Using a Token Approach for the MAC Layer of Linear Sensor Networks

Impact of the Redundancy on the Throughput

El Hadji Malick Ndoye

1,2

, Fr´ed´erique Jacquet

, Michel Misson

and Ibrahima Niang

Clermont Universit´e / LIMOS CNRS - Complexe Scientiﬁque des C´ezeaux, 63172 Aubi`ere cedex, France

Laboratoire d’Informatique, Universit´e Cheikh Anta Diop de Dakar (UCAD), B.P. 5005 Dakar-Fann, S´en´egal

Keywords:

Wireless Sensor Network, Linear Topology, Throughput, Mac Protocol, Token Passing.

Abstract:

Wireless Sensor Networks (WSNs) consist of a large number of sensor nodes deployed in a wide area for

monitoring applications. For some of these applications, as pipeline or road monitoring, wireless sensor nodes

have to be deployed in a linear manner. We refer to these WSNs as Linear Sensor Networks (LSNs). Due to

speciﬁcity of LSNs, MAC protocol designed for WSNs, as contention or TDMA based protocols, are often

not suitable. Furthermore, wireless node deployment can provide a certain form of redundancy to prevent

link or node failures. In this paper, we propose a token based MAC protocol for linear sensor networks in

order to improve the network performance. We evaluate the effect of the redundancy on the number of packets

delivered to the sink. We show that the redundancy induces a signiﬁcant improvement both on the delivered

trafﬁc and on the FIFO queue size of the nodes.

1 INTRODUCTION

Wireless sensor networks (WSNs) are used to observe

and react to eventsand phenomena in order to perform

environmental monitoring. Very often, the geometry

of the WSN is linear du to the monitored objects. A

such geometry can be found in applications like de-

tection of the presence of workers in a gallery of an

underground mine (Li and Liu, 2009) or ﬂuid leaks in

water or oil pipelines (Jawhar et al., 2007)(Ted et al.,

2012)(SunHee et al., 2009), in transmission of data

between the wagons of a freight train (Zimmerling

et al., 2008)(Wang et al., 2011)(Berlin and Van Laer-

hovenand, 2013), in the monitoring of roads, bridges

and tunnels (Meng et al., 2008). In this paper, we refer

to these networks as Linear Sensor Networks (LSNs).

We focus on an application where packets have to

reach a sink station arbitrarily located at the right end

of the network. In LSNs each node has some neigh-

bors to its left (they are called left neighbors) and to

its right (they are called right neighbors). When sen-

sor nodes are aligned on a straight line strictly form-

ing a line, or a thin LSN as deﬁned in (Jawhar et al.,

2008), the network has to deal with two identiﬁed

drawbacks: the link or node failure impacting the con-

nectivity and the hidden terminal problem impacting

the frame loss rate. This is why we are considering

both the use of a token passing mechanism and the

use of a light redundancy for LSN topology, provid-

ing alternative paths for the trafﬁc. We assume a uni-

form placement of nodes, where the Sink is located at

one end and a ﬁrst node named Allocator at the other

end. The allocator is in charge of providing tokens

periodically.

Enhancing throughput and robustness, is one of

the most important challenge in LSN. The density of

the topology (number of neighbors of a current node)

is low and depending on the propagation conditions

and on the stability of the wireless links. This density

will be used to introduce the redundancy of a LSN

topology. If each node has only two neighbors, ie one

on its right and one its left, the LSN is strictly lin-

ear. We called it 1-Redundancy LSN topology. When

each node has 2*R neighbors, both R on its right and

on its left, we are dealing with a R-Redundancy LSN

topology allowing routing facilities. In this paper, we

evaluate the impact of R on the trafﬁc reaching the

sink. The LSN topologies expose packets to the ef-

fect of the hidden problem and induce high latency

(Noori and Arkani, 2008). So, the need of a suitable

MAC protocol for network performance improvement

is a real challenge in LSN. We propose a token based

MAC protocol for LSNs deployed according a R-

Redundancy topology (R varying from 1 to 3). The

authorization to transmit is granted via the reception

of a token ie a speciﬁc short frame. Tokens are pro-

122

Ndoye E., Jacquet F., Misson M. and Niang I..

Using a Token Approach for the MAC Layer of Linear Sensor Networks - Impact of the Redundancy on the Throughput.

DOI: 10.5220/0005243501220129

In Proceedings of the 4th International Conference on Sensor Networks (SENSORNETS-2015), pages 122-129

ISBN: 978-989-758-086-4

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

vided periodically by the ﬁrst node named Allocator

and then propagated from node to node. When a node

is token holder it is allowed to transmit uplink traf-

ﬁc toward the sink and/or downlink trafﬁc toward the

Allocator. Packets are sent to the targeted node in a

multi-hop manner following a path depending on the

redundancy factor R.

