102 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 4, NO. 2, MAY 2008
Recent and Emerging Topics in Wireless
Industrial Communications: A Selection
Andreas Willig, Member, IEEE
Invited Paper
Abstract—In this paper we discuss a selection of promising and
interesting research areas in the design of protocols and systems
for wireless industrial communications. We have selected topics
that have either emerged as hot topics in the industrial commu-
nications community in the last few years (like wireless sensor net-
works), or which could be worthwhile research topics in the next
few years (for example cooperative diversity techniques for error
control, cognitive radio/opportunistic spectrum access for mitiga-
tion of external interferences).
Index Terms—Cognitive radio, IEEE 802154, industrial-QoS,
spatial and cooperative diversity, ultra-wideband, wireless sensor
networks, ZigBee.
I. INTRODUCTION
F
OR most people the significance of wireless technologies
comes from its ability to provide services like voice/video
transmission or Internet access at places without cabled net-
working infrastructure or while being on the move. Wireless
technologies have also been identified as a very attractive op-
tion for industrial and factory automation, distributed control
systems, automotive systems and other kinds of networked em-
bedded systems [1], [2], with mobility, reduced cabling and in-
stallation costs, reduced danger of breaking cables, and less
hassle with connectors being important benefits. Some poten-
tially interesting classes of industrial applications are closed-
loop control involving mobile subsystems, coordination among
mobile robots or autonomous vehicles, health monitoring of
machines, tracking of parts and many more. An important char-
acteristic in these application areas is that (wireless) data com-
munications must satisfy tight real-time and reliability require-
ments at the same time, otherwise loss of time and money or
even physical damage can result. To achieve this goal, on the one
hand certain functionalities that are specific for wireless com-
munications (like mobility management, quick handovers) must
be considered, and on the other hand the unfriendly error prop-
erties of the wireless channel significantly challenge real-time
and reliability. Consequently, significant research is needed to
adapt existing wireless technologies and protocols to industrial
settings, or, when this is not sufficient, to develop new ones. This
research has been done with significant intensity for more than
Manuscript received February 19, 2008; revised March 14, 2008. Paper no.
TII-07-12-0198. R1.
The author is with the Telecommunication Networks Group (TKN), Technical
University of Berlin, Berlin, Germany (e-mail: awillig@ieee.org).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TII.2008.923194
one decade now, and in [1] a selective review and tutorial on re-
search issues and approaches has been given.
This paper is a followup to [1]. Our main goal is to discuss
a selection of promising and interesting research areas that
received no or only limited coverage in [1] and in the industrial
communications community. We start with a brief overview on
the quality of service (QoS) features that have been added to the
ubiquitous IEEE 802.11 wireless LAN standard (and resulting
in IEEE 802.11e) and which are also interesting for use in
industrial scenarios. Following this, the paper covers wireless
sensor networks. Wireless sensor networks have recently re-
ceived increased attention in the industrial communications
community. They differ considerably from wireless LANs.
Sensor networks support much lower data rates and much
smaller transmit powers. More fundamental to the design of
sensor networks, however, is that sensor nodes have a severely
limited energy budget and consequently energy-efficiency is
the single most important figure of merit. One consequence of
this is that sensor nodes should spend most of their time in a
sleep state in which they are not able to transmit or receive data.
These properties do not favor the adoption of sensor networks
in tight control loops. Instead, they are mostly considered for
less time-critical monitoring tasks like for example monitoring
machine health or leakage monitoring. We provide an introduc-
tion to important concepts of sensor networking and discuss
a number (by far not all) of protocol design issues that are
relevant for industrial applications.
In the second main part of this paper we discuss approaches
that we believe can have a significant impact on the future de-
sign of wireless industrial communication protocols. In partic-
ular, we introduce recent techniques to mitigate channel fading
and external interferences (two of the main reasons for the bad
quality of the wireless channel!) that are currently hot topics in
the wireless communications community and from which the
industrial networking community can significantly benefit. For
the sake of completeness we have also included a brief discus-
sion of an existing commercial systems for wireless industrial
communications: the WISA system from ABB.
