Improving NS-2 Network Simulator to Evaluate

IEEE 802.15.4 Wireless Networks Under Error Conditions

Andr´e Guerreiro, Jeferson L. R. Souza and Jos´e Ruﬁno

University of Lisbon - Faculty of Sciences, Large-Scale Informatics System Lab. (LaSIGE), Lisbon, Portugal

Keywords:

NS-2 Simulator, Wireless Communications, Network Inaccessibility, Timeliness, Dependability.

Abstract:

The behaviour of wireless networks in the presence of error conditions is still being studied by the research

community. Improvements on the evaluation methods and tools are crucial to acquire a better knowledge,

and understanding of the network operation under such conditions. This paper presents enhancements on the

network simulator NS-2 to support the evaluation of the IEEE 802.15.4 standard, used as a case study. We

are specially interested to evaluate the temporal behaviour of the network operation under errors conditions,

considering the applicability of the IEEE 802.15.4 standard in safety-critical environments such as industrial

and vehicular.

1 INTRODUCTION

The applicability of wireless technologies on envi-

ronments with temporal restrictions has been attract-

ing interest of the real-time research community in

the last decade (

Akerberg et al., 2011; Stone et al.,

2012). The main advantages offered by wireless net-

works are: the reduction of size, weight, and power

(SWaP) consumption; the ability to have mobile enti-

ties; and the possibility to establish networking com-

munications where the use of wires is extremely dif-

ﬁcult or impractical (Kandhalu and Rajkumar, 2012).

There are many studies in wireless communica-

tions addressing the temporal behaviour of commu-

nication services at the lowest level of the protocol

stack (Han et al., 2011; Shuai and Zhang, 2010; Hou

and Bergmann, 2010). These studies pay little or no

attention to the dependability aspects of medium ac-

cess control (MAC) sublayer and its services, which

are essential to assure the timeliness and resilience of

the network when operating under error conditions.

This paper evaluate the network simulator NS-2

to identify its limitations, proposing enhancements

to provide a better knowledge and understanding of

wireless network operation under such conditions.

The NS-2 was chosen due to their native support to

simulate wireless networks based on IEEE 802.15.4

standard, which is used as case study.

Our research achievements are organised as fol-

lows: Section 2 describes our system model, which

comprises the assumptions utilised through the pa-

per; Section 3 presents a brief overview of the IEEE

802.15.4 standard; Section 4 addresses the main tem-

poral issues of the network operation under error con-

ditions; Section 5 presents a brief overview of the

NS-2 simulator, including its limitations; Section 6

presents the improvements in the IEEE 802.15.4 NS-

2 module, including a fault injector that complements

the existent NS-2 mechanisms, and a new component

to perform the temporal analysis of the network oper-

ation; Section 7 presents the simulation setup of the

Inaccessibility Scenarios and a simulation script de-

scription; Section 8 presents the results obtained in

the simulation of IEEE 802.15.4 networks, allowing

an enhanced temporal evaluation of such networks;

Finally, Section 9 presents some conclusions and fu-

ture directions of this work.

2 SYSTEM MODEL

Our system model is formed by a set of wireless

nodes

X = {x

,...,x

}, being 1 < n ≤ #A, where

A is the set of all wireless nodes using the same com-

munication channel. The set X itself represents a net-

work entity dubbed wireless network segment (WnS),

as depicted in Figure 1. A WnS establishes a wire-

less network where each node can sense one another

within one-hop of distance, being more complex net-

A wireless node is a networked device capable to commu-

nicate with other nodes

213

Guerreiro A., L. R. Souza J. and Ruﬁno J..

Improving NS-2 Network Simulator to Evaluate IEEE 802.15.4 Wireless Networks Under Error Conditions.