This paper is organized as follows: Section 2 jus-

tiﬁes the choice of our MAC protocol and presents

the state of art on MAC protocol based on token

for wireless sensor networks for throughput improve-

ment. Section 3 introduces how the density allows us

to deﬁne a R-Redundancy LSN and for each value of

R, how to deﬁne the minimal distance between two

consecutive tokens. Section 4 presents details about

our token based MAC protocol. We show the mecha-

nism allowing a token holder node to transmit trafﬁc

both toward the sink and toward the Allocator. Sec-

tion 5 evaluates theoretically the impact of the redun-

dancy on the optimum number of packets delivered to

the sink. Section 6 conﬁrms the impact of factor R on

the optimal delivery rate, via NS2 simulation results.

Section 7 discuss about the theoretical and simulation

results obtained. This paper is concluded in Section

Section 8.

2 STATE OF ART

The usual MAC protocol based on contention or

TDMA are not suitable for LSNs. In the case of

contention MAC protocols such as protocols based

on Carrier Sense Multiple Access (CSMA) with or

without RTS/CTS (Ndoye et al., 2013), 802.15.4

CSMA, SMAC, TMAC, encounter collision problems

and induce congestion areas due to the hidden termi-

nal problem. Thus, packet retransmissions decrease

throughput at the sink. In the case of MAC proto-

cols based on slotted access method, such as protocols

based on Time Division Multiple Access (TDMA),

TRAMA, FLAMA, they require a strict synchroniza-

tion and/or a complex scheduling between nodes to

avoid large unproductive time guard intervals. The

drift of the clocks of the nodes decreases highly the

throughput due to the loss of packets and induces en-

ergy wasting.

In the research literature related to the speciﬁci-

ties of linear wireless sensor networks, LSNs are of-

ten used to drain data from a set of nodes (raw con-

vergecast) in order to aggregate it in a sink. It is well

known that a hop by hop routing of data packets to

one or several speciﬁc sink nodes induces a kind of

concentration of trafﬁc along the path followed by the

forwarded frames. At each hop along the path con-

ducting to a sink, an additional local trafﬁc is gath-

ered to the data to be forwarded. This concentration

of trafﬁc increases progressively along this path and

it may cause locally an overload of the medium. In

(Noori and Arkani, 2008) authors show that the trafﬁc

increases progressivelywhen it gets closer to the sink.

Such areas are called congestion areas to point out the

fact that the MAC has to deal with a local load exceed-

ing the usual capacity of the medium. This causes

several drawbacks:

• An overload of the FIFO of node within the con-

gestion areas, inducing delays in the forwarding

process and a risk of frame dropping due to FIFO

overﬂow,

• An increasing of busy medium status returned by

channel sensing (CCA operation),

• An increasing rate of collisions increased by the

hidden terminal problem.

These two last phenomena cause losses of frames

(collision or dropping). In the literature, this spe-

ciﬁc behavior has been identiﬁed early for solutions

based on linear deployment of a wireless infrastruc-

ture made of 802.11 Access Points (Moutairou et al.,

2009). For WSN applications, many research focus

on the propriety of MAC protocol in order to avoid

congestion and thus improve the throughput. In re-

cent years, a lot of research focuses on token based

protocols due to the limitation of classic MAC proto-

cols (contention and TDMA). Nevertheless, most of

these proposed protocols do not match in linear sen-

sor network. ToKeN-TWiNs (TKN-TWN) described

in (Liu et al., 2013a) is a high throughput data col-

lection exploiting the advantages of TDMA. It eases

the scheduling burden by using two tokens to arbi-

trate transmission activities. Access to the medium

is organized by a centralized token passing mecha-

nism. The sink node generates two types of tokens

which are passed to two different top-subtrees, and a

multi-channel approach is used to avoid interference.

In (Liu et al., 2013b) an implementation and evalua-

tion of TKN-TWN in term of throughput is presented.

Authors show that in a binary-tree-formed network

ToKeN-TWiNs throughput outperforms the through-

put of collection protocol (Incel et al., 2012). This

proposition was designed for WSNs deployed accord-

ing to a tree topology and it loses a part of its in-

terest for linear topology. In (Fan et al., 2012) au-

thors present an Enhanced Dynamic Token Protocol

(EDTP) using TPQ as described in WDTP (Xianpu

et al., 2007). They show that in underwater acoustic

WSN, EDTP performs better throughputthan TDMA.

