Upload
phoenixbyte
View
220
Download
2
Embed Size (px)
Citation preview
Computer Networks xxx (2013) xxx–xxx
Contents lists available at ScienceDirect
Computer Networks
journal homepage: www.elsevier .com/ locate/comnet
A lightweight distributed scheme for mitigating inter-userinterference in body sensor networks
1389-1286/$ - see front matter � 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.comnet.2013.09.013
⇑ Corresponding author at: Department of Electrical and ComputerEngineering, National University of Singapore, Singapore. Tel.: +65 65161076.
E-mail addresses: [email protected], [email protected] (W.Sun).
Please cite this article in press as: W. Sun et al., A lightweight distributed scheme for mitigating inter-user interference in bodynetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.2013.09.013
Wen Sun a,b,⇑, Yu Ge b, Wai-Choong Wong a
a Department of Electrical and Computer Engineering, National University of Singapore, Singaporeb Sense and Sense-abilities Programme, Institute for Infocomm Research, Singapore
a r t i c l e i n f o a b s t r a c t
Article history:Received 11 January 2013Received in revised form 23 May 2013Accepted 24 September 2013Available online xxxx
Keywords:Inter-user interferenceInterference mitigationBody sensor networksScheduling algorithm
Inter-user interference deteriorates reliable communication in body sensor networks(BSNs) when multiple BSNs are transmitting simultaneously in close proximity to eachother. This paper presents a lightweight and distributed inter-user interference mitigation(IIM) scheme, that can be easily integrated with the IEEE 802.15.4 protocol stack. The pro-posed scheme takes into consideration the generic property of low channel utilization inBSNs and enables affected BSNs to adaptively reschedule their transmission time or switchchannels. Based on the detected information from neighboring BSNs, BSNs reschedule theirtransmissions in a distributed and coordinated manner, so that wireless channels can beeffectively utilized by multiple BSNs. Moreover, the IIM scheme is performed only whenthe performance of the BSN is degraded to an unacceptable level due to severe interferenceto reduce the rescheduling cost. Simulation results show that the proposed schemeimproves the network throughput by 18% and reduces the energy consumption by 22%as compared with the existing beacon schedule scheme.
� 2013 Elsevier B.V. All rights reserved.
1. Introduction
Body sensor network (BSN) enables wireless communi-cations between several miniaturized body sensors and asingle coordinator worn on the human body. Since a BSNprovides remote and continuous health monitoring for pa-tients without constraining their movements, it plays acrucial role in next generation healthcare applications[1,2].
As physiological information is transmitted in BSNs,reliable data transmission is crucial. Inter-user interfer-ence, which is incurred by simultaneous transmissions ofmultiple BSNs in the same vicinity, deteriorates reliablecommunication of BSNs. Natarajan et al. [3] highlighted
the existence of inter-user interference, and found thatsuch interference reduces packet delivery rate by 35% inthe presence of eight or more interfering BSNs. In ourprevious work [4], we investigated the prevalence andseverity level of inter-user interference in a realistic BSNdeployment in a hospital scenario, and showed that only68.5% of data transmission can meet the reliability require-ment even in the off-peak period. Such a situation is aggra-vated when more BSN applications are deployed [4].Therefore, it is imperative to devise an effective inter-userinterference mitigation scheme for BSNs.
The existing interference mitigation methods designedfor other networks are inappropriate for BSNs due to thefollowing reasons:
First, the BSN communication has stringent require-ments such as reliability and energy efficiency in health-care applications. The commonly used contention-basedcarrier sense multiple access with collision avoidance(CSMA/CA) method [5] may not be able to satisfy the com-munication requirement of BSNs due to its unreliable clear
sensor
2 W. Sun et al. / Computer Networks xxx (2013) xxx–xxx
channel assessment (CCA), traffic correlation, and severecollisions [6].
Second, exchange of messages exists only within thecluster of sensor nodes in a BSN and there is no messageexchanges among BSNs. Without inter-BSN communica-tion, it is challenging for BSNs to collect the surroundinginformation and take actions in a coordinated manner toreduce interference. For example, the mesh election algo-rithm is effectively used in the IEEE 802.16 wireless meshnetwork [7,8] to avoid collisions among nodes by using theneighborhood information obtained by inter-device coor-dination. However, such coordination is inapplicable inBSNs.
Third, BSNs are usually mobile which differentiatesthem from most other wireless sensor networks (WSNs).In WSNs, inter-cluster interference can be minimized bya self-organizing medium access control (MAC) allocationscheme based on the feedback derived from collisionsexperienced by the local nodes within a cluster [9]. Thismethod is difficult to apply in BSNs, as the delay for thefeedback becomes intolerable when mobility is involved.Another similar example is the cluster scheduling and col-lision avoidance problem in IEEE 802.15.4 beacon-enabledcluster-tree WSNs [10], in which beacons from differentclusters are assigned to transmit in their dedicated timeslots using time division method. However, a static andpredefined deployment of wireless nodes is assumed, mak-ing it inapplicable in the mobile BSN scenarios.
The above mentioned challenges motivate us to designa lightweight and distributed inter-user interference miti-gation (IIM) scheme explicitly for BSNs. In this paper, wepropose an IIM scheme, which takes into considerationthe generic property of low channel utilization in BSNsand enables BSNs to adaptively reschedule their transmis-sion time or channel if interference occurs. It includes dy-namic detection of inter-user interference, collection ofneighboring information, and rescheduling of transmissionaccordingly.
The main contributions of this paper are as follows.Firstly, we model the inter-user interference in BSNs andcalculate the rescheduling time of a BSN in two dimen-sions, target to mitigate the interference with the shortestlatency. Secondly, we propose a scheme to mitigate inter-user interference in a distributed manner without relyingon any centralized synchronization mechanism. The pro-posed scheme is reservation-based, while the benefits ofreservation-based and contention-based schemes are com-bined to reduce the rescheduling cost. Thirdly, we conductextensive performance evaluation through simulations andprove that the proposed scheme significantly improvesthroughput and energy consumption, as spectrum utiliza-tion is improved by rescheduling the transmissions of mul-tiple BSNs on the same channel for the low loading BSNscenarios.
The remainder of this paper is organized as follows. InSection 2, we review the related work in interference mit-igation schemes for BSNs. Section 3 describes the networkmodel and then formulates the inter-user interference.Section 4 introduces our scheme to mitigate inter-userinterference. Then the performance of the proposed miti-gation scheme is analyzed in Section 5. Section 6 compares
Please cite this article in press as: W. Sun et al., A lightweight distribunetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
the performance of the proposed scheme with the existingapproaches through simulations. Finally, Section 7 con-cludes this paper.
2. Related work
The existing interference mitigation schemes for BSNsmainly fall into several categories: frequency division mul-tiple access (FDMA), power control, and time divisionschemes.
