See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/220948730

TEEN: ARouting Protocol for Enhanced Efficiency in Wireless Sensor Networks.

Conference Paper · January 2001

DOI: 10.1109/IPDPS.2001.925197 · Source: DBLP

CITATIONS READS
1,608 2,065

2 authors, including:

Dharma Agrawal
University of Cincinnati
683 PUBLICATIONS   17,049 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A cluster based secure routing scheme for wireless ad hoc networks View project

Wireless Body Area Sensor Network View project

All content following this page was uploaded by Dharma Agrawal on 30 May 2014.

The user has requested enhancement of the downloaded file.

TEEN: A Routing Protocol for Enhanced Efficiency in Wireless Sensor Networks

Arati Manjeshwar and Dharma P. Agrawal

Center for Distributed and Mobile Computing, ECECS Department,
University of Cincinnati, Cincinnati, OH 45221-0030

Abstract (i.e., require no installation cost etc). Protocols for these

networks must be designed in such a way that the limited
Wireless sensor networks are expected to find wide ap- power in the sensor nodes is efficiently used. In addition,
plicability and increasing deployment in the near future. In environments in which these nodes operate and respond are
this paper, we propose a formal classification of sensor net- very dynamic, with fast changing physical parameters. The
works, based on their mode of functioning, as proactive and following are some of the parameters which might change
reactive networks. Reactive networks, as opposed to passive dynamically depending on the application:
data collecting proactive networks, respond immediately to
 Power availability.
changes in the relevant parameters of interest. We also in-
troduce a new energy efficient protocol, TEEN (Threshold  Position (if the nodes are mobile).
sensitive Energy Efficient sensor Network protocol) for re-
 Reachability.
active networks. We evaluate the performance of our proto-
col for a simple temperature sensing application. In terms  Type of task (i.e. attributes the nodes need to operate
of energy efficiency, our protocol has been observed to out- on)
perform existing conventional sensor network protocols.
So, the routing protocol should be fault-tolerant in such a
dynamic environment. The traditional routing protocols de-
fined for wireless ad hoc networks [1] [9] are not well suited
1. Introduction due to the following reasons:
1. Sensor networks are “data centric” i.e., unlike tradi-
In recent years, the use of wired sensor networks is being
tional networks where data is requested from a specific
advocated for a number of applications. Some examples
node, data is requested based on certain attributes such
as, which area has temperature > 50 F ?
include distribution of thousands of sensors and wires over
strategic locations in a structure such as an airplane, so that
conditions can be constantly monitored both from the inside 2. The requirements of the network change with the ap-
and the outside and a real-time warning can be issued when plication and so, it is application-specific [3]. For ex-
the monitored structure is about to fail. ample, in some applications the sensor nodes are fixed
Sensor networks are usually unattended and need to be and not mobile, while others need data based only on
fault-tolerant so that the need for maintenance is mini- one attribute (i.e., attribute is fixed in this network).
mized. This is especially desirable in those applications
where the sensors may be embedded in the structure or 3. Adjacent nodes may have similar data. So, rather than
are in inhospitable terrain and are inaccessible for any ser- sending data separately from each node to the request-
vice. The advancement in technology has made it possi- ing node, it is desirable to aggregate similar data and
ble to have extremely small, low powered devices equipped send it.
with programmable computing, multiple parameter sensing 4. In traditional wired and wireless networks, each node
and wireless communication capability. Also, the low cost is given a unique id, used for routing. This cannot be
of sensors makes it possible to have a network of hundreds effectively used in sensor networks. This is because,
or thousands of these wireless sensors, thereby enhancing these networks being data centric, routing to and from
the reliability and accuracy of data and the area coverage as specific nodes is not required. Also, the large number
well. Also, it is necessary that the sensors be easy to deploy of nodes in the network implies large ids [2], which
 This work is supported by the Ohio Board of Regents’ Doctoral En- might be substantially larger than the actual data being
hancement Funds transmitted.

