Distributed Forwarder Selection on Beaconless Routing for Real-time

Services in Wireless Sensor Networks

Yongbin Yim, Hosung Park, Jeongcheol Lee, Seungmin Oh and Sang-Ha Kim

Department of Computer Engineering, Chungnam National University,

220 Gung-dong, Yuseong-gu, Daejeon, Republic of Korea

Keywords:

Beaconless, Real-time, Geographic Routing, Routing Protocol, Wireless Sensor Networks.

Abstract:

In wireless sensor networks (WSNs), real-time service is one of the important issues. Typically, existing

studies for the service are relied on beacons. Recently, beaconless routing has been proposed to overcome

control overhead for resource constrained environment of WSNs. Thus, real-time communication based on

the beaconless scheme could give great advantages with less energy consumption. To do that, however, it

brings new challenges. For the real-time communication, each node needs to be aware of single hop delay for

data forwarding within desired time. In conventional approaches, it is based on delay in neighbor information

through beacons. An appropriate next forwarder is selected by a sender. However, in the beaconless scheme,

a sender could not select a next forwarder since that is determined by a receiver. Also, contention delay is

included in single hop delay for beaconless routing. Thus, the delay estimation principle should be renovated.

In this paper, we present a receiver-based real-time routing protocol, called RBRR. We design a novel delay

estimation strategy. The single hop delay from a sender to a receiver could be calculated by a receiver but

not a sender. Therefore, the receiver itself makes a decision whether desired time requirement is satisﬁed.

Simulation experiments show that the strategy achieves higher performance in terms of energy consumption.

1 INTRODUCTION

Wireless sensor networks (WSNs) are composed of a

large number of sensor nodes with low power battery.

Besides, most of power resources are dissipated in

transmission. Therefore, design of energy conserving

routing protocol is one of the important issues in these

WSNs. Though earlier position-aware routing proto-

cols are considered as energy efﬁcient, they still suffer

periodic beacons to maintain knowledge of presence

of neighboring nodes. Nodes even not taking part in

any routing process should emit beacons. Also, it is

may not be appropriate for WSNs in highly dynamic

scenarios where network topology changes frequently

due to nodes availability.

Recently, beaconless routing protocols have been

proposed in an effort to overcome such drawbacks

(H. Fuessler and Hartenstein, 2003; B. Blum and

Stankovic, 2003). When data packets are needed to

transmit, next forwarder is selected by contention.

Contention is completed through waiting function,

which is uniquely assigned to each candidate for a

next forwarder selection. The winner for the con-

tention has permission to deliver data. Typically, a

node with the largest progress to a destination has the

shortest waiting time. This obtains an outcome of en-

ergy saving effect. Also, since actual existent neigh-

bors participate in the contention, it could avoid selec-

tion of non-valid next forwarder caused by outdated

neighbor information.

One of essential application categories that are in-

dispensable in WSNs is real-time service. Existing

real-time communication protocols are mostly gov-

erned by the conventionalgeographic routing scheme,

which each node has to periodically broadcast bea-

cons. Therefore, real-time communications based on

the beaconless scheme could give great advantages

with less energy consumption.

To do that, however, it brings new challenges. In

conventional real-time communication, a sender se-

lects an appropriate node which satisﬁed real-time

constraints and forwards data to the node by main-

taining information of neighboring nodes. For decid-

ing whether a neighbor node satisﬁes real-time con-

straints, a sender calculates delay between entering

time to output queue and sending time of the last

bit for a packet. Namely, delay information main-

tained by each node is duration while data is for-

202

Yim Y., Park H., Lee J., Oh S. and Kim S..

Distributed Forwarder Selection on Beaconless Routing for Real-time Services in Wireless Sensor Networks.

DOI: 10.5220/0004315902020207

In Proceedings of the 3rd International Conference on Pervasive Embedded Computing and Communication Systems (PECCS-2013), pages 202-207

ISBN: 978-989-8565-43-3

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

Broadcasts data

with sender’s average

processing delay

Sender’s

processing delay

Waiting

delay

single hop delay (S ĺn

)

Figure 1: Delay estimation in RBRR.

warded from the node itself to a neighbor node. How-

ever, in receiver-based beaconless schemes, a receiver

itself should decide to become a next forwarder. The

receiver requires delay of duration from a sender to

the receiver itself in such routing schemes. Thus,

existing schemes could not support this delay infor-

mation. Also, the receiver-based beaconless schemes

have contention delay to select a next forwarder with-

out neighbor information. Each candidate has unique

waiting time to avoid transmission collision. The

node having the shortest waiting time wakes up ﬁrstly

and wins permission to forward data. Therefore, this

contention delay should be included to the single hop

delay but existing schemes could not support this de-

lay estimation principle.

