Hoai Hoang and Jerker Bengtsson
CERES - Centre for Research on Embedded Systems, Halmstad University, Halmstad, Sweden
Real-time communication, wireless network, QoS.
IEEE 802.15.4 is a new enabling technology for low data rate wireless personal networks. This standard was
not specifically designed for wireless sensor networks, but it has shown to be a good match with necessary
requirements on low data rate, low power consumption and low cost. Unlike the former 802.11 standard,
the MAC protocol specified in IEEE 802.15.4 can operate in two different modes: beacon-enabled mode or
non-beacon enable mode. In beacon-enabled mode, nodes can exclusively allocate a number of guaranteed
time slots, similar to a resource reservation scheme. Hence, the IEEE 802.15.4 MAC protocol have sufficient
capabilities for supporting real-time communication. This paper presents the key features of IEEE 802.15.4
which makes it an attractive standard to use for real-time wireless sensor networks. Two real-time protocols
extending the IEEE 802.15.4 standard are reviewed. The purpose of this paper is to present the state of the
art on real-time support over IEEE 802.15.4 for wireless sensor networks and to discuss the possibilities on
improvements on both the standard and the real-time protocols extending the standard.
Supporting real-time (RT) communication over Wire-
less Sensor Networks (WSNs) has attained an increas-
ing interest from both industry and academia. During
the recent decade, advances in WSN enabling tech-
nology has opened up for many proposals of new
innovative applications. However, many of the pro-
posed applications require support for RT commu-
nication; guarantees for data being delivered from a
source node to a destination node within its given end-
to-end deadline. For example, a surveillance system
needs to alert the detection of an intruder within sec-
onds for the authorities to be able to initiate pursuing
actions in time (He et al., 2006). Depending on the
specific application, different kinds of data have dif-
ferent deadline requirements due to its validity inter-
val. Data registering the movement of a tank in a mili-
tary surveillance system have shorter update intervals
than data registering transport pallets being moved in
and out from a goods storage. Similarly, the loca-
tion registration of an intruder have a shorter valid-
ity interval than the temperature measurement in au-
tomatic room temperature control system (Lu et al.,
2002). Thus, there is a need for WSNs to support
time-sensitive communication by guaranteeing pack-
ets end-to-end deadlines or minimizing the deadline
miss ratio, i.e., the number of packets that meet their
end-to-end deadline.
Supporting RT communication in WSNs is a great
challenge due to many reasons (Stankovic et al.,
2003). The communication in WSN per se is less
reliable and less predictable than in wired networks
because of its wireless nature. Moreover, a WSN
in general consists of an arbitrarily number of nodes
equipped with highly limited resources; at the one
hand batteries need to be small and at the other hand
they need to be operable during a long time. There-
fore a major factor that must be considered in the de-
velopment of a communication protocol for WSNs is
power-efficiency. Most of the earlier reported work
on RT communication protocols for WSNs have to
a large an extent been focused on the trade-off be-
tween energy efficiency and end-to-end delay guaran-
tees (He et al., 2003; Chipara et al., 2006). Still it re-
mains as one of the most important issues for research
community. Other characteristics that complicates RT
support is network topology. In many proposed ap-
plications, the topology is required to be highly dy-
namic: nodes can relocate, join or leave the network.
Besides making routing more complicated, the work-
loads in such dynamic topologies become highly un-
predictable. Clearly, there is a need for new schedul-
ing methods which can adapt efficiently to such dy-
Hoang H. and Bengtsson J. (2008).
In Proceedings of the International Conference on Wireless Information Networks and Systems, pages 82-87
DOI: 10.5220/0002023800820087
namic environments. A more detailed view of RT
challenges in WSNs can found in (Stankovic et al.,
2003; Li et al., 2007).
In December 2003, IEEE introduced the IEEE
802.15.4 standard (IEEE, 2003) - covering the
Medium Access Control (MAC) layer and the phys-
ical layer - for Low-Rate Wireless Personal Area Net-
works (LR-WPAN). The IEEE 802.15.4 protocol is
particularly suitable for WSNs because it is mainly
specified for: low data rates, providing reliable data
transfer, low cost and low power consumption. Espe-
cially interesting for RT applications is that the IEEE
802.15.4 MAC protocol has the capability of provid-
ing Guaranteed Time Slots (GTS) for time-sensitive
data, similar to traditional TDMA allocation. The
IEEE 802.15.4 protocol has been widely adopted as
one of the more interesting candidates for develop-
ing RT communication protocols for WSNs. The
GTS transmission scheme allows each node in the
network to send data, using a certain number of time
slots according to the first-come-first-served (FCFS)
algorithm. However, due to its simplicity, only static
schedules can be implemented using FCFS based
scheduling. This problem has motivated a number of
researchers to work on more efficient RT scheduling
methods, enabling dynamic allocation of time slots.