We had to make choices on what to include in the paper. In
terms of protocols we have mostly favored topics related to the
lower layers of the industrial protocol stack (i.e., the MAC and
the link-layer with its error control functionality) and their prop-
erties in terms of real-time and reliability. In terms of technolo-
gies we selected topics that have received only limited or no
coverage at all in [1]. For example, we have included WSN tech-
nologies and IEEE 802.11e, but we have mostly left out Blue-
tooth or plain IEEE 802.11 technologies. We have furthermore
1551-3203/$25.00 © 2008 IEEE
WILLIG: RECENT AND EMERGING TOPICS IN WIRELESS INDUSTRIAL COMMUNICATIONS: A SELECTION 103
favored a tutorial-style exposition discussing fundamental is-
sues and solution approaches over the detailed discussion and
comparison of specific solutions from the literature. To help the
reader to delve further into these solutions and approaches, we
provide a fair number of references. We also aim to point out
interesting research questions.
The paper is structured as follows: in Section II we provide
a broad overview on the general research areas that need to be
addressed for wireless industrial networking. Since the focus of
the remaining paper is mostly on real-time and reliability prop-
erties, we also discuss appropriate performance measures. Fol-
lowing this, in Section III we describe the QoS enhancements
to the IEEE 802.11 standard and point to an interesting research
issue. In Section IV we begin our discussion of wireless sensor
networks by explaining their fundamentals. In Section V we
present the IEEE 802.15.4, ZigBee and ISA SP-100 standards
for wireless sensor networking, since these can be expected to
have significant impact in the industrial field. In Section VI we
briefly look at the vast problem of providing real-time and relia-
bility in multihop wireless sensor networks. The first part of the
paper concludes with Section VII, in which the existing wireless
industrial communication systems WISA is briefly reviewed.
In the second part we discuss more general research issues
and selected topics from the field of wireless communications
that are probably relevant for wireless industrial communica-
tion systems as well: spatial/cooperative diversity techniques in
Section IX, the general issue of industrial QoS provisioning and
analysis in Section X, the usage of cognitive radio techniques for
mitigating external interferences in Section XI and the adoption
of ultra-wideband technologies in Section XII. The paper ends
with our conclusions in Section XIII.
II. O
VERVIEW ON RESEARCH
AREAS FOR WIRELESS
INDUSTRIAL NETWORKING
To properly design networks and protocols for wireless indus-
trial networking, several issues have to be addressed, including
the following ones:
• Providing the required QoS in terms of reliability and
real-time to applications: design of protocols and of
wireless channel resource allocation schemes (frequen-
cies, transmit power, rate [as determined by modulation
and coding scheme], time budgets), as well as analysis
schemes that evaluate achievable QoS over wireless fading
channels. The relevant industrial-QoS measures are dis-
cussed in more depth in Section II-A.
• Engineering and network planning: a comprehensive set
of methodologies and tools for network planning, dimen-
sioning and configuration, as well as run-time fault and per-
formance monitoring needs to be developed. This is tightly
coupled to resource allocation. Some references for net-
work planning in the industrial context are [3] and [4], a
platform-based protocol framework allowing to adjust pa-
rameters according to pre-specified QoS levels is presented
in [5].
• MAC protocol design: the MAC layer is a key functionality
for (wireless) industrial communication systems, since it
directly impacts the timeliness of packets. The goal is to
find deterministic protocols that can support packet prior-
ities or which allow fine-grained channel scheduling. We
discuss in Section III-B briefly the specific issue of pri-
ority enforcement on wireless channels, a technique that is
needed to implement CAN-like protocols and to leverage
existing work on schedulability analyses for CAN.