In (Na et al., 2009) authors describe a Data Filtration-

Aware MAC protocol for wireless Sensor Networks

UsingaTokenApproachfortheMACLayerofLinearSensorNetworks-ImpactoftheRedundancyontheThroughput

123

(DF-MAC). It is a token-based MAC protocol based

on TR-MAC (Na et al., 2007). Indeed, after nodes se-

lection and the formation of a logical ring, data trans-

mission is performed by using a token. It was shown

that DF-MAC provides a better throughput than TR-

MAC. Unfortunately, DF-MAC is designed for clus-

tered network and cannot be relevant for LSNs. Even

if the token based protocols presented above improve

performance of classic or clustered WSNs, they are

not designed for LSNs and they do not take advantage

of the speciﬁcities of such a topology. The design of a

MAC protocol for a LSN must consider a trade-offbe-

tween the speciﬁcities of this kind of networks: topol-

ogy, low density, small processing power, energy lim-

itations, etc. We propose a MAC protocol based on a

slotted access method needing a soft synchronization

and taking advantage of linearity of the deployment

of nodes. The use of a token to propagate the right

to transmit is a way to avoid the constraints of a strict

and global synchronization.

Before sending data frames, a current node has to

wait the reception of a token from one of its neigh-

bors. When this token reception occurs, this current

node becomes Token Holder for a given amount of

time SD (Shuttle Duration). Then this node can trans-

mit uplink trafﬁc toward the sink or downlink trafﬁc

toward the Allocator, details will be given in the next

paragraph. The token frame contains information as

the token generationperiod, the sleep and wakeup cal-

endar. We evaluate our protocol in terms of through-

put in order to show the impact of redundancy of lin-

ear network sensor on its behavior.

3 HYPOTHESES AND NETWORK

TOPOLOGY

We outline now the details about the density of LSNs

we are studying, the timing of MAC protocol we pro-

pose and the trafﬁc model we use in this paper. In

the following, we consider a one-dimensional LSN of

N nodes uniformly placed over a length-L with equal

distance d between two adjacent nodes. The kind of

LSN we are going to use is depending on the density

of the topology. If the radio range p and the distance

d between nodes are such that d < p < 2d, each node

has two neighbors (one to its left and one to its right).

We are dealing with a strictly LSN having a density

of 2 as shown in Fig. 1a. As deﬁned previously it is a

1-Redundant LSN.

When p > 2d, each node has at least 4 neighbors

(two or more to its left and two or more to its right).

For such topology, if a link is broken or a node is

out of order it is still possible to forward the frames

(mainly to the sink) by skipping the defective device.

This kind of LSN is called R-redundant where R is the

number of neighbors in each direction (right and left).

A 2-redundant as shown in Fig. 1b. In the following,

we focus on this kind of redundantLSNs, and we refer

to them as 2-Redundant LSNs or 3-Redundant LSNs

according to their density.

Fig. 1b: Nodes of a 2−Redundant LSN

Fig. 1a: Nodes of a strictly LSN

Figure 1: LSN topologies.

Each node has its own ID. In this paper we sup-

pose the following points: (i) the node located at left

end of the LSN is Node 1 and is the Allocator provid-

ing tokens, (ii) the node at the right side of Node I is

Node I+1, (iii) the node located at the right end of the

LSN is the Sink. The routing (or forwarding) scheme

works as follows. Each node transmits data frames to

its one-hop or R-hop neighbor according to the LSN

topology.

In our simulation, the model we use for a current

node is given in Fig. 2. A queue for packets is at-

tached to each node; this queue is connected to the

next node. In the following, we show the impact of

the FIFO size and of the Shuttle Duration length on

the number of packets delivered to the sink. Pack-

ets ﬂow from the left to the right and a local load is

injected into the queue of each node. For many rea-

sons such as (i) retransmission credit exceeded (oc-

curs when a node has transmitted ﬁve times the same

data frame without receiving any acknowledgment),

(ii) queue overload (occurs when the queue of the

node is requested to keep too many packets), packets

may not be received by the sink. We deﬁne the ag-

gregate throughput as the number of packets received

by the sink per unit of time. Throughput increases as

the offered load provided by nodes (packets locally

produced) increases. When the offered load reaches

a certain threshold, the throughput does not increase

any more. Sometimes, it can even start to decrease.

We denote by t a period of time including several to-

ken periods, we can deﬁne the aggregate throughput

of the LSN as:

S =

n∗PacketsLength

where n is the number of pack-

ets received by the sink during t. This throughput will

be used for the evaluation of the efﬁciency of the To-

ken MAC protocol according to the redundancy of the

LSN.

