Performance Analysis of the Actuator Discovery Protocol ... - Fethi Filali
Performance Analysis of the Actuator Discovery Protocol ... - Fethi Filali
Performance Analysis of the Actuator Discovery Protocol ... - Fethi Filali
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Institut Eurécom<br />
Department <strong>of</strong> Mobile Communications<br />
2229, route des Crêtes<br />
B.P. 193<br />
06904 Sophia-Antipolis<br />
FRANCE<br />
Research Report RR-06-159<br />
<strong>Performance</strong> <strong>Analysis</strong> <strong>of</strong> <strong>the</strong> <strong>Actuator</strong> <strong>Discovery</strong> <strong>Protocol</strong><br />
for Mobile Sensor and <strong>Actuator</strong> Networks<br />
March 30 th , 2006<br />
Muhammad Farukh Munir and <strong>Fethi</strong> <strong>Filali</strong><br />
Tel : (+33) 4 93 00 82 03<br />
Fax : (+33) 4 93 00 82 00<br />
Email : {Muhammad-Farukh.Munir, <strong>Fethi</strong>.<strong>Filali</strong>}@eurecom.fr<br />
1 Institut Eurécom’s research is partially supported by its industrial members: Bouygues Télécom,<br />
Fondation d’entreprise Groupe Cegetel, Fondation Hasler, France Télécom, Hitachi, Sharp, ST<br />
Microelectronics, Swisscom, Texas Instruments, Thales.
<strong>Performance</strong> <strong>Analysis</strong> <strong>of</strong> <strong>the</strong> <strong>Actuator</strong> <strong>Discovery</strong> <strong>Protocol</strong><br />
for Mobile Sensor and <strong>Actuator</strong> Networks<br />
Muhammad Farukh Munir and <strong>Fethi</strong> <strong>Filali</strong><br />
Abstract<br />
This paper presents <strong>the</strong> performance analysis <strong>of</strong> our proposed energydelay<br />
aware coordination protocol, ADP [1] (<strong>Actuator</strong> <strong>Discovery</strong> <strong>Protocol</strong>)<br />
for mobile SANETs (Sensor and <strong>Actuator</strong> Networks). We have presented<br />
our analysis on ADP for static SANETs in an earlier work and it is shown<br />
that ADP provides optimal attachment for a sensor node to <strong>the</strong> actuator with<br />
minimum cost. For <strong>the</strong> sensor-actuator coordination, <strong>the</strong> proposed ADP can<br />
guarantee ordering, synchronization and eliminates <strong>the</strong> redundancy <strong>of</strong> actions<br />
[1]. In this work, we present our analysis on <strong>the</strong> dynamics <strong>of</strong> ADP<br />
to handle mobile sensor and actuator applications. The interval chosen to<br />
validate <strong>the</strong> acquired paths through ADP due to mobility can be a direct<br />
mapping <strong>of</strong> application’s energy and delay constraints. Our NS-simulations<br />
show that <strong>the</strong> total-energy consumed to validate <strong>the</strong> paths for time-stringent<br />
traffic is actually less compared to <strong>the</strong> overall energy consumed during arbitrary<br />
transmission <strong>of</strong> sensor data to-wards <strong>the</strong> sink. We have utilized a<br />
random way-point mobility model in our ns-simulations and analyze <strong>the</strong> performance<br />
<strong>of</strong> ADP in terms <strong>of</strong> energy estimates for maintaining paths (toward<br />
<strong>the</strong> optimal-actuator) in order to keep a promised QoS for time-stringent applications.<br />
The proposed ADP has also achieved a good balance for sensor<br />
energy-consumption while keeping a low end-to-end delay (optimal number<br />
<strong>of</strong> ’message-exchange’ to keep <strong>the</strong> paths updated).<br />
Keywords: SANETs (Sensor and <strong>Actuator</strong> Networks), self-organization,<br />
coordination protocols, energy and delay aware coordination protocols.<br />
ii
Contents<br />
1 Introduction and Related Work 1<br />
2 The ADP for Mobile SANETs 2<br />
2.1 Overview <strong>of</strong> <strong>the</strong> ADP . . . . . . . . . . . . . . . . . . . . . . . . 2<br />
2.2 ADP Dynamics for Mobile SANETs . . . . . . . . . . . . . . . . 4<br />
3 Experimental analysis <strong>of</strong> <strong>the</strong> ADP 5<br />
4 Conclusion and Future Work 8
List <strong>of</strong> Figures<br />
1 AttachRequest by sensors at <strong>the</strong> start <strong>of</strong> ADP . . . . . . . . . . . . . . . . . . 3<br />
2 <strong>Actuator</strong>-replies (AttachReply) for corresponding AttachRequest messages . . . . . . 4<br />
3 The occurrence <strong>of</strong> Failure due to mobility and Recovery (exchange <strong>of</strong> ’Hello’ msgs) Procedure. 4<br />
4 CDF for Attachment-Delay for Static Complex Network Configuration . . . . . . . . 6<br />
5 Mean Path Length for Static Complex Network Configuration . . . . . . . . . . . 6<br />
6 Average Cluster Size over time for Static and Mobile SANETs . . . . . . . . . . . 7<br />
