TOPOLOGICAL DEPENDENCE AND FAULT TOLERANCE IN

TDMA BASED POWER CONSERVATION FOR WSNS

Dimitrios J. Vergados

School of Electrical and Computer Engineering, National Technical University of Athens

Heroon Polytechniou 9, Zographou, GR-157 73, Greece

Nikolaos A. Pantazis

Department of Information and Communication Systems Engineering, University of the Aegean

Karlovassi, Samos, GR-83200, Greece

Department of Electronics, Technological Educational Institute (T.E.I.) of Athens

Ag. Spyridonos 21, Aigaleo, GR-122 10, Greece

Dimitrios D. Vergados

Department of Information and Communication Systems Engineering, University of the Aegean

Karlovassi, Samos, GR-3200, Greece

Department of Informatics, University of Piraeus, Karaoli & Dimitriou St. 80, GR-185 34 Piraeus, Greece

Christos Douligeris

Department of Informatics, University of Piraeus, Karaoli & Dimitriou St. 80, GR-185 34 Piraeus, Greece

Keywords: Wireless Sensor Networks, Energy Efficiency, TDMA.

Abstract: Energy conservation is a very critical issue in energy-constrained wireless sensor networks that introduces

sleep-mode related delay. Since a long time delay can be harmful for either large or small wireless sensor

networks, a TDMA-based scheduling scheme has been proposed, that achieves the reduction of the end-to-

end delay caused by the sleep mode operation while at the same time it maximizes the energy savings.

However, the performance of this system has not been studied with respect to the topology of the network,

and taking into consideration node failures. In this paper, we evaluate the TDMA-based energy conservation

scheme, and compare it to the S-MAC and the adaptive listening schemes, on various random topologies. In

addition, we examine the performance when node failures occur, and introduce a schedule update criterion.

1 INTRODUCTION

Wireless Sensor Networks (WSNs) are increasingly

used in a great number of applications nowadays.

The environmental, medical and military sectors are

some of the most important areas that the recent

developments have been applied in. WSNs consist

of an adequate number of tiny, cheap and low-power

sensor nodes, which collect and disseminate critical

data (Akyildiz, Weilian, Sankarasubramaniam, &

Cayirci, 2002). Each sensor node has limited power

capabilities due to the various limitations arising

from the need for inexpensive device, its limited

size, small weight, and ad-hoc method of

deployment. Various energy-efficient schemes have

been proposed in the literature in order to guarantee

the WSNs’ survivability and to increase the network

lifetime in such special-purpose environments.

Significant energy savings can be accomplished by

allowing the sensor nodes to enter sleep mode.

However, since nodes in sleep mode cannot detect

any transmissions in their vicinity, they cannot

participate in packet forwarding. Thus

synchronization schemes are needed, that can let the

J. Vergados D., A. Pantazis N., D. Vergados D. and Douligeris C. (2008).

TOPOLOGICAL DEPENDENCE AND FAULT TOLERANCE IN TDMA BASED POWER CONSERVATION FOR WSNS.

In Proceedings of the International Conference on Wireless Information Networks and Systems, pages 53-58

DOI: 10.5220/0002027600530058

 SciTePress

sensors’ transceivers remain in sleep mode as long

as possible, and at the same time to keep the network

from partitioning.

Several power conservation mechanisms have

been proposed for WSNs (Pantazis, Vergados &

Vergados, 2006; Pantazis & Vergados, 2007;

Srisathapornphat & Chien-Chung, 2002; Trigoni,

Yao, Demers, Gehrke, & Rajaraman, 2004;

Vergados, Vergados, & Douligeris, 2005; Yang &

Vaidya, 2004). S-MAC (Ye, Heidemann, & Estrin,

2002) is the most well-known sleep mode

synchronization scheme. It saves energy by creating

a common wakeup period for the nodes in the

network, followed by a sleep period. Nodes buffer

their packets until the next wakeup period takes

place. This strategy however increases the end-to-

end delay for multihop communication paths. This

happens, because each forwarder, after having

received a packet, must wait for the next scheduled

wakeup time and must then perform the

transmission, otherwise the following node will

probably be in the sleep mode, and there will be no

reception. Thus, the end-to-end delay, caused by the

sleep mode, is an increasing function of the

intermediate forwarders.