Sergio and Chen [11] proposed a FDMA approach,where each BSN is assigned a different frequency channelat the network initialization phase. This approach allowsmonitoring as many patients as available channels, butradio channels are hard to be reused in a dynamic way.To increase the number of monitored BSNs and enablethem to move freely, an alternative approach is to allocatechannels dynamically in the small-scale deployment ofBSNs. Silva et al. [12] developed an infrastructure-basedscheme, where BSNs are reallocated channels by a fixedinfrastructure when they move into the radio range of eachother. This scheme reduces interference effectively if thenumber of congregated BSNs within the interference rangeis fewer than the number of available channels. However,besides the infrastructure cost, this approach leads to fre-quent channel switching, which incurs much overheadand is thus unsuitable for occasional and short-terminterference.
Power control is another approach to reduce the inter-ference in multi-user environments. Wu et al. [13] pro-posed a power control approach for interferencemitigation, where each BSN measures the interferencefrom other BSNs and then selects a suitable channel andtransmission power by utilizing non-cooperative gametheory and a no regret learning algorithm. A major draw-back of this method is the long iteration period (more than20 iterations) to reach the optimal point. As such, the uti-lized channels and transmission powers may be changedfrequently during the long computing period which makesthe system unstable. Power control improves spatial utili-zation of channels, but it may compromise transmissionperformance with a reduced transmission power [4].
In healthcare applications, the channel utilization of aBSN is usually low for energy conservation. Kim et al.[14] proposed a distributed flexible beacon schedulescheme to reduce the interference. By employing carriersensing before each beacon transmission, collisions canbe avoided if other BSNs attempt to access the channel atthe same time. This scheme consumes additional energyin the channel access as multiple carrier sensing iterationsare possibly conducted before each beacon transmission.Considering the periodic data characteristics in most BSNapplications, a reservation-based scheme outperforms acontention-based scheme in terms of energy conservationand throughput enhancement, because overhead and colli-sions are significantly reduced in the reservation-based ap-proach. As such, our proposed IIM scheme utilizes thereservation-based approach, where reservations are madedynamically based on the information acquired from chan-nel listening. The simulation results in Section 6 show that
ted scheme for mitigating inter-user interference in body sensor13.09.013
Inactive period
: Beacon : Data transmission
2SOSD aBaseSuperframeDuration
(Active period)
W. Sun et al. / Computer Networks xxx (2013) xxx–xxx 3
the proposed scheme improves network throughput by18% and reduces energy consumption by 22% as comparedwith the work of Kim et al. [14]. Incorporated with channelswitching, IIM can be used in many application scenarioswith both light and heavy loads.
2BOBI aBaseSuperframeDurationCAP CFP
(Superframe)
Fig. 2. An example of the IEEE 802.15.4 MAC superframe structure(SO = 0, BO = 1).
t
iSD
t
iBI delayT
BSN i
BSN j
it
jt
0it
iSD : Active period length of BSN iiBI : Superframe length of BSN i
: Transmission : Collision : Transmission if no rescheduling
Fig. 3. The inter-user interference between BSN i and BSN j.
3. Network model and problem description
3.1. Network model and assumptions
Fig. 1 illustrates the common architecture of BSNs. Thephysiological information collected by sensor nodes is firstdelivered to a coordinator on the body, which then for-wards the information to the local or remote server for fur-ther processing [15–17]. In this paper, we only consider thecommunication between the sensor nodes and the coordi-nator for inter-user interference analysis. The interferenceincurred by the communication between the coordinatorand the server is out of the scope and is handled by theexisting technology such as wireless local area network(WLAN) and cellular network [18,19].
Currently the most widely used radio standard for BSNcommunication is the IEEE 802.15.4 (ZigBee) standard,which is a cost-effective technology [20]. The IEEE802.15.4 protocol usually operates in star topology in aBSN, i.e., the sensor node either transfers data to the coor-dinator, or polls the coordinator to receive data. There aretwo modes designed for the IEEE 802.15.4 multiple accessscheme: non-beacon enabled and beacon enabled, depend-ing on whether the network supports the transmission ofbeacons. In a non-beacon enabled scheme, a sensor nodesimply transmits data using unslotted CSMA/CA. In a bea-con enabled mode, beacons are utilized to synchronizethe transmission of sensor nodes in a superframe.
In this study, we consider the IEEE 802.15.4 wirelesstechnology with beacon-enabled mode [21] for the BSNcommunication. Fig. 2 shows the superframe structure ofthe beacon-enabled mode, which contains an active periodand an inactive period. The active period is further dividedinto a contention access period (CAP) and an optional con-tention free period (CFP). At the beginning of the activeperiod, the coordinator of a BSN synchronizes its sensornodes by broadcasting a beacon packet. The beacon packetcontains schedule information of the BSN. According to thebeacon information, a sensor node that wishes to commu-
BSN2BSN1 BSN3
Sensor nodeCoordinator
Cloud
Medical databaseEmergency
BSN application field
Interference range
Fig. 1. The common architecture of BSNs.
Please cite this article in press as: W. Sun et al., A lightweight distribunetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
nicate during the CAP competes with other nodes using aslotted CSMA/CA mechanism. In the CFP, nodes transmitin a TDMA mode in their allocated slots without competi-tion. Sensor nodes are in sleep mode in the inactive periodto save energy. The length of a superframe is equal to abeacon interval (BI), and the length of the active period isequal to a superframe duration (SD). BI and SD are set bythe beacon order (BO) and the superframe order (SO)(0 6 SO 6 BO 6 14) respectively. As shown in Fig. 2, aBase-SuperframeDuration denotes the minimum duration of thesuperframe, which is 15.36 ms assuming 250 kbps trans-mission rate in 2.4 GHz frequency band.
BSN communication remains within the cluster of sen-sor nodes worn on a human body. Inter-user interference issignificantly affected by BSN mobility when the BSNs moveinto the interference range of each other and transmitsimultaneously. Mobility models of the mobile users areconsidered in the IIM design and performance evaluation.1
The pedestrian characteristics of the mobility pattern ofBSNs are set according to [22].
In this study, we make the following assumptions:
(1) All the BSNs have the same priority for their datatransmission.
(2) The length of the superframe (BI) is the same for allBSNs, while the starting time of the superframesmay be different.
(3) Periodic and low duty cycle data transmissions areassumed in BSNs, which is the case in most health-care applications.
3.2. Problem description
In this subsection, we first formulate the inter-userinterference problem, and then propose our solution.
1 In the simulations, the random waypoint model is utilized for themovement of the BSNs because of its simplicity and wide adoption.
ted scheme for mitigating inter-user interference in body sensor13.09.013
4 W. Sun et al. / Computer Networks xxx (2013) xxx–xxx
The schedule information of BSN i is expressed as (SDi, -BIi,ci, ti), where SDi and BIi are the SD and BI of BSN i, ci is thechannel index and ti is the next transmission time of BSN i.As shown in Fig. 3, BSN i starts transmission at ti, thetransmission lasts for SDi and repeats every period of BIi
on channel ci if no collision occurs. Collision happens whena neighboring BSN j has overlapping transmission withBSN i on the same channel, i.e., kti � tjk 6 SDi. Suchoverlapping transmission is referred to as inter-user inter-ference. The interference lasts until the departure of theinterferer BSN j. The notations involved are summarizedin Table 1.