7 Cluster Densities avg_speed = 1-m/s, pause_time 20s, Tx_range = 11 m, distance d (between<br />
any two nodes) = 10 m . . . . . . . . . . . . . . . . . . . . . . . . 8<br />
8 Cluster Densities avg_speed = 2-m/s, pause_time 10s, distance d (between any two nodes)<br />
= 10 m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9<br />
9 Average Cluster Size for Mobile SANET Configuration Vs. Time . . . . . . . . . . 10<br />
10 Delay Comparison for Static and Mobile SANET Configuration . . . . . . . . . . 11<br />
11 Energy-Consumption for Static and Mobile SANETs . . . . . . . . . . . . . . . 11<br />
iv
1 Introduction and Related Work<br />
The advent in technology led to <strong>the</strong> emergence <strong>of</strong> SANETs giving a distributed<br />
control to <strong>the</strong> management, communication and coordination aspects <strong>of</strong> <strong>the</strong> network<br />
functioning formerly referred to as WSNs (Wireless Sensor Networks). A<br />
SANET consists <strong>of</strong> a group <strong>of</strong> sensors and actors (actuators), that are deployed to<br />
perform distributed sensing and actuation tasks, linked up by a wireless medium.<br />
The sensor nodes (small, cheap devices with limited computation) are deployed for<br />
<strong>the</strong> collection <strong>of</strong> data through a sensing mechanism, while actuators (resource rich,<br />
better processing capabilities and stronger transmission power) take decisions and<br />
<strong>the</strong>n perform appropriate actions upon <strong>the</strong> environment. Most <strong>of</strong> <strong>the</strong> sensor network<br />
research tends to focus on specific hardware with efficient, ingenious communication<br />
protocol algorithms and system control architectures that address <strong>the</strong><br />
specific resource constraints, which did not lead to <strong>the</strong> emergence <strong>of</strong> a generic architecture<br />
for <strong>the</strong> sensor networks [2, 3, 4]. Despite <strong>of</strong> <strong>the</strong> sufficient work done<br />
in SANETs [5, 6, 7, 8], an effective and distributed coordination mechanism for<br />
efficient sensing, dissemination <strong>of</strong> information, and to perform right and timely<br />
actions (actuated by <strong>the</strong> actuators) is required. These networks can be <strong>the</strong> integral<br />
part <strong>of</strong> systems such as: disaster/crime prevention, real-time military applications,<br />
environmental and health monitoring to smart spaces [9].<br />
This paper analyzes our proposed self-organizing framework for <strong>the</strong> mobile<br />
SANETs. In this regard, we have implemented ADP (<strong>Actuator</strong> <strong>Discovery</strong> <strong>Protocol</strong>)<br />
as an application layer coordination protocol. which provides optimal attachment<br />
to a sensor node in <strong>the</strong> network with <strong>the</strong> nearest and most relevant actuator<br />
node. A sensor node anywhere in <strong>the</strong> network finds <strong>the</strong> nearest actuator using <strong>the</strong><br />
proposed ADP during <strong>the</strong> initial deployment phase. Afterwords <strong>the</strong> data is relayed<br />
to <strong>the</strong> attached-actuator through a multi-hop discrete path unless <strong>the</strong> network undergoes<br />
a remarkable topology change, which covers <strong>the</strong> basics <strong>of</strong> sensor-actuator<br />
coordination. We have employed a periodic exchange <strong>of</strong> “Hello” messages to keep<br />
<strong>the</strong> paths updated. The update interval can be modified according to <strong>the</strong> applications’<br />
requirements. Even in case <strong>of</strong> mobile SANETs, <strong>the</strong> mobility <strong>of</strong> <strong>the</strong> nodes<br />
is not expected to be very high as opposed to <strong>the</strong>ir sensing duration during monitoring<br />
a zone, which would be a realistic deployment situation for most mobile<br />
applications. ADP is very simple to accommodate <strong>the</strong> limitations <strong>of</strong> small sensor<br />
nodes and is energy efficient. The functionality <strong>of</strong> ADP is independent <strong>of</strong> <strong>the</strong><br />
routing protocol and is distributed to cope with large scale deployments. For <strong>the</strong><br />
actuator-actuator coordination, we employ <strong>the</strong> use <strong>of</strong> <strong>the</strong> assumed separate wireless<br />
interface to communicate with <strong>the</strong> neighboring actuators so that <strong>the</strong>y can perform<br />
long-range data communication without any involvement <strong>of</strong> sensor nodes to effectively<br />
coordinate and to perform <strong>the</strong> actuation process. The examples <strong>of</strong> actuation<br />