This paper presents the most relevant characteris-
tics of the IEEE 802.15.4 - the de facto standard for
time-sensitive WSN applications - for RT scheduling
methods. Two of the more interesting RT schedul-
ing methods developed for this standard are reviewed
(Koubaa et al., 2006b; Mishra et al., 2007). The
rest of the paper is organized as follows: section
2 introduces to real-time concepts and the main re-
quirements for developing a RT communication pro-
tocol. Section 3 provides a brief description of the
IEEE 802.15.4 standard. In section 4, we discuss
two related allocation methods: one method based
on fix-priority scheduling and one based on dynamic-
priority scheduling. Finally, we make a concluding
remark of our review findings and discuss possibili-
ties for future work in section 5.
The characteristics of RT communication differ from
the characteristics of traditional best-effort communi-
cation. For example, in best-effort communication,
the performance measurement of a communication
protocol is the network throughput. In RT communi-
cation the performance of the protocol is measured by
the number of messages successfully delivered before
their deadlines. RT communication is usually divided
into two main categories: hard RT and soft RT. In soft
RT communication, some amount of packet loss can
be tolerated, while in hard RT communication there is
no toleration for any packet loss. In hard RT commu-
nication, all messages have strict timing constraints.
For example, if any packet is delivered late, it is con-
sidered as being lost. In hard RT communication, an
upper bound on end-to-end message delays or a de-
terministic bound is required. In contrast, for sup-
porting soft RT communication over the network, a
probabilistic end-to-end delay bound should be pro-
For WSNs, it is more realistic to offer a probabilis-
tic end-to-end guarantees, due to facts that: the net-
work conditions varies with time, the traffic is unpre-
dictable and constraints on power-efficiency requires
i.e., limited transmit power. A common approach
taken for supporting end-to-end guarantees in multi-
hop networks in general, and in WSNs specifically, is
that the MAC protocol should provide a delay bound
for a single-hop, and the network layer should take the
responsibility for multi-hop delay bounds. Schedul-
ing methods are commonly used in the MAC protocol
(scheduled-based MAC protocols). Finding efficient
scheduling algorithms for the MAC protocol is one of
the most important issues for RT communication in
3 IEEE 802.15.4
The IEEE 802.15.4 protocol specifies the MAC layer
and the physical layer for LR-WPANs. This proto-
col is often used in conjunction with ZigBee proto-
col (Zigbee Alliance, 2004) to provide a full protocol
stack. This section presents the relevant features of
the IEEE 802.15.4 standard related to RT communi-
cation over WSNs. Further details of the standard can
be found in (IEEE, 2003; Koubaa et al., 2005).
3.1 Network Devices and Topologies
Network Devices. The IEEE 802.15.4 standard
specifies two types of network devices that are sup-
ported: a Full Function Device (FFD) and a Reduced
Function Device (RFD). An RFD is a simple de-
vice operating with a minimal implementation of the
IEEE 802.1.5.4 protocol. An FFD is a device that
has the capability of operating in two modes; either
as a coordinator, providing a synchronization service
through the transmission of beacons, or, as a simple
device but implementing the complete protocol set.
An FFD can act as both a simple coordinator and
a PAN-coordinator. A Coordinator is a network de-
vice that provides synchronization services through
the transmission of beacon frames. If a coordinator
has the function of identifying and controlling the net-
work, it is called the PAN coordinator. A network
should include at least one FFD performing as PAN-
Network Topologies. The IEEE 802.15.4 protocol
supports networks organized in both star-topology
peer-to-peer topology. In star topology, communi-
cation is established between network devices (FFD
and RFD) and the PAN-coordinator. After an FFD
is activated the first time, it may establish its own
network and become the PAN coordinator acting as
the network’s controller. A PAN coordinator can al-
low other devices to join its network. The communi-
cation paradigm in star network topology is central-
ized. In peer-to-peer topology, any network device
can communicate with any other devices within its
radio range. There is also a PAN coordinator in the
peer-to-peer topology. In contrast to the star topol-
ogy, a device can be nominated as PAN coordinator
by another device in the peer-to-peer topology. A net-
work with peer-to-peer topology can be ad hoc, self-
organizing, self-healing, and allows multiple hops to
route data from any source node to any destination
3.2 IEEE 802.15.4 MAC Layer
The MAC protocol specified in IEEE 802.15.4 can
operate in two different modes: beacon-enabled
mode and non beacon-enabled mode.