This Adaptive listening (Ad_Li) provides an

extension to S-MAC by trying to reduce the end-to-

end delay caused by the periodic listen-and-sleep

(Ye et al., 2002). The basic concept is that the node

which overhears its neighbors’ transmissions

(ideally only the RTS or CTS packets) will wake up

for a short period of time after the transmission,

allowing its neighbor to immediately transmit the

packet, without waiting for the next scheduled

transmission time.

A TDMA scheduling scheme for energy

efficiency has been proposed in (Pantazis, Vergados

& Vergados, 2006; Pantazis, Vergados, Vergados &

Douligeris, 2008) which constructs an appropriate

transmission schedule that achieves high levels of

power conservation and at the same time reduces the

end-to-end transmission time delay. This is achieved

by dividing the wakeup period into slots, and

carefully assigning slots to nodes, in a way that

transmissions from any node can be forwarded to the

gateway in a single wakeup period. In case a node

needs to reach the gateway, a WakeUP (WU) packet

is transmitted in the appropriate slot, and repeatedly

forwarded until it reaches the gateway. Nodes that

receive this packet will remain active anticipating

the reception of data. Assuming a target end-to-end

delivery time, this strategy achieves a reduction of

the power consumption, since only one wakeup

period is needed for reaching the gateway.

The operation of the TDMA schemes is closely

related to the network topology. Thus, the

performance of the TDMA power conservation

scheme needs to be investigated in various diverse

configurations (number of nodes, density, and

range). Also, when node failures occur, that may not

be instantly reflected in the schedule, some nodes

become unreachable. Thus, in this paper, we will

compare the performance of the aforementioned

power conservation mechanisms, on various random

topologies, and for different traffic loads, in terms of

delay and power conservation. Also, we will study

this effect of faulty nodes, and introduce a schedule

update criterion, that will be used for deciding when

the TDMA schedule needs to be updated, in order to

prevent degraded performance.

This paper is organized as follows: Section 2

summarizes the operation of the TDMA scheduling

energy conservation scheme. The performance

evaluation is presented in section 3, while section 4

concludes the paper.

2 THE TDMA SCHEDULING

ALGORITHM FOR ENERGY

EFFICIENCY IN WIRELESS

SENSOR NETWORKS

The main goal of the algorithm is to reduce the sleep

mode delay in WSNs. In order to achieve this goal,

the algorithm builds the schedule using previously

collected information which consists of the total

number of nodes, the one-hop neighbors and the

next hop of every node. This schedule assigns a

number of receiving and transmitting slots to every

sensor. The assignment procedure takes place in

such a way so that a transmission from any sensor

can be forwarded to the gateway within a single

frame. Energy savings are accomplished by turning

off the transceivers of every sensor in the network

during the idle operation, and only periodically

entering wake up periods. During these periods, the

sensors wake up according to a specified schedule.

In case there is no need for communication, no

packets are exchanged during this phase. On the

contrary, if a sensor needs to transmit information to

the gateway, it uses its transmission opportunity in

the WakeUP phase, and transmits a WU message.

This message is repeatedly forwarded, until it

reaches the gateway. The nodes that have received

and forwarded the wakeup message do not turn off

their transceivers during the following sleep-period,

until the exchange of information has been

WINSYS 2008 - International Conference on Wireless Information Networks and Systems

completed. This procedure retains the amount of idle

listening time at a low value, but at the same time it

limits the end-to-end delivery time.

The following procedures are executed: Initially

a setup phase takes place, which builds the

transmission schedule, followed by an energy-saving

phase. During the setup phase, the sensors do not

achieve the maximum level of energy conservation,

because their wireless transceivers do not enter the

sleep mode. The setup phase consists of the

following steps: (1) the exchange of hello and

routing messages among the sensors, (2) the

transmission of the needed information to the

gateway, (3) the schedule calculation, and (4) the

flooding of the schedule back to the sensors.

Thus, every sensor becomes aware of the other

sensors (neighbors) in its communication range, and

of the next node in its path towards the gateway.

Afterwards, each sensor in the network transmits the

above information to the gateway, which uses the

algorithm described in the following sections to

calculate the appropriate TDMA schedule. Finally,

this schedule is distributed to the sensors, which in

turn use it during the energy-saving phase for

determining the sleep and wake-up periods.