0-7695-0990-8/01/$10.00 (C) 2001 IEEE

 Thus, sensor networks need protocols which are applica- 4. Classification of Sensor Networks
tion specific, data centric, capable of aggregating data and
optimizing energy consumption. An ideal sensor network Here, we present a simple classification of sensor net-
should have the following additional features: works on the basis of their mode of functioning and the type
Attribute based addressing is typically employed in sen- of target application.
sor networks. The attribute based addresses are composed
of a series of attribute-value pairs which specify certain Proactive Networks
physical parameters to be sensed. For example, an attribute
address may be (temperature > 100F , location = ??). So, The nodes in this network periodically switch on their
all nodes which sense a temperature greater than 100F sensors and transmitters, sense the environment and trans-
should respond with their location. mit the data of interest. Thus, they provide a snapshot of
Location awareness is another important issue. Since the relevant parameters at regular intervals. They are well
most data collection is based on location, it is desirable that suited for applications requiring periodic data monitoring.
the nodes know their position whenever needed.
Reactive Networks
2. Related Work
In this scheme the nodes react immediately to sudden
In this section, we provide a brief overview of some re- and drastic changes in the value of a sensed attribute. As
lated research work. such, they are well suited for time critical applications.
Intanagonwiwat et. al [7] have introduced a data dissem-
ination paradigm called directed diffusion for sensor net- 5. Sensor Network Model
works. It is a data-centric paradigm and its application to
query dissemination and processing has been demonstrated We now consider a model which is well suited for these
in this work. sensor networks. It is based on the model developed by
Estrin et. al [3] discuss a hierarchical clustering method Heinzelman et. al. in [5]. It consists of a base station(BS),
with emphasis on localized behavior and the need for asym- away from the nodes, through which the end user can access
metric communication and energy conservation in sensor data from the sensor network. All the nodes in the network
networks. are homogeneous and begin with the same initial energy.
A cluster based routing protocol (CBRP) has been pro- The BS however has a constant power supply and so, has no
posed by Jiang et. al in [8] for mobile ad-hoc networks. It energy constraints. It can transmit with high power to all
divides the network nodes into a number of overlapping or the nodes. Thus, there is no need for routing from the BS to
disjoint two-hop-diameter clusters in a distributed manner. any specific node. However, the nodes cannot always reply
However, this protocol is not suitable for energy constrained to the BS directly due to their power constraints, resulting
sensor networks in this form. in asymmetric communication.
Heinzelman et. al [5] introduce a hierarchical clustering This model uses a hierarchical clustering scheme. Con-
algorithm for sensor networks, called LEACH. We discuss sider the partial network structure shown in Fig. 1. Each
this in greater detail in section 6.1. cluster has a cluster head which collects data from its clus-
ter members, aggregates it and sends it to the BS or an upper
3. Motivation level cluster head. For example, nodes 1.1.1, 1.1.2, 1.1.3,
1.1.4, 1.1.5 and 1.1 form a cluster with node 1.1 as the clus-
In the current body of research done in the area of wire- ter head. Similarly there exist other cluster heads such as
less sensor networks, we see that particular attention has not 1.2, 1 etc. These cluster-heads, in turn, form a cluster with
been given to the time criticality of the target applications. node 1 as their cluster-head. So, node 1 becomes a second
Most current protocols assume a sensor network collecting level cluster head too. This pattern is repeated to form a
data periodically from its environment or responding to a hierarchy of clusters with the uppermost level cluster nodes
particular query. We feel that there exists a need for net- reporting directly to the BS. The BS forms the root of this
works geared towards responding immediately to changes hierarchy and supervises the entire network. The main fea-
in the sensed attributes. We also believe that sensor net- tures of such an architecture are:
works should provide the end user with the ability to con-  All the nodes need to transmit only to their immediate
trol the trade-off between energy efficiency, accuracy and cluster-head, thus saving energy.
response times dynamically. So, in our research, we have
focussed on developing a communication protocol which  Only the cluster head needs to perform additional com-
can fulfill these requirements. putations on the data. So, energy is again conserved.