As shown in Fig. 3, our proposed IIM scheme resched-ules BSN is transmission from ti
0 to ti in order to avoid col-lisions. Each rescheduling operation incurs a delay Tdelay
from the original transmission time ti0 to the rescheduled
transmission time ti. The minimum delay Tdelay is desirablefor achieving optimal throughput. In the case that N BSNscongregate together, denoting the neighbor set of BSN ias {Xi: (SDj,BIj,cj, tj)j1 6 j 6 N,j – i}, BSN i needs to avoidcollisions with all the BSNs in its neighbor set Xi inrescheduling. Moreover, when multiple channels can beutilized, the transmission can be rescheduled in twodimensions (ci, ti). Fig. 4 shows the transmission status ofneighboring BSNs around BSN i in a multiple-channel sce-nario, which depicts whether it is collision or collision-freefor BSN i to start transmission at (ci, ti). Similar to the sce-nario in Fig. 3, each reschedule decision (ci, ti) has a delayTdelay(ci, ti) from the original transmission time to therescheduled transmission time. In general, the objectiveof our IIM scheme is to select a collision-free duration,starts with ti on channel ci and lasts for SDi, with minimumlatency Tdelay,
Table 1The notations of the terms.
i, j BSN indexSDi Superframe duration of BSN iBIi Beacon interval of BSN ici Channel index of BSN iti Next transmission time of BSN it0
i Transmission time of BSN i if no collisionsXi The neighbor set of BSN iN The number of neighboring BSNsTdelay Rescheduling delay(al, . . . , an) The time slots, n is the number of time slots in a
superframe(cl, . . . , cm) Available channel set, m is the number of channels
0
1c
2c
3c
16a8a 24a 32a 40a 48a 56a 64a
: Collision: Collision-free
Cha
nnel
inde
x
Time slot index
Fig. 4. Transmission status of neighboring BSNs around BSN i in amultiple channel scenario.
Please cite this article in press as: W. Sun et al., A lightweight distribunetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
ðci; tiÞ ¼ arg minðcl ;akÞ
ðTdelayðcl; akÞÞ; ð1Þ
where cl is the channel index in the available channel set,ak is the time slot index in the rescheduled superframe. Be-cause of the slotted access in the IEEE 802.15.4 beacon-en-abled mode, ti needs to be selected from one of the timeslots ak. After selecting (ci, ti), BSN i reserves the successiveSDi for transmission by announcing the schedule in thenext beacon.
It is challenging to obtain such a transmission statustable in the absence of global synchronization. Note thatthe beacon contains the schedule information (SDi,BIi,ci, ti).The transmission status table of BSN i can be obtained byoverhearing the beacon packets of its neighboring BSNs.In the next section, we will describe the IIM scheme indetail.
4. IIM scheme
4.1. IIM procedure
The IIM scheme has the following modules: (1) inter-user interference detection, (2) transmission status collec-tion from other BSNs, in this module, beacons are collectedfrom other BSNs, and (3) rescheduled time and channelselection module, as shown in Fig. 5. In particular, foursteps are followed:
Step 1 (Initiation): The IIM scheme is initiated when aBSN (e.g. BSN i) experiences significant performancedegradation. That is the coordinator identifies signifi-cant decrease in throughput or packet reception ratio,while the received signal strength does not obviouslydrop. In this case, the performance degradation is prob-ably due to BSN congestion instead of bad channel.Upon initiation, sensor nodes of BSN i fall asleep asusual during the inactive period, and wake up at thenext active period to await the beacon packet fromthe coordinator. The coordinator enters the listeningperiod instead of falling asleep after this active period.Step 2 (Listening): The coordinator of BSN i listens for asuperframe length (BI) to collect its neighbors’ informa-tion by overhearing their beacon packets. BSN i canthen decode the overheard beacons and get the sche-dule information from them. Based on the overheardbeacon information, BSN i establishes its neighbor listwith their schedule information. The problem of incom-plete neighbor list problem will be addressed in Step 4.Step 3 (Rescheduling): After the listening period, thecoordinator of BSN i executes the reschedulingalgorithm (described in Section 4.2) to determine thepossible rescheduled transmission time ti
temp. This step
Inter-userInterference
detection
Beacon collectionfrom other BSNs
Selectingrescheduled time
and channel
Fig. 5. Modules of the IIM scheme.
ted scheme for mitigating inter-user interference in body sensor13.09.013
Regular transmission
Performancedegradation?
IIM Initiation
Listening
Current channelfully-utilized?
Switch channel
Transmit with carriersensing and backoff
Collision?Transmit at the nexttransmission time
Calculate the possibletransmission time
Yes
No
Yes
No
YesNo
Rescheduling
Avoidance of incomplete neighbor list
Fig. 6. Flow chart of the IIM scheme.
W. Sun et al. / Computer Networks xxx (2013) xxx–xxx 5
Plene
ensures that overlapping transmissions can be allevi-ated, while the existing transmissions of other BSNsare not affected.Step 4 (Collision avoidance in case of incomplete neigh-bor information): To avoid collisions caused by anincomplete neighbor list, the BSN transmits at therescheduled transmission time ti
temp with carrier sensingand backoff (see Section 4.3). The coordinator informsits sensor nodes of the new schedule through its bea-con. After that, the sensor nodes of BSN i transmit peri-odically in the assigned time slots until the next IIMinitiation.
BSN1:(not initiate)
BSN2:
BSN3:
BSN4:
Initiate
Initiate
Initiate
Listening period
Listening period
Listening period
1t
Inactive period
2t
Resch
: Beacon : Data transmission : Collision
Fig. 7. Example of th
ase cite this article in press as: W. Sun et al., A lightweight distributworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
In Step 3, when the rescheduling algorithm fails to ob-tain ti
temp on the current channel, the channel is deemedfully-occupied. Accordingly, the coordinator switches toanother channel and executes from Step 2 again. The flowchart of the IIM scheme is shown in Fig. 6. Fig. 7 gives anexample of the IIM procedure where four co-located BSNsare transmitting on the same channel. In Fig. 7, BSN 1maintains its regular transmission in the absence of colli-sions, while BSN 2, BSN 3, and BSN 4 are initiated sequen-tially to reschedule their transmissions due to collisions.After a listening period, they are rescheduled to avoidcollisions without interrupting the transmission of BSN 1.The schedule is maintained periodically thereafter untilthe next initiation.