• Error-control schemes: error-control schemes directly im-
pact the achievable reliability. Recent approaches to error
control are addressed in more detail in Section IX.
• Routing and transport protocols: especially in multihop
networks like wireless sensor networks (soft) real-time
guarantees have to be provided over multiple hops. This
will be discussed in more detail in Section VI.
• Application-layer protocols: application support protocols
and the applications themselves must be designed with ex-
plicit consideration of the wireless channel properties in
mind. One example is the research area of networked con-
trol systems (see [6] and [7]). An alternative view of appli-
cation-layer protocols is taken for example in [8], where
the authors argue that on top of commercial wireless hard-
ware like IEEE 802.11 WLAN it is the application layer
that must ensure appropriate real-time and reliability prop-
erties through specifically designed protocols.
• Hybrid wired/wireless systems: in many applications it is
beneficial to adjoin wireless stations to existing wired net-
works and therefore to create
hybrid networks. This has in
general been discussed in [1] and [9], whereas specific net-
works and protocols have for example been considered in
[10]–[13].
• Mobility support and handovers under real-time and reli-
ability constraints. The design of suitable schemes, espe-
cially for hybrid systems, depends on the underlying MAC
protocols. One example reference is [14], in which mo-
bility and handovers are considered for hybrid wired/wire-
less PROFIBUS systems.
• Security and privacy: Security in general and wireless se-
curity in particular are vast research topics [15], security
aspects for industrial networks are discussed in [16]. A
brief account on some issues is provided in Section II-B.
• Energy consumption and energy-efficient design: To fully
achieve the benefits of wireless communications, no ca-
bling at all should be used, and this includes also the energy
supply. This issue is discussed in more detail in Section IV.
• Scalability: in a factory plant both the number of wire-
less networks as well as the number of wireless nodes per
network might become large. This is especially true for
wireless sensor networks which must be significantly over-
provisioned in order to give individual nodes enough op-
portunity to enter a low-energy sleep state while ensuring
that the overall network has still enough awake nodes to
achieve its task.
In the remainder of this section we discuss relevant quality of
service measures for wireless industrial networks, followed by
a brief account of security issues.
A. Quality of Service Measures for Wireless Industrial
Networks
Industrial networks are in general designed to carry traffic
that is dominated by exchanges of sensor readings and actu-
ator commands between sensors/actuators on the one hand and
(often centralized) controllers on the other hand [2], [17], [18].
Important characteristics of industrial traffic are the presence of
104 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 4, NO. 2, MAY 2008
deadlines, high reliability requirements and the predominance
of short packets [19]. For these exchanges two different interac-
tion patterns have emerged: the master/slave model and the pro-
ducer-(distributor)-consumer model, and for these patterns dif-
ferent QoS parameters are important. Since the wireless channel
is random and time-varying, the classical deterministic perfor-
mance measures like for example the worst-case transmission
times should be replaced by probabilistic measures.
In the master/slave model (for example adopted in the
PROFIBUS [20], [21]) information is exchanged between
controller and sensors/actuators by using a unicast commu-
nication mechanism. In this setting the key aspect of service
quality is to enable transmission of periodic or sporadic mes-
sages within pre-specified
deadlines and in a reliable fashion.
Reliability is especially important for critical alarm mes-
sages. There are different performance measures that jointly
account for real-time and reliability. Such measures can be
defined on different timescales: in terms of long-term averages
or on much shorter timescales. One measure belonging to
the first category is the success probability,defined as the
long-term probability that a message can be successfully (i.e.,
acknowledged) transmitted within its deadline. Another one
is the capacity-vs-outage-probability measure [22], denoting
the transmission rates (and therefore the delay) so that the
probability of not achieving this rate is below a pre-specified
threshold. A third measure is the delay-limited capacity,itis
defined as the capacity-vs-outage probability measure for a
pre-specified outage probability of zero (see also the discussion
in [23]).