The question arising is which slot should each

sensor node use to transmit its WU messages

(originated or forwarded) and to which slots should

each sensor node listen to. Path WU requires that the

first sensor nodes in the path should be assigned in

earlier timeslots than the sensor nodes that follow.

On the other hand, collisions can be avoided if the

sensor nodes, which receive packets at the same

time, are not one-hop neighbors. Moreover,

transmissions to the same destination should be

assigned in different time slots. Ideally, sensor nodes

which are not one-hop neighbors should receive at

the same time in order to achieve the reduction of

the total frame length to the minimum possible

(Vergados et al., 2005). Therefore, timeslot

scheduling should take into account both the routing

paths and the neighboring information. The

proposed algorithm creates a TDMA schedule

appropriate for WU transmissions in WSNs.

The TDMA scheduling algorithm assigns a

transmission slot to every node in the sensor

network, and a number of reception slots for every

forwarding sensor node, one for each corresponding

transmitting sensor node. The algorithm uses the

collected information in order to calculate the

TDMA schedule. Based on this information, the

number of time slots, which each node has to receive

prior to transmitting, is calculated.

In the Energy-saving phase the sleep and wake-

up periods are determined using the schedule from

the setup phase. During the energy-saving phase

Path WakeUP instead of Node WakeUp and Path

WU Message Aggregation techniques are used for

saving energy and maintaining a low end-to-end

delay. Since the reduction of the end-to-end delay is

required, Path WakeUp is used instead of Node

WakeUp. Node WakeUp has longer end-to-end

delays than Path WakeUp, as the wakeup messages

cannot reach the gateway in a single frame. Also, the

application of the Path WU Message Aggregation is

important, since, by requiring only one sending slot

for each sensor, it reduces the number of receiving

slots and, therefore, it reduces the energy spent for

idle listening.

This energy conservation scheme is designed for

static sensor deployments, where the majority of the

traffic is directed to the gateway. The centralized

nature of the scheduling algorithm makes it suitable

for static deployments with no or very infrequent

topological changes.

Most Sleep–WakeUp schemes operate in a per-

hop basis. Thus, the transmitting sensor node should

wait for the arrival of the appropriate wakeup time

of its next-hop destination before it transmits the

data. Then, the forwarding sensor node should wait

for the wakeup time of the next sensor node in the

path, before it transmits the data, and so on, until the

message reaches the sensor gateway. This strategy

leads to end-to-end time delays that are related to the

product of the number of intermediate forwarders

times the length of the wakeup interval (Ye et al.,

2002).

slot

SLEEP ID L E RECV SEND

Figure 1: The path WakeUp.

If the source sensor node transmits a short WU

packet to the next hop in the transmission path, and

this WU packet is forwarded until it reaches the final

destination, then all the related sensor nodes will be

in the ACTIVE mode, anticipating the reception of

the information. Thus, the intermediate sensor nodes

can forward the message immediately. Through this

technique, the Path WakeUp strategy can allow

longer wakeup intervals for the forwarding sensor

nodes while keeping the end-to-end delay as short as

possible.

TOPOLOGICAL DEPENDENCE AND FAULT TOLERANCE IN TDMA BASED POWER CONSERVATION FOR

WSNS

In order to take advantage of the Path WakeUp

mechanism, TDMA scheduling should be performed

in such a way as to ensure that the transmitting

timeslot of each intermediate sensor be scheduled

before the transmitting timeslots of the next

forwarder in the path. Fig. 1 illustrates the power

mode transitions of Path WakeUp for a simple 3-hop

network. In Fig. 1, A, B, and C are the nodes in a

simple 3-hop network, with node D being the

gateway. The length of the total period is

, the

WakeUp frame duration is

and the slot length

slot

3 PERFORMANCE ANALYSIS

AND EVALUATION OF THE

TDMA SCHEDULING

ALGORITHM

The performance of power-aware MAC protocols

depends on many different parameters, such as the

traffic arrival rates, the channel congestion, the

topology of the sensor network and the routing

algorithms. In this section, we quantify the power

savings achieved by the proposed protocol, under

various network and traffic conditions and the

obtained results are compared to other power-saving

approaches in the literature. Thus, a simulation tool

was developed, that allowed an accurate estimation

of the sensors’ variables that helped to calculate the

values for the entire WSN. Repeated executions on

random networks produced the required mean values

and the intervals.