In beacon-enabled mode, the coordinatorgenerate
beacons periodically in order to synchronize all nodes
in the PAN and to identify the PAN. Data exchanged
between all nodes and the PAN coordinator has a spe-
cial format, which is defined by the PAN coordina-
tor, called superframe structure. The length of a super
frame is equal to the interval between two consecutive
beacons, and it is divided into 16 equally sized time
slots. Figure 1 illustrates some parameters embedded
in a superframe.
A superframe can either include both an active-
and an inactive period, or just an active period. Dur-
ing the inactive period, the coordinator do not have
any interaction with other nodes and it stays in a low
power-mode. The first slot in a superframe is used for
the beacon frame.
CAP. The contention access period (CAP) starts
immediately after the beacon frame. Transmis-
sions occurring during the CAP period must fol-
low a slotted-CSMA/CA mechanism. Nodes
wishing to transmit must compete for the com-
munication medium based on CSMA/CA with
some back-off period, aligned with superframe
slot boundaries.
CFP. In difference to other IEEE standards
for wireless communication, such as the IEEE
802.11, the IEEE 802.15.4 do not only offer the
CSMA/CA mechanism for the MAC protocol.
The IEEE 802.15.4 MAC protocol may also in-
clude a contention free period (CFP) within the
superframe. CFP is an optional field, which has
been defined for QoS requirements on for exam-
ple low-latency or a specific transmission band-
width. A PAN coordinator can allocate a cer-
tain number of time slots called Guaranteed Time
Slots (GTS), during the CFP upon on request. All
transmissions occurring during the CFP do not
follow the CSMA/CA mechanism. The PAN co-
ordinator can assign up to 7 GTSs and can only be
used for communication between a node and the
PAN coordinator.
GTS. A GTS scheme is a kind of resource reser-
vation in WPANS. It is a portion of the super-
frame, which is allocated exclusively to a given
node in the PAN. When a node have data with
timing constraints to transmit, it sends a request
frame to the PAN coordinator asking for GTSs.
The PAN coordinator will do admission control to
check whether the request can be accepted or not.
The GTSs are allocated following the basic first-
come-first-served (FCFS) rule. A node is assigned
GTSs if there are resources available, i.e. if the
number of requested slots are less than or equal
to the number of free slots in the superframe du-
ration. Node with allocated GTSs can access the
communication medium without contention dur-
ing the CFP. A node transmitting during a GTS
must guarantee that its transmission end before
the next GTS start. There are maximum seven
GTSs that can be used at the same time.
In non beacon-enabled mode, there are neither
beacons nor superframes. Nodes communicate with
each other freely according to the unslotted CSMA
mechanism. Both the non beacon-enabled and the
beacon-enabled mode use the CSMA/CA with back-
off periods, but in the former the back-off period for
each node is independent from other nodes back-off
WINSYS 2008 - International Conference on Wireless Information Networks and Systems
Free Period
Contention Access
Beacon frame
Figure 1: Structure of a supperframe.
OVER IEEE 802.15.4
The GTS scheme in beacon-enabled mode is more in-
teresting when it comes to supporting RT communi-
cation over the IEEE 802.15.4. However, there are
still some limitations that might make it less effi-
cient (Koubaa et al., 2006b). Firstly, since the GTS
allocation is managed using simple FCFS, there is
no priority support. All types of traffic are treated
equally. Secondly, it might also happen that a node
with low arrival rate will be assigned unnecessary
high bandwidth and thereby reducing the network uti-
lization. Moreover, the GTS allocation is made ex-
plicitly, which means that the number of nodes that
can have access to GTSs is limited to the maximum
number of GTSs.
During the last two years, there are few works re-
ported on improving the GTS allocation scheme in or-
der to provide a more efficient way of utilizing GTS to
support RT communication. The most common way
is to apply some kind of scheduling method, either
static scheduling algorithms or dynamic scheduling
algorithms. In this section, we review two of the more
recent and interesting GTS allocation methods. One
method uses static scheduling based on round-robin
scheduling algorithm (called i-GAME ) (Koubaa et
al., 2006b) and the other applies dynamic scheduling
based on the earliest deadline first (EDF) scheduling
algorithm (Mishra et al., 2007) (called GSA).