DOI: 10.5220/0004734102130220

In Proceedings of the 3rd International Conference on Sensor Networks (SENSORNETS-2014), pages 213-220

ISBN: 978-989-758-001-7

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: The graphical representation of a wireless net-

work segment.

works composed by more than one WnS. For simpliﬁ-

cation purposes, our analyses assume a network with

one WnS, being its behaviour supported by the fol-

lowing assumptions:

1. The communication range of X, i.e. its broadcast

domain, is given by: B

j=1

(x), ∀x ∈ X,

where B

(x) represents the communication range

of a node x;

2. ∀x ∈ A, x ∈ X ⇐⇒ B

(x)

= B

or, as

a consequence of node mobility, x /∈ X ⇐⇒

(x)

6= B

;

3. ∀x ∈ X can sense the transmissions of one another;

4. ∃x ∈ X which is the coordinator, being unique and

with responsibility to manage the set;

5. A network component (e.g. a node x ∈ X) ei-

ther behaves correctly or crashes upon exceed-

ing a given number of consecutive omissions (the

component’s omission degree, f

) in a time inter-

val of reference

, T

;

6. failure bursts never affect more than f

transmis-

sions in a time interval of reference, T

;

7. omission failures may be inconsistent (i.e., not ob-

served by all recipients).

For a given WnS, assumptions 1, 2, and 3 deﬁne

the physical relationship between nodes, assumption

4 deﬁnes the existence of a coordinator, and assump-

tions 5, 6, and 7 deﬁne how the occurrence of com-

munication errors are modelled and handled within

the WnS. All communications and relations between

For instance, the duration of a given protocol execution.

Note that this assumption is concerned with the total num-

ber of failures of possibly different nodes.

Figure 2: Superframe structure (Standard and Society,

2011) .

nodes are established at MAC level, which are rein-

forced by assumption 3. As a consequence of mobil-

ity, nodes may be driven away of a given WnS (as-

sumption 2). All communication errors within WnS

are transformed into omissions (assumption 5), and in

the context of network components an omission is an

error that destroys a data or control frame.

3 IEEE 802.15.4 OVERVIEW

IEEE 802.15.4 is a standard for wireless sensor and

actuator networks (WSANs), which support two op-

erating modes: beacon-enabled and non beacon-

enabled. In this paper we only address the beacon-

enabled mode, which supports trafﬁc with temporal

restrictions. The network coordinator controls the

network access through the superframe structure de-

picted in Figure 2.

This superframe structure comprises an active pe-

riod, and optionally, an inactive period. In the ac-

tive period there are a contention access period (CAP)

and a contention free period (CFP). CAP is used to

transmit trafﬁc without any temporal guarantee and

in a best effort approach. In CFP nodes can allo-

cate time slots to transmit trafﬁc with temporal restric-

tions (i.e., time division multiple access (TDMA) ap-

proach), where such slots are dubbed guarantee time

slots (GTSs). The inactive period is used for power

saving purposes (when needed).

As depicted in Figure 2, the duration of CAP and

CFP are deﬁned by two parameters: the beacon or-

der (BO); and the superframe order (SO). The value

of BO deﬁnes the superframe duration (i.e., the bea-

con interval, T

), and the value of SO the duration

of the active period (CAP+CFP). The duration of the

beacon interval is T

= T

BSD

, where T

BSD

is the

base value of T

when BO = 0, as deﬁned within

the IEEE 802.15.4 standard. The real length of CFP

depends on the number of GTSs actually allocated.

There is no inactive period when BO = SO, being the

duration of the active period equal to T

SENSORNETS2014-InternationalConferenceonSensorNetworks

214

Table 1: Easy-to-use formulas deﬁning the durations of periods of network inaccessibility .