In this paper, we present a receiver-based beacon-

less real-time routing protocol, called RBRR. The sin-

gle hop delay is calculated at a receiver. When a

sender broadcasts data, the data includes average sin-

gle hop delay. After a neighbor wakes up, the node

decides to become a next forwarder by the average

single hop delay information and waiting time con-

sumed for contention. Simulation results show that

our protocol has less energy consumption and high

performance for real-time service.

The remainder of this paper is organized as fol-

lows. In section II, we review related work. The pro-

posed protocol is presented in Section III and the ex-

periment result of the proposed protocol is Section IV.

Finally, the paper is concluded in Section V.

2 RELATED WORK

Beaconless routing reduces energy consumption by

eliminating beacon message exchange. Beaconless

routing protocols are categorized into two types ac-

cording to selection scheme of relay node: sender-

based and receiver-based.

There are IGF (B. Blum and Stankovic, 2003),

Forwarding

Area

Sub-time zone 0

- 1

Figure 2: Waiting delay assignment by sub-time zone for

contention.

CBF (H. Fuessler and Hartenstein, 2003) in sender-

based scheme. These studies exploit RTS/CTS hand-

shaking. A sender broadcasts RTS and then neighbors

receiving this message have unique waiting time re-

spectively. A node waking up ﬁrst sends CTS to the

sender. The sender selects the node as a relay node

and delivers data to the node.

BLR (M. Heissenbuttel and Wachli, 2004), BGR

(Turau, 2005) could be deﬁned as receiver-based

scheme. In these studies, a sender deliver data ﬁrst

rather than RTS/CTS messages. Neighbors receiving

data also wait for contention. A node waking up ﬁrst

resends data for next forwarding. In other words, re-

ceivers themselves whether to decide deliver data or

not.

SPEED (T. He and Abdelzaher, 2005) exploits

feedback control technique to maintain delay of

neighbors. The study selects relay nodes by non-

deterministic geographic forwarding scheme. Each

sensor node sends 1-hop delivery speed information

to neighbors which want the information. It maintains

nodes which meet desired speed among its neigh-

bors into routing list. Then a sender transmits pack-

ets to one of the nodes in the list. At this time, se-

lected node as a relay node reports its own delivery

speed to surroundingnodes after transmitting a packet

by the feedback control technique. SPEED of this

kind method is the ﬁrst spatiotemporal communica-

tion scheme.

CBRR (Huang and Wang, 2010) is recent one of

studies for providing real-time service based on bea-

conless routing. CBRR exploits sender-based relay

node selection scheme. Since sender-based schemes

need additional control process such as RTS/CTS

message exchanging, sender-based schemes have

longer delay than receiver-based schemes have. Thus,

deliverysuccess ratio within desired time could be de-

crease.

DistributedForwarderSelectiononBeaconlessRoutingforReal-timeServicesinWirelessSensorNetworks

203

3 RECEIVER-BASED

BEACONLESS REAL-TIME

ROUTING

Unlike convention real-time routing protocols which

maintain neighbor list and select a next relay node

utilizing the list, in RBRR, a receiver itself deter-

mines whether to be a relay node. Also, because there

is no exchanging beacon messages, sensors are not

aware of existence of each other. Therefore, a sender

broadcasts data and a next forwarder acquires trans-

mission permission through contention process in the

receiver-based schemes. Waiting time is consumed

for the contention so it should be considered that this

time is included. Thus, we need to redesign delay

estimation principle to adapt beaconless-routing. In

this section, we describe delay estimation method on

receiver-based scheme and explain how data is deliv-

ered by utilizing this method.

3.1 Spatiotemporal Approach for

Real-time Services

The conventional schemes (T. He and Abdelzaher,

2005) for real-time data dissemination mainly ex-

ploit the spatiotemporal approach in order to deliver

data from a source to a static sink within a desired

time deadline T

setdeadline

. While in multi-hop wireless

sensor network, since communication is physically

bounded, the end-to-end delay depends not only on

single hop delay (temporal), but also on the distance a

packet travels (spatial). To achieve this, source nodes

initially calculate a desired delivery speed S

setspeed

with the time deadline and the end-to-end distance

d(source, sink) from the source to the sink as follows:

setspeed

D(source, sink)

setdeadline

(1)

In the protocols, each node on the dissemination

route selects a node as its next-hop node which is

nearer to the sink and provides a better relay speed

than the desired delivery speed S

setspeed

. The relay

speed means the advance in distance to each next node

dividing by the delay to forward a packet to the each

next node. The end-to-end real-time data dissemina-

tion is achieved by maintaining the desired delivery

speed from sources to the sink.

