[ieee icc 2012 - 2012 ieee international conference on communications - ottawa, on, canada...

MAC Protocols for Wireless Sensor Networks overRadio-over-Fiber Links

Tiago P. C. de Andrade, Nelson L. S. da FonsecaUnicamp, Brazil

[email protected], [email protected]

Leonardo B. OliveiraUFMG, Brazil

[email protected]

Omar C. BranquinhoPUCC, Brazil

[email protected]

Abstract—In this paper, two Medium Access Control (MAC)protocols exclusively tailored to WSNs over RoF (RWSNs)namely SPP-MAC (Scheduled Priority Polling Medium AccessControl) (polling-based) and HMARS (Hybrid Medium AccessControl for Hybrid Radio-over-Fiber Wireless Sensor NetworkArchitecture) (hybrid-based) are proposed. They deal with themain problems in RWSNs i.e. the delay imposed by optical fiberand the existence of two collision domains: the wireless and thefiber optical links. The performance of these two protocols evincetheir effectiveness for the connection of WSNs by RoF links.

Index Terms—MAC Protocols; Radio-over-Fiber (RoF); Wire-less Sensor Networks (WSNs); Hybrid Systems; New Architec-ture.

I. INTRODUCTION

Wireless Sensor Networks (WSNs) are ad-hoc networks

comprised mainly of small sensor nodes (SNs) with limited

resources and one or more base stations (BSs), which connect

the sensor nodes to the rest of the world [1][2]. They are used

for monitoring environments and provide fine grained sensing

to users. Application areas range from battlefield reconnais-

sance and emergency rescue operations to surveillance and

environmental protection. As any other technology, however,

they are not a panacea and suffer from problems such as rapid

attenuation of signal strength [3].

Radio-over-Fiber (RoF), has been employed in network

infrastructures due to its large capacity and low signal attenu-

ation. For instance, RoF has been employed for monitoring

cities [4], mines [5], pipe of gas and oil. In this type of

infrastructure, there are dark fibers that can be used for the

transmission of radio signals [6].

In RoF, radio signals are transmitted on optical links by

Remote Antenna Unit (RAU) while more complex signal

processing is carried out at the Base Station Controller (BSC).

In this way, operational costs are reduced and coverage area

enlarged. In addition, it also implies on access with fine cap-

illarity. Moreover, the large amount of underutilized fibers in

the world makes the RoF technology of paramount importance

for new uses of underutilized fibers. Such idleness is caused

mainly by the wide spread of this type of infrastructure in the

past few years and by the advent of the Wavelength-Division

Multiplexing (WDM) [7] that has optimized the use of the

capacity of the optical fiber.

The use of RoF links for connecting WSNs implies on

low delay when compared with ad-hoc and mesh WSNs and

on high reliability when compared with traditional non-RoF

connectivity [4]. Moreover, the employment of RoF links to

connect WSNs reduces significantly the cost of deploying

WSNs in areas with underutilized fibers since there is no need

to install several base stations. However, the delay introduced

by the fibers and the existence of two collision domains (the

wireless and the optical one) makes such integration specially

challenging.

In this work, we propose two MAC protocols exclusively

tailored to WSNs over RoF (RWSNs) namely SPP-MAC

(polling-based) and HMARS (hybrid-based). They deal grace-

fully with two collisions domains and achieve high perfor-

mance by decreasing collisions, specially the HMARS pro-

tocol. It is also shown that pure collision based protocols

degrades considerably the performance of RWSN.

The remainder of this work is organized as follows. In Sec-

tion II, we discuss related work. In Section III, we present the

architecture concept. In Section IV and Section V, we present

the SPP-MAC and the HMARS MAC protocols respectively.

In Section VI we describe simulation setup and results. Finally,

in Section VII we draw conclusions.

II. RELATED WORK

There is a significant number of studies on MAC protocols

for conventional WSNs (e.g. [8], [9], [10], [11], [12], [13],

[14], [15], [16]). Ye et al., proposed the Sensor MAC (S-MAC)

protocol [8] that provides a tunable periodic active/sleep cycle

for energy conservation. During sleeping periods, nodes turn

off the radio to save energy and during active periods, nodes

turn on the radio to Tx/Rx frames. Active periods are of fixed

duration whereas the duration of sleep periods depends on a

IEEE ICC 2012 - Ad-hoc and Sensor Networking Symposium

978-1-4577-2053-6/12/$31.00 ©2012 IEEE 254

predefined duty-cycle parameter. Besides, neighboring nodes

can form virtual clusters to set up a common sleep schedule.