Scenario Equation

Single Beacon Frame Loss T

ina←sb fl

= T

BSD

. (2

+ 1)

Multiple Beacon Frame Loss T

ina←mb fl

= T

BSD



+ 1



. nrLost

Synchronisation Loss T

ina←nosync

= T

BSD



+ 1



. nrLost

Orphan Node T

ina←orphan

= T

ina←nosync

+ T

MLA

(Orphan) +

nrchannels

∑

j=1



MAC

(Orphan) + nrWait . T

BSD



+ T

MAC

ack

(Realign)

Coordinating Realignment T

ina←realign

= T

MLA

(Realign) + T

MAC

ack

(Realign)

Coordinator Conﬂict Detection T

ina←C

Detection

= T

MAC

ack

Conflict)

Coordinator Conﬂict Resolution T

ina←C

Resolution

= T

MLA

(Conflict) +

nrchannels

∑

j=1



MAC

(Beacon

R)+nrWait.T

BSD



+ T

MLA

(Realign) + T

MAC

(Realign)

Extract Request T

ina←extReq

= T

MAC

ack

(ExtReq) +T

wait

GTS request T

ina←GTS

= T

MAC

ack

(GTS)

Association T

ina←assoc

nrchannels

∑

j=1



MAC

(Beacon

R) + nrWait.T

BSD



+ T

MLA

(Beacon)+ T

ina←extReq

+ T

MLA

(AssocReq) + T

MAC

ack

(AssocReq)

Re-Association T

ina←reAssoc

= T

ina←nosync

+ T

ina←assoc

4 CHARACTERISING IEEE

802.15.4 NETWORK

OPERATION UNDER ERROR

CONDITIONS

The utilisation of IEEE 802.15.4 WSANs is emerg-

ing in environments such as industrial and vehicu-

lar, where some networking communications must re-

spect restrict temporal constrains. Wireless communi-

cations may be affected by different sources of inter-

ferences, such as electromagnetic waves, obstacles on

the communication path, or even by the mobility of

nodes. Communication errors may occur as a conse-

quence of such interferences, disturbing the commu-

nication services and the network operation itself.

The occurrence of such communication errors

may affect two different types of operations, which

are related to transmit data trafﬁc and to control

and maintain the network operation. The literature

presents different works (Wang et al., 2012; Saifullah

et al., 2011), which are only focused on the character-

isation and presence of errors on data transmissions,

disregarding the negative effects of such conditions in

the MAC management operations.

In the contextof MAC managementoperations, an

already known severe consequence of communication

errors is dubbed network inaccessibility (Souza and

Ruﬁno, 2009). A network inaccessibility period is

characterised by the occurrence of ”blackouts“ within

networking communications, where the network re-

mains inaccessible by a temporary period of time.

A research study performed by (Souza and Ruﬁno,

2009) presents formulas to specify the duration of

network inaccessibility for the IEEE 802.15.4 stan-

dard. Those communication ”blackouts“ may have a

huge impact on the timeliness and dependability of

the whole networking system, where a better evalua-

tion may suggest the incompatibility of the guarantees

offered by the communication service, and the tempo-

ral requirements of the target environment.

In Table 1 we present a summary of the worst case

duration (represented by the superscript (

)) for each

network inaccessibility scenario. As an example, we

will brieﬂy explain the characterisation of the most

evident network inaccessibility scenarios, which are

related with the loss of beacon frames. Three different

inaccessibility scenarios may occur if such frames are

not received correctly.

A single beacon frame loss

occurs when only one

beacon is lost. The duration of such scenario is equal

to T

+ T

BSD

, where T

BSD

is utilised as a tempo-

ral compensation to accommodate possible clock de-

viations between network nodes. The loss of mul-

tiple and consecutive beacons characterises the oc-

currence of the multiple beacon frame loss scenario

where a correct beacon is received after the loss of the

previous nrLost beacons. The synchronisation loss

is a special case of the multiple beacon frame loss

scenario where after the loss of nrLost beacons the

next beacon is also lost. The duration of both

multiple beacon frame loss

and synchronisation loss

is a multiple of the single beacon frame loss, which

is nrLost .(T

+ T

BSD

). For simpliﬁcation purposes

we replace the (T

BSD

) by T

, as indicated in sec-

tion 3. The complete network inaccessibility charac-

terisation for the IEEE 802.15.4 is present in (Souza

and Ruﬁno, 2009).

