Efficient Aggregate Computations in Large-Scale Dense WSN.
ABSTRACT We focus on large-scale and dense deeply embedded systems where, due to the large amount of information generated by all nodes, even simple aggregate computations such as the minimum value (MIN) of the sensor readings become notoriously expensive to obtain. Recent research has exploited a dominance-based medium access control(MAC) protocol, the CAN bus, for computing aggregated quantities in wired systems. For example, MIN can be computed efficiently and an interpolation function which approximates sensor data in an area can be obtained efficiently as well. Dominance-based MAC protocols have recently been proposed for wireless channels and these protocols can be expected to be used for achieving highly scalable aggregate computations in wireless systems. But no experimental demonstration is currently available in the research literature. In this paper, we demonstrate that highly scalable aggregate computations in wireless networks are possible. We do so by (i) building a new wireless hardware platform with appropriate characteristics for making dominance-based MAC protocols efficient, (ii) implementing dominance-based MAC protocols on this platform, (iii) implementing distributed algorithms for aggregate computations (MIN, MAX, Interpolation) using the new implementation of the dominance-based MAC protocol and (iv) performing experiments to prove that such highly scalable aggregate computations in wireless networks are possible.
- SourceAvailable from: Aloysius Mok[show abstract] [hide abstract]
ABSTRACT: P u b l i s h e d b y t h e I E E E C o m p u t e r S o c i e t y Opportunities and Obligations for Physical Computing Systems T he recent confluence of embedded and real-time systems with wireless, sensor, and networking technologies is creating a nascent infrastructure for a technical, economic, and social revolution. Based on the seamless integration of computing with the physi-cal world via sensors and actuators, this revolution will accrue many benefits. Potentially, its impact could be similar to that of the current Internet. We envision data and services that will be available any place, any time, to all people, not just technically sophisticated organizations and individu-als. Major systems such as those in transportation, manufacturing, infra-structure protection, process control, and electricity distribution networks will become more efficient and capable. People will be safer and experience an improved standard of living. New applications not even imagined today will become a reality. Although ingredients of this vision have existed for several years and the concept of pervasive computing does not differ radically from the work we describe, we believe developers must focus on the physical, real-time, and embedded aspects of pervasive computing. We refer to this domain as phys-ical computing systems. For pervasive computing to achieve its promise, developers must create not only high-level system software and application solutions, but also low-level embedded systems solutions. To better understand physical computing's advantages, we consider three application areas: assisted living, emergency response systems for natural or man-made disasters, and protecting critical infrastructures at the national level.Computer 12/2005; · 1.68 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: This article addresses the challenges and opportunities of instrumenting the physical world with pervasive networks of sensor-rich, embedded computation. The authors present a taxonomy of emerging systems and outline the enabling technological developments.IEEE Pervasive Computing 02/2002; 1(1):59- 69. · 2.06 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Sensor networks differ from traditional networks in several ways: sensor networks have severe energy constraints, redundant low-rate data, and many-to-one flows. The end-to-end routing schemes that have been proposed in the literature for mobile ad-hoc networks are not appropriate under these settings. Data-centric technologies are needed that perform in-network aggregation of data to yield energy-efficient dissemination. In this paper we model data-centric routing and compare its performance with traditional end-to-end routing schemes. We examine the impact of source-destination placement and communication network density on the energy costs, delay, and robustness of data aggregation. We show that data-centric routing offers significant performance gains across a wide range of operational scenarios.02/2003;
EFFICIENT AGGREGATE COMPUTATIONS IN LARGE-SCALE DENSE WSN
Abstract— This work addresses scenarios where even
a small area may contain several tens of nodes, in this
context, we addresses the problem obtaining aggregate
quantities (e.g., the minimum, maximum or median of
the values proposed by all nodes) with a time-complexity
that is independent of the number of nodes, or grows
very slowly as the number of nodes increases. This is
achieved by co-designing the distributed algorithms for
obtaining aggregate quantities and the underlying com-
munication system.In this paper, we summarize the
results of the research developed around this problem.
These results are significant because often networks of
nodes that take sensor readings are designed to be large
scale, dense networks and it is exactly for such scenarios
that the proposed distributed algorithms for obtaining
aggregate quantities excel. The implementation and test
of these distributed algorithms in a hardware platform
developed has shown that aggregate quantities in large-
scale, dense wireless sensor systems can be obtained ef-
Keywords—Medium Access Control (MAC), Data Pro-
cessing, Wireless Sensor Networks, Cyber-Physical Sys-
tems, Data aggregation.