4.1 i-GAME
Motivated by having a more efficient GTS alloca-
tion in IEEE 802.15.4, A. Koubaa et. al proposed
an implicit GTS Allocation Mechanism (i-GAME)
for time-sensitive wireless sensor networks (Koubaa
et al., 2006b), instead of using explicit GTS alloca-
tion. The i-GAME method allows several nodes in
the network to share the same GTS under the round-
robin scheduling algorithm. The node that wants to
have a guaranteed time slot, sends its traffic specifica-
tion and its delay requirement to the PAN coordinator.
The PAN coordinator will do the admission control
to check whether the amount of available GTSs sat-
isfy the communication requirements. The authors in
(Koubaa et al., 2006b) have adopted the (b,r) traffic
model, in which b denotes the maximum burst size,
r denotes the average arrival rate. The network is a
star topology network with a unique PAN coordinator.
Each node i in the network generates a flow F
sented by three parameters (b
, r
, D
), where D
is the
delay requirement of flow F
. It has been addressed
in (Koubaa et al., 2006a; Koubaa et al., 2006b) that,
N flows F
, i = 1, 2, .., N are allowed to share a GTS
allocation consisting of k slots if
the sum of all arrival rates is less than or equal
to the bandwidth of k time slots:
k × R
where R
is bandwidth guaranteed per time slot
the delay bound guaranteed by the allocation,
, does not exceed the delay requirement:
, 1 i N
The calculation of guaranteed delay bounds per time
slot for original explicit GTS allocation and for the
implicit method has been presented in (Koubaa et al.,
2006a) and (Koubaa et al., 2006b) respectively. Net-
work calculus have been used in both cases. By ap-
plying a round-robin scheduling algorithm, i-GAME
allows each flow to equally share the bandwidth of k
time slots in GTS.
Although the round-robin algorithm provides a
fair GTS allocation scheme, it is not effective when
different traffic flows have different arrival rates.
Thus, it would be even more interesting to find a more
flexible allocation scheme that can support different
classes of traffic.
4.2 GTS Allocation with On-line
Scheduling Algorithm - GSA
A. Mishra et. al (Mishra et al., 2007) recently
presented a new GTS allocation scheme in IEEE
802.15.4, with an on-line scheduling algorithm sup-
porting both periodic and aperiodic real-time traffic
(named GSA). GSA is applied in the same network
configuration as i-GAME, in beacon-enabled mode,
but a different traffic model has been used. Each node
i in the network generates its traffic flow F
, which is
represented by three parameters: time constraint d
total payload p
and the number of requested slots r
For the GSA method, the authors have presented
a formula to calculate the estimated end-to-end delay
required for each node to send one frame to the PAN
coordinator. Base on this delay estimate, the PAN co-
ordinator will find the minimum GTSs needed to be
assigned for each node, in such way that the estimated
delay must be less than or equal to the its requested
delay. All the traffic flows in the network will be
sorted according to the order of increasing deadlines
(EDF scheduling). A set of traffic flows is schedula-
ble with the GSA method, if for each flow, the cumu-
lative transmission delay is less than or equal to its
upper bound. The GSA can be described as follows:
1. Input: S = F
, p
, r
2. Calculate estimated delay ED
3. Assign minimum number of GTSs s
for each flow
4. Do a feasibility analysis of the set S
5. If not feasible the last flow will be rejected
The biggest advantages of GSA is that it applies
an on-line scheduling algorithm, providing a better
support for different traffic classes. However, GSA
is composed of a complicated analysis which cause
overhead for the PAN coordinator. It is obvious that
the analysis used in GSA could be improved. One of
the very well known techniques used for EDF feasi-
bility analysis - that have been widely adopted in the
research community, processor demand test (Baruah
et al., 1990), could be applied in GSA. The authors
in (Mishra et al., 2007) also showed the improvement
of GSA in comparison with basic FCFS scheduling.
However, the complexity of the algorithm has been
overlooked. Another drawback of GSA is that it may
waste a portion of the GTSs since it always allocates
the GTSs from the first slot of the CFP.
In this paper, we have presented the most relevant
characteristics of the IEEE 802.15.4 standard for sup-
porting RT communication in WSNs. We have cho-
sen to review two of the more interesting RT protocols
that have been proposed for IEEE 802.15.4: one pro-
tocol applying a static scheduling method (i-GAME)
and the other applying a dynamic scheduling method
(GSA). The i-GAME protocol has low complexity, is
easy to implement and introduces fairly small over-
head for the PAN coordinator. Moreover, iGAME
provides a fair allocation for all nodes in the network,
which is very efficient in the specific case where all
traffic flows have a similar arrive rate. On the other
hand, the GSA protocol has significant advantages
compared to iGAME since scheduling is dynamic and
performed on-line. Thus, a broader range of traf-
fic classes are supported, such as for applications re-
quiring communication with unbalanced arrival rates.