3.2 Delay Estimation

In order to calculate relay speed discussed above at

each hop, we need to be aware of the single hop delay

(radio range)

Figure 3: Basic principle of data forwarding in RBRR.

between a sender and a receiver, and the distance be-

tween them. Single hop delay means hold-up time in a

hop. In RBRR, we deﬁne single hop delay as the time

for that data from a current forwarder is broadcasted

and a next forwarder is determined among neighbors

receiving the data. Therefore, single hop delay in-

cludes processing delay due to buffering, MAC pro-

tocols and so on such backoff, and waiting delay for

contention.

Existing single hop delay estimation is computed

by a sender because next forwarder selection is per-

formed on sender side. However, since a next for-

warder is determined by neighbors after receiving

data in RBRR, it should provide single hop delay in-

formation for receivers. Conventional single hop de-

lay estimation principle could not be directly adapted.

When a sender broadcasts data, it includes aver-

age processing delay of the sender into the data packet

header. Calculation of the processing delay is ob-

tained by difference between the entering time into

output queue of a data packet (T

arr

) and the actual

sending time of last bit of the data (T

dprt

) as well as

the conventional real-time schemes. This time is ac-

cumulated to previous average processing delay and

then the average processing delay is updated as fol-

lows.

Delay

new

= ε·(T

dprt

− T

arr

)+ (1− ε) · Delay

(2)

As shown in Fig. 1, when a candidate node wakes

up after certain waiting time, this node adds average

processing delay of the sender to consumed time for

waiting (waiting delay) and decides to satisfy desired

speed.

3.3 Forwarding in RBRR

Since communication is physically bounded in wire-

less networks, minimizing the number of hops brings

better performance in terms of delivery deadline suc-

cess ratio. Therefore, the closer node is to a destina-

tion, the shorter waiting time becomes in contention

process for beaconless routing. Waiting time is as-

signed according to distance to a destination between

PECCS2013-InternationalConferenceonPervasiveandEmbeddedComputingandCommunicationSystems

204

Overhear data

transmission of

another candidate?

Timer is

expired?

Receiving

data

Discard the data

Waiting for contention

Satisfy

desired speed?

Broadcast the data

yes

Figure 4: Decision ﬂow of becoming a next forwarder.

0 and D

MAX

. D

MAX

represents maximum time given

by an application to wait for contention.

For contention process, each candidate must have

unique waiting time to avoid collision with the oth-

ers. It can be assigned according to progress toward

a destination. However, some nodes might have same

progress. Also, adding simply random values incurs

duplicate wait time as well in case of some nodes hav-

ing distance differences as much as a random value.

To deal with this, we divide forwarding area into sub-

time zones (J. A. Sanchez and Ruiz, 2007). Each sub-

time zone has mutually different waiting time. In a

same sub-time zone, each candidate has random time

respectively. Therefore, all candidates could avoid

transmission collision. To construct sub-time zones,

a forwarding area is divided into the number of sub-

time zones (N

) given according to applications. A

zone index afﬁliated candidate neighbor n is obtained

as follows:





1−

d(s, d) − d(s, n)



(3)

where s and d are the current forwarding node and

the destination. The function d(c, d) represents the

Euclidean distance between the positions of the nodes

c and d. In order to give shortest waiting time to the

closest node toward the destination, it reversely assign

the time in proportion to radio range r.

Each sub-time zone could wait until for time

MAX

/ N

). Thus, we could compute delay of each

candidate neighbor as follows:

wait



MAX



+ rand



MAX



(4)

here, the function rand(x) returns a random value be-

tween 0 and x. Therefore, a node in the farthest group

from a sender node starts wake up in order. If there is

no qualiﬁed node, chance passes to the next group by

equation (4). Fig. 2 shows an example of the sub-time

zone.

Movement speed could be calculated by single

hop delay and progressed distance. Satisfying this

speed represents that real-time data is successfully

delivered within desired time. Thus, data is re-

broadcasted by the receiver and this means namely

relaying data is successfully fulﬁlled. This procedure

is continuously repeated until data reach to a destina-

tion. Fig. 3 shows data delivery process described

above. Sender S broadcasts data. Then candidate

neighbor n

, n

and n

receive data and wait for as-

signed time respectively. Since n

is the closest to

destination D, n

wakes up ﬁrst and decides whether

to be able to satisfy desired speed. If n

satisﬁes the

desired speed, n

broadcasts data as the next relay

node. Other candidates overhear the broadcasting of

and cancel its own waiting timer.