There are a few design considerations in IIM:
� Carrier sensing and backoff are performed only beforethe rescheduled beacon transmission.� Rescheduling transmission time is preferred for inter-
ference mitigation as compared with switching chan-nels for spectrum efficiency when the channelutilization is low [23].� If switching channel is desirable, the coordinator of BSN
i first informs its sensor nodes of the schedule using theoriginal channel. Then the coordinator and sensor nodesswitch to the new channel to start transmissions.
4.2. Rescheduling algorithm
In Step 3 of the workflow described in Section 4.1, therescheduling algorithm is executed to determine the possi-ble transmission time ti
temp for BSN i. It consists of estab-lishing the transmission status table and selecting ti
temp
accordingly.We define two superframes in the IIM scheme: (1) lis-
tening superframe, where BSN i listens to the beaconpackets from other BSNs; and (2) rescheduling super-frame, where BSN i establishes its transmission status ta-ble and reschedules its transmission. The transmissionstatus table indicates whether transmission on time slotak of the rescheduling superframe will be interfered ornot. Fig. 8 shows BSN i establishes its transmission statustable in the presence of an interferer BSN j. In Fig. 8, the
Reschedule
3t
Reschedule
4t
edule
4't
2 1listenT t t
3 2waitT t t
4 3'competeT t t
4 4'backoffT t t
The rescheduling delaycomponents for BSN4:
e IIM scheme.
ted scheme for mitigating inter-user interference in body sensor13.09.013
t
t
Listening superframe(BI)
Rescheduling superframe(BI)
1jt 1
j jt SD jt j jt SD
BSN i:
BSN j:
0t 0t BI
: Beacon : Data transmission: Receive beacon from BSN j
0t : Beginning of the reschedule period
1jt : Reception time of beacon from
BSN j
: Next transmission time of BSN j
itempt
: Collision-free : CollisionIn the transmission status table:
itempt : The possible rescheduled
transmission time
0t BI
jt
Fig. 8. The procedure of establishing the transmission status table of BSN i in the presence of an interferer BSN j.
6 W. Sun et al. / Computer Networks xxx (2013) xxx–xxx
listening superframe of BSN i is from (t0 � BI) to t0, and therescheduling superframe is from t0 to (t0 + BI). During thelistening superframe, BSN i receives BSN j’s beacon packetat tj
�1. From the information obtained from the beacon(SDj,BIj,cj, tj), it is known that the current transmission ofBSN j lasts for a period of SDj, and the next transmissionof BSN j will start at tj. The transmission status table ofBSN i is established in the rescheduling superframe ofBSN i by marking the time slots that has overlapping trans-mission with BSN j as collision.
When multiple BSNs congregate together, within theinterference range of one another, their transmissionstatus tables are established similarly, except that all theBSNs in the neighbor set {Xi: (SDj,BIj,cj, tj)j1 6 j 6 N, j – i}are considered. In general, two principlesare followed:
(1) If the current transmission of BSN j falls in therescheduling superframe of BSN i, i.e.,
tj�1 6 t0 6 tj
�1þSDj, mark the corresponding time
slots ak 2 t0; tj�1 þ SDj
� �as collision.
(2) If the next transmission of BSN j is in the reschedul-ing superframe of BSN i, i.e., t0 6 tj
6 t0 + BI, mark thecorresponding time slots ak 2 (tj, min[t0 + BI, tj + SDj])as collision.
The algorithm is implemented over all the members inthe neighbor list Xi. The other time slots in the reschedul-ing superframe are set to be collision-free in the transmis-sion status table.
From the transmission status table, the next possibletransmission time ti
temp of BSN i is determined as thetime slot whose subsequent collision-free time slots aresufficiently long for BSN i’s transmission. We record allthe possible transmission time slots, and sort them bytheir corresponding rescheduling delays TdelayðqÞ ¼ti
tempðqÞ � t0. It is always preferred to transmit at the ear-liest possible transmission time (q = 1). For example, inFig. 8, BSN i selects ti
temp as its earliest possible transmis-sion time to avoid collision with BSN j. However, if thetransmission fails at the first trial (q = 1) due to anincomplete neighbor list (see Section 4.3), the transmis-sion is attempted at the next possible transmission timeti
tempðqþ 1Þ. The procedure of the rescheduling algorithmis shown in Algorithm 1.
Please cite this article in press as: W. Sun et al., A lightweight distribunetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
Algorithm 1. Rescheduling algorithm
Input: Neighbor list of BSN i {Xi:(SDj,BIj,cj, tj)j1 6 j 6 N, j – i}.
Output: Collision-free transmission opportunity titemp.
1: At the end of the listening period, BSN i performsthe following:
2: (a1, . . . , an) collision - free;3: for BSN j in the neighbor list Xi do4: if there is a transmission going on,
tj�1 6 t0 6 tj
�1 þ SDj then
5: ak 2 t0; tj�1 þ SDj
� � collision;
6: end if7: if the next transmission of BSN j is in the
reschedule superframe of BSN i, i.e., t0 6 tj6 t0 + BI
then8: ak 2 (tj, min[t0 + BI, tj + SDj]) collision;9: end if
10: end for11: for ak in the rescheduling superframe do12: if there is a collision-free time duration starting
with ak for the transmission of BSN i then13: ti
tempðqÞ ak;14: q q + 1;15: k k + 1;16: end if17: end for18: if collision-free duration is not found, q = 0 then19: Switch channel;20: end if
4.3. Collision avoidance in case of incomplete neighboringinformation
In Step 4 of the workflow, to avoid the collision causedby an incomplete neighbor list, carrier sensing and backoffare performed before the rescheduled beacon transmis-sion. As the neighbor list is obtained from the received bea-cons, the derived neighbor list may be incomplete due tocollision of beacons or hidden terminals. When a BSN withan incomplete neighbor list transmits according to ti
temp, itmay collide with the undetected BSNs, whose information
ted scheme for mitigating inter-user interference in body sensor13.09.013
Earliest possibletransmission time
Holding time Backoff
Rescheduled starttime
Listening period
t
: Beacon : Data transmission
Fig. 9. Carrier sensing and backoff before the rescheduled beacontransmission.
W. Sun et al. / Computer Networks xxx (2013) xxx–xxx 7
is not updated in the neighbor list as expected. For in-stance, in Fig. 7, BSN 3 decides its ti
temp as t3 and postponesits transmission until t3. Because of the deferment, BSN 4,which is initiated after it, misses BSN 3s beacon and alsodecides on t3. As a result, both BSN 3 and BSN 4 wouldtransmit at t3 causing packet collisions.