1
For short timescales formulations based on the notion
of
-firm deadlines [24], [25] can be used, in which at
most
out of any consecutive packets are allowed to be lost,
otherwise a deadline violation occurs.
In the producer-(distributor)-consumer model (for example
adopted in WorldFIP [26], [27]) communication is based on
unacknowledged broadcasts of data identifiers (by the distrib-
utor), to which the station possessing the identified data item
(the producer) responds by broadcasting its current value. All
nodes interested in this data (the consumers) copy the received
value into an internal link-layer buffer for later delivery to the
higher layers. This can be regarded as an instance of a pub-
lish/subscribe interaction pattern [28]. Here the most important
performance measures are related to the degree by which all the
consumers are able to simultaneously capture the data and to
maintain consistent buffer states—this is referred to as spatial
consistency.
2
Such a consistency requirement can be captured
by the agreement probability, i.e., the probability that all con-
sumers have reached agreement within a certain, pre-specified
time window.
1
Technically, these measures are defined for stationary and frequency-flat
block fading channels. In a block fading channel, the packet duration is smaller
than the channel coherence time, i.e., the time during which the channel does
not change its characteristics (appreciably). The term frequency-flat refers to a
situation in which all the involved frequencies of a transmitted signal have the
same attenuation level. For industrial networks with predominant small packet
sizes a block fading channel is a reasonable assumption, and when the transmis-
sion rates are reasonably small then the channel is also frequency-flat.
2
A similar consistency requirement is called relative temporal consistency:
Many control applications require that the relevant sensors sample the environ-
ment nearly simultaneously in a pre-specified time window. In some industrial
communication systems this is achieved by using explicit triggering signals,
broadcast by the central controller. Again, the degree to which all the relevant
sensors receive the trigger packet is important.
Those previously defined performance measures represent
the prime performance metrics for wireless industrial com-
munication systems—we refer to them as industrial-QoS.A
range of secondary metrics can be devised that measure the
efficiency of protocol mechanisms attempting to improve the
primary performance measures. Examples of such secondary
metrics are the protocol or memory overhead, computational
overheads, additional interference created, and others.
B. Security
Today’s automation networks tend to be more and more inte-
grated with other networks, for example to allow cost-effective
remote monitoring and maintenance of machine plants. There
are many techniques to protect a network against attackers from
outside, for example firewalls [15]. But when the network uses
wireless transmission, an attacker that is close enough to the
network can eavesdrop, it can insert malicious packets, or it can
simply jam the wireless medium and distort any other transmis-
sion, this way challenging reliable and timely transmission.
Encryption can be used to prevent eavesdropping. To prevent
insertion of malicious packets, mechanisms for ensuring au-
thentication (“who sent this message?”) and message integrity
(“is this the message originally sent?”) are needed [15], [29] to
create mutual trust relationships between wireless stations. Such
mechanisms are often implemented using shared secrets and
public key cryptography, calling in turn for proper key distri-
bution schemes. To avoid replay attacks proper sequence num-
bers/session keys have to be used. Some challenges for imple-
menting security mechanisms in wireless industrial networks
and wireless sensor networks are the following [30, Sec. 1.2]:
• Ensuring authentication and message integrity for each
message requires message integrity check (MIC) fields in
each message. To be effective, this field should have a rea-
sonable minimal length, for example 16 bytes. However,
since in most fieldbus systems the maximum allowable
frame size is small (and many packets have only a few
bytes anyway), the MIC fields account for significant
fraction of overhead.
• Key distribution schemes introduce significant protocol
and administrative overhead.
• In the case of hybrid systems, one has to take into account
that often the legacy fieldbus protocols running in the wired
stations do not have any security mechanisms.
III. IEEE 802.11E
The IEEE 802.11 family of wireless LAN (WLAN) standards
is certainly predominant in the realm of WLAN technologies,
and it has also been considered extensively in the context of
wireless industrial communications, see for example [31]–[38].