SENSORNETS2015-4thInternationalConferenceonSensorNetworks

124

MAC

address

filter

MAC

Mechanism

Physical layer

(Traffic generator)

Local application

Next node

Current node

local FIFO

Previous node

Uplink traffic Downlink traffic

Forwared traffic

Figure 2: Node Model.

4 OUR PROPOSAL: TOKEN

BASED MAC PROTOCOL

In this section we describe our proposal token based

MAC protocol for LSNs.

4.1 Access Control

The token generation governing the access to the

medium is initiated by the Allocator. It produces

a token which circulates from node to node until it

reaches the sink. In the following, the path followed

by a token is always A, B, C, , Sink. Each node which

receives a token is allowed to transmit during a given

time called shuttle duration. So, a node has two major

states: it is either token holder, or waiting for a token.

In token holder state, the node can transmit differently

data frames during the shuttle duration, according to

three consecutive periods. (i) T

: during this amount

of time the node transmit data frames to its neighbor

toward the Allocator. This trafﬁc is called downlink

trafﬁc and comes from the sink. (ii) T

: during this

amount of time the node transmit data to the sink.

This trafﬁc is called uplink trafﬁc. (iii) At the begin-

ning of period T

, the node being token holder passes

the token to its neighbor at its right side and reaches

the waiting for a token state. When there is no pend-

ing downlink trafﬁc, a token holder node uses also T

for uplink trafﬁc. When a node is in the waiting token

state, it can either listen for uplink or for downlink

packets coming from the sink, it can also switch off

its radio to save energy. The temporal pattern of the

activity of a current node is given on Fig. 3.

4.2 Periodicity of the Token Production

In this section, we deﬁne the minimal distance in term

of nodes between two consecutive tokens in order to

have several tokens circulating in the network at the

same time. This distance must be calculated accord-

ing to the transmission range of each node and on the

downlink traffic

Listening for

receptionuplink traffic uplink traffic

Transmitting ofTransmitting ofTransmitting ofThe token Listening for

downlink trafficthe token

Shuttle

T’

’

Shuttle Duration (ShuDur)

Figure 3: Example of token passing.

possibility of having downlink trafﬁc. In Fig. 4a, it

is shown that in a strictly LSN with downlink trafﬁc,

the distance between two token holders is three nodes

(four hops). The part of the LSN corresponding to

nodes (A, B, C and D) as shown in can be consid-

ered as a cluster moving from node to node at each

expiration of the shuttle duration. So, in the case of

strictly LSN the size of the cluster (CluSize) is set to

four hops. So if a reverse trafﬁc is possible, a node is

token holder at most only a quarter of its time. Dur-

ing three quarters of its time, it has to queue the trafﬁc

locally produced.

During the remaining quarter of its time, the node has

to queue (in an interleaved manner) the local trafﬁc in

addition to the trafﬁc forwarded by the previous node

of the LSN. If the radio range allows the possibility to

exchange with its two-hop neighbors, the spatial reuse

becomes less efﬁcient, as it requires an increase in the

distance between two nodes being simultaneously in

the token holding state.

As shown in Fig. 4b, for a 2-redundant LSN, two

token holder nodes have to be separated by at least

four nodes when the trafﬁc is only for the sink, and

by six nodes when a reverse trafﬁc is possible. The

size of the linear cluster (CluSize) is respectively 5

and 7. The minimal distance between two successive

token holder nodes has a strong impact on the network

performance and on the token production activity of

the allocator. For a given token holding time ( T

+ T’

as deﬁned in Fig. 3), the minimal period of

token production is given T

Token

(min) = ShuDur x

CluSize.

The time separating two consecutivetokens T

Token

must be greater than T

Token

(min); the choice of this

period has to be done by considering the following

factors: the energy autonomy of nodes, the proﬁle of

the offered trafﬁc load, and so on.

UsingaTokenApproachfortheMACLayerofLinearSensorNetworks-ImpactoftheRedundancyontheThroughput

125

A B

C D

A B

C ED

Ack

Uplink Traffic

Fig. 4a: Distance between token holders for a strictly LSN

Uplink Traffic A−>C

Ack F−>H

Downlink Traffic H−>FAck C−>A

Fig. 4b: Distance between token holders for a 2−Redundant LSN

Token holder node

Downlink Traffic

Figure 4: Distance between token holders.