To avoid collisions caused by an incomplete neighborlist, before ti
temp, the coordinator of BSN i senses the channelto check if there is an on-going transmission of the unde-tected BSNs. If the channel is idle for a certain period,called holding time, the coordinator transmits after a back-off period as planned. Otherwise, if the channel is sensedbusy (either immediately or during the holding time), thecoordinator monitors the channel at the next possibletransmission time until it is sensed to be idle for one hold-ing time, and then transmits after a backoff period. Theholding time is configured to be a two-slot duration.Fig. 9 shows the case when the channel is idle and the af-fected BSN transmits after a holding time and backoff time.Therefore, the final rescheduled transmission time is thetime that BSN i successfully accesses the channel afterchannel sensing and backoff, while ti
temp obtained in therescheduling algorithm is just the time that BSN i triesto access the channel and starts channel sensing.
Because most collisions have been alleviated by therescheduling algorithm (Section 4.2), the exponential back-off mechanism is not necessary and a short fixed backoffwindow is sufficient to avoid collisions.
In the presence of interference from other networks(e.g. WLAN), IIM explores the utilization of the currentchannel and schedules transmissions adaptively. As anexample, Fig. 10 depicts the transmission scheme of BSNi in the presence of WLAN. As shown in Fig. 10, BSN ireschedules its transmission through carrier sensing afterthe WLAN transmission to avoid collisions, although it isunable to collect the transmission status of the WLAN dur-ing the listening period in the absence of beacon packets.In the case of severe interference from other networks,
BackoffListening period
WLANtransmission:
BSN i:
Rescheduled time
Carrier sense
: Beacon : Data transmission : WLAN transmission
Fig. 10. Example of IIM scheme in the presence of WLAN.
Please cite this article in press as: W. Sun et al., A lightweight distribunetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
IIM detects significant performance degradation whileidentifies very few neighboring BSNs and cannot obtainthe transmission status of other networks during listeningperiod. In such a case, the channel utilization is high so thatrescheduling transmission in the same channel does notmake much sense, and only channel switching can improvethe system performance.
5. Performance analysis and discussion
In this section, we first calculate the maximum numberof BSNs that can be supported by IIM Nmax. Then we ana-lyze the rescheduling delay Tdelay, throughput, and conver-gence of IIM.
If the number of neighboring BSNs is below Nmax(cl) onchannel cl, BSN i can find a suitable schedule. Otherwise,switching channel is desirable.
Assume all the BSNs have the same superframe length(BI) and active period length (SD). When BSNs transmitexactly after each other, the upper bound of Nmax(cl) onchannel cl is achieved as
NUBmaxðclÞ ¼
BISD
� �; ð2Þ
where b�cis the integer floor function. This equation indi-cates that NUB
maxðclÞ is the multiplicative inverse of the dutycycle (SD/BI). For example, if the duty cycle of a BSN is 1/5,the theoretical maximum BSN number that can be sup-ported in a channel is 5.
However, Nmax(cl) is reduced when BSNs do not trans-mit exactly after one another. With carrier sensing andbackoff, the channel utilizing period of a BSN is slightlylonger than the actual transmission time. The worst casehappens when the interval between two transmissions isslightly less than the transmission period of a BSN. In thiscase, the lower bound of Nmax(cl) is expressed as
NLBmaxðclÞ ¼
BI2SDþW
� �; ð3Þ
where W is the fixed backoff window size and (SD + 0.5 W)is the transmission period of a BSN. In general, the Nmax(cl)is between the upper bound and lower bound. Thus wehave
NmaxðclÞ 2 NLBmaxðclÞ;NUB
maxðclÞh i
: ð4Þ
In practice, Nmax(cl) is decided dynamically according tothe neighbor list (see Section 4.2). In particular, if BSN i isunable to find a possible transmission time ti
temp aftersearching its transmission status table, the number of con-gregated BSNs is deemed to exceed Nmax(cl) on channel cl.In this case, switching channel is desirable, as shown inthe last section of Algorithm 1. When multiple channelsare available, the maximum number of BSNs that can besupported by IIM is
Nmax ¼Xm
cl¼1
NmaxðclÞ; ð5Þ
where m is the channel number. If the number of congre-gated BSNs exceeds Nmax, IIM decides that some of the
ted scheme for mitigating inter-user interference in body sensor13.09.013
Interferenceoccurence
IIM initiation
Interferencemitigation
Regulartransmission
IIMExecution
Interferedtransmission
Fig. 11. System states of the IIM scheme.
2 The throughput of the interfered transmission state is negligible due toits short duration.
8 W. Sun et al. / Computer Networks xxx (2013) xxx–xxx
BSNs have to change channels for effective transmissions.Alternatively, each BSN has to reduce its duty cycle to sup-port more BSNs operating in the common space.
The rescheduling delay Tdelay of IIM is defined from theoriginal transmission time to the rescheduled transmissiontime. It has four components:
� Listening delay (Tlisten): it is introduced by channel lis-tening. While the length of the listening period is asuperframe length (BI), the length of Tlisten is only anactive period (SD) because the other part of the listeningperiod falls in the inactive period.� Waiting delay (Twait): it is introduced to avoid overlap-
ping transmissions with the neighboring BSNs. Its valueis determined by the specific interference situation ofBSN i.� Competition delay (Tcompete): it happens when BSN i fails
to transmit at the first trial titempðqÞ and waits until the
next transmission opportunity titempðqþ 1Þ to access
the channel.� Backoff delay (Tbackoff): this delay is for the carrier sens-
ing and backoff before the rescheduled beacontransmission.
Hence Tdelay is expressed as
Tdelay ¼ Tlisten þ Twait þ Tcompete þ Tbackoff : ð6Þ
In Eq. (6), when Tdelay is less than (SD + BI), the BSN is ableto find a suitable schedule on the current channel. (SD + BI)is the summation of the listening delay and the reschedul-ing superframe length. The four components of reschedul-ing delay Tdelay are indicated in Fig. 7.
Assume a BSN scans n channels before choosing a suit-able schedule, Tdelay is expressed as
Tdelay ¼ ðn� 1ÞBI þ Tlisten þ Twait þ Tcompete þ Tbackoff :
The first term of the right-hand side of the above equationrepresents the time for scanning (n � 1) channels (Eq. (6) isthe special case when n = 1).
In summary, each BSN has three states, namely theinterfered transmission state, the IIM execution state, andthe regular transmission state. The above three statesmay transit to one another, as shown in Fig. 11. Whenthe interference occurs, the BSN transmission is adverselyaffected. The average length of the interfered transmissionis half of the active period (0.5SD), because the interferersarrive during the active period with equal probability andIIM is always initiated at the end of the active period. Uponinitiation, IIM is executed to find a suitable schedule. The
Please cite this article in press as: W. Sun et al., A lightweight distribunetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
length of the execution period is Tdelay(k), where k indicatesthe kth initiation. After interference mitigation, the regulartransmission is maintained for a length of Treg(k) until thenext interference occurrence.2 Let throughput of the regulartransmission after the kth initiation be u(k), the overallthroughput uis expressed as
u ¼XK
k¼1
TregðkÞTregðkÞ þ TdelayðkÞ þ 0:5 � SD
�uðkÞ; ð7Þ
where K is the total initiation times during the BSN operat-ing duration.