Motivation and Objectives
Microprocessors are everywhere. Nowadays, we can
find computing capabilities in everyday physical objects
as diverse as mobile phones, digital personal assistants,
gaming platforms, household appliances or cars, just to
name a few examples.
Computing-enabled physical objects often have to
deal with physical processes and tightly integrate com-
puting with the physical world via sensors and actua-
tors. The integration of physical processes and comput-
ing is not a new problem. Embedded systems, which
have been in place long ago, often combine physical
processes with computing. However, with the massive
deployment of networked embedded computing devices,
we are observing the next step in the evolution of em-
bedded computing. The term Cyber-Physical Systems
(CPS) has been used to describe these pervasive com-
puting systems, where emphasis is put on the physical,
real-time and embedded aspects .
Such large-scale, sensor-rich networked systems will
generate an enormous amount of sensor data , and
handling such amounts of data introduces significant
challenges.One approach to deal with the amount
of data generated in these systems is to perform in-
network data aggregation. Instead of collecting data
from all nodes to a central point, in-network data ag-
gregation applies a data-reduction function to the data
traveling through the network such that the total num-
ber of messages transmitted is reduced .
Despite the previous research developed in the field
of data aggregation, its performance is limited by the
fact that, nodes in the same radio broadcast range can-
not transmit in parallel, hence the time-complexity still
depends on the number of sensor nodes. If we envision
scenarios where even a small area may contain several
tens of sensor nodes, the advantages of typical data
aggregation solutions found in the literature are signif-
It is in this context that the research work described
addresses the problem of performing scalable and ef-
ficient aggregate quantities (e.g., the minimum, maxi-
mum or median of the values proposed by all nodes)
in dense networks.Here “efficient” means that the
desired computation should be performed while con-
suming very little resources (such as energy, commu-
nication links, memory and processor) and “scalable”
means that the consumption of resources should in-
creases slowly or not at all as the number of sensor
readings to be processed and/or the number of embed-
ded computer nodes increases.
To illustrate this concept, consider a large-scale dense
networked sensor system, whose nodes have a common
sensing goal to measure temperature. Now consider the
problem of computing a simple aggregate quantity: the
minimum (MIN) sensed temperature among the nodes
at some given moment. Computing MIN seems triv-
ial, but for dense and large-scale systems, it poses an
important problem: communicating sensor data indi-
vidually makes the time-complexity of computing MIN
a function of the number of nodes. This is true for any
data aggregation mechanism employed.
This research work aims at being able to validate and
explore the following hypothesis:
Is it possible to efficiently obtain aggregate quantities
with a time-complexity that is independent of the
number of sensor nodes?
In other words, and taking the example of MIN, we
aim at computing MIN with a time-complexity that is
equivalent to the time of transmitting a single message,
even if tens or thousands of nodes share the same radio
Obtaining scalable and efficient aggregate quanti-
ties in large-scale dense networked sensor systems re-
quires tight integration between the data aggregation
techniques and communication mechanisms. This is a
key observation underlying this research work, where
Fig. 1. Data Aggregation by Exploiting a Prioritized MAC.
the approach to obtain scalable and efficient aggregate
quantities in large-scale dense networked sensor sys-
tems is co-designing (i) distributed algorithms to obtain
aggregate quantities and (ii) the underlying communi-
The main objective of this thesis is to demonstrate
that is possible to obtain aggregate quantities efficiently
by co-designing distributed algorithms for data aggre-
gation with the underlying communication services.
The approach to achieve this includes developing a
prioritized MAC protocol and design distributed algo-
rithms that exploit this MAC protocol. The next sub-
section presents a summary of the contributions made
by this research work.
This research work explores mechanisms for obtain-
ing aggregate quantities that are efficient, even in very
dense networks. The efficiency of traditional data ag-
gregation mechanisms results from applying data re-
duction functions to data coming from different sources,
and from exploiting the opportunities for parallel trans-
missions. In the extreme case where all nodes are in
the same broadcast domain, nodes cannot transmit in
parallel and there are no opportunities for traditional
data aggregation techniques to apply a data reduction
The novel approach explored in this thesis is based on
the adaptation to wireless media of a family of medium
access control (MAC) protocols. This family of proto-
cols is known as dominance or binary countdown pro-
tocols  and is already present in wired networking
solutions: the Controller Area Network (CAN) tech-
nology . We then design distributed algorithms that
exploit the MAC protocol to efficiently obtain aggre-
Dominance/binary countdown protocols can be ex-
ploited to efficiently obtain a range of aggregate quan-
tities. Let us briefly exemplify, to give further intuition,
the case of MIN, which can be obtained with a time-
complexity that is equivalent to the time of transmit-
ting a single message. This is illustrated in Figure 1,
where all nodes are in the same broadcast domain. Sup-
pose that the temperature values are coded as n-bit
integers. Starting with the most significant bit first,
let each node send the temperature reading bit by bit.