5 NS-2 SIMULATOR OVERVIEW

The NS-2 is a discrete-event simulation tool, widely

ImprovingNS-2NetworkSimulatortoEvaluateIEEE802.15.4WirelessNetworksUnderErrorConditions

215

used to study the dynamics of communication net-

works. It is developed in a collaborative effort by

many institutions, containing contributions from dif-

ferent researchers.The simulation library and network

protocols are written using the C and C++ languages.

The simulation environment is described and modi-

ﬁed using the OTcl script language, without the ne-

cessity to recompile the whole NS-2 source code.

Every action in NS-2 is associated with events

rather than time. An event comprises an execution

time, a set of tasks, and a reference to the next event.

These events are connected to each other, and form a

chain of events on the simulation time line. The se-

quential execution of this chain of events is controlled

and managed by a scheduler component, the brain and

execution engine of the NS-2. It is possible to deﬁne

its own procedures and variables to facilitate the inter-

action. The member procedures and variables in the

OTcl domain are called instance procedures.

The IEEE 802.15.4 NS-2 module is provided in

the form of methods of each layer speciﬁed in the

IEEE 802.15.4 standard. The module came with dif-

ferent functionalities, and support different network

topologies (star and point-to-point), two types of op-

eration (beacon and non-beacon enabled), and ba-

sic MAC management actions such as Association,

Channel Scan, energy model, etc.

6 ENHANCING NS-2 SIMULATOR

The evaluation of the network operation under error

conditions needs components capable to inject faults

in the simulation, which cause the network inacces-

sibility scenarios described previously. The NS-2 al-

ready provides components to perform such fault in-

jection, but these components using an error model

not capable to affect speciﬁc MAC frames, utilised by

the IEEE 802.15.4 to control the access to the net-

work.

To overcame the current error model limitation,

we complement the existing NS-2 components with

a new fault injector component, which is capable to

generate faults in speciﬁc MAC frames. We also in-

corporate in NS-2 a a temporal analysis component,

which is needed to account and measure the effects

of faults generated by the fault injector component.

These two components are independent of the type

of network, being separated from the IEEE 802.15.4

module, as represented in the ﬁgure 3.

6.1 Fault Injector Component

Our fault injector is capable to use a fault pattern to

Figure 3: New Features in 802.15.4 module.

Figure 4: Fault Injector scheme.

inject errors during the simulation. The criteria to de-

ﬁne the fault pattern is totally conﬁgurable, allowing

the deﬁnition of deterministic or probabilistic fault

patterns. An illustration of the fault injection scheme

is shown in the ﬁgure 4.

A fault pattern can be deﬁned to generate trans-

mission errors randomly in time (random noise or in-

terference) or be localized in speciﬁc time intervals

(deterministic noise). On both of these patterns, the

fault injector can be customized regarding the type of

frame to affect, the rate and the duration of the fault

injection.

Patterns with long duration are discouraged for

deterministic error models, since such long duration

may cause a permanent inaccessibility to the network

access. For example, if we are corrupting a beacon

frame injecting deterministic faults successively over

a long period we may cause the loss of synchroniza-

tion by the node and consequently this becoming un-

able to access the network again. However this type

of pattern is beyond the scope of this work that is to

analyse accidental faults where such pattern does not

happen.

To perform the random noise or interference is

possible to simulate aleatory errors on the network

communication, injecting faults between the MAC

and the PHY. A random function implemented in

the fault injector allows inserting random corrup-

tion events in the NS-2 scheduler as described in

Algorithm:1. In case of random noise the instant

when the corruption occurs is totally aleatory, and is

generated through a seed given by argument as de-

SENSORNETS2014-InternationalConferenceonSensorNetworks

216

Algorithm 1: Fault Injector - A random function.