0-7695-0990-8/01/$10.00 (C) 2001 IEEE

 Base Station 3.1
3.2
the cluster-head. The cluster-head aggregates this data and
sends it to the base station or the higher level cluster-head,
as the case may be. This ensures that the user has a com-
2
3
plete picture of the entire area covered by the network.
2.3
3.3
Parameters
2.2
2.1

1.0.1 1.2.5 Cluster Formation

1 Report Time
Cluster Change Time
1.0.2 1.2
1.2.4

1.0.3 Figure 2. Time line for proactive protocol

1.2.1

1.2.3
1.1.2 1.2.2
1.1.3
1.1
Simple Sensor Node Important Features
1.1.4 First Level Cluster Head
The important features of this scheme are mentioned be-
1.1.1 1.1.5 Cluster Second Level Cluster Head low:

Figure 1. Hierarchical Clustering 1. Since the nodes switch off their sensors and transmit-
ters at all times except the report times, the energy of
the network is conserved.
 Cluster-heads at increasing levels in the hierarchy need
to transmit data over correspondingly larger distances. 2. At every cluster change time, TR and A are transmitted
Combined with the extra computations they perform, afresh and so, can be changed. Thus, the user can de-
they end up consuming energy faster than the other cide what parameters to sense and how often to sense
nodes. In order to evenly distribute this consumption, them by changing A and TR respectively.
all the nodes take turns becoming the cluster head for
a time interval T, called the cluster period. This scheme, however, has an important drawback. Be-
cause of the periodicity with which the data is sensed, it is
6. Sensor Network Protocols possible that time critical data may reach the user only after
the report time. Thus, this scheme may not be very suitable
The sensor network model described in section 5 is used for time-critical data sensing applications.
extensively in the following discussion of sensor network
protocols.
LEACH
6.1. Proactive Network Protocol LEACH (Low-Energy Adaptive Clustering Hierarchy) is
a family of protocols developed in [5]. LEACH is a good
In this section, we discuss the functionality and the char- approximation of a proactive network protocol, with some
acteristics expected in a protocol for proactive networks. minor differences.
Once the clusters are formed, the cluster heads broad-
Functioning cast a TDMA schedule giving the order in which the cluster
At each cluster change time, once the cluster-heads are members can transmit their data. The total time required
decided, the cluster-head broadcasts the following parame- to complete this schedule is called the frame time TF . Ev-
ters : ery node in the cluster has its own slot in the frame, during
which it transmits data to the cluster head. When the last
Report Time(TR ): This is the time period between succes- node in the schedule has transmitted its data, the schedule
sive reports sent by a node. repeats.
The report time discussed earlier is equivalent to the
Attributes(A): This is a set of physical parameters which
frame time in LEACH. The frame time is not broadcast by
the user is interested in obtaining data about.
the cluster head, though it is derived from the TDMA sched-
At every report time, the cluster members sense the pa- ule. However, it is not under user control. Also, the at-
rameters specified in the attributes and send the data to tributes are predetermined and are not changed midway.

0-7695-0990-8/01/$10.00 (C) 2001 IEEE

 Parameters Attribute > Threshold
Example Applications

This network can be used to monitor machinery for fault Cluster Formation
detection and diagnosis. It can also be used to collect data Cluster Change Time
Cluster-head receives message

about temperature change patterns over a particular area.

Figure 3. Time Line for TEEN
6.2. Reactive Network Protocol: TEEN
Important Features
In this section, we present a new network protocol
called TEEN (Threshold sensitive Energy Efficient sensor The main features of this scheme are as follows:
Network protocol). It is targeted at reactive networks and
is the first protocol developed for reactive networks, to our 1. Time critical data reaches the user almost instanta-
knowledge. neously. So, this scheme is eminently suited for time-
critical data sensing applications.
Functioning
2. Message transmission consumes much more energy
than data sensing. So, even though the nodes sense
In this scheme, at every cluster change time, in addition
continuously, the energy consumption in this scheme
to the attributes, the cluster-head broadcasts to its members,
can potentially be much less than in the proactive
Hard Threshold (HT ): This is a threshold value for the
network, because data transmission is done less fre-
quently.
sensed attribute. It is the absolute value of the attribute
beyond which, the node sensing this value must switch
on its transmitter and report to its cluster head. 3. The soft threshold can be varied, depending on the crit-
icality of the sensed attribute and the target application.
Soft Threshold (ST ): This is a small change in the value of
4. A smaller value of the soft threshold gives a more ac-
the sensed attribute which triggers the node to switch
curate picture of the network, at the expense of in-
on its transmitter and transmit.
creased energy consumption. Thus, the user can con-
The nodes sense their environment continuously. The trol the trade-off between energy efficiency and accu-
first time a parameter from the attribute set reaches its hard racy.
threshold value, the node switches on its transmitter and
sends the sensed data. The sensed value is stored in an inter- 5. At every cluster change time, the attributes are broad-
nal variable in the node, called the sensed value (SV). The cast afresh and so, the user can change them as re-
nodes will next transmit data in the current cluster period, quired.
only when both the following conditions are true:
The main drawback of this scheme is that, if the thresh-
1. The current value of the sensed attribute is greater than olds are not reached, the nodes will never communicate,
the hard threshold. the user will not get any data from the network at all and
will not come to know even if all the nodes die. Thus,
this scheme is not well suited for applications where the
2. The current value of the sensed attribute differs from
user needs to get data on a regular basis. Another possible
SV by an amount equal to or greater than the soft
problem with this scheme is that a practical implementation
threshold.
would have to ensure that there are no collisions in the clus-
ter. TDMA scheduling of the nodes can be used to avoid
Whenever a node transmits data, SV is set equal to the cur-
this problem. This will however introduce a delay in the re-
rent value of the sensed attribute.
porting of the time-critical data. CDMA is another possible
Thus, the hard threshold tries to reduce the number of
solution to this problem.
transmissions by allowing the nodes to transmit only when
the sensed attribute is in the range of interest. The soft
threshold further reduces the number of transmissions by Example Applications
eliminating all the transmissions which might have other-
wise occurred when there is little or no change in the sensed This protocol is best suited for time critical applications
attribute once the hard threshold. such as intrusion detection, explosion detection etc.