Vamdam et al. presented Timeout MAC (T-MAC) [9] that

follows up on the basic idea introduced by S-MAC. The

novelty of their solution is an adaptive duty-cycle in which

the duration of active periods is no longer fixed but varies

according to the traffic. The key idea is to make a node predict

the channel activity during an active period so that it can

switch its radio off before the active period ends, in case it

does not expect any traffic.

Polastre et al. came up with Berkeley MAC (B-MAC) [11],

which uses a technique based on outliers detection to improve

the quality of Clear Channel Assessment (CCA) comparing

with IEEE 802.15.4 [14] standard. In this technique, a node

searches for outliers in the received signal such that the

channel energy is significantly below the noise floor. If the

node detects outliers during channel sampling, then it declares

the channel is clear. If the node does not find any outliers in

fives samples, then it declares the channel to be busy.

The Framelet MAC (F-MAC) [15] essentially tries to reduce

collisions and interference. It decomposes a packet into several

framelets and sends them at a given frequency. As each

potentially interfering node operates at a different frequency,

the probability of collisions is reduced.

The Sparse Topology and Energy Management (STEM) [16]

protocol uses two channels: a wakeup channel and a data

channel. The wakeup channel is used to organize a meeting

between the transmitter and the receiver to avoid deafness,

whereas the data channel is used only for data exchange

once the meeting occurs. To ensure a meeting between the

transmitter and the receiver, nodes follow a preamble sampling

approach.

Among the works that consider RoF as a backhaul for

radio applications, Gomes et al. [17] present an analysis of

the use of RoF technology in IEEE 802.16 networks. The

propagation delay introduced by the fiber length impacts the

tuning of MAC and physical layers parameters. For an effec-

tive tuning, this work presents a comprehensive study of the

performance degradation of WiMAX networks employing RoF

infrastructure indicating the feasibility of RoF scenarios with

degradation bounded to 20% at the physical layer when using

fiber links with maximum length of 115km and degradation

bounded to 20% at the application layer for fiber length of up

to 80km.

Tang et al. [5] proposed a hybrid architecture that is

adequate to monitor temperature, humidity, gas and to locate

employees in environment such as mine. In this scenario,

the radio technology is not feasible, considering the high

attenuation of radio signals. Hossen et al. [4] pointed out

the advantages of RWSNs for metropolitan areas considering

radius of around 10 km to create a smart environment for

monitoring. The application, without the RoF as a backhaul,

would need several ad-hoc network hops to transmit from

distant nodes to the central station.

The work in [?] proposed solutions for the increase of

fiber extension based on the manipulation of the RTS/CTS

timeout. Simulation results show that fiber up to almost 8 km is

feasible, but leads to degradation of 15%. In [18], an extensive

analysis has been carried out using real testbed experiments,

simulations and analytical approximations. Results showed

that the insertion of fibers decreases the throughput less than

15%. However, fiber delays that exceed the defined timeouts

of acknowledgment and RTS/CTS mechanisms cause commu-

nication fails when using fibers longer than 8.1 km.

III. HYBRID RADIO-OVER-FIBER WIRELESS SENSOR

NETWORK ARCHITECTURE

A. Overview of RoF Systems

Optical fibers can provide infrastructures capable of trans-

mitting the whole usable RF signals and, therefore, it is the

preferable medium for the distribution of wide-area wireless

network data. When implementing an optical infrastructure

for wireless network all signal processing functions can be

performed in a centralized way. In this scenario, the antenna

system can cover several kilometers since fiber links have very

low losses (typically 0.2 dB/km) and simple and non expensive

electrical/optical (E/O) converters are available in the market.

In RoF systems, the optical link is an analogue transmission

medium that does not modify the nature of radio signal format,

the fiber links deliver the signals at the RAUs, which act as

transparent O/E converters.

B. Proposed Architecture

��

��

��

��

��

��

��

��

��

��

��

Figure 1. Proposed Architecture

The proposed architecture considers the use of an optical

fiber pair to connect WSNs, using RoF to enlarge the coverage

areas, building a large sensor network.