Consider also that, for each transmitted bit, nodes read
the resulting value in the channel (something straight-
forward in a wired medium) and the channel imple-
ments a logical AND of the transmitted bits. Further-
more, if a node reads ’0’ and is transmitting a ’1’, it
stops transmitting. Then, at the end of the transmis-
sion of the n bits, the “observed” value in the channel
will correspond to the MIN. It is as if all temperature
readings were transmitted in parallel at the same time,
and the resulting value of this non-destructive collision
is a useful aggregate quantity.
It is based on this concept that the novel dis-
tributed algorithms developed in this work are designed
upon. To accomplish this proposal, first it is neces-
sary to design a MAC protocol that implements domi-
nance/binary countdown in wireless environments, and
then develop the algorithms to exploit that MAC proto-
col. However, designing and implementing such MAC
protocol for wireless media is not trivial. For this rea-
son, a substantial part of this work is focused on the
development of such MAC protocol.
First, the problem is tackled assuming that all nodes
belong to a single broadcast domain (SBD). Nodes are
in a SBD when (i) a wireless broadcast made by one
node reaches all other nodes in the same broadcast
domain and (ii) if a node transmits a packet, then it
can be correctly received by another node in the same
broadcast domain only if the transmission of the packet
does not overlap in time with another packet trans-
mission. Note that, in this research work, there is no
assumption about regular propagation patterns of the
Achieving dominance in the wireless domain is chal-
lenging. To begin with, it is not possible to directly
translate the behavior of wired protocols, as these re-
quire that nodes are able to transmit and receive at
the same time. This is not possible in common radio
transceivers, because the transmitted energy is much
higher than the received energy. For this reason, dom-
inance in wireless systems was achieved using a sim-
ple principle: when the transmitted bit is dominant, a
pulse of a carrier wave is transmitted and there is no
need to sense the medium. Conversely, when the bit to
transmit is recessive, nothing has to be effectively sent,
instead only the medium state has to be sensed.
Although the concept and approach sounds simple,
a number of difficulties must be solved when proposing
the design of a correct dominance protocol for wire-
less networks. These include achieving proper synchro-
nization between the nodes, defining the parameters of
the protocol such that clock inaccuracies, time-of-flight
and other real-world effects are dealt with and how to
perform reliable carrier detection. These aspects are
addressed in this work and an implementation of dom-
inance protocols in wireless media − WiDom (short for
wireless dominance) − is presented.
This work also deals with the case of networks with
multiple broadcast domains (MBD). Considering MBD
is important because it will be difficult to make the
SBD assumption hold in a large number of networks
deployed in the real-world. Nodes are in a MBD net-
work if it holds that a wireless broadcast made by one
node cannot reach all nodes in the network. Such net-
works (MBD networks) suffer from the well known hid-
den node problem. This is a challenge that needs to
be solved when considering the extention of WiDom to
While a significant effort of this research work is put
into designing novel distributed algorithms to obtain
aggregate quantities in a SBD, local aggregation be-
tween nodes in geographic proximity can be used as
an intermediate step to compute aggregated quantities
among all nodes in a multihop network; hence the solu-
tion to the problem of computing aggregated quantities
in a SBD forms a relevant building block for large-scale
data aggregation in multihop networks. This challenge
is also tackled in this work.
A final note on the MAC protocol that imple-
ments dominance/binary countdown in wireless media
(WiDom). One important property of WiDom is that it
allows enforcing static priorities. Therefore, it enables,
for the first time, static priority scheduling over wireless
media. This is also a relevant characteristic in emerging
embedded systems because these systems deal with the
physical world, therefore one important requirement to
be met is that their data services are able to meet tim-
ing constraints . The research approach also takes
this property into account, and a response-time analy-
sis for the proposed MAC protocol is also developed.