The trade-off in the GSA compared to iGAME, is the
higher complexity of the scheduling algorithm and a
higher overhead for the PAN coordinator. It should
also be mentioned that iGAME has been implemented
in a real sensor network while GSA has been evalu-
ated by a simulation study.
As future work, we aim to investigate improved
scheduling methods to be applied for the IEEE
802.15.4. We are in favor of dynamic scheduling
methods, such as in the GSA, and we believe that
scheduling complexity can be reduced, for example
by applying a more simple feasibility analysis such
as the processor demand function. We are also inter-
ested in investigating some specific application such
as voice or media over WSNs, where QoS is highly
The authors would like to thank Dr. Anis Koubaa and
Dr. Elisabeth Uhlemann for valuable input to this pa-
per. This work has been funded by the CERES re-
search profile grant from The Knowledge Foundation.
T. He, P. A. Vicaire, T. Yan, L. Gu, G. Zhou, R. Stoleru, Q.
Cao, J. A. Stankovic, and T. Abdelzaher Achivieing
Real-time Target Tracking Using Wireless Sensor
Networks, Proc of the 12th IEEE Real-Time and
Embedded Technology and Applications Symposium,
C. Lu, B. M. Blum, T. F. Abdelzaher, J. A. Stankovic, and
T. He. “RAP: A Real-Time Communication Archi-
tecture for Large-Scale Wireless Sensor Networks”,
Proc. of the 8th IEEE Real-Timeand Embedded Tech-
nology and Applications Symposium, 2002.
J. A. Stankovic, T. Abdelzaher, C. Lu, L. Sha, and J. Hou.
“Real-Time Communication and Coordination in Em-
bedded Sensor Networks”, Proc. of the IEEE, 2003.
T. He, J. A. Stankovic, C. Lu, and T. F. Abdelzaher.
“SPEED: A Stateless Protocol for Real-Time Com-
munication in Sensor Networks”, Proc. of the Interna-
tional Conference on Distributed Computing Systems,
WINSYS 2008 - International Conference on Wireless Information Networks and Systems
O. Chipara, Z. He, G. Xing, Q. Chen, X. Wang, C. Lu, J.
A. Stankovic and T. Abdelzaher. “Real-Time Power-
Aware Routing in Sensor Networks”, Proc of the 14th
Workshop on QoS, 2006.
Y. Li, C. Chen and Y-Q. Song, “Real-Time QoS Support
in Wireless sensor Networks: A survey, Proc. of
7th IFAC International Conference on Fieldbuses and
Networks in Industrial and Embedded Systems, 2007.
IEEE 802.15.4 Standard-2003 “Part 15.4: Wireless medium
Access Control (MAC) and Physical Layer (PHY)
Specification for Low-Rate Wireless Personal Area
Networks (LR-WPANs)”, IEEE-SA Standards Board,
Zigbee Alliance, Zigbee Specification, December 2004:
A. Koubaa, M. Alves and E. Tovar. “IEEE 802.15.4 for
Wireless Sensor networks: A Technical Overview”,
IPP-HURRAY Technical Report (TR-050702), 2005.
A. Koubaa, M. Alves, and E. Tovard. “GTS Allocation
Analysis in IEEE 802.15.4”, Proc. of the 20th Inter-
national Parallel and Distributed Processing Sympo-
sium, 2006.
A. Koubaa, M. Alves, and E. Tovard. “i-GAME: An Im-
plicit Allocation mechanism in IEEE 802.15.4 for
Time-Sensitive Wireless Sensor Networks”, Proc. of
the 18th Euromicro conference on Real-Time Systems,
A. Mishra, C. Na, and D. Rosenburgh. “On Scheduling
Guaranteed Time Slots for Time Sensitive Transac-
tions in IEEE 802.15.4 Networks”, Proc. of the IEEE
Military Communication Conference (MILCOM07),
S. K. Baruah, R. R. Howell, and L. E. Rosier. Algorithms
and Complexity Concerning the Preemptive Schedul-
ing of Periodic Real-Time Tasks on One Processor”,
Real-Time Systems, 1990.