1: Begin.

2: randomTime ← randomGenerator(seed);

3: NewRandomEvent ← faultInjector(frameToCorrupt);

4: Scheduler.schedule(NewRandomEvent,randomTime);

5: CorruptNode.Update();

6: End.

scribed in line 2. A new eventis created and the action

associated with it is a frame corruption performed by

the fault injector as indicated in line 3. Finally the

NewRandomEvent which will perform the corruption

is inserted in the NS-2 scheduler and executed at the

deﬁned instant as in line 4. An information about the

corruption occurred in a speciﬁc node is recorded as

described in line 5.

The fault injector achieves the frame corruption

as described in Algorithm:2, accessing the command

header of the frame as represented in line 5, and

changing a bit in the frame content, implying the drop

of these frames in the MAC level of the receiving

nodes. When the frame is received if the fault in-

jector is active, we can decide if a speciﬁc frame is

affected or any frame that a node receives will be

corrupted. The parameter frameToCorrupt repre-

sented on line 3 is previously deﬁned and if desired

all the received frames can be affected deﬁning the

frameToCorrupt to a speciﬁc value. An informa-

tion about the corruption occurred in a speciﬁc node

is recorded as described in line 6. This information

is used for a better control of the simulation events.

The fault injection may be performed in the coordi-

nator, which implies, depending on the type of frame

affected, that the whole network may be inaccessible,

in the speciﬁc case of affecting a MAC control frame.

In case we decide to affect a MAC control frame, af-

fecting speciﬁc network points, the fault injection can

be performed for example at non-coordinator nodes

tracking the reception of beacon frames. In the spe-

ciﬁc case, when we perform corruption in a MAC

control frame such as the beacon in the coordinator,

none of the nodes receives the beacon and therefore

the whole network will be inaccessible. Otherwise,

when the corruption is performed in the nodes that

should receive beacon frames, only the node that has

the fault injector component activated, i.e. beacon

corruptions occurring, cannot access the medium and

becomes inaccessible. The corruption of the frames

can be disabled, through the deactivation of the fault

injector on the tcl script, and the normal behaviour of

the network restored at any time.

6.2 Temporal Analysis Component

The temporal analysis component was designed to

Algorithm 2: Fault Injector Mechanism.

1: Begin.

2: MAC.Receive( f rame);

3: if f rame = frameToCorrupt then

4: for selected Fault Pattern do

5: CommandHeader( frame)− > error() = 1;

6: CorruptNode.Update();

7: end for

8: end if

9: End.

measure the temporal aspects of received frames,

from their duration to the effects of frames received

with errors during the network operation. In this pa-

per, we instrumented the temporal analysis compo-

nent to measure network inaccessibility events. As

the occurrence of network inaccessibility is related

with the MAC control frames (e.g., beacons), this

component was conﬁgured to monitor and capture in-

formation about such kind of frames.

Let us present an example to characterise how the

temporal analysis component works. We use the bea-

con frame as the frame to be monitored in such exam-

ple. When the received beacon is corrupted, the tem-

poral analysis component starts a timer to account the

related network inaccessibility period. When a new

beacon is received successfully, the timer is stopped

and the duration of such period is registered.

When the simulation is ﬁnished, the temporal

analysis component generates a report, regarding all

events captured during the simulation. The report is

utilised as input to a gnuplot script, which transform

the raw data within the report to a graphical represen-

tation of the captured events.

7 SIMULATING

INACCESSIBILITY

SCENARIOS

The simulation is deﬁned in an OTcl script (Listing:1)

and is carried out in an one-hop star topology, where

all the nodes are within the range of each other.