The inter-user interference occurs for two reasons incontention-based rescheduling schemes: (1) a BSN movesinto the interference range of BSN i; (2) an existing neigh-boring BSN reschedules to transmit simultaneously withBSN i. In IIM, interference occurs only due to the first rea-son, because a BSN always reserves time slots with theawareness of the transmissions of others when it getsrescheduled. As a result, K in Eq. (7) is minimized. Whenthe total operating duration is fixed, the interference dura-tion Treg(k) is maximized as K is minimized. Moreover, Tdelay
is chosen as the shortest delay to find a collision-free timeslot. Therefore throughput can be maximized by the pro-posed IIM scheme.
As long as the number of the congregated BSNs is lessthan Nmax, the BSN state will finally converge to the regulartransmission state. The regular transmission state ismaintained until the next occurrence of inter-userinterference.
6. Simulation results
We implemented the proposed IIM scheme in theQualNet 5.0.2 simulator [24]. The performance of IIMis compared with the basic scheme of IEEE 802.15.4 [20]and the flexible beacon scheduling scheme [14] wherebeacons are scheduled using the contention-basedscheme.
6.1. Simulation settings
In the simulations, a coordinator and a sensor nodeform a BSN. The radio settings are configured accordingto the IEEE 802.15.4 standard and the profiles of TelosBmotes [25]. As we focus on the investigation of inter-userinterference, we consider an inter-body path loss modelwith a path loss exponent of 2.4 and a shadowing standarddeviation of 6.2 dB [26]. We choose a radio data rate of250 kbps and a superframe length (BI) of 0.1 s. For health-care applications, the data rate requirements of commonlyused sensor nodes are 5 kbps for Electrocardiograph (ECG)and Electroencephalography (EEG), and 1 kbps for temper-ature sensor, respiratory sensor, and pulse sensor [27].Consider the combined usage of those sensors in a BSN,the traffic load per BSN varies from 5 to 25 kbps in the sim-ulations. For simplicity, the traffic load for all the BSNs areset the same in a specific scenario.
ted scheme for mitigating inter-user interference in body sensor13.09.013
Table 2The parameter settings of the simulation.
Transmission power (dBm) �10Mobile speed interval (m/s) [0.2, 2.2]BSN walk interval (s) [2,6]BSN pause interval (s) [0, 6]Direction interval (degree) [�180,180]Simulation duration (s) 1000
5 10 15 20 250
2
4
6
8
10
Traffic Load (kbps)
Rat
io o
f un
dete
cted
BSN
s (%
)
# of BSNs=2# of BSNs=4# of BSNs=6# of BSNs=8# of BSNs=10
Fig. 12. The ratio of the undetected BSNs to the actual neighboring BSNs.
5 10 15 20 250
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Traffic Load (kbps)
Res
ched
ulin
g D
elay
(s)
5 BSNs single channel10 BSNs single channel20 BSNs double channel
Fig. 13. The rescheduling delay of IIM for three scenarios.
5 10 15 20 250
2
4
6
8
10
Traffic Load (kbps)
Rat
io o
f de
lay
per
supe
rfra
me
(%)
Basic schemeBeacon scheduleIIM
Fig. 14. The ratio of average delay per superframe to the superframelength.
W. Sun et al. / Computer Networks xxx (2013) xxx–xxx 9
We consider the typical BSN deployment in a hospitalscenario,3 where the average area occupied by each patientis from 5 to 10 square meters. Considering the general casewhere only partial patients utilize BSNs, we start with 5 BSNwearers moving randomly and freely within the space30 m � 30 m. Later the results are compared with othertwo scenarios with 10 BSNs and 20 BSNs, respectively. EachBSN moves according to the random waypoint model[28,29]. The pedestrian characteristics of each BSN wearerare listed in Table 2 according to [22]. Initially, all BSNsare uniformly deployed and then they move independently.To remove the effect of differing initial conditions on perfor-mance, we run the simulation fifty times with different ini-tial conditions and then calculate the average results.
6.2. Simulation results
In this section, we investigate the performance of theIIM scheme in terms of ratio of the undetected BSNs,rescheduling delay, collision probability, throughput, andenergy consumption.
6.2.1. Ratio of the undetected BSNsFig. 12 shows the ratio of the undetected BSNs to the ac-
tual neighboring BSNs over various traffic loads. Each curverepresents a scenario with a certain number of BSNs. It canbe seen that the undetected ratio increases when either thetraffic load of each BSN or the number of neighboring BSNsincreases. The reason is that when interference becomesmore severe, more beacons are lost. It is noticed that even
3 The configuration settings of Changi General Hospital and Tan TockSeng Hospital in Singapore.
Please cite this article in press as: W. Sun et al., A lightweight distribunetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
in the worst case (10 BSNs with the traffic load of 25 kbpseach), the neighbor list contains more than 91% of theinformation of all neighboring BSNs. Hence, the neighborlist can provide sufficient information for rescheduling.
6.2.2. Rescheduling delayFig. 13 displays the average rescheduling delay Tdelay in
three scenarios: (1) 5 BSNs in a single channel scenario, (2)10 BSNs in a single channel scenario, (3) 20 BSNs in a dou-ble channel scenario. The result converges well with theanalysis in Section 5. In particular, Tdelay in the single chan-nel scenario is less than 0.11 s, which is (SD + BI) in Eq. (6).In the double channel scenario, BSNs suffer longer delays(less than (2BI + SD)). It is also noticed that Tdelay increaseswhen the traffic load increases, as it is more challenging fora BSN to find rescheduled time in the channel with a higherload. However, even the highest delay of 0.134 s in thethird scenario is tolerable to most applications [30].
Fig. 14 shows the ratio of the average delay per super-frame to the superframe length in the scenario of 10 BSNsoperating in a single channel. For IIM, it is obtained byaveraging Tdelay into superframes over the relatively longinterference duration. The result of IIM is compared withthat of the basic scheme and beacon schedule scheme. Asshown in Fig. 14, the delay of IIM is quite low (lower than2% of superframe length), because rescheduling is
ted scheme for mitigating inter-user interference in body sensor13.09.013
5 10 15 20 250
0.1
0.2
0.3
0.4
Traffic Load (kbps)
Col
lisio
n Pr
obab
ility
of
Bea
cons Basic scheme
Beacon scheduleIIM
Fig. 15. The collision probability of beacons.
5 10 15 20 250
5
10
15
20
25
Traffic Load (kbps)
Thr
ough
put (
kbps
)
Basic schemeBeacon scheduleIIM
(a) 5 BSNs in single channel scenario.
5 10 15 20 250
5
10
15
20
Traffic Load (kbps)
Thr
ough
put (
kbps
)
Basic schemeBeacon scheduleIIM
(b)10 BSNs in single channel scenario.