In this work we reasoned on the design and im-
plementation of a prioritized MAC protocol (WiDom)
which enforces strict priorities over wireless channels
and on the design of algorithms that efficiently ex-
ploit this MAC protocol to obtain, in a SBD, the min-
imum (MIN), maximum (MAX) and interpolation of
sensor values with a time-complexity that is indepen-
dent of the number of sensor nodes (it depends only
on the sensor value range). These techniques also en-
able to efficiently obtain estimates of the number of
nodes (COUNT) and the MEDIAN. For MBD, the
time-complexity of the proposed distributed algorithms
developed also depends on the network diameter.
Adaptation of Dominance Protocols to Wire-
less Media. This research work introduced an adapta-
tion of a dominance protocols for wireless media, which
existed previously only for wired media. The imple-
mentation of a dominance protocol for wireless media
was named WiDom and was initially proposed under
the assumption of a SBD , . WiDom can be ex-
ploited to efficiently obtain aggregated quantities, and
it is also useful to provide pre-runtime guarantees for
sporadic messages streams. A schedulability analysis
for WiDom was developed accordingly.
Extension of WiDom to Support Multiple
Broadcast Domains. To cope with larger geo-
graphical areas, networks with multiple broadcast do-
mains (MBD) need to be considered. An extension of
WiDom for wireless networks with MBD was also pro-
posed. The proposed solution is the first prioritized
and collision-free MAC protocol designed to success-
fully deal with hidden nodes without relying on out-of-
band signaling .
Improving the Reliability of WiDom in SBD.
The techniques employed to solve the hidden node
problem in  can also be adapted to improve the re-
liability of the protocol in a SBD. The proposed solu-
tion has a result that is similar to a cooperative relay-
ing scheme, where several nodes can participate in the
transmission of the priority bits .
Scalable and Efficient Aggregate Quantities.
By exploiting dominance protocols it is straightforward
to demonstrate that, in a SBD, the minimum value
(MIN) can be obtained with a time-complexity that
is O(npriobits), where npriobits is the number of bits
used to represent the sensor data (the same technique
can be applied to obtain the maximum value (MAX);).
techniques to efficiently compute more complex aggre-
gate quantities such as the number of nodes (COUNT),
MEDIAN and interpolation by exploiting dominance
protocols were also implemented in the wireless do-
main . Finally, the techniques employed to obtain
aggregate quantities in a SBD were also extended for
multihop networks . The algorithms proposed for
MBD have a time-complexity that only depends on the
network diameter and on the value range of the sensor
There are two important sets of contributions in this
research: (i) the development and implementation of
WiDom and the support for static priority scheduling
over wireless links; and (ii) the algorithms for efficient
data aggregation based on WiDom. Let us now review
and briefly discuss each of them.
WiDom Development and Implementation.
WiDom was proposed, a novel wireless MAC proto-
col inspired in dominance/binary countdown protocols
which existed previously only for wired media . This
achievement is non-trivial. Firstly, implementations of
dominance protocols for a wired medium are based on
a wired–AND behavior of the bus, where the domi-
nant signal overwrites the recessive signal. Secondly,
these implementations require that nodes are able to
monitor the medium while transmitting. Clearly this
does not easily extend to the case of wireless channels.
Moreover, due to non-idealities of transceivers and the
nature of the wireless medium, it was not obvious how
a dominance protocol could be achieved.
WiDom supports a large number of priorities. Al-
though this number of priorities introduces overhead,
the application developer has the freedom to choose
the number of priority levels required, and thus possi-
bly reduce the overhead introduced. Nevertheless, such
a large number of priorities can be supported by other
prioritized protocols (see e.g., , ) but at the cost
of an overhead several orders of magnitude higher.
The initial design of WiDom was created under the
assumption of a SBD. An extension of WiDom for
wireless networks with MBD was also developed. In
such scenario, the hidden node problem must be dealt
with. The proposed solution is the first prioritized
and collision-free MAC protocol designed to success-
fully deal with hidden nodes without relying on out-of-
The idea of retransmitting priority bits used to solve
the hidden node problem can also be adapted to im-
prove the reliability of WiDom in a SBD. In the pres-
ence of several nodes in the same broadcast domain,
various nodes can cooperate in the transmission of the
priority bits. This simple modification of the protocol
can result in a substantial gain in the reliability, as the
number of nodes increases.
WiDom was implemented and evaluated experimen-
tally using Commercial-Off-The-Shelf (COTS) technol-
ogy. The implementation of WiDom using COTS tech-
nology suffered however from a significant overhead.