5 MECHANISM OF TOKEN

REDUNDANCY

The creation and the forwarding of a new token by

the Allocator, allows the trafﬁc of the nodes to be

drained in a multi-hop manner. Each time a node

is token holder, it can transmit (i) ﬁrst the trafﬁc to-

wards the Allocator (downlink trafﬁc), (ii) and during

the remaining holding time, the trafﬁc having the sink

as destination (uplink trafﬁc). Our choice to allow

uplink and downlink trafﬁc has an impact on the fre-

quency of the token production. The distance between

two consecutive nodes being token holder, expressed

in number of hops, depends on the range of the ra-

dio link and on the distance between two consecutive

nodes.

We can notice that the case of a strictly linear net-

work, it is easy to show (Fig. 4) that the distance be-

tween two nodes being token holder is equal to four

hops. That is to say 3 times the radio range expressed

in number of hops plus 1. For a 2-redundant LSN,

this distance increases to 7 hops (3 times the radio

range plus 1) and for a 3-redundant LSN this distance

reaches 10 hops (3 times the range plus 1).

Fig. 5.a and Fig. 5.b are used to introduce how

the token passing is managed and how the path of

data packets takes advantage of redundancy.In the-

seﬁgures, dashed arrows indicate the path systemat-

ically followed by the tokens: A −→ B −→ C ... −→

Sink, blue arrows indicate the links which are simul-

taneously active at a given point in time. Fig. 5.a is for

the 2-Redundant case, the trafﬁc generated by node A

is transmitted directly to node C to be concatenated

with the trafﬁc locally generated. This trafﬁc is then

forwarded to node E and so on. In such a network it

can be seen that there is two branches (A, C, E, G)

and (B, D, F,H) which converge towards the sink.

Fig. 5.b is for the 3-redundant case. The trafﬁc

generated by node A is transmitted directly to node

D to be concatenated with the trafﬁc locally gener-

ated. This trafﬁc is then transmitted to node G and

so on. In such a network, three branches converging

towards the sink S, can be identiﬁed :(A, D, G, J),

(B, E, H, K) and (C, F, I, L).This way of spreading

the LSN trafﬁc over these three branches has an im-

pact on performance in terms of throughput. Let us

consider Fig. 5.b when node A and node K are token

holder, Node A is allowed to transmit data packets to

D, and then its token to B, Node K will directly trans-

mit its data packets to the sink and its token to L.We

can notice two important things:

• The sink can receive data packets every time one

of its neighbors is token holder. So for a 3-

Redundant LSN, the same token gives 3 oppor-

tunities to the sink to receive data packets. Each

token allows the sink S of Fig. 5.b to receive con-

secutively from J, K and L.

• The FIFO of each node being a neighbor of the

sink must be large enough to store and forward the

trafﬁc of its branch. So for a 3-Redundant LSN,

the evaluation of the FIFO size of a sink neighbor

node is only governed by the trafﬁc of its branch.

The FIFO size of node K of Fig. 5.b is supposed

to be larger enough to concatenate the trafﬁc of

nodes of the branch (A, D, G, J).

The two previous points will be used to evaluate the

throughput capacity of a LSN according to the redun-

dancy factor R. For R equals from 1 to 3, we want to

estimate the optimal number of packets delivered to

the sink for a given bit rate.

The word optimal is used here to point out the fact

the FIFO size is adjusted to the shuttle duration to

maximize the throughput and to avoid the loss of data

packets. For this evaluation we chose to deal with

medium sized data packets and we suppose that all the

nodes of the same FIFO size for generality purpose.

Each neighbor of the sink is token holder during just

enough time to empty its FIFO. A direct consequence

of this hypothesis is that the size of the FIFO equals

the maximum number of data packets a token holder

can sent before passing the token. In others words,

the capacity of the shuttle is equal to the size of the

node FIFO.

Let SC (Shuttle Capacity) be the number of pack-

ets a shuttle can carry. For a 1-Redundant LSN, the

sink receives P packets by token, ie P packets for each

time period of 4 ShuttleDurations(SD). The delivery

rate of data packets to the sink is

Theoretical

Throughput(1)=

1∗SC

4∗SD

expressed in

packets per second

For a 2-redundant network and for each token, the

sink receives S packets from its one hop neighbor but

SENSORNETS2015-4thInternationalConferenceonSensorNetworks

126

also S packets from its two hop neighbor. For such a

case two consecutive tokens are separated by 7 Shut-

tle Durations. The delivery rate of data packets to the

sink is

Theoretical

Throughput(2) =

2∗SC

7∗SD

For the case of Fig. 5.b, this delivery rate becomes

Theoretical

Throughput(3) =

3∗SC

10∗SD

For an R-redundant LSN, this can be generalized

by:

Theoretical

Throughput(R) =

R∗SC

((3∗R+1)∗SD)

Concerning

is common ratio independent of R, but depend-

ing on the value of parameters of physical and MAC

layer used for the LSN. The implementation follow-

ing this study will be based on the physical layer of

802.15.4, it is why this evaluation will be done us-

ing a bit rate of 250 Kbps. The throughput evaluation

is based on 100 byte data-frames; the length of the

data-frames has not a signiﬁcant impact on through-

put. The use of short frames reduces slightly the

throughputintroducing more overheadbut it improves

the end-to-end delay.