The network topology (number and location of

sensors) may become a conclusive factor to affect

the effectiveness of the power conservation scheme,

especially during the TDMA scheduling phase

which is the one most heavily influenced by the

sensor nodes’ topology. This section highlights the

effect of the topology on the power conservation

mechanisms, both in terms of end-to-end delay and

of average power consumption.

In order to accurately determine the behavior of

each energy conservation scheme, we must examine

various types of network topologies. An object-

oriented simulator has been created, which

implements a random topology generator, a

scheduler and a trace analysis module. Their values

are used to calculate the average value of the end-to-

end delay for each sensor transmission to the

gateway, and the average power consumption of the

network, both in standby modes and under various

traffic conditions, according to the following

equations:

idleTofi

recvTsendToMACS

PLLTT

LPLPP

)21)(/(

)(

λλ

λλλ

−−+

++=

−

(1)

idleTofi

recvTsendToadaptive

PLLTT

LPLPP

)21)(/(

)(

λλ

λλλ

−−+

++=

(2)

idleWaitWtidlefslot

recvTsendToTDMA

PLLPTTn

LPLPP

++=

)/(

)(

λλλ

(3)

txcsffMACS

ttTTNNDE ++−=

−

2/)}({

(4)

2/222/)}({

ftxcsfadaptive

TttTNNDE −++=

(5)

2/)()}({

ftxcsTDMA

TttNNDE ++=

(6)

Where:

is the arrival rate of the through

traffic, n is the number of TDMA slots each sensor

listens to, and

is the average number of times

node packets are retransmitted, until they reach the

gateway.

The above equations were used by the analysis

module of the simulator tool. The T

value for the S-

MAC scheme was given by the following equations:

(

)( )

∑

−−−⋅= 5.0/

_ icstxSMACf

nttDNT

(7)

(

)( )

∑

−−−⋅= 1/222

_ icstxadaptivef

nttDNT

(8)

where D is the desired delay, and n

is the number of

retransmissions required for sensor i’s packets to

reach the gateway. For the rest of the parameters, the

following values were used: P

send

= 1.5, P

recv

= 1.3,

idle

= 1, λ = 0.0000 (or otherwise stated), T

0.001, T

= 0.5, T

= 10 (or otherwise stated), T

0.001. We generated a 1000 node network and

varied the originating arrival rate (λ

) from 10

-9

packets/second/sensor, to 1 packet/second /sensor.

Fig. 2 illustrates the average power consumption of

each energy conservation scheme, for a single 1000

node topology. We observe that the TDMA power

conservation scheme has a lower idle power

conservation, but when the traffic increases S-MAC

becomes better. This is due to the fact that in the

TDMA scheme, all forwarding sensor nodes must be

awaken as soon as the packet is transmitted from the

originating node, resulting in an increased idle

listening time, when the arrivals are more frequent.

Nevertheless, the ability to use a large interval

for the same delay, results in lower idle power

consumption. For the comparison of the

performance of the three schemes, under different

network topologies, we generated topologies ranging

from 198 nodes to 1500 nodes. For each number of

nodes, a = 0.02 confidence intervals were produced

by generating 50 different topologies. Fig. 3(a) and

WINSYS 2008 - International Conference on Wireless Information Networks and Systems

3(b) illustrate the resulting performance. The TDMA

scheme consumes much less power under all

circumstances (with no traffic), when the system is

tuned to give the same delay for the three schemes.

1.0E‐05

1.0E‐04

1.0E‐03

1.0E‐02

1.0E‐01

1.0E+00

1.0E+01

1.0E‐10 1.0E‐07 1.0E‐04 1.0E‐01



ArrivalRate

TDMAPower SMACPower ada

tivePower

Figure 2: Power Consumption for the S-MAC, Ad_Li, and

the TDMA schemes, as a function of packet arrival rate

for a random 1000 node topology.