covers a wide range <strong>of</strong> possibilities and is typically application dependent.<br />
For example, targeting an enemy holding a snipper in a battle field can be an<br />
interesting case to consider. The actuation process has to localize <strong>the</strong> position <strong>of</strong><br />
<strong>the</strong> enemy and actuate <strong>the</strong> destruction process. But <strong>the</strong> important constraint in this<br />
1
case is <strong>the</strong> latency because <strong>the</strong> sensor data can be no more valid at <strong>the</strong> time <strong>of</strong><br />
actuation in case <strong>of</strong> increased latency. Which makes <strong>the</strong> assumption <strong>of</strong> a separate<br />
wireless interface for actuator-actuator coordination quite relevant.<br />
Note that in this work, we study <strong>the</strong> generic architecture which is robust and entirely<br />
distributed with a little varying-degree <strong>of</strong> centralized control for <strong>the</strong> actuators<br />
on <strong>the</strong>ir local clusters. In prior work done for SANET coordination protocols, an<br />
excess <strong>of</strong> energy is consumed whenever sensors have some sensed data to transmit<br />
to <strong>the</strong> nearest sink, even at <strong>the</strong> cost <strong>of</strong> variable delay and regardless <strong>of</strong> <strong>the</strong> fact that<br />
events were rarely detected or an any-cast transmit strategy is followed. Instead<br />
<strong>of</strong> trying to achieve improvements for a particular application, we addressed both<br />
<strong>the</strong> static and dynamic network configurations with different natures <strong>of</strong> mobility,<br />
changes to <strong>the</strong> discrete paths, energy and delay efficient path-recovery procedures,<br />
and avoid using <strong>the</strong> GPS system for position localizations. In addition, <strong>the</strong> energy<br />
consumption by <strong>the</strong> sensors (in order) to keep <strong>the</strong> updated-paths is highly dependent<br />
on mobility. In this regard, <strong>the</strong> nodes can adaptively tune <strong>the</strong>ir update-interval<br />
based on <strong>the</strong>ir mobility, which makes <strong>the</strong> ADP highly adaptive to any kind <strong>of</strong> network<br />
configuration with optimum energy consumption and a promised QoS. It<br />
also has <strong>the</strong> ability to adapt any nature <strong>of</strong> mobility at <strong>the</strong> expense <strong>of</strong> minimum total<br />
network energy-consumption. This feature has <strong>the</strong> tendency to lean <strong>the</strong> SANETs<br />
to-wards minimum residual energy sensor and actuator networks. We believe that<br />
<strong>the</strong> proposed ADP leads toward optimal total network energy consumption providing<br />
a guaranteed QoS for both densely deployed and stringent real-time traffic<br />
networks.<br />
The remainder <strong>of</strong> <strong>the</strong> paper is structured as follows. In Section 2, <strong>the</strong> dynamics<br />
<strong>of</strong> <strong>the</strong> proposed ADP for mobile SANETs are explained in detail. Section 3<br />
presents <strong>the</strong> experimental results. Section 4 summarizes <strong>the</strong> paper with a brief<br />
conclusion and outlines <strong>the</strong> future work.<br />
2 The ADP for Mobile SANETs<br />
As we have already explained in <strong>the</strong> earlier section that <strong>the</strong>re is no sensor/actuator<br />
network structure and architecture, and <strong>the</strong> basic goals <strong>of</strong> <strong>the</strong> previously done work<br />
relied mostly on <strong>the</strong> considered application. In this work, we propose an architecture<br />
which is tailored toward a standard behavior for most deployment scenarios,<br />
aiming to satiate <strong>the</strong> time-stringent requirements and effective energy utilization in<br />
a purely distributed fashion.<br />
2.1 Overview <strong>of</strong> <strong>the</strong> ADP<br />
The learning-phase starts when <strong>the</strong> sensors during <strong>the</strong> initial deployment stage locate<br />
<strong>the</strong> nearest actuator using a one-hop broadcast. The deployment <strong>of</strong> <strong>the</strong> sensor<br />
and actuator nodes can be randomly chosen or follow any pre-defined distributions.<br />
The finding <strong>of</strong> <strong>the</strong> "optimal-actuator attachment" for each sensor node is<br />
2
done through <strong>the</strong> proposed novel protocol called ADP.<br />
When a sensor node is turned on, it should first determine to which actuator<br />
node it has to be associated. For this end, a sensor node transmits a broadcast message<br />