All sensor nodes within the coverage area of the RAU form

a cluster, with the RAU as the Cluster Head (CH). Clusters

are distributed along an optical fiber link and the SNs in

each cluster transmit only to the BSC located in Data Center

(DC). Thus, all clusters are in the same collision domain. In

this architecture, the MAC protocol needs to deal with the

challenge of contention in both wireless and optical channels.

IV. PROTOCOLO SPP-MAC

The SPP-MAC combines polling and prioritization tech-

niques to allocate the right amount of slots to each sensor

node.

To save energy, SSP-MAC enables transceivers only when

nodes need to transmit frames. Another source of concern is

overhearing. SPP-MAC performs overhearing avoidance from

MAC headers alone. In the protocol, a receiver examines the

destination address of a frame as soon as it receives the MAC

255

header – even before completely receiving the frame. If it

is a unicast frame addressed to any other node, the receiver

immediately ceases the reception of the frame.

A. MAC Frame Format

Figure 2 shows the three types of SPP-MAC’s frames,

namely the Poll Frame, the Data Frame, and the ACK Frame.

• Frame Control field which carries the information about

the frame type and other control flags;

• Target Cluster and Target Node fields which stores the

target cluster and node IDs, respectively. These two fields

compose the Destination Address;

• Source Cluster and Source Node fields which carries

the source cluster and node IDs, respectively. These two

fields compose the Source Address;

• Sequence Number field specifies the sequence identifier

for the frame.

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

� ��

��

��

��

��

��

��

��

��

��

��

��

��

Figure 2. The SPP-MAC’s frames format

B. Priority Assignment

In WSNs, nodes typically have different roles and thus

different levels of priority. As a result, a node might have

more or less opportunities to transmit than others. In other

words, nodes that play critical roles in the network must have

higher probability to transmit. To address this, SPP-MAC uses

a priority policy based on the role of the network nodes. Note

that these priorities are not related to QoS but actually with

the probability of transmission.

The BSC keeps up to date a list with nodes priorities. Based

on this list, the BSC then builds the polling list. Only nodes

in the list are eligible to receive polls.

C. The Polling Scheduler

The polling algorithm provides one more polling slot to

sensor nodes with priority i than those to sensor nodes with

priority i+ 1.

Slots are organized into cycles. For each cycle, nodes with

priority i (1 ≤ i ≤ m) will get m − i + 1 poll slots. This

means that nodes with priority m will get exactly one polling

slot during the cycle. Therefore, the total number of polling

slots in each cycle is

m∑

i=1

ni ∗ (m− i+ 1) (1)

The next step is to sort the poll slots. One way would

be to scan all nodes and assign each station m − i + 1consecutive poll slots. The strategy does not perform well

since nodes would have to wait a considerable amount of time

to retransmit. Instead, we divide a cycle into m rounds and

for each round i all sensor nodes with priorities from 1 to iwill be polled. The round counter is initially set to m and it

is monotonically decreased after a transmission round until it

reaches 0. Whenever this happens, i.e. the round counter is set

to 0, ending one transmission and the scheduling process.

Algorithm 1 Poll Scheduling Algorithm

Input: mOutput: list L of scheduling

round = 1while round ≤ m do

for each priority ∈ (1,m− round + 1) doadd in L all nodes with priority priority

end forround = round + 1

end while

D. Data Transference

Whenever a sensor needs to transmit data to the BSC, it

enables its transceivers and waits for a polling frame. As soon

as the frame is received, the node transmits. On the arrival

of a frame, the BSC checks if the sender has asked for an

acknowledgment and, if so, it sends back an acknowledgement

to the node. Otherwise, it directly sends a polling frame to the

next node on the L list.

V. HMARS PROTOCOL

By combining the reservation and the contention techniques,

the HMARS Protocol uses Time Division Multiple Access

(TDMA) to avoid collisions in the optical channel and Carrier

Sense Medium Access (CSMA) to avoid collisions in the

wireless channel.

This approach requires knowledge of the network topology

and network synchronization to establish a schedule that

allows each cluster to access the optical channel and commu-

nicate with the BSC. In TDMA, time is divided into frames

and each frame is divided into slots. During a frame, each

cluster is assigned a unique slot during which it has the right to

transmit. As a consequence, transmissions of different clusters

do not collide. In CSMA, a node having backlogged frames

first senses the wireless channel before actually transmitting

the frames. In case the node finds the wireless channel busy,

it postpones its transmission to avoid interfering with ongoing

transmission. In case the node finds the wireless channel idle,

256

it transmits the frame (after possibly having waited a random

time).