In this paper we focus on aspects of the current standard that
have gained increased importance since publication of [1] and
which at the same time are especially interesting for industrial
applications, namely the QoS support that is now part of the
2007 version of the standard [39] and which was formerly spec-
ified in a separate amendment [40].
3
We therefore give a brief
3
The 2007 version of the standard supersedes the 1999 version and its 2003
reaffirmation, and includes also previous amendments to the IEEE 802.11 stan-
dard like IEEE 802.11a (an OFDM physical layer for the 5.2-GHz ISM band),
IEEE 802.11b (11 Mb/s physical layer for the 2.4-GHz band), IEEE 802.11g (a
high-rate physical layer for the 2.4-GHz ISM band, including 54 Mb/s OFDM),
IEEE 802.11i (security enhancements) and IEEE 802.11e (quality of service).
WILLIG: RECENT AND EMERGING TOPICS IN WIRELESS INDUSTRIAL COMMUNICATIONS: A SELECTION 105
introduction into the new QoS mechanisms and point to some
research issues.
A. IEEE 802.11E Quality-of-Service Support
The QoS functions are available in infrastructure IEEE
802.11 networks. They consist of a hybrid coordination func-
tion (HCF) that operates on top of a (modified) distributed
coordination function (DCF) as it is known from the orig-
inal IEEE 802.11 standard. For some parts of the hybrid
coordination function a centralized control entity called hy-
brid coordinator (HC) is required which is co-located with
an access point. Amongst other functionalities, the HC can
perform admission control. More precisely, the stations in a
QoS-enabled IEEE 802.11 network (henceforth simply referred
to as network) can send a request to the HC asking him to
schedule transmit opportunities (TXOP). A TXOP is a con-
tiguous window of time which a station can use exclusively
(i.e., without having to expect parallel transmissions from other
network members) to transmit one or more frames including
MAC layer acknowledgements to any station in the network. It
is possible that a station requests the HC to grant these TXOPs
periodically (for example for voice or video data streams) and
upon receiving such a request the HC has to decide whether the
new flow can be admitted without breaking guarantees for the
already present flows. The precise admission control algorithm
is not prescribed by the standard but left to the implementers.
The standard provides two different access schemes for pro-
viding QoS support: the enhanced distributed coordination ac-
cess (EDCA) and the hybrid coordination function controlled
channel access (HCCA). We describe both of them brieflybut
start with a reminder of the basic distributed coordination func-
tion (DCF) upon which both access schemes are built. Many
details are left out, for example the RTS/CTS scheme.
1) DCF and EDCF: The (E)DCF belongs to the class
of carrier-sense multiple access with collision-avoidance
(CSMA-CA) protocols. It relies on the physical carrier-sense
function of the underlying physical layer, which indicates the
presence or absence of signals or ongoing transmissions on
the wireless medium. In addition, a station performs a virtual
carrier-sense operation. In this operation, the station maintains
a special variable called network allocation vector (NAV).
Most of the packets in IEEE 802.11 contain a duration field,
which denotes the remaining time that is needed until the
ongoing transaction (for example data plus acknowledge) is
finished. Whenever a station receives a packet, it updates its
NAV variable with the packets duration field in order to prevent
own transmissions during the remaining transaction time. In
summary, a carrier-sense operation indicates an idle channel
only if the NAV is zero and the physical carrier-sense does not
indicate any transmission activity.
For the following description please refer also to Fig. 1. When
a new packet shall be transmitted, the station performs a carrier-
sense operation. When the medium is idle for a certain amount
of time called inter-frame space, the station starts to transmit.
The packet at hand is associated with one of four pre-defined
access classes and the inter-frame space that the station de-
pends on this access class. For the
-class the inter-frame space
Fig. 1. Timing of the IEEE 802.11 DCF.
is called . These values can be configured, but
are required to be distinct.
When the medium is busy, the station enters the backoff mode.