In the script (Listing:1) we deﬁne that the ﬁrst

node to start was the coordinator, specifying the val-

ues of its BO and SO in line 1. After the WnS is es-

tablished we start the nodes in line 2. Our temporal

analysis component is enabled on line 3, given the se-

lected scenario. The periodic beacon transmission is

initiated at the coordinator on line 4, taking the BO

and SO as arguments. At line 5 we enable the GTS

transmission for the node(1), which means that each

time this node have data to transmit will use the GTS

mechanism. Finally, at line 6 we start our fault injec-

ImprovingNS-2NetworkSimulatortoEvaluateIEEE802.15.4WirelessNetworksUnderErrorConditions

217

tor to, in this example, corrupt beacon frames for a

certain number of rounds.

1 Event at 0 .0 node (0)

st a rtW nSCoor dinat or $beaconOrder

$superFrameOrder ” ;

2 Event at 20 .0 node ( 1 ) & node ( 2)

s t a r t D e v i ce ”

3 Event at 20 .0 node ( 1 )

enableTemporalAccount $Scenario ”

;

4 Event at 30 .0 node ( 0 )

st a rtBeaconTran sm i s si on

$beaconOrder $superFrameOrder ”

5 Event at 30 .0 node ( 1 ) GTS On”

6 Event at 30 .0 node ( 1 ) S t a r t F aul t

I n j e c t i o n $Beacon $Rounds ”

7 Event at $stopTime ” st op ”

Listing 1: Example of NS-2 Simulation Script.

To simulate a network operation under error con-

ditions, implying the occurrence of network inacces-

sibility, we conﬁgure our fault injector component to

generate deterministic faults. For each addressed sce-

nario we set our fault injector to corrupt a speciﬁc

frame at a given number of times, on a chosen node.

The fault injector can corrupt one of each frame type

present in the Table: 2.

To achieve the Single Beacon Frame Loss

(SBFL) scenario we executed the following schedule

of Events:

1 Event at 30 .0 node ( 1 ) S t a r t F aul t

I n j e c t i o n $Beacon $SBFL”

In this scenario the beacon frame will be corrupted

SBFL number of times (i.e., only one time) at the

Node(1), after 30 simulation seconds.

The Multiple Beacon Frame Loss (MBFL)

occurs when we change the number of corrupting

rounds on the fault injector depending on the value

that MBFL assumes in order to achieve the loss

of nrLost beacons. The synchronization loss is

a special case of the MBFL scenario where after

the loss of nrLost beacons the next beacon is also lost.

1 Event at 30 .0 node ( 1 ) S t a r t F aul t

I n j e c t i o n $Beacon $MBFL”

The Orphan notiﬁcation and Coordinator

realignment are achieved when the fault injector

corrupts NOSYNC beacon frames, corresponding to

the current scenario, and the node lose the synchro-

nization. The Orphan notiﬁcation is observed on the

node and the Coordinator realignment is transmitted

by the coordinator on response.

Table 2: MAC frame types.

Frame type value Command frame ID Standard Reference

0 Beacon

1 Data

2 Ack

3 MAC Control Frame

01 Association Request

02 Association Response

03 Disassociation

04 Data Request

05 Coordinator conﬂict

06 Orphan

07 Beacon request

08 Coordinator realignment

09 GTS request

1 Event at 30 .0 node ( 1 ) S t a r t F aul t

I n j e c t i o n $Beacon $NOSYNC”

So that Coordinator Conﬂict Detection can oc-

cur, this event has to be forced on the simulator. Once

every time a node becomes a coordinator it assumes

its ID as the networkID, so a coordinator conﬂict is

impossible because every coordinator assumes a dis-

tinct ID. To force that event we oblige the coordinator

to use the same identiﬁer with the following line.

1 Event at 0 . 0 node (1) Coord inator

Co nf li ct 1”

2 Event at 0 . 0 node (0) Coord inator

Co nf li ct 1”

When the GTS mechanism is previously acti-

vated from the script, and the node has data to trans-

mit, a GTS Request will occur. This request will be

send to the coordinator by the node to perform an allo-

cation of a GTS slot for exclusive transmission time.