5 10 15 20 250
5
10
15
20
25
Traffic Load (kbps)
Thr
ough
put (
kbps
)
Basic schemeBeacon scheduleIIM
(c) 20 BSNs in double channel scenario.
Fig. 16. Throughput versus the traffic load for three scenarios.
10 W. Sun et al. / Computer Networks xxx (2013) xxx–xxx
performed only when the interference situation changes.For the beacon schedule scheme, the delay is longerbecause multiple carrier sensing iterations are possiblyconducted before each beacon transmission. Regular trans-mission of a BSN can be interrupted by its neighboringBSNs operations, resulting in more rescheduling times.For the basic scheme, the delay is near zero becausepackets are transmitted directly at the cost of much higherenergy consumption and lower throughput. When thenumber of the congregated BSNs increases, the delay ofthe beacon schedule method obviously increases, becausethe BSNs experience long backoff time on the currentfully-occupied channel in the absence of an effective chan-nel switching approach.
6.2.3. Collision probabilityFig. 15 shows the collision probability of beacons in the
scenario of 10 BSNs operating in a single channel. It can beseen that the collision probability of IIM is the lowest,because collisions are effectively eliminated throughrescheduling. For the basic scheme, BSNs transmit inde-pendently according to fixed schedules, resulting in severecollisions. For the beacon schedule method, a BSN has tocompete with all its neighboring BSNs to access the chan-nel, thus collision is more severe than that of IIM. In addi-tion, once collision happens, there will be collisions everysuperframe for the whole interference duration becauseof the periodic data characteristic of BSNs.
6.2.4. ThroughputFig. 16 depicts the throughput of various traffic loads in
three scenarios. As discussed in Section 5, IIM achieves thehighest throughput among the three schemes. The averagethroughput improvement is around 30% compared to thebasic scheme and 18% compared to the beacon schedulemethod with all traffic loads. This is because collisionsare effectively eliminated by IIM with minimum resched-uling delay, shown in Eq. (7). For the basic scheme, thefixed schedule leads to severe collisions. For the beaconschedule scheme, the throughput is lower than IIM, be-cause collisions are more severe and the delay is longer.
When the density of BSNs increases from 5 BSNs to 10BSNs, the performance of the basic scheme decreases sig-nificantly because there is higher possibility that a BSN
Please cite this article in press as: W. Sun et al., A lightweight distribunetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
interferes with one another (see Fig. 16(a) and (b)). Thecomparative performance of IIM and the beacon schedulemethod are still acceptable. When the traffic load of BSNsincreases to a certain level, channel switching should beincorporated to ensure the network throughput.
Fig. 16(c) shows the throughput in the scenario of 20BSNs with two available channels. It can be seen that withchannel switching the throughput of IIM improves evenwhen the average number of BSNs per channel is the same
ted scheme for mitigating inter-user interference in body sensor13.09.013
5 10 15 20 25
0.075
0.08
0.085
0.09
0.095
0.1
Traffic Load (kbps)
Ene
rgy
Con
sum
ptio
n (m
J)
Basic schemeBeacon scheduleIIM
Fig. 17. Energy consumption per successfully delivered packet for the 10BSNs in single channel scenario.
Table 3The power consumptions of TelosB radio.
Idle mode 0.8 mWCCA mode 40 mWReceive mode 40 mWTransmit mode 30 mW
W. Sun et al. / Computer Networks xxx (2013) xxx–xxx 11
as that in Fig. 16(b). The reason is that channels can beeffectively utilized in two dimensions by IIM. For the bea-con schedule scheme, without the effective channelswitching method, the BSNs continuously compete on thecurrent channel resulting in low throughput.
6.2.5. Energy consumptionFig. 17 shows the energy consumption per successfully
delivered packet in the single channel scenario of 10 BSNs.It is calculated from the power consumption of TelosBmote, shown in Table 3. As expected, the energy cost ofdelivering a packet using IIM is much lower than that ofthe other two methods, i.e. around 16% lower than the ba-sic scheme and 22% lower than the beacon schedule meth-od. This is because IIM effectively avoids collision, andhence reduces the number of retransmissions. For the basicscheme, much energy is wasted by collisions and retrans-missions. For the beacon schedule method, much energyis consumed for carrier sensing before each beacon trans-mission. When the traffic load increases, the energy costgets higher. Especially for beacon schedule method, muchenergy is consumed as multiple carrier sensing iterationsare possibly conducted before each beacon transmission.
7. Conclusion
In this paper, we have proposed a lightweight and dis-tributed scheme (IIM) for mitigating inter-user interfer-ence in body sensor networks (BSNs). The proposedscheme considers the generic property of low channel uti-lization of BSNs and enables BSNs to adaptively rescheduletheir transmission time or channel when the interferenceoccurs. Based on the neighboring information, two actionsare conducted: (1) BSNs reschedule their transmissions in
Please cite this article in press as: W. Sun et al., A lightweight distribunetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
a coordinated manner to reduce the rescheduling costwhen the channel utilization is low. (2) When the channelis fully-occupied, channel switching decision can be madepromptly so that wireless channels can be effectively uti-lized. The simulation results showed that the proposedIIM scheme outperforms the basic scheme and the beaconschedule scheme in terms of throughput and energy effi-ciency. The results confirm that when the density of BSNsbecomes higher, the performance improvement of theIIM scheme are more significant.
Although IIM is designed for the beacon-enabled IEEE802.15.4 protocol in this paper, it can also be implementedin other BSN communication schemes as long as a beaconpacket is involved to synchronize the transmissions of thesensor nodes. For future work, the analysis will be ex-tended to other mobility patterns. We will also implementIIM in an actual BSN system and evaluate the performancewith extensive experiments.
Acknowledgement
This research was carried out at the NUS-ZJU Sensor-Enhanced Social Media (SeSaMe) Centre. It is supportedby the Singapore National Research Foundation under itsInternational Research Centre @ Singapore FundingInitiative and administered by the Interactive DigitalMedia Programme Office.
References
[1] G.Z. Yang, Body Sensor Networks, Springer, 2006.[2] A. Pantelopoulos, N.G. Bourbakis, A survey on wearable sensor-based
systems for health monitoring and prognosis, IEEE Transaction onSystem, Man, and Cybernetics 40 (2010) 1–12.
[3] A. Natarajan, M. Motani, B.d. Silva, K. Yap, K. Chua, Investigatingnetwork architectures for body sensor networks, in: Proc. ACMSIGMOBILE, San Juan, Puerto Rico, 2007.
[4] W. Sun, Y. Ge, W.C. Wong, Inter-user interference in body sensornetworks: a case study in moderate-scale deployment in hospitalenvironment, in: Proc. IEEE HealthCom, Beijing, China, 2012.
[5] G. Bianchi, Performance analysis of the IEEE 802.11 distributedcoordination function, IEEE Journal on Selected Areas inCommunications 18 (2000) 535–547.