0.00E+00

2.00E+00

4.00E+00

6.00E+00

8.00E+00

1.00E+01

1.20E+01

0 500 1000 1500 2000

NumberofNodes

0.00E+00

2.00E‐04

4.00E‐04

6.00E‐04

8.00E‐04

1.00E‐03

1.20E‐03

0 500 1000 1500 2000



NumberofNodes

TDMA

ower SMACPower ada

tivePower

Figure 3: Delay (a) and power consumption (b) for the S-

MAC, Ad_Li and TDMA schemes, for topologies with

increasing nodes, and no traffic.

Figure 4 was produced using the same

configuration, but keeping the total traffic constant,

at N λo = 0.005. In this configuration, the TDMA

scheme is always better than the S-MAC, when the

system is tuned to give the same delay for the three

schemes.

Since all the previous results were produced

assuming there are no sensor failures, in this section

we study how the performance of the proposed

algorithm is affected by sudden topological changes,

if the computed schedule has not been updated in a

timely manner. Even when no sensor movement

takes place, the topology will probably change as a

result of node failures cased by various reasons (i.e.

exhaust of battery, DoS etc). These nodes, referred

to as dead nodes, cannot participate in

communication and therefore in the network. In such

a case, the network may become partitioned and

some nodes or groups of nodes may become isolated

and therefore, they cannot reach the gateway, since

the required forwarding nodes have become dead.

These nodes, together with the dead nodes, are

referred to as disconnected nodes. Moreover, the

non-disconnected nodes that their transmissions

cannot reach the gateway, because at least one of the

nodes in the scheduled path towards the gateway is

dead, are referred as unreachable. The transmissions

of these nodes will not be able to reach the gateway

until the schedule is updated. The nodes that don’t

require rescheduling to transmit to the gateway are

referred to as active.

0.00E+00

2.00E+00

4.00E+00

6.00E+00

8.00E+00

1.00E+01

1.20E+01

0 50 0 1000 1500 2000

NumberofNodes

‐1.00E‐03

1.00E‐03

3.00E‐03

5.00E‐03

7.00E‐03

0 500 1000 1 500 2000



NumberofNodes

TDMA

ower SMACPower ada

tivePower

Figure 4: Delay (a) and power consumption (b) for the S-

MAC, Ad_Li, and TDMA schemes, for topologies with

increasing nodes, and some traffic (N λo = 0.005).

As the WSN operation evolves, and nodes

gradually become dead, an increasing number of

nodes will become either disconnected or

unreachable, leaving a reduced number of active

nodes, causing the network performance to degrade.

In order to quantify the above observation,

simulation runs were carried out on a 1000 node

network, for various numbers of dead nodes, and the

number of unreachable and disconnected nodes was

calculated, and depicted in Fig. 5(a). Statistical

significance was obtained by averaging 50 different

simulation runs for every number of dead nodes, on

the same topology. The number of disconnected

nodes is close to the number of dead ones for most

numbers of dead nodes, but the number of

unreachable ones reaches significant levels even

when the number of dead nodes is relatively low.

Thus, the network performance will be decreased

drastically, even when only limited nodes become

dead.

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0% 20% 40% 60% 80% 100%



PercentageofNodesthatare Dead

Unreachable

Disconn ec ted

Dead

10%

15%

20%

25%

30%

35%

0% 20% 40% 60% 80% 100%



PercentageofNodesthatare Dead

10%

20%

30%

40%

50%

60%

0% 5% 10% 15%



RelativeDelay decrease

Figure 5: (a) The percentage of disconnected and

unreachable nodes in the networks for varying numbers of

dead nodes (1000 node topologies, 98% confidence

interval). (b) The relative decrease of the end-to-end

sensor to gateway delay (of the active nodes) in the

networks for varying numbers of dead nodes (1000 node

topologies). (c) The percentage of unreachable nodes in

the network versus the relative decrease of the end-to-end

delay (1000 node topologies).

TOPOLOGICAL DEPENDENCE AND FAULT TOLERANCE IN TDMA BASED POWER CONSERVATION FOR

WSNS

Since the disconnected nodes cannot become

active again, while the unreachable ones can, the

performance of the network will be improved by

using a timely rescheduling, that will increase the

number of active nodes in the network. Depending

on the percentage of unreachable nodes that can be

tolerated by the specific application, a schedule

update criterion can be defined as the percentage of

the dead nodes in the network. Fig. 5(b) illustrates

the end-to-end delay of the previous simulation, for

different numbers of dead nodes. This is more

obvious in Fig. 5(c), where the percentage of

unreachable nodes in the network is plotted against

the relative decrease of the end-to-end delay.