It station first waits until the medium is idle for a time of at least
distributed inter-frame space (DIFS). At this time the backoff
slots start. If the station was not in backoff mode before, it draws
a random number out of the current contention window and sets
a counter with this number. In the following, during each idle
slot the counter is decremented by one and if it reaches zero, the
station starts to transmit. If the carrier-sense mechanism indi-
cates a busy medium, the process of decrementing the backoff
counter is suspended, and it is resumed later on once the medium
becomes idle again.
The contention window size is dynamic. It is initialized with
a pre-configured value
. Whenever a packet transmis-
sion is not successful, the contention window size is doubled,
until a maximum value
is reached. This is useful when
transmission failures are interpreted as resulting from channel
collisions, since an increase in contention window size leads to
increased average backoff times and therefore to a reduction of
the pressure on the channel.
We are now in the position to briefly explain the difference
between the classical DCF and the EDCF. In the DCF there
are no access classes and all stations use the DIFS instead of
the
inter-frame spaces. Furthermore, all stations use the
same values for
and . This means that all sta-
tions have the same chance to access the channel. In the EDCF
for each access class
separate values of , and
can be configured at the access points, which then dis-
tributes these values to all stations in its beacons. By choosing
proper values for these parameters it is possible to ensure with
high probability that a packet of a better access class wins over
a contender having a packet of lower access classes. In other
words: it is possible to achieve stochastic prioritization or ser-
vice differentiation. A further difference between the DCF and
the EDCF is the following: in the DCF winning contention ac-
quires the right to transmit one packet. In the EDCF the winner
receives a TXOP which, as explained above, may encompass
several packets. The maximum duration of a TXOP is a pre-con-
figured value. However, only packets belonging to the access
class for which the TXOP was won are allowed to be transmitted
during a TXOP.
A number of performance analyses (e.g., regarding
throughput) of the DCF and EDCF are available in the lit-
erature [34], [41]–[45]. These studies confirm also that the
EDCA mechanism is indeed capable of achieving stochastic
prioritization.
106 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 4, NO. 2, MAY 2008
2) HCCA: In the HCCA a central entity, the HC, is respon-
sible for coordinating access to the medium. The HC receives
reservation requests for TXOPs, grants or rejects these requests
(admission control) and is responsible for actually scheduling
the TXOPs for all attached stations. The channel is, however,
not all the time under full control of the HC. Instead, scheduled
TXOPs alternate with phases in which stations contend for the
medium using the EDCF. The reservation requests can either be
sent as separate packets (using the EDCF) or they can be piggy-
backed onto data packets that are sent during a scheduled TXOP.
To gain control over the channel the HC uses also the basic
DCF mechanism, but the medium has only to be idle for PIFS
time (which is smaller than DIFS and all the
values)
and after that time the HC starts with its transmissions. Specifi-
cally, to start a TXOP the HC sends a poll packet to the owner of
the TXOP and the owner then uses the TXOP to transmit one or
multiple packets according to the results of a local scheduling
policy (which, however, typically prefers better access classes
over lower ones). The admission control policy and the sched-
uling policies applied in the HC and the stations are not specified
in the standard.
The HCCA is rather complex as there are rich interactions
with other features of the IEEE 802.11 protocol like for example
the rate adaptation feature of some of the physical layers. It is
similarly complex as the point coordination function of the orig-
inal IEEE 802.11 standard, which is still present in the 2007 ver-
sion and which can be operated jointly with the HCCA. The au-
thor is not aware of any existing implementation of the HCCA.
One weakness of the HCCA that prevents it from achieving
perfectly periodic services is that at the scheduled time of
TXOPs the medium might still be busy from a previous TXOP
obtained by the EDCF mechanism and the HC has to wait a
random time until the packet end before he gains access to
the medium. A key research issue in the HCCA is the design
of appropriate admission control scheduling policies, which
has for example been done in [46]–[48]. Its usage in industrial
scenarios has been considered in [33] and [49].