HMARS suppresses the RTS/CTS exchange and employs

random backoff to reduce overhead and collisions. It also

suppresses acknowledgment frames.

A. MAC Frame Format

Figure 3 shows the two types of HMARS’s frames, the Data

Frame and the Schedule Frame.

• Frame Control field which carries the information about

the frame type and other control flags;

• Target Cluster and Target Node fields which stores the

target cluster and node IDs, respectively. These two fields

compose the Destination Address;

• Source Cluster and Source Node fields which carries

the source cluster and node IDs, respectively. These two

fields compose the Source Address;

• Sequence Number field specifies the sequence identifier

for the frame.

In the Schedule Frame, the Start field contains the time

when the first active period will begin, the Active field contains

the duration (in seconds) of the active period, the Sleep field

contains the duration (in seconds) of the inactive period and

the Schedule field contains the time when the next scheduling

will occur.

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

Figure 3. The HMARS’s Frames format

B. CSMA access to the Wireless Channel

Three variables are maintained: CW, NA, and MAXAttempt.CW is the contention window size, which defines the number

of backoff periods without channel activity before transmission

can start; its value is initialized to 2 before each transmission

attempt and reset to 2 each time the channel is busy. NA is

the number of attempts of transmission; its value is initialized

to 2 before each transmission. MAXAttempt is the maximum

number of transmission attempts.

The MAC sublayer delays transmission for a random num-

ber of complete backoff periods in the range 0 to 2 (Figure 4,

Step 2) and then requests the physical layer to perform a Clear

Channel Assessment (CCA) (Figure 4, Step 3).

If the channel is busy, NA is incremented by one and CWis set to two (Figure 4, Step 5). If NA is less than or equal

to MAXAttempt, a new transmission is attempted after 0 to 2

backoff periods later (Figure 4, Step 2). If NA is greater than

MAXAttempt, the MAC sublayer reports a transmission error

to the layer above.

If the channel is idle, the MAC protocol makes sure the

contention windows has expired. To do that, CW is decre-

mented monotonically (Figure 4, Step 4) and, subsequently, it

is checked whether its value is zero. If so, the MAC sublayer

immediately requires a new CCA to the physical layer. If this

value is null, transmission is restarted. By doing that, we limit

the number of simultaneous transmissions and decrease the

number of intra-cluster collisions.

��

��

��

��

��

��

��

��!��

"!##�$$

%

�

�

�

%

%

&��'� ��()*�!��+�#,� ��$

(�*

(�*

()*

(-*

(.*

(/*

(0*

Figure 4. The HMARS’s CSMA algorithm

C. TDMA access to the Optical Channel

We assume that the BSC knows the network topology and

the delays to each cluster, which are realistic assumptions.

As the optical channel is the common collision domain to

all nodes in the clusters, the TDMA protocol tries to avoid

collisions among them.

Time is divided into frames and each frame is divided into

slots. The number of slots is the number of clusters in the

network and each slot has the same duration. The number of

frames depends on the synchronization interval.

During a frame, each cluster is assigned a unique slot

according to its location along the optical fiber, during which

257

the clusters has the right to transmit. As a consequence, colli-

sions among clusters are avoided, which guarantees finite and

predictable scheduling delays and also increases the overall

network throughput under heavy loaded networks.

Scheduling is the crucial part of the protocol. As it requires

great computational power, it is carried out by the BSC.

VI. SIMULATION AND PERFORMANCE EVALUATION

A. Simulation Setup

The NS-2 Simulator with the CMU Wireless extension was

used in the simulations. An RoF module was implemented as

well as the SPP-MAC and HMARS protocols.

Physical radio characteristics of each sensor node and the

BSC (such as antenna gain, transmit power and receiver

sensitivity) are shown in Table I.

In our simulation, each cluster contains 10 sensor nodes

distributed along the optical fiber link separated by 5 km from

one another. Each node in the cluster is 10 meters away from

the RAU and sends their data only to the BSC. Impairments

caused by RoFs has not been implemented in the simulation

and only the attenuation of RF in the optical fiber link has

been considered, which was set to 0.4 dBm/km.

Each sensor generates one Poisson flow to the BSC; the

packets size was 18 bytes.

Finally, all simulations were run independently using 5

different seeds.