0-7695-0990-8/01/$10.00 (C) 2001 IEEE

7. Performance Evaluation both the protocols is the attribute to be sensed, which is the
temperature.
7.1. Simulation The performance of TEEN is studied in two modes, one
with only the hard threshold (hard mode) and the other with
To evaluate the performance of our protocol, we have both the hard threshold and the soft threshold (soft mode).
implemented it on the ns-2 simulator [10] with the LEACH The hard threshold is set at the average value of the low-
extension [4]. Our goals in conducting the simulation are as est and the highest possible temperatures, 100 F . The soft
follows: threshold is set at 2 F for our experiments.

 Compare the performance of the TEEN and LEACH 7.2. Results

protocols on the basis of energy dissipation and the
longevity of the network. We executed 5 runs of the simulator for each protocol
 Study the effect of the soft threshold ST on TEEN.
and for each mode of TEEN. The readings from these 5
trials were then averaged and plotted. A lower value of
The simulation has been performed on a network of 100 the energy-dissipation metric and a higher number of nodes
nodes and a fixed base station. The nodes are placed ran- alive at any given time indicates a more efficient protocol.
domly in the network. All the nodes start with an initial en-
2.5
ergy of 2J. Cluster formation is done as in the leach protocol
[5] [6]. However, their radio model is modified to include
idle time power dissipation (set equal to the radio electron-
ics energy) and sensing power dissipation (set equal to 10% 2

of the radio electronics energy). The idle time power is the

same for all the networks and hence, does not affect the per-
mte
formance comparison of the protocols.
Energy Dissipated(J)

1.5

leach
Simulated Environment
1
For our experiments, we simulated an environment with leach−c

varying temperature in different regions. The sensor net-

work nodes are first placed randomly in a bounding area
of 100x100 units. The actual area covered by the network 0.5

is then divided into four quadrants. Each quadrant is later

assigned a random temperature between 0 F and 200F static−sluster
every 5 seconds during the simulations. It is observed that 0
0 200 400 600 800 1000 1200 1400 1600 1800
most of the clusters have been well distributed over the four Time(s)

quadrants.
Figure 4. Energy dissipation: LEACH
Experiments
Figures 4 and 5 show the behavior of the network in
We use two metrics to analyze and compare the perfor- proactive mode. This comparison was originally done in
mance of the protocols. They are: LEACH [6]. It is repeated here taking into account the mod-
ified radio energy model. Of the four protocols [6], mte
Average energy dissipated: This metric shows the aver- (minimum transmission energy) lasts for the longest time.
age dissipation of energy per node over time in the However, we observe from Fig. 5 that only one or two
network as it performs various functions such as trans- nodes are really alive. As such, leach and leach-c (a variant
mitting, receiving, sensing, aggregation of data etc. of leach) can be considered the most efficient protocols, in
Total number of nodes alive: This metric indicates the terms of both energy dissipation and longevity.
overall lifetime of the network. More importantly, it In Figures 6 and 7, we compare the two protocols. We
gives an idea of the area coverage of the network over see that both modes of TEEN perform much better than
time. leach. If the cluster formation is based on the leach-c pro-
tocol, the performance of the TEEN protocol is expected to
We now look at the various parameters used in the im- be correspondingly better.
plementation of these protocols. A common parameter for As expected, soft mode TEEN performs much better than