Therefore, a platform with better characteristics to im-
plement dominance protocols was also studied and de-
veloped. This platform is the proof of concept that
highly scalable aggregate computations in wireless net-
works are competitive in practice.
The experimental evaluation of WiDom shows that
the probability that a message is transmitted collision-
free, correctly prioritized and received (neither lost nor
corrupted) by all other nodes is high and this reliability
justifies the study of schedulability analysis techniques
for sporadic messages in wireless networks; WiDom is
an enabling technology allowing schedulability analy-
sis (for example to exercise in practice the analysis
proposed by ) in wireless multihop networks with
multiple broadcast domains. For the case of SBD, a
response-time analysis for WiDom was developed and
tested as well.
Efficient Data Aggregation.
The research on efficient data aggregation in this
work is motivated by scenarios where even a small
broadcast domain may contain several tens of sensor
nodes. In these scenarios, the advantages of data ag-
gregation solutions found in previous research are lost,
since it is neither possible to apply data reduction func-
tions to data coming from different sources nor is it
possible to exploit the opportunities for parallel trans-
missions. In this thesis was demonstrated that it is
possible to exploit a dominance-based MAC protocol
to efficiently compute aggregate quantities with a time-
complexity that is equivalent to the time of transmit-
ting a single message, even if hundreds of nodes are in
the same broadcast domain.
Concretely, the present work demonstrated that, in
a single broadcast domain (SBD), the minimum value
(MIN) can be obtained with a time-complexity that
is O(npriobits), where npriobits is the number of bits
used to represent the sensor data. In this case, the
message complexity (and thus, the time-complexity) is
independent of the number of sensor nodes. The same
technique can be applied to obtain the maximum value
Based on these techniques (of obtaining MIN or
MAX), more elaborate aggregated quantities can be
obtained. In this work, useful examples such as the
number of nodes (COUNT) and MEDIAN were also
Often, it is required to know how physical quantities
(such as temperature) vary over an area. Clearly, the
physical location of each node must be known then.
For such systems, an algorithm that produces an in-
terpolation of the sensor data as a function of space
coordinates was proposed. The resulting interpolation
is a compact representation of sensor data at a moment
and is obtained efficiently.
The algorithms (to obtain aggregate quantities) were
initially designed with the assumption of a SBD net-
work.Nevertheless, in practice, most networks are
not SBD networks. Therefore, solutions for MBD net-
works were also studied and proposed. An algorithm
for computing the MIN (or MAX) of sensor readings in
a multihop network was proposed. That algorithm has
the particularly interesting property of having a time-
complexity that does not depend on the number of sen-
sor nodes; only on the network diameter and the range
of the value domain of sensor readings matter. Other
more sophisticated algorithms were demonstrated also
feasible for MBD networks.
These results are significant because often networks
of nodes that take sensor readings are designed to be
large scale, dense networks and it is exactly for such sce-
narios that the proposed algorithms (designed in close
articulation with the MAC protocols) excel. The imple-
mentation of these algorithms in the hardware platform
developed shows that such highly scalable aggregate
computations in wireless networks are indeed competi-
tive in practice.
Discussion on Important Assumptions
There are two scenarios in this research work that
are important to fully grasp the relevance of the con-
Dense Wireless Networks
The algorithms designed in this research are devel-
oped for large-scale, dense networks. In particular, it
is in the presence of many nodes in the same broad-
cast domain that the advantages of an algorithm whose
complexity does not depend on the number of nodes be-
comes evident. Note that, it was show that we can see
advantages with as few as two nodes .
Some may object accepting such scenario. Indeed,
there are a few results which might advise against
deploying dense networks. For instance, previous re-
sults on the capacity of ad-hoc wireless networks 
show that, under certain assumptions, the capacity per
node approaches zero as the number of nodes increases.
However, several facts show that this is not the case
in the context of WSN , , and other solutions
can be devised (see, for example, the results reported
One reason to search for other solutions is that data
in sensor network is correlated, and this can be explored
in several ways (e.g., , ). In this work, the spa-
tial correlation between sensor readings was exploited
to perform a weighted-average interpolation and select
only a subset of nodes.
A second reason differentiating WSN is that the com-
munication pattern is often from several source nodes
to a sink. This enables several techniques such as in-
network data aggregation or other techniques such as
antenna sharing .