1 Event at 30 .0 node ( 1 ) GTS On”

8 RESULTS

8.1 Simulation Setup

The network was simulated with seven nodes, where

one of these nodes, in the center, was the coordinator.

All other nodes are in the radio transmission range

of the coordinator, and in the range of each other. A

BO = 3 was utilised to specify the superframe dura-

tion within simulations.

The characteristics of the simulation setup sce-

nario are shown in Table 3. To evaluate the network

behaviour under error conditions we applied different

SENSORNETS2014-InternationalConferenceonSensorNetworks

218

Table 3: Simulation Parameters.

Simulation Parameters

NS-2 Version 2.35 updated with GTS, Fault Injector,

and Temporal Analysis features

Network Topology Star Topology

Nodes 7

Trafﬁc Constant Bit Rate (CBR)

Reception range 15m

Carrier Sense range 15m

Packet Size 8, 67, 127 kbytes

CAP Transmission Type Direct, using CSMA/CA

CFP Transmission Type GTS transmission

Transmission/Reception Power 30mW

Beacon Enabled

Beacon Order 3

Superframe Order 3

Maximum CSMA/CA Attempts 4

Simulation Time 600 seconds

error patterns on the simulation through fault injec-

tion.

The fault injection can be performed using three

different durations: short, medium, or long. The short

had the duration of a normal frame transmission, the

medium had the duration of the transmission of 3

frames, and the long had the duration of half a bea-

con.

The trafﬁc generator is set to produce Constant

Bit Rate trafﬁc (CBR), which means data frames are

transmitted at a constant rate from the nodes to the

coordinator. The payload of the sent data was also

varied, being the smallest payload of 8 kilobytes, the

medium of 67 kilobytes, and the large of 127 kilo-

bytes. The characteristics of the simulation scenario

are shown in Table 3.

8.2 Simulation Results

After the environment setup, the simulation was per-

formed to obtain the best and worst case duration of

the inaccessibility scenarios.

Figure 5 shows the graphic that represents the du-

ration of each network inaccessibility scenario, com-

paring the results of the previous theoretical study

in (Souza and Ruﬁno, 2009) and the obtained simula-

tion results. The results presented in Figure 5 clearly

shows that periods of inaccessibility may have a huge

duration, which represents a non-negligible impact

for networks with temporal restricted trafﬁc.

In the Figure 6 we can observe the Packet Error

Rate (PER) between different error conditions. As

expected, longer periods of fault injections implies a

higher PER. In comparison with the Control frame,

a frame that was transmitted without errors, we can

see an increasing PER related to the higher periods of

fault injection.

Single

beacon

frame loss

Multiple

beacon

frame loss

Synchonisation

loss

Orphan Coordinator

realignment

GTS

Request

Coordinator

Conflict

Detection

Coordinator

Conflict

Resolution

Association Re-

Association

Normalised duration (T

times)

Network inaccessibility scenarios (BO=3)

Beacon interval (T

)

Theoretical

Simulated

Figure 5: Normalized Inaccessibility Scenarios comparison

between Theoretical and Simulated worst case (BO=SO=3

and T

= 0.123s).

8 67 127

Packet Error Rate (PER) (%)

Payload length (Bytes)

Fault Injection duration:

Control

Short

Medium

Long

Figure 6: Packet Error Rate comparison between different

error patterns.

The result with the greatest impact is related with

the bigger payload, achieves a PER of more than 10%.

An important result of this study is that the inﬂu-

ence of network errors, causing periods of inacces-

sibility in the network, cannot be overlooked if pre-

dictability and real-time operation is a system require-

ment, under the risk of jeopardize the safety and time-

liness of the entire system. The effects of network

inaccessibility incidents should be controlled by the

deﬁnition of strategies for the reduction of the peri-

ods of inaccessibility, which is achieved in (Souza and

Ruﬁno, 2013).