TABLE IFIXED MODEL PARAMETERS

PHY Module ParametersData Rate 250kbpsFrequency 915 MHzTransmitter Power 10 dBmCarrier Sense Sensitivity -95 dBm

Propagation ModelModel ShadowingAttenuation Factor (β) 3.41Standard Deviation (σ) 5Reference Distance (d0) 1

Others Module ParametersTransmit Power 90 x 10−3WReceive Power 45 x 10−3WIdle Power 15 x 10−3WSleep Power 20 x 10−6W

B. Performance Metrics

Three metrics to evaluate the performance of the protocols

were assessed, (i) Delivery Ratio which is the ratio of

successfully delivered packets to the total packets originating

from all sources; (ii) Aggregate Throughput which is the

aggregate throughput of the traffic that goes to the BSC;

and (iii) Successful Poll Rate which is the percentage of

successfully polls over the total number of polls.

If the BSC sends a poll frame to node i and receives

data from i before polling another node j, the poll is then

considered a successful one; otherwise, it is considered a

missed poll.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

2 3 4 5 6 7 8 9 10

Del

iver

y R

atio

Number of Clusters

AlohaCSMA 0.85 persistent

HMARSSPP-MAC

Figure 5. Delivery Ratio

40

60

80

100

120

140

160

2 3 4 5 6 7 8 9 10

Agg

rega

te T

hrou

ghpu

t (kb

ps)

Number of Clusters

25 frames/second50 frames/second


Figure 6. HMARS Aggregate Throughput

C. Simulation Results

Figure 5 compares the Delivery Ratio of SPP-MAC and

HMARS with two widely deployed protocols; the Aloha and

the CSMA p-persistent. Flows with 500 frames per second

were generated to all protocols.

The delivery ratio of the Aloha and CSMA degrades with

the increase of the number of clusters due to collisions.

Conversely, both HMARS and SPP-MAC can achieve 100%

delivery ratio since these protocols decrease collisions intra-

clusters.

Figure 5 shows that nodes using HMARS or SPP-MAC can

obtain almost dedicated bandwidth, achieving delivery ratio

equals to one, regardless of the number of nodes which does

not happen with the contention based protocols.

HMARS has a constant aggregate throughput regardless of

the number of clusters (Figure 6). This is because the protocol

is TDMA-based and each node and cluster issue the same

amount of traffic under high loads.

As the number of clusters increases the distance between

the BSC and the last cluster also increases. As a consequence,

258

45

50

55

60

65

70

75

80

85

90

2 3 4 5 6 7 8 9 10

Agg

rega

te T

hrou

ghpu

t (kb

ps)

Number of Clusters



Figure 7. SPP-MAC Aggregate Throughput

the optical fiber delay gets longer and the duration of timeouts

increases. The larger the delay, the lower is the number of poll

frames transmitted by the BSC and, therefore, there are less

transmissions of data frames, which decreases the aggregate

throughput (Figure 7). With the load increase, the polling

overhead has a lower impact on the throughput. However,

such overhead prevents the SPP-MAC of achieving the same

throughput achieved by HMARS.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

2 3 4 5 6 7 8 9 10

Suc

cess

ful P

oll R

ate

Number of Clusters



Figure 8. SPP-MAC Successful Poll Rate

Figure 8 shows the percentage of successful polls as a

function of the number of clusters. Successful poll rate in-

creases as the number of active nodes increases. Note that

successful poll rate is important since it increases the wireless

medium utilization. A missed poll means waste of bandwidth

and medium resources.

VII. CONCLUSION

In this work, we proposed two MAC protocols exclusively

tailored to RWSNs, namely SPP-MAC and HMARS. They

deal with the main problems of RWSNs i.e. the delay im-

posed by the optical fiber and the existence of two collision

domains. The performance of different MAC protocols were

evaluated. Enhanced delivery ratio and aggregate throughput

were achieved by two proposed protocols. Besides, results

indicate that SPP-MAC and HMARS perform considerably

better than the others MAC protocols as Aloha and CSMA in

the analyzed scenarios.

REFERENCES

[1] D. Estrin, R. Govindan, J. Heidemann, and S. Kumar, “Next centurychallenges: scalable coordination in sensor networks,” in Proceedings ofthe 5th annual ACM/IEEE international conference on Mobile comput-ing and networking, MobiCom ’99, (New York, NY, USA), pp. 263–270,ACM, 1999.