Another question is why deploy a dense sensor net-
work if we know that we will be gathering a lot of re-
dundant data? Deploying a dense sensor network might
be convenient for several reasons. First, when consid-
ering the case where the individual cost of each node is
negligible, then deploying redundant nodes might not
be a primary factor in the cost of the system. Redun-
dant nodes are useful for fault-tolerance (under certain
fault assumptions) and noise immunity. Deploying re-
dundant nodes also allows for a very fine spatial resolu-
tion in the sampling of the phenomena being observed.
More importantly, deploying a dense network allows a
better resolution of how the physical world is perceived;
for example, when we are interested in high-resolution
sampling in a certain region of interest, but we do not
know in advance where that region is. This is expected
to be essential in forthcoming innovative applications
in cyber-physical systems.
Timeliness Guarantees in Wireless and Reliability
An important contribution of this research work
is also a MAC protocol that supports static priority
scheduling in wireless networks. This contribution en-
compasses its full impact, when assuming that it is rele-
vant to analyze the problem of providing timing require-
ments within a hard real-time context.
It is often indicated that wireless links are unreliable,
thus it is not meaningful to approach the problem from
a guaranteed timeliness perspective.
It is true that designing a protocol with an upper
bound on queuing time is not sufficient to guarantee
that hard real-time deadlines are satisfied in practice.
However, it is a necessary step towards that goal. It is
important to note that part of this problem is a tech-
nological one. The reliability of wireless networks has
evolved noticeably over the last few years; it is safe to
assume this evolution will continue. The experimen-
tal evaluation performed in this research work suggests
that deadline misses due to noise are rare, so this pro-
vides evidence that it is still useful to consider hard
real-time bounds on queueing delays. Obviously, any
kind of guarantees are subject to some assumptions.
For the case of WiDom in SBD networks, there is a
range of techniques originally developed for the CAN
bus that can be applied, and this is a very interest-
ing research path to explore.
clude the schedulability analysis as presented in , ,
which are easily extendable to consider acknowledge-
ments and retransmissions, and also other techniques
These techniques in-
such as stochastic approaches to model faults , for
example. Moreover, the scheme to improve the reliabil-
ity of WiDom in a SBD results in a substantial gain in
the reliability, as the number of nodes increases. This
fact alone can enable hard real-time deadlines to be
satisfied with a very high probability in practice.
Even if not employing a guaranteed framework, be-
ing able to enforce strict priorities is useful in general.
One recent example can be found in , where the
authors built a streaming audio application employing
a prioritized MAC protocol previously proposed in the
literature . It was found that, although the over-
head of this MAC protocol is high, it still offers better
throughput than normal CSMA/CA protocols because
such prioritized MAC protocols eliminate the overhead
of back-off after collisions. The scalability of the work
reported in  is partially limited to the small number
of priorities supported by the MAC protocol adopted
and it is exactly in this point that WiDom stands out.
Compared to that MAC protocol, WiDom offers a much
higher number of priorities for a given similar overhead.
There are several opportunities that can be identified
for further research:
• development of other mechanisms to obtain other ag-
gregate quantities or perform other distributed compu-
• enclosing in a query processing system;
• integration with other communication protocols;
• power management schemes/duty cycling;
• improvement of current implementations;
• further development of radio hardware;
• maximize the number of parallel transmissions of
The following paragraphs briefly discuss each one of
these possible future research topics.
The fact that MIN can be obtained efficiently by ex-
ploiting WiDom can serve as a building block for other
computations like COUNT, MEDIAN or Interpolation.
It is possible to foresee that other computations may
eventually be devised out of similar ideas and this is a
research topic to be explored.
The computations enabled by WiDom could be
encapsulated in a query processing system, similar
TinyDB . The query processing would receive the
query specifications and map these into primitives that
exploit WiDom adequately. Eventually, queries that
cannot be more efficiently carried out by exploiting
WiDom could be mapped into other primitives that
do not work by directly exploiting WiDom.
One interesting possibility is to integrate WiDom in
other communication protocols. For example, a TDMA
protocol can benefit from WiDom by allowing several
nodes to share the same timeslot. Inside that timeslot,
contention is resolved using WiDom. This combination
would allow reducing the TDMA cycle. Another similar
example is to integrate WiDom in the IEEE 802.15.4
beacon-enabled mode, where one timeslot could be re-
served for WiDom. In this case, during that times-
lot, all nodes are allowed to access the medium using
WiDom. Both examples can improve the schedulability
of the system.
