming to the

    most popular industrial communica-tion protocols. In the context of these

    activities, he is an expert of the IEC

    SC 65 Standardization Subcommittee,

    and he has been the scientific respon-

    sible of the Italian Profibus Compe-

    tence Center.

    References[1] J.P. Thomesse, “Fieldbus technology in indus-

    trial automation,” Proc . IEEE , vol. 93, no. 6, pp.1073–1101, June 2005.

    [2] J.D. Decotignie, “Ethernet-based real-time and

    industrial communications,”  Proc . IEEE , vol.93, no. 6, pp. 1102–1117, June 2005.[3] A. Willig, K. Matheus, and A. Wolisz, “Wireless

    technology in industrial networks,” Proc. IEEE ,vol. 93, no. 6, pp. 1130–1150, June 2005.

    [4]  Prof ibus-DP Standard, Translation of the German National Standard DIN 19245 Part 3 , DeutschesInstitut fuer Normung, 1994.

    [5]  Low-Voltage Switchgear and Controlgear— Controller-Device Interfaces (CDIs)—Part 3: DeviceNet , IEC Standard 62026-3, July 2000.

    [6] Open DeviceNet Vendor Association, EtherNet/  IP Adaptation of CIP, Edition 1.3  (CIP NetworksLibrary, Vol. 2). Ann Arbor, MI: ODVA, 2007.

    [7] PROFIBUS International, “Application layerprotocol, application layer services for decen-tralized periphery and distributed automa-tion, specification for Profinet,” ver. 2.2, Oct.

    2007.[8]  IEEE Standards for Information Technology— 

    Telecommunications and Information Exchange Between Systems—Local and Metropolitan Area Networks—Specific Requirements—Par t11 Wireless LAN Medium Access Control (MAC)and Physical Layer (PHY) Specifications, IEEEStandard 802.11-2007, June 2007.

    [9]  IEEE Standard for Information Technology— Telecommunications and information exchangebetween systems—Local and metropolitanarea networks—Specific requirements—Part15.4: Wireless Medium Access Control (MAC)and Physical Layer (PHY) Specifications for Low—Rate Wireless Personal Area Networks(WPANs), IEEE Standard 802.15, Sept. 2003.

    [10] L. Rauchhaupt, “System and device architec-ture of a radio based fieldbus—The RF Field-

    bus system,” in  Proc . IEEE WFCS 2002 , Vast-eras, Sweden, Sept. 2002.

    [11] S. Lee, K.C. Lee, M.H. Lee, and F. Harashima,“Integration of mobile vehicles for automatedmaterial handling using Profibus and IEEE802.11 networks,”  IEEE Trans. Ind. Electron.,vol. 49, no. 3, pp. 693–701, June 2002.

    [12] A. Willig, M. Kubisch, C. Hoene, and A. Wolisz,“Measurements of a wireless link in an industri-al environment using an IEEE 802.11-compliantphysical layer,” IEEE Trans. Ind. Electron., vol.49, no. 6, pp. 1265–1282, Dec. 2002.

    [13] Bluetooth Special Interest Group, [Online].Available: www.bluetooth.com

    [14] D. Dzung, J. Endresen, C. Apneseth, and J.-E.Frey, “Design and implementation of a real-time wireless sensor/actuator communicationsystem,” in  Proc. IEEE ETFA 2005 , Catania,Italy, pp. 433–442, Sept. 2005.

    [15]  Indust rial Communication Networks—Part 2: Additional Fieldbus Prof iles for Real-Time Net- works Based on ISO/IEC 8802-3 , IEC Standard61784-2, Dec. 2007.

    [16]  IEEE Standard for Local and Met ropolitan Area Networks Vir tual Bridged Local Area Networks ,IEEE Standard 802.1Q-2005, 2006.

    [17] Open DeviceNet Vendor Association, Common Indust rial Protocol (CIP) Specif ication, Edition3.2  (CIP Networks Library, Vol. 1). Ann Arbor,MI: ODVA, 2007.

    [18] Open DeviceNet Vendor Association,  Network

     Infrastructure for EtherNet/ IP: Introduction andConsiderations (publication no. PUB00035R0).[Online]. Available: http://www.odva.org/Por-tals/0/Library/Publications_Numbered/PU-B00035R0_Infrastructure_Guide.pdf

    [19] F. Halsall,  Data Communications, Computer Networks and Open Systems. Reading , MA:Addison Wesley, 1996.

    [20] C. Koulamas, S. Koubias, and G. Papadopoulos,“Using cut-through forwarding to retain thereal-time properties of Profibus over hybridwired/wireless architectures,”  IEEE Trans. Ind. Elect ron., vol. 51, no. 6, pp. 1208–1217, Dec.2004.

    [21] D. Brevi, D. Mazzocchi, R. Scopigno, A. Boniven-to, R. Calcagno, and F. Rusinà, “A methodologyfor the analysis of 802.11a links for safety-critical

    industrial communications,” in Proc. IEEE WFCS2006 , Torino, Italy, pp. 165–174, June 2006.

    [22] Siemens AG, “SCALANCE W788 RR AccessPoint” [Online]. Available: www.automation.siemens.com/net

    [23] G. Cena, I. Cibrario Bertolotti, A. Valenzano,and C. Zunino, “Evaluation of response timesin industrial WLANs,”  IEEE Trans. Ind. Infor- mat., vol. 3, no. 3, pp. 191–201, Aug. 2007.

    [24] G. Baumann, “Controlling the wireless field,” Indust . Wireless Book, vol. 9, no. 3, June 2006.

    [25] OMRON, (2002, Sept.) WD30-ME/-SE/-ME01/-SE01 DeviceNet Wireless Units—OperationManual. [Online]. Available: http://www.om-ron.com/

    [26] Media Access Control (MAC) Bridges , IEEE Stan-dard 802.1D, June 2004.

    [27] User Datagram Protocol , DARPA Standard RFC768, Aug. 1980.

    [28] Ethernet Powerlink Standardization Group,“Ethernet Powerlink communication profilespecification,” ver. 2.0, 2003.

    [29] G. Santandrea, “A Profinet IO applicationimplemented on wireless LAN,” presented at6th IEEE Int. Workshop on Factory Communica- tion Systems (Industry Day), Torino, Italy, June2006. (Slides available: http://wfcs2006.ieiit.cnr.it/)

    [30] 3Com. (2006, June 19). 3Com Wireless 776011a/b/g PoE Access Point User’s Guide. [On-line]. Available: http://www.3com.com/

    [31]  Logical Link Control   (with amendments 3, 6and 7), IEEE Standard 802.2, 1998.

    [32] The ZigBee Alliance (2006, Dec.) “ZigBee specifi-

    cation,” [Online]. Available: www.zigbee.org/en/spec_download/zigbee_downloads.asp