0-7695-0990-8/01/$10.00 (C) 2001 IEEE

 100

90

100
80
TEEN (soft)
90
70
leach−c

Number of nodes alive

80
60

70
50
leach−c
Number of nodes alive

60 leach
40
static−cluster
50
30 TEEN (hard)
leach
40
20

30
10

20 mte
0
0 200 400 600 800 1000 1200
Time(s)
10

0 Figure 7. Comparison of the no. of nodes alive

0 200 400 600 800 1000 1200 1400 1600 1800
Time(s)

Figure 5. No. of nodes alive: LEACH hard mode TEEN because of the presence of the soft thresh-
old.

8. Conclusions

In this paper, we present a formal classification of sen-

sor networks. We also introduce a new network protocol,
2.5
TEEN for reactive networks. TEEN is well suited for time
critical applications and is also quite efficient in terms of
energy consumption and response time. It also allows the
2 user to control the energy consumption and accuracy to suit
the application.
TEEN (hard)
Energy Dissipated(J)

leach−c
1.5 TEEN (soft) Acknowledgment
leach

We would like to thank Wendi Heinzelman for her valu-

1 able suggestions and for letting us use her LEACH extension
to the ns simulator for our experiments.

0.5
References

[1] J. Broch, D. Maltz, D. Johnson, Y. Hu, and J. Jetcheva.

0
0 200 400 600 800 1000 1200 “A Performance Comparison of Multi-Hop Wireless Ad
Time(s)
Hoc Network Routing Protocols”. In Proceedings of the
Fourth Annual ACM/IEEE International Conference on Mo-
Figure 6. Comparison of average energy dissipation bile Computing and Networking(MOBICOM), ACM, Octo-
ber 1998.
[2] J. Elson and D. Estrin. “An Address-Free Architecture
for Dynamic Sensor Networks”. Technical Report 00-724,
Computer Science Department, USC, January 2000.

0-7695-0990-8/01/$10.00 (C) 2001 IEEE

 [3] D. Estrin, R. Govindan, J. Heidemann, and S. Kumar. “Next
Century Challenges: Scalable Coordination in Wireless Net-
works”. In Proceedings of the 5th Annual ACM/IEEE In-
ternational Conference on Mobile Computing and Network-
ing(MOBICOM), pages 263–270, 1999.
[4] W. Heinzelman, A. Chandrakasan, and H. Balakrish-
nan. “uAMPS ns Code Extensions”. http://www-
mtl.mit.edu/research/icsystems/uamps/ leach.
[5] W. Heinzelman, A. Chandrakasan, and H. Balakrishnan.
“Energy-Efficient Communication Protocols for Wireless
Microsensor Networks”. In Proceedings of Hawaiian In-
ternational Conference on Systems Science, January 2000.
[6] W. B. Heinzelman. “Application-Specific Protocol Archi-
tectures for Wireless Networks”. PhD thesis, Massachusetts
Institute of Technology, June 2000.
[7] C. Intanagonwiwat, R. Govindan, and D. Estrin. “Di-
rected Diffusion: A Scalable and Robust Communication
Paradigm for Sensor Networks ”. In Proceedings of the
6th Annual ACM/IEEE International Conference on Mobile
Computing and Networking(MOBICOM), pages 56–67, Au-
gust 2000.
[8] M. Jiang, J. Li, and Y. C. Tay. “Cluster Based Routing Pro-
tocol”. Internet Draft, 1999.
[9] E. M. Royer and C.-K. Toh. “A Review of Current Rout-
ing Protocols for Ad-Hoc Mobile Wireless Networks”. In
IEEE Personal Communications Magazine, pages 46–55,
April 1999.
[10] UCB/LBNL/VINT. “Network Simulator-ns”. http://www-
mash.cs.berkeley.edu/ns.

0-7695-0990-8/01/$10.00 (C) 2001 IEEE

View publication stats