WiDom is energy efficient as it can avoid packet col-
lisions and enable very efficient computations, but it
lacks energy efficiency in the sense that it requires that
nodes constantly monitor the medium. However, this
does not need to be the case. WiDom is compatible
with power management schemes proposed in WSN,
such as coordinating activity/sleep schedules between
the nodes (e.g. ). This is an important topic for
At this point, the implementations of WiDom (and
distributed algorithms) available are considered only
as proof-of-concept prototypes. This means that the
implementations were not developed for general use and
they require some effort of understanding the details of
the implementation in order to be used. This is not a
topic for research, but it is important to consider this
fact in order to asess the amount of effort needed to
experiment with WiDom and/or further develop it.
The development of an efficient implementation of
WiDom is still an important subject for future work.
While a platform specifically designed for implement-
ing WiDom efficiently in provides an indication that the
overhead of the protocol can be low such that the algo-
rithms based on exploiting WiDom can be very com-
petitive , there is still a lot of work to be done in
this aspect. The platform presented is a prototype de-
veloped from components commercially available at the
time. The development and research of better radios
for the execution of WiDom is a topic that would ben-
efit the algorithms for obtaining aggregate quantities
presented and would also enable low overhead static
priority scheduling in wireless systems.
Finally, the proposed protocol for WiDom in a net-
work with MBD , does not maximize the number of
parallel transmissions. This is a problem that can be
difficult to solve, but, in practice, strategies as planning
the priorities in the nodes such that multihop compet-
ing is avoided can provide interesting solutions.
This paper presented the research efforts developed
around the following hypothesis: Is it possible to com-
pute aggregate quantities with a time-complexity that is
independent of the number of sensor nodes?
To achieve this goal, WiDom was designed in close
articulation with the data aggregation mechanisms. By
exploiting the properties of a prioritized MAC protocol,
the formulated hypothesis can be supported in SBD
networks. Effectively, the time-complexity of the algo-
rithms for obtaining aggregate quantities in SBD net-
works only depends on the sensor value range. In the
case of a MBD network, the time-complexity of the
algorithms developed also depends on the network di-
One important part of this research was dedicated
to make this approach effective in wireless networks.
This included the design and implementation of a pri-
oritized MAC protocol in wireless media. Several im-
plementations , , ,  have shown that such
approach is viable. While evidence was presented that
the approach is feasible and the algorithms based on
exploiting WiDom can be very competitive, there is a
big gap to fulfill until such mechanisms are useful in
more general settings. Filling this gap involves a wide
range of aspects, from development of a framework to
ease the use of such techniques by application develop-
ers, maturation of radio hardware or integration with
other communication protocols.
This research work has demonstrated innovative
mechanisms for obtaining certain aggregate quantities
in dense wireless sensor networks. The work included
the development of a prioritized MAC protocol that en-
ables these mechanisms and can also efficiently schedule
sporadic messages. Despite the dificulties yet to over-
come, these constitute an attractive set of solutions for
emerging Cyber Physical Systems.
This research work was partially developed at the
Real-Time Computing Systems Research Centre (CIS-
TER), from the School of Engineering of the Polytech-
nic Institute of Porto (ISEP).
John A. Stankovic, Insup Lee, Aloysius Mok, and Raj Ra-
jkumar. Opportunities and obligations for physical comput-
ing systems. Computer, 38(11):23–31, 2005.
 D. Estrin, D. Culler, K. Pister, and G. Sukhatme. Con-
necting the physical world with pervasive networks. IEEE
Pervasive Computing, pages 59–69, January-March 2002.
 Bhaskar Krishnamachari, Deborah Estrin, and Stephen
Wicker. Modelling data-centric routing in wireless sensor
networks. In 21st Annual Joint Conference of the IEEE
Computer and Communications Societies INFOCOM’02,
New York, USA, 2002.
 Aloysius K. Mok and Steve Ward. Distributed broadcast
channel access. Computer Networks, 3:327–335, 1979.
 CAN specification, ver. 2.0, 1991. Bosch GmbH, Stuttgart,
 N. Pereira, B. Andersson, and E. Tovar. Implementation of
a dominance protocol for wireless medium access. In 12th
IEEE International Conference on Embedded and Real-
Time Computing Systems and Applications (RTCSA’06),
pages 162–169, Sydney, Australia, 2006.
N. Pereira, B. Andersson, and E. Tovar. WiDom: A dom-
inance protocol for wireless medium access. IEEE Trans-
actions on Industrial Informatics (TII), 3(2):120–130, May
Nuno Pereira, Bjrn Andersson, Eduardo Tovar, and An-
thony Rowe. Static-priority scheduling over wireless net-
works with multiple broadcast domains. In 28th IEEE Real-
Time Systems Symposium (RTSS’07), pages 447–456, Tuc-
son, Arizona, USA, 2007.