Simulations allow estimating the average time be-

tween the transmissions of two consecutive acknowl-

edged frames. This time is useful to estimate the ca-

pacity of the shuttle that is to say the number of pack-

ets a node is able to send while being token holder.

If each node is token holder during 250 ms (SD) and

if transmitting period is 4.5 ms, the shuttle capacity

(SC) is about 55 packets. So

is equal to 44 kbps

Concerning

3∗R+1

is an increasing function. It starts at 0.25

and converges to 0.33. The optimum throughput ob-

tained for a 3-redundant LSN is

higher than that

of a strictly linear network, so this function can be

considered as a redundancy gain. This throughput in-

crease has no impact on the FIFO size of the nodes.

Trafﬁc of a branch of such a network should be cal-

culated so that the node of this branch that is neigh-

bor of the sink contains at most SC packets when it

becomes token holder. In this theoretical evaluation

of the impact of factor R on the delivery packet ra-

tio, the time needed for the token passing mechanism

was neglected. This choice has no real impact if each

node remains token holder in order to send a signif-

icant number of data frames. In the following is on

this theoretical approach that simulations have been

carried out. Simulation results conﬁrm the expected

gain in term of throughput.

Fig 5.a: Clusters formation in a 2−Redundant LSN

Fig 5.b: Clusters formation in a 3−Redundant LSN

Figure 5: Clusters formation in LSNs.

6 EVALUATION

We perform our simulations on NS2 (version 2.32).

Our results are given for a linear network of sixteen

nodes: node 1 (the Allocator), is on the left side, and

node 16 (the Sink), is on the right side. Local trafﬁc

is produced pseudo-periodically and starts randomly

between 0 and 1 second, and independently for each

node. A given current node might receive trafﬁc to be

forwarded when one of its neighbors becomes token

holder. We suppose in the following that all the nodes

of the LSN have the same type of queue managed in

a ﬁrst in ﬁrst out (FIFO) manner. The capacity of this

queue can be expressed by the number of packets (Fi-

FoSize) it can contain. The existence of a channel for

a downlink trafﬁc is an interesting capability allowing

the possibility of changing the period of token gener-

ation for example, but, in the following we suppose

that the number of packets targeted to the sink repre-

sents the dominant trafﬁc. The size of the reverse traf-

ﬁc is ignored in our simulations. The capture model is

deﬁned as in 802.15.4 and the new reception capture

threshold is set to 10 dBm. The propagation model is

Tworayground and we suppose also that all the pack-

ets have the same size. In this paper, we focus on the

throughput as described above.

Table 1 presents the simulation parameters.

Fig. 6, Fig. 7 and Fig. 8 present the performances

parameters in a 1-Redundant, 2-Redundant and 3-

Redundant LSN in term of throughput in order to

show the impact of the redundancy. They evaluate

the evolution of the throughput as a function of data

transmission rate for a given shuttle duration. For

each shuttle duration, two phases can be identiﬁed:

between 8 and 48 Kbps the throughput evolution in-

creases. Indeed, in this phase, the load of the network

is low and thus FIFO contain a few number of data

packets for uplink trafﬁc. The number of packets re-

ceived by the sink from its direct neighbors per sec-

ond is low. In this phase, nodes do not transmit any

UsingaTokenApproachfortheMACLayerofLinearSensorNetworks-ImpactoftheRedundancyontheThroughput

127

Table 1: Simulation parameters.

Parameter Value

Downlink trafﬁc time 10 ms

Shuttle duration [50-300]ms

Token packet size 11 bytes

Data packet size 100 bytes

Number of repetitions 50

Physical Layer 802.15.4

Transmission Power -5 dBm

FiFoSize 60

Distance between two nodes in 1-

Redundant LSN

90 meters

Distance between two nodes in 2-

Redundant LSN

45 meters

Distance between two nodes in 3-

Redundant LSN

30 meters

Data transmission rate [8-80] Kbps

uplink trafﬁc most of the shuttle duration. Between

48 and 80 Kbps the throughput evolution is station-

ary. During this phase the load of the network induces

saturation. The FIFOs contain the maximal number

of packets that can be sent during the shuttle duration.