When the ﬁrst node ﬁnished waiting time trans-

mits data as a relay node, other candidates give up

the contention by overhearing the transmission. How-

ever, some nodes might not be included within ra-

dio range of the relay node. This may cause du-

plicate transmission. To avoid the duplications, we

bring Reuleaux Triangle described in BLR. All nodes

within the area can overhear each other. Basic process

of next forwarder decision is described in Fig. 4.

4 EXPERIMENTAL RESULTS

We compare the performance of RBRR with that of

the representative real-time routing scheme in WSNs,

SPEED (T. He and Abdelzaher, 2005). We imple-

mented three protocols in Network Simulator Qual-

net 4.0 (QUALNET, 2008). Sensor nodes follow the

speciﬁcation of MICA2 (Hill and Culler, 2002) and

their transmission range is about 50m. IEEE 802.11b

was used as the MAC layer protocol. The size of

the sensor network is set to 250m 250m where 2500

nodes are randomly distributed. For all simulations,

we use one source-sink pair for performance evalua-

tion. We use the following metrics for performance

analysis and evaluations: the success delivery ratio is

deﬁned as the ratio of the number of data packet suc-

cessfully received within desired time by the sink to

the number of data packets generated by the source.

The control overhead is deﬁned as the total number

of control packets. The average energy consumption

is deﬁned as average consumption of twenty times of

transmission. Transmitting and receiving power con-

sumption rates of the sensors are 21mW and 15mW,

respectively. The beacon interval is set to 1.5s and ac-

DistributedForwarderSelectiononBeaconlessRoutingforReal-timeServicesinWirelessSensorNetworks

205

300 400 500 600 700 800 900

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Delivery deadline success ratio

Desired delivery speed (m/s)

SPEED

RBRR

60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360 380 400

1500

2000

2500

3000

3500

4000

4500

5000

5500

Average Energy Consumption (mW)

Simulation Time (ms)

SPEED

RBRR

8 10 12 14 16

20000

25000

30000

35000

40000

The number of packets

Node density (node numbers / radio range)

SPEED

RBRR

Figure 5: Success delivery ratio im-

pacted by sink speed.

Figure 6: Average energy consumption

according to elapsed simulation time.

Figure 7: The number of total packets

impacted by node density.

cordingly the time-out interval to 4.5 1.5s = 6.75s af-

ter which a node is deleted from the neighbor table if

no beacon is received. We use constant bit rate (CBR)

trafﬁc. To test performance at one time, a node is ran-

domly chosen from left side of the terrain and sends

periodic data to the right side of the terrain. The node

generates 1 CBR ﬂow with a rate of 1 packet/second.

The results in Fig. 5 show network lifetime ac-

cording to elapsed simulation time. For comparing

network lifetime, we estimate the energy consump-

tion of the whole sensor nodes based on MICA2 mote

energy model. We set battery capacity of a sensor

node to 3000mW. In SPEED, since each sensor node

periodically exchanges beacon messages, depletion

time of a node is shorter than one of RBRR.

In Fig. 6, deadline delivery success ratio due to

desired delivery speed is depicted. We vary a de-

sired delivery speeds between 400m/s and 1000m/s

to compare an effect on the deadline delivery success

ratio against different speeds. In this simulation, we

set that SPEED is so operated that the fastest path is

selected for transmission. Also, we do not consider

trafﬁc distribution. SPEED has no delay for selec-

tion relay nodes so delivery deadline success ratio is

relatively higher than one of RBRR. However, since

success ratio keeps over 0.9 to desired speed 700m/s,

it could say that RBRR is effective.

Fig. 7 shows communication overhead accord-

ing to density. Packet overhead means total number

of packets sent at the radio layer. In SPEED, packet

overhead rapidly increases due to congestion caused

by beacons to update neighbor table. Therefore, re-

transmission or additional control is more required ac-

cording to be dense. However, in RBRR, the density

does not affect the communication overhead. Since

RBRR does not require beaconing, congestion proba-

bility is almost constant. So, RBRR has advantage in

terms of scalability.

Fig. 8 shows the number of node alive accord-

ing to elapsed simulation time. Periodic beacon mes-

sages in SPEED quicken dissipation of node. There-

fore, the number of node alive dramatically decreases

in the middle of simulation time. At the end of the

simulation time, since many nodes are dissipated, the

number of beacons decreases and then nodes dissipa-

tion speed shows a decreasing trend. On the other

hand, since RBRR extremely reduces control mes-

sages, nodes endure long time in comparison with

SPEED.