[2] G. J. Pottie and W. J. Kaiser, “Wireless integrated network sensors,”Commun. ACM, vol. 43, pp. 51–58, May 2000.

[3] A. Fanimokun and J. Frolik, “Effects of Natural propagation environ-ments on Wireless Sensor Network coverage area,” in Proceedings ofthe 35th Southeastern Symposium, pp. 16–20, 2003.

[4] M. Hossen, B.-J. Jang, K.-D. Kim, and Y. Park, “Extension of wirelesssensor network by employing rof-based 4g network,” in AdvancedCommunication Technology, 2009. ICACT 2009. 11th InternationalConference on, vol. 01, pp. 275 –278, feb. 2009.

[5] J. Tang, X. Jin, Y. Zhang, X. Zhang, and W. Cai, “A hybrid radio overfiber wireless sensor network architecture,” in Wireless Communications,Networking and Mobile Computing, 2007. WiCom 2007. InternationalConference on, pp. 2675 –2678, sept. 2007.

[6] X. N. Fernando and A. Anpalagan, “On the Design of Optical Fiberbased Wireless Access Systems,” pp. 3550–3555, 2004.

[7] O. Gerstel, B. Li, A. McGuire, G. N. Rouskas, K. Sivalingam, andZ. Zhang, “Special issue on protocols and architectures for next gen-eration Optical WDM Networks,” in IEEE Journal Selected Areas inCommunications, October 2000.

[8] W. Ye, J. Heidemann, and D. Estrin, “An Energy-Efficient MACprotocol for Wireless Sensor Netwoks,” in Annual Koint Conference ofthe Computer and Communicaton Societies (INFOCOM), vol. 3, (NewYork, NY, USA), pp. 1567–1576, November 2002.

[9] T. van Dam and K. Langendoen, “An adaptive energy-efficient macprotocol for wireless sensor networks,” in Proceedings of the 1st in-ternational conference on Embedded networked sensor systems, SenSys’03, (New York, NY, USA), pp. 171–180, ACM, 2003.

[10] G. Lu, B. Krishnamachari, and C. S. Raghavendra, “An AdaptiveEnergy-Efficient and Low-Latency MAC for Data Gathering in SensorNetworks,” in WMAN, (Santa Fe, NM, USA), April 2004.

[11] J. Polastre, J. Hill, and D. Culler, “Versatile low power media accessfor wireless sensor networks,” in Proceedings of the 2nd internationalconference on Embedded networked sensor systems, SenSys ’04, (NewYork, NY, USA), pp. 95–107, ACM, 2004.

[12] P. Lin, C. Qiao, and X. Wang, “Medium Access Control with a DynamicDuty Cycle for Sensor Networks,” in IEEE Wireless Communicationsand Networking Conference (WCNC), vol. 3, (Atlanta, GA, USA),pp. 1534–1536, March 2004.

[13] M. Burttner, E. Yee, G. V. Anderson, and R. Han, “X-MAC: A ShortPreamble MAC Protocol for Duty-Cycled Wireless Sensor Networks,”in 4th International Conference on Embedded Networked Sensor Systems(SenSys), 2006.

[14] “802.15.4-2006 IEEE Standard for Information Technology,” 2006.[15] U. Roedig, A. Barroso, and C. J. Sreenan, “F-MAC: A Deterministic

Media Access Control Protocol without Time Syncronization,” inEuropean Conference on Wireless Sensor Networks (EWSN), (Zurich,Switzerland), pp. 276–291, February 2006.

[16] C. Schurgers, V. Tsiatsis, S. Ganeriwal, and M. Srivastava, “OptimizingSensor Networks in the Energy-Latency-Density Design Space,” inIEEE Trans. Mobile Comput., pp. 70–80, January 2002.

[17] P. Gomes, N. da Fonseca, and O. Branquinho, “Analysis of perfor-mance degradation in Radio-over-Fiber systems based on IEEE 802.16protocol,” in LATINCOM ’09. IEEE Latin-American Conference onCommunications 2009, pp. 1–6, sept. 2009.

[18] A. Das, M. M. amd A. Nkansah, and N. J. Gomes, “Effects on IEEE802.11 MAC Throughput in Wireless LAN Over Fiber Systems,” inJournal of Lightwave Technology, vol. 25, pp. 3321–3328, November2007.

259

[ieee icc 2012 - 2012 ieee international conference on communications - ottawa, on, canada...

Documents