So, the number of packet received by the sink from its

direct neighbor is constant. During this phase, nodes

send uplink trafﬁc most of the shuttle duration.

The impact of the redundancy can be shown also

in Fig. 6, Fig. 7 and Fig. 8. Indeed, the throughput

at the sink for each shuttle duration and for a given

data transmission rate is higher for the 3-Redundant

Shuttle duration = 50 ms

Shuttle duration = 100 ms

Shuttle duration = 150 ms

Shuttle duration = 200 ms

Shuttle duration = 250 ms

Shuttle duration = 300 ms

10 20 30 40 50 60 70 80

Throughput (Kbps)

Data Transmission Rate (Kbps)

Figure 6: Throughput comparison by Shuttle duration in a

1-Redundant LSN.

Shuttle duration = 50 ms

Shuttle duration = 100 ms

Shuttle duration = 150 ms

Shuttle duration = 200 ms

Shuttle duration = 250 ms

Shuttle duration = 300 ms

10 20 30 40 50 60 70 80

Throughput (Kbps)

Data Transmission Rate (Kbps)

Figure 7: Throughput comparison by Shuttle duration in a

2-Redundant LSN.

Shuttle duration = 50 ms

Shuttle duration = 100 ms

Shuttle duration = 150 ms

Shuttle duration = 200 ms

Shuttle duration = 250 ms

Shuttle duration = 300 ms

10 20 30 40 50 60 70 80

Throughput (Kbps)

Data transmission Rate (Kbps)

Figure 8: Throughput comparison by Shuttle duration in a

3-Redundant LSN.

3−Redundant LSN

2−Redundant LSN

1−Redundant LSN

50 100 150 200 250 300 350 400

Maximal Throughput (Kbps)

Shuttle duration (ms)

Figure 9: Maximal throughput for data transmission rate of

80 Kbps for various shuttle duration. The maximal through-

put is achieved with a shuttle duration of 250 ms.

LSN and followed by the 2-Redundant LSN. For ex-

ample for a shuttle = 50 ms, data transmission rate

= 80 Kbps we have for3-Redundant: 45.5 Kbps, for

2-Redundand: 35.5 Kbp and for 1-Redundant: 31.1

Kbps. This conﬁrms our theoretical analysis. Fig. 9

shows the maximal throughput for a data transmis-

sion rate of 80 Kbps for various values of the shuttle

duration. This ﬁgure shows that maximal throughput

is obtained for a shuttle duration of 250 ms. Indeed,

direct neighbors from the sink transmit maximal num-

ber of packets from thier FIFO. Also, they are in the

most time of the shuttle in transmitting state. Con-

trarily for the shuttle duration of 300 where nodes are

more often in waiting state due to the fact that the

FIFO are empty before the end of the shuttle. That is

why the throughput decreases for a shuttle of 400 ms.

The impact of the redundancy can be shown. The

maximal throughput is for the 3-Redundant LSN (

trafﬁc from 3 direct neighbors of the sink) is 51.27

Kbps , 46.13 Kbps for the 2-Redundant LSN and

40.37 Kbps for the 1-Redundant LSN.

7 DISCUSSION

The maximal throughput given by simulation is al-

ways slightly lower than the theoretical throughput.

Some frames must be repeated on the path followed

SENSORNETS2015-4thInternationalConferenceonSensorNetworks

128

Table 2: Throughput comparison between theoretical and

simulation analysis.

R Theoretical

Throughput (Kbps)

Throughput from

simulation (Kbps)

1 44 40.37

2 50.3 46.13

3 53 51.27

to reach the sink. This is decided after the expiration

of the timer used to detect a no-acknowledgement. It

is the main reason of this slight difference. Table 2

presents the comparison between theoretical and sim-

ulations results.

8 CONCLUSION

Linear Sensor Networks (LSNs) have a large interest

for monitoring applications. In this paper , we pro-

pose a token based MAC protocol to manage the ac-

cess to the medium. We study the behavior of LSN

in the case of three topologies. Thus, we deﬁne a R-

redundant LSN where R is the number of neighbors

in each direction for a given node. Speciﬁcally, we

study the impact of the redundancy on the throughput

at the sink. We show that by theoretical and simu-

lation analysis that more the factor of redundancy R

is great more the throughput at the sink is also great.

We show also that the redundant allows nodes to have

an equitable distribution of the trafﬁc by dividing the

network into branches.