9 CONCLUSIONS AND FUTURE

DIRECTIONS

The paper addressed the behaviour of IEEE 802.15.4

networks in the presence of network errors, leading to

periods of network inaccessibility.

ImprovingNS-2NetworkSimulatortoEvaluateIEEE802.15.4WirelessNetworksUnderErrorConditions

219

Signiﬁcant improvementsand modiﬁcationsin the

NS-2 simulator IEEE 802.15.4 module were pre-

sented, which includes the speciﬁcation of two addi-

tional components capable to inject and measure the

effects of errors during the network operation. The

presence and use of these two component were es-

sential to perform the simulation and evaluation of all

network inaccessibility scenarios.

The results obtained by our simulations evidence

the relevant temporal aspects of the IEEE 802.15.4

beacon-enabled networks operating under error con-

ditions. Such results can be utilised in the speciﬁca-

tion of a robust timeliness model of the network, in

order to achieve an effective support to real-time op-

eration in IEEE 802.15.4 networks.

ACKNOWLEDGEMENTS

This work was partially supported: by the

EC, through project IST-FP7-STREP-288195

(KARYON); by FCT/DAAD, through the transna-

tional cooperation project PROPHECY; and by

FCT, through the project PTDC/EEI-SCR/3200/2012

(READAPT), the Multiannual Funding Program, and

the Individual Doctoral Grant SFRH/BD/45270/2008.

REFERENCES

Han, S., Zhu, X., Mok, A., Chen, D., and Nixon, M. (2011).

Reliable and Real-Time Communication in Industrial

Wireless Mesh Networks. 2011 17th IEEE Real-Time

and Embedded Technology and Applications Sympo-

sium.

Hou, L. and Bergmann, N. (2010). System Requirements

for Industrial Wireless Sensor Networks The Univer-

sity of Queensland.

Kandhalu, A. and Rajkumar, R. (2012). Qos-based resource

allocation for next-generation spacecraft networks. In

IEEE 33rd Real-Time Systems Symposium (RTSS).

Akerberg, J., Gidlund, M., and Bj¨orkman, M. (2011). Fu-

ture research challenges in wireless sensor and actu-

ator networks targeting industrial automation. In 9th

IEEE Internation Conference on Industrial Informat-

ics (INDIN).

Saifullah, A., Xu, Y., Lu, C., and Chen, Y. (2011). End-

to-end delay analysis for ﬁxed priority scheduling in

wirelesshart networks. In 17th IEEE Real-Time and

Embedded Technology and Applications Symposium

(RTAS).

Shuai, X. and Zhang, Z. (2010). Research of Real-Time

Wireless Networks Control System MAC Protocol.

Journal of Networks.

Souza, J. L. R. and Ruﬁno, J. (2009). Characterization of

inaccessibility in wireless networks - a case study on

IEEE 802.15.4 standard. In IFIP 3th International

Embedded System Simposium IESS.

Souza, J. L. R. and Ruﬁno, J. (2013). Analysing and reduc-

ing network inaccessibility in IEEE 802.15.4 wireless

communications. In 38th IEEE Conference on Local

Computer Networks (LCN).

Standard, I. and Society, I. C. (2011). IEEE Standard for

Local and metropolitan area networks, Part 15 . 4

: Low-Rate Wireless Personal Area Networks ( LR-

WPANs ) IEEE Computer Society S ponsored by the.

Stone, T., Alena, R., Baldwin, J., and Wilson, P. (2012).

A viable COTS based wireless architecture for space-

craft avionics. In IEEE Aerospace Conference, pages

1–11.

Wang, J., Dong, W., Cao, Z., and Liu, Y. (2012). On the

delay performance analysis in a large-scale wireless

sensor network. In IEEE 33rdReal-Time Systems Sym-

posium (RTSS).

SENSORNETS2014-InternationalConferenceonSensorNetworks

220