Fig. 9 plots delivery deadline success ratio in

network dynamics. In this experiment, we set de-

sired delivery speed to 500m/s. As simulation time

passes, network topology frequently changes due to

dissipated sensor nodes. In the middle of the simu-

lation time between 140s and 240s, sensor nodes are

rapidly dissipated and then topology of the network

variesfast. In SPEED, since forwarder selection relies

on the neighbor table, information about some nodes

might not valid anymore on the situation. Thus, trans-

mission failure might occur frequently until the tables

are updated by beacons. However, RBRR needs not

to maintain neighbor tables. It always operates among

actually existent nodes. RBRR is tolerant of network

dynamics in comparison with conventional real-time

scheme based on beacon messages.

Fig. 10 shows end-to-end delay according to net-

work dynamics. As we discussed above experiment,

frequent topology changes might cause retransmis-

sion or additional control. It affects to end-to-end

packet delivery delay. In SPEED, delay rapidly in-

creases due to the reason. As simulation time passes,

many nodes are dissipated. Therefore, end-to-end

path is longer so that the delay proportionally in-

creases. Although RBRR has typically larger delay

than that of SPEED, it shows stable delay in high dy-

namics of networks.

5 CONCLUSIONS

In this paper, we propose a receiver-based beaconless

PECCS2013-InternationalConferenceonPervasiveandEmbeddedComputingandCommunicationSystems

206

60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360 380 400

500

1000

1500

2000

2500

The number of nodes alive

Simulation time (s)

SPEED

RBRR

60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360 380 400

0.6

0.7

0.8

0.9

1.0

Delivery Deadline Success Ratio

Simulation time (s)

SPEED

RBRR

60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360 380 400

100

120

140

160

180

200

220

240

End-to-end Delay (ms)

Simulation time (s)

SPEED

RBRR

Figure 8: The number of nodes alive ac-

cording to elapsed simulation time.

Figure 9: Delivery deadline success

ratio according to elapsed simulation

time.

Figure 10: End-to-end delay according

to elapsed simulation time.

real-time routing scheme. Beaconless routing with-

out periodic exchanging beacon messages is attrac-

tive approach. For providing real-time services with

receiver-based beaconless, delay estimation principle

should be renovated focusing on receivers. To do that,

we propose new delay estimation principle, called

RBRR. A sender includes its own single hop delay

into data for receiver to receive. Also, it adds waiting

time for contention to single hop delay. Experimental

results for performance evaluation show that RBRR

well provides real-time services while achieves high

energy efﬁciency.

REFERENCES

B. Blum, T. He, S. S. and Stankovic, J. (2003). Igf: A

statefree robust communication protocol for wireless

sensor networks. Technical report.

H. Fuessler, J. Widmer, M. K. M. M. and Hartenstein, H.

(2003). Contention-based forwarding for mobile ad

hoc networks. AdHoc Networks, 1(4):351–369.

Hill, J. and Culler, D. (2002). Mica: a wireless platform for

deeply embedded networks. IEEE Micro, 22(6):12–

24.

Huang, C. and Wang, G. (2010). Contention-based bea-

conless real-time routing protocol for wireless sensor

networks. Wireless Sensor Network, 2(7):528–537.

J. A. Sanchez, R. M.-P. and Ruiz, P. M. (2007). Boss:

beacon-less on demand strategy for geographic rout-

ing in wireless sensor networks. In MASS’07, 4th

IEEE Int. Conf. on Mobile Ad-hoc and Sensor Sys-

tems, pages 1–10. IEEE.

M. Heissenbuttel, T. Braun, T. B. and Wachli, M. (2004).

Blr: Beacon-less routing algorithm for mobile ad-

hoc networks. Elsevier’s Computer Communications

Journal (ECC), 27(1):1076–1086.

QUALNET (2008). Scalable network technologies. http://

www.scalable-networks.com.

T. He, J. Stankovic, C. L. and Abdelzaher, T. (2005). A

spatiotemporal communication protocol for wireless

sensor networks. IEEE Trans. Parallel and Distrib.

Syst.

Turau, M. W. V. (2005). Bgr: Blind geographic routing for

sensor networks. In WISES’05, the Third Workshop

on Intelligent Solutions in Embedded Systems. IEEE.

DistributedForwarderSelectiononBeaconlessRoutingforReal-timeServicesinWirelessSensorNetworks

207