B. Research Issues
Assuming that HCCA implementations will not become
widely available during the next few years (PCF implemen-
tations actually never did), it makes sense to concentrate on
improvements of the DCF and the EDCF.
One very interesting and relevant problem is the determin-
istic priority enforcement on the wireless channel. The problem
is defined as follows: given two devices
and which have
packets ready for transmission at the same time and whose trans-
mission ranges overlap, it should be deterministically ensured
that
’s packet can be transmitted before ’s when ’s packet
has a higher priority and
’s receiver is in the overlap area of
’s and ’s ranges. In other words: ’s less important packet
should not block
’s transmission. In a probabilistic variant of
priority enforcement the protocol must make it more probable
that
wins than that wins. This probabilistic variant is indeed
implemented by the EDCF, supporting four different priorities.
Some industrial and automotive communication systems like
for example CAN [50] employ a MAC layer technique in which
to each packet a priority value is assigned and these priorities are
then used to resolve contention among different stations. When
the priority enforcement is deterministic, a proper assignment
of priorities to packets allows to perform deterministic schedu-
lability analyses [51], [52].
In the CAN protocol, a bitwise priority-arbitration technique
is used for collision resolution: a contending station awaits the
end of an ongoing transmission (if any) and then enters con-
tention phase. The contention phase is driven by the priority
of the packets ready in the contenders and proceeds bit-by-bit.
A contender transmits the value of the current priority bit and
simultaneously receives feedback from the channel (which is
guaranteed to have a well-defined level and can be converted
back into a bit). When the own transmitted bit and the bit read
back from the channel differ, the station has lost contention and
defers, otherwise the station proceeds with the next bit.
This approach cannot be directly implemented with commer-
cial wireless transceivers, since it requires full-duplex operation
of the transceiver. Wireless transceivers are in general half-du-
plex: they are not able to transmit and receive simultaneously
on the same channel because their own signals would drown all
signals from other stations. Because of this fact, most wireless
transceivers share some circuitry between transmit and receive
path, which naturally prevents that they work in parallel.
Therefore, alternative mechanisms are needed to enforce
packet priorities on the channel. There are several options for
(almost) deterministic priority enforcement in fully meshed
networks. By the rules of (persistent) CSMA protocols, a
contention cycle starts at the end of a previous packet. One
possibility is to let contending nodes send jamming signals
of length according to the priority of their current packet.
Afterwards, a node switches to receive mode and performs a
carrier-sense operation to check whether any other contender
emits a longer jamming signal. If so, the listening station defers
and the other station has won the contention. This approach,
while having been used in the HIPERLAN-I standard [53], is
unfortunately not implementable with commercial IEEE 802.11
transceivers as these do not offer the generation of jamming
signals. A complementary and more practical approach is to
let nodes listen on the channel for a time proportional to their
packets priority (the more important the packet, the shorter a
stations listens for other stations and the earlier the station starts
to transmit its packet). In both cases, the maximum length of
bursts/listening periods is linear in the number of priorities that
can be supported. In the wireless dominance protocol (WiDom)
approach presented in [54] a bitwise priority arbitration scheme
is mimicked by providing one time slot for each priority bit,
i.e.,
slots for priority bits. During such a timeslot a station
having a dominant bit transmits, while stations with recessive
bits receive. When a recessive station receives a signal, it has
lost contention and gives up. With this approach, the number of
priorities that can be distinguished is
, while the duration of
the contention resolution period is linear in
.
However, all these approaches share some problems and
therefore need additional research:
• The above discussed deterministic schemes all rely on car-
rier-sensing and are thus vulnerable to external interfer-
ences—any unwanted signal that is detected by the car-
rier-sense algorithm leads to deferral of transmissions and
higher probabilities of deadline misses.
• They do not work in hidden-terminal situations, i.e., in
settings where two stations having packets of different