[6] S.J. Marinkovic, E.M. Popovici, C. Spagnol, S. Faul, W.P. Marname,Energy-efficient low duty cycle MAC protocol for wireless body areanetworks, IEEE Transaction on Information Technology inBiomedicine 13 (2009) 915–925.
[7] M. Cao, W. Ma, Q. Zhang, X. Wang, Analysis of IEEE 802.16 meshmode scheduler performance, IEEE Transaction on WirelessCommunications 6 (2007) 1455–1464.
[8] Y. Ge, C. Tham, P. Kong, Y. Ang, Dynamic end-to-end capacity in IEEE802.16 wireless mesh networks, Computer Networks Journal 54(2010) 2147–2165.
[9] T. Wu, S. Biswas, Minimizing inter-cluster interference by self-reorganizing MAC allocation in sensor networks, Wireless Networks13 (2007) 691–703.
[10] A. Koubaa, A. Cunha, M. Alves, A time division beacon schedulingmechanism for IEEE 802.15.4/Zigbee cluster-tree wireless sensornetworks, in: Proc. Euromicro Conference on Real-Time Systems,Pisa, Italy, 2007.
[11] G.V. Sergio, M. Chen, Mobility support for health monitoring at homeusing wearable sensors, IEEE Transaction on Information Technologyin Biomedicine 15 (2011) 539–549.
[12] B.d. Silva, A. Natarajan, M. Motani, Inter-user interference in bodysensor networks: preliminary investigation and an infrastructure-based solution, in: Proc. International Workshop on Wearable andImplantable Body Sensor Networks, Berkeley, CA, 2009.
[13] G. Wu, J. Ren, F. Xia, L. Yao, Z. Xu, DISG: decentralized inter-userinterference suppression in body sensor networks with non-cooperative game, in: Proc. ATC, Xian, China, 2010.
ted scheme for mitigating inter-user interference in body sensor13.09.013
12 W. Sun et al. / Computer Networks xxx (2013) xxx–xxx
[14] S. Kim, J.W. Kim, D.S. Eom, Flexible beacon scheduling scheme forinterference mitigation in body sensor networks, in: Proc. IEEESECON, Seoul, Korea, 2012.
[15] B. Latre, B. Braem, I. Moerman, C. Blondia, P. Demeester, A survey onwireless body area networks, Wireless Networks 17 (2011) 1–18.
[16] S. Marinkovic, E. Popovici, C. Spagnol, S. Faul, W. Marnane, Energy-efficient low duty cycle MAC protocol for wireless body areanetworks, IEEE Transaction on Information Technology inBiomedicine 13 (2009) 915–925.
[17] O. Omeni, A. Wong, A. Burdett, C. Toumazou, Energy efficientmedium access protocol for wireless medical body area sensornetworks, IEEE Transaction on Biomedical Circuits and Systems 2(2008) 251–259.
[18] T. Rappaport, Wireless Communications: Principles and Practice,Prentice-Hall, NJ, 2001.
[19] M. Karakayali, G. Foschini, R. Valenzuela, Network coordination forspectrally efficient communications in cellular systems, IEEETransaction on Wireless Communications 13 (2006) 56–61.
[20] M. Chen, S. Gonzalez, A. Vasilakos, H. Cao, V.C.M. Leung, Body areanetworks: a survey, ACM/Springer Mobile Networks andApplications 16 (2011) 171–193.
[21] P. Baronti, P. Pillai, V. Chook, S. Chessa, A. Gotta, Y. Hu, Wirelesssensor networks: a survey on the state of the art and the 802.15.4and ZigBee standards, Computer Networks Journal 30 (2007) 1655–1695.
[22] R. Knoblauch, M. Pietrucha, M. Nitzburg, Field studies of pedestrianwalking speed and start-up time, Transportation Research Record:Journal of the Transportation Research Board 1538 (1996) 27–38.
[23] W. Sun, Y. Ge, W.C. Wong, A lightweight inter-user interferencemitigation method in body sensor networks, in: Proc. IEEE WiMob,Barcelona, Spain, 2012.
[24] S.N. Technologies, Simulator, Qualnet Network, 2011. <http://www.qualnet.com>.
[25] C.T. Inc, Data Sheet of TelosB Mote, 2003. <http://www.willow.co.uk>.
[26] A. Molisch, et al., IEEE 802.15.4a Channel Model-Final Report, IEEEP802.15.4: 0662, 2004.
[27] S. Ullah, H. Higgins, B. Braem, B.B. Latre, I. Moerman, Acomprehensive survey of wireless body area networks on PHY,MAC, and network layer solutions, Journal of Medical System 36(2010) 1065–1094.
[28] F. Bai, A. Helmy, A survey of mobility models in wireless adhocnetworks, Wireless Adhoc Networks (2004) 1–30.
[29] X. Li, V. Jilkov, Survey of maneuvering target tracking-part 1:dynamic models, IEEE Transaction on Aerospace and ElectronicSystems 39 (2003) 1333–1364.
[30] M. Patel, J. Wang, Applications, challenges, and prospective inemerging body area networking technologies, IEEE WirelessCommunications 17 (2010) 80–88.
Please cite this article in press as: W. Sun et al., A lightweight distribunetworks, Comput. Netw. (2013), http://dx.doi.org/10.1016/j.comnet.20
Wen Sun received her B.E. degree from Har-bin Institute of Technology, China. Since 2009,she has been working towards the Ph.D.degree in the Department of Electrical andComputer Engineering, National University ofSingapore (NUS). Since 2011, she has been anintern student at Institute for InfocommResearch (I2R), Singapore. Her researchinterests are transmission schemes andresource management in wireless sensornetworks.
Yu Ge is now Scientist in the Institute forInfocomm Research (I2R), A-Star, Singapore.She received her MEng and Ph.D. degrees fromNational University of Singapore and NanyangTechnological University, all in wireless com-munication networks area. She joined I2R in2001 and worked in various research areasincluding VoIP in heterogeneous wirelessnetworks, wireless mesh/ad hoc networks,and wireless sensor networks. She is currentlyleading a research team in the area of wirelessbody sensor networks (WBSNs) for human-
centric sensing. Her current research interests are transmission andsensing technologies in wireless communication networks for end-to-endhuman-centric service provisioning.
Wai-Choong Wong received the B.Sc. (firstclass honors) and Ph.D. degrees in electronicand electrical engineering from Loughbor-ough University, Loughborough, UK. He is aProfessor in the Department of Electrical andComputer Engineering, National University ofSingapore (NUS). He is currently DeputyDirector (Strategic Development) at theInteractive and Digital Media Institute (IDMI)in NUS. He was previously Executive Directorof the Institute for Infocomm Research (I2R)from November 2002 to November 2006. His
research interests include wireless networks and systems, multimedianetworks, and source-matched transmission techniques.
ted scheme for mitigating inter-user interference in body sensor13.09.013