In future works, we plane to reverse channel in

order to master the token production frequency ac-

cording to the spatial reuse and energy saving con-

straints. Another way to improve the capacity of such

a network is to add a priority policy to the node FIFO

management by allowing highest priorities to the data

frames coming from the farthest nodes. Finally, we

plane to use a Log-normal Shadowing model in order

to model the path loss due to the environment ﬂuctu-

ations.

REFERENCES

Berlin, E. and Van Laerhovenand, K. (2013). Sensor net-

works for railway monitoring: Detecting trains from

their distributed vibration footprints. In IEEE Interna-

tional Conference on Distributed Computing in Sen-

sor Systems.

Fan, G., Li, J., Liu, W., and Zhao, Y. (2012). An enhanced

dynamic token protocol for underwater accoustic sen-

sor networks. In International journal on smart sens-

ing and intelligent systems, volume 4.

Incel, O. D., Ghosh, A., Krishnamachari, B., and Chinta-

lapudi, K. (2012). Fast data collection in tree-based

wireless sensor networks. In IEEE Transactions on

Mobile Computing, volume 11, pages 86–99.

Jawhar, I., Mohamed, N., Shuaib, K., and Kesserwan, N.

(2008). An efﬁcient framework and networking pro-

tocol for linear wireless sensor networks. Ad Hoc &

Sensor Wireless Networks, 7(1-2):1–9.

Jawhar, I., Nader, M., and Shuaib, K. (2007). A framework

for pipeline infrastructure monitoring using wire-

less sensor networks. In The Sixth Annual Wire-

less Telecommunications Symposium (WTS 2007),

Pomona, California, USA.

Li, M. and Liu, Y. (2009). Underground coal mine monitor-

ing with wireless sensor networks. ACM Trans. Sen.

Netw., 5(2):10:1–10:29.

Liu, J., Suzuki, M., Lee, D. ., and morikawa, H. (2013a). A

token scheduled high throughput multi-channel data

collection protocol for wireless sensor network. In

Proc. IEEE Veh. Technol. Conf.

Liu, J., Suzuki, M., Lee, D., and morikawa, H. (2013b). Im-

plementation and evaluation of token-scheduled col-

lection protocol in wireless sensor network. In Proc.

IEEE Veh. Technol. Conf.

Meng, S., Kunqing, X., Xiujun, M., and Guojie, S. (2008).

An on-road wireless sensor network approach for ur-

ban trafﬁc state monitoring. In IEEE Conference on

Intelligent Transportation Systems.

Moutairou, M. M., Aniss, H., Delisle, G. Y., and Hakem,

N. (2009). Global optimization of multi radio mesh

access point location in underground area. Journal of

Networks, 4(7):630–640.

Na, L., Yuanan, L., Bihua, T., and Fan, W. (2009). A data

ﬁltration-aware mac protocol for wireless sensor net-

works. In IEEE Transactions on Mobile Computing.

Na, L., Yuanan, L., Bihua, T., Fan, W., Lidong, Z., and

Gang, X. (2007). A new mac protocol based on to-

ken ring in wireless sensor networks. In Interna-

tional Symposium on Communications and Informa-

tion Technologies.

Ndoye, E. H. M., Jacquet, F., Misson, M., and Niang, I.

(2013). Evaluation of rts/cts with unslotted csma/ca

algorithm in linear sensor networks. In NICST 2013.

Noori, M. and Arkani, M. (2008). Characterizing the trafﬁc

distribution in linear wireless sensor networks. IEEE

Communications Letters, 12:554–556.

SunHee, Y., Wei, Y., Jhon, H., Brian, L., and Cyrus, S.

(2009). Swats: Wireless sensor networks for steam-

ﬂood and waterﬂood pipeline monitoring.

Ted, T.-T. L., Kuei-Han, L., Polly, H., and Hao-Hua, C.

(2012). Triopusnet: Automating wireless sensor net-

work deployment and replacement in pipeline moni-

toring. In IPSN’12, Beijing, China.

Wang, N., Meng, Q., Li, T., and Ma, Q. (2011). Research on

linear wireless sensor networks used for on-line mon-

itoring of rolling bearing in freight train. In 9th Inter-

national Conference on Damage Assessment of Struc-

tures.

Xianpu, S., Yanling, Z., and Jiandong, L. (2007). Wireless

dynamic token protocol for manet. In International

Conference on Parallel Processing Workshops.

Zimmerling, M., Dargie, W., and M. Reason, J. (2008). Lo-

calized power-aware routing in linear wireless sensor

networks. In The 2nd ACM Workshop on Context-

Awareness for Self-managing Systems, Sydney, Aus-

tralia.

UsingaTokenApproachfortheMACLayerofLinearSensorNetworks-ImpactoftheRedundancyontheThroughput

129