 N. Pereira, B. Andersson, E. Tovar, and P. Carvalho. Im-
proving the reliability of widom in a single broadcast do-
main. In IEEE Symposium on Industrial Embedded Sys-
tems (SIES’09), Work-in-Progress Session, pages 144–147,
Lausanne, Switzerland, 2009.
 N. Pereira, R. Gomes, B. Andersson, and E. Tovar. Efficient
aggregate computations in large-scale dense WSN. In 15th
IEEE Real-Time and Embedded Technology and Applica-
tions Symposium (RTAS’09), pages 317–326, San Francisco,
California, USA, 2009.
 N. Pereira, B. Andersson, E. Tovar, and P. Carvalho. Ef-
ficient computation of min and max sensor values in mul-
tihop networks - by exploiting a prioritized MAC protocol.
Springer Lecture Notes on Electrical Engineering (LNEE),
38, Intelligent Technical Systems:233–246, 2009.
 Joo L. Sobrinho and A.S. Krishnakumar. Quality-of-service
in ad hoc carrier sense multiple access networks. IEEE Jour-
nal on Selected Areas of Comm., 17(8):1353–1368, 1999.
 B. D. Bui, R. Pellizzoni, M. Caccamo, C. F. Cheah, and
A. Tzakis. Soft real-time chains for multi-hop wireless ad-
hoc networks. In 13th IEEE Real Time and Embedded Tech-
nology and Applications Symposium (RTAS’07), pages 69–
 T. F. Abdelzaher, S. Prabh, and R. Kiran. On real-time ca-
pacity limits of multihop wireless sensor networks. In IEEE
Real-Time Systems Symposium (RTSS’04), pages 359–370,
Lisbon, Portugal, 2004.
 P. Gupta and P. R. Kumar.
IEEE Transactions on Information Theory,
 Daniel Marco, Enrique J. Duarte-melo, Mingyan Liu, and
David L. Neuhoff. On the many-to-one transport capacity
of a dense wireless sensor network and the compressibility
of its data. In 2nd International Workshop on Information
Processing in Sensor Networks (IPSN’03), Lecture Notes in
Computer Science, pages 1–16, Palo Alto, CA, USA, 2003.
 H.E. Gamal.On the transport capacity of the many-to-
one dense wireless network. In IEEE 58th Vehicular Tech-
nology Conference (VTC’03-Fall), pages 2881–2885, Vol.5,
 Himanshu Gupta, Vishnu Navda, Samir Das, and Vishal
Chowdhary. Efficient gathering of correlated data in sensor
networks. In 6th ACM Intl. symposium on Mobile ad hoc
networking and computing (MobiHoc’05), pages 402–413,
 Ian Broster, Alan Burns, and Guillermo Rodr´ ıguez-Navas.
Probabilistic analysis of can with faults. In 23rd IEEE Inter-
national Real-Time Systems Symposium (RTSS’02), pages
269–278, Austin, TX ,USA, 2002.
 Joo L. Sobrinho and A. S. Krishnakumar. Real-time traffic
over the IEEE 802.11 medium access control layer. Bell Labs
Technical Journal, 1(2):172–187, 1996.
 Samuel R. Madden, Michael J. Franklin, Joseph M. Heller-
stein, and Wei Hong. TinyDB: an acquisitional query pro-
cessing system for sensor networks. ACM Trans. Database
Syst., 30(1):122–173, 2005.
 Wei Ye, John Heidemann, and Deborah Estrin. Medium ac-
cess control with coordinated adaptive sleeping for wireless
sensor networks. IEEE/ACM Transactions on Networking
(TON), 12(3):493–506, 2004.
The capacity of wireless
Nuno Pereira is a researcher at CIS-
TER/ISEP and a Lecturer at the Com-
puter Engineering Department of ISEP,
the School of Engineering of the Polytech-
nic of Porto. He received a licentiate de-
gree in Computer Engineering from the
School of Engineering of the Polytechnic
Institute of Porto and an MSc Degree from
the University of Minho in 2002 and 2005
respectively, and has also recently pre-
sented his doctoral thesis. Nuno worked
in a number of research projects related to several aspects of
real-time systems and has authored and co-authored more than
20 research papers in international refereed conferences and jour-