Therefore, in case sensor failures occur in our

network, which will cause significant nodes to

become unreachable, the rapid decrease in delay will

be detected by the gateway, triggering a schedule

update. This procedure retains the number of active

nodes which are close to the highest possible value,

by removing the dead nodes from the schedule, and

thus improving the network performance.

4 CONCLUSIONS

Power control and Energy efficiency are major

issues in WSNs, since they determine the network

lifetime. Several energy-efficient schemes have been

proposed in the literature to prolong the lifetime of

sensor networks, by periodically putting the sensor

nodes to sleep mode. This introduces a sleep-related

access delay that increases with the achieved power

saving. The Path-WakeUp and the wakeup message

aggregation strategies presented in this paper can be

used for minimizing the sleep-related end-to-end

delay and for minimizing the idle listening time, in

order to decrease the power consumed for given

delay levels. Simple analytic models were developed

for quantifying the power consumption of several

schemes. The performance evaluation showed that

the TDMA scheme achieves higher power

conservation than other relevant schemes, when the

traffic generation rate is low, and thus it can be used

for WSNs that monitor rare events and are expected

to operate for a long period of time, maximizing the

energy conservation.

ACKNOWLEDGEMENTS

This paper is part of the 03ED485 - “Design and

Development Models for QoS Provisioning in

Wireless Broadband Networks” research project,

implemented within the framework of the

“Reinforcement Programme of Human Research

Manpower” (PENED) and co-financed by National

and Community Funds (25% from the Greek

Ministry of Development-General Secretariat of

Research and Technology and 75% from E.U.-

European Social Fund).

REFERENCES

Akyildiz, I. F., Weilian, S., Sankarasubramaniam, Y., &

Cayirci, E. (2002). A survey on sensor networks.

Communications Magazine, IEEE, 40(8), 102-114.

Pantazis, N. A., Vergados, D. J. & Vergados, D. D.

(2006). Increasing Intelligent Wireless Sensor

Networks Survivability by Applying Energy-Efficient

Schemes. In AIAI 2006, Artificial Intelligence

Applications and Innovations Conference (Vol. 247,

pp. 657-664). Springer IFIP Series.

Pantazis, N. A., & Vergados, D. D. (2007). A survey on

power control issues in wireless sensor networks.

Communications Surveys & Tutorials, IEEE, 9(4), 86-

107.

Pantazis, N. A., Vergados, D. J., Vergados, D. D. &

Christos Douligeris, C. (2008). Energy Efficiency in

Wireless Sensor Networks using Sleep Mode TDMA

Scheduling. Ad-hoc Networks Journal, Elsevier, DOI:

10.1016/j.adhoc.2008.03.006.

Srisathapornphat, C., & Chien-Chung, S. (2002).

Coordinated power conservation for ad hoc networks.

In ICC 2002, IEEE International Conference on

Communications.

Trigoni, N., Yao, Y., Demers, A., Gehrke, J., &

Rajaraman, R. (2004). WaveScheduling: energy-

efficient data dissemination for sensor networks. In the

1st international workshop on Data management for

sensor networks: in conjunction with VLDB 2004.

Vergados, D. D., Vergados, D. J., & Douligeris, C. (2005).

A New Approach for TDMA Scheduling in Ad-Hoc

Networks. In PWC'05, 10

IFIP International

Conference on Personal Wireless Communications.

Yang, X., & Vaidya, N. H. (2004). A wakeup scheme for

sensor networks: achieving balance between energy

saving and end-to-end delay. In RTAS 2004, 10th

IEEE Real-Time and Embedded Technology and

Applications Symposium.

Ye, W., Heidemann, J., & Estrin, D. (2002). An energy-

efficient MAC protocol for wireless sensor networks.

In INFOCOM, 21

Annual Joint Conference of the

IEEE Computer and Communications Societies.

WINSYS 2008 - International Conference on Wireless Information Networks and Systems