named AttachRequest(cost, M j , C ) to all its one-hop neighbors as shown in<br />
Figure 1. A neighboring node upon receiving an attach-request message checks<br />
that it has sent an attach-request in <strong>the</strong> period T n (application specific), if it has<br />
already sent a broadcast to its neighbors, it has to wait for a reply until timeout.<br />
O<strong>the</strong>rwise it does <strong>the</strong> same unless <strong>the</strong> probe reaches <strong>the</strong> nearest actuator. The reply<br />
message named AttachReply i (cost, M j , A i ) from <strong>the</strong> actuator follows <strong>the</strong> probe<br />
and terminates at its origin defining a discrete path to <strong>the</strong> sensor node. If a node<br />
receives multiple paths <strong>the</strong>n it chooses <strong>the</strong> shortest one (hop-count), and certainly<br />
if it receives more than one path containing <strong>the</strong> same hop-count, it chooses <strong>the</strong><br />
earliest reply.<br />
ACTOR NODE<br />
Sensor Nodes<br />
AttachRequest<br />
Figure 1: AttachRequest by sensors at <strong>the</strong> start <strong>of</strong> ADP<br />
We don’t take into account <strong>the</strong> distance between any node and its associated<br />
neighbors with <strong>the</strong> reason being that <strong>the</strong> energy required to transmit to a node<br />
in its sensing radius is a constant (no power control assumed for transmissions).<br />
ADP produces loop-free paths to <strong>the</strong> actuator nodes, as stated below. The next-hop<br />
selected by a sensor with ADP has a defined optimal path to <strong>the</strong> actuator node.<br />
As depicted by Figure 2, now a sensor node has a specified path to route its<br />
sensed data to <strong>the</strong> actuator nodes by simply forwarding it to only one <strong>of</strong> its neighbors<br />
(immediate next node in <strong>the</strong> path to <strong>the</strong> actuator), and <strong>the</strong> actuator also keeps<br />
<strong>the</strong> defined path to <strong>the</strong> node (building its tree structure for <strong>the</strong> localized cluster).<br />
The background for this consideration takes us to <strong>the</strong> lack <strong>of</strong> a GPS system for<br />
position localizations in our study, which consumes a lot <strong>of</strong> sensor nodes’ energy.<br />
In a similar fashion all <strong>the</strong> nodes reserve an optimal path to <strong>the</strong>ir nearest actuators,<br />
forming a local cluster, thus giving us <strong>the</strong> initial deployment in <strong>the</strong> form <strong>of</strong> welldistributed<br />
small clusters.<br />
3
Attach−Replies<br />
Actor−Node<br />
Sensor−Node<br />
Attach−Reply<br />
Paths<br />
not chosen due<br />
more hop−count<br />
Figure 2: <strong>Actuator</strong>-replies (AttachReply) for corresponding AttachRequest messages<br />
2.2 ADP Dynamics for Mobile SANETs<br />
The Failure and Recovery phase monitors a periodic update <strong>of</strong> <strong>the</strong> optimal paths<br />
and if any particular path is not valid anymore, <strong>the</strong> sensors send a one-hop broadcast<br />
to <strong>the</strong>ir neighbors and acquire ano<strong>the</strong>r optimal actuator to route <strong>the</strong>ir fur<strong>the</strong>r<br />
sensed data. The sensor <strong>the</strong>n informs <strong>the</strong> optimal actuator to join its local cluster<br />
and similarly <strong>the</strong> actuator informs <strong>the</strong> neighboring actuators <strong>of</strong> <strong>the</strong> modification for<br />
effective actuation process. The process <strong>of</strong> updating a path in case <strong>of</strong> node failure<br />
or mobility is shown in Figure 3.<br />
At T2, an Actor updating <strong>the</strong> neighboring Actors<br />
New Cluster<br />
found at T2<br />
Node moving from<br />
One Cluster to ano<strong>the</strong>r at T1<br />
Figure 3: The occurrence <strong>of</strong> Failure due to mobility and Recovery (exchange <strong>of</strong> ’Hello’ msgs) Procedure.<br />
In this work, we have taken into consideration two possible deployment scenarios.<br />
(i) Both sensors/actuators are static: For <strong>the</strong> static-case, if <strong>the</strong>re is a node<br />
failure in <strong>the</strong> middle <strong>of</strong> <strong>the</strong> path, <strong>the</strong> far<strong>the</strong>r nodes can do a one-hop broadcast<br />
to all <strong>the</strong> neighbors in <strong>the</strong>ir sensing range, which can reply with <strong>the</strong>ir respective<br />
4
hop-counts toward <strong>the</strong> actuator. The sensor node selects <strong>the</strong> shortest path to <strong>the</strong><br />
actuator and an update message is sent (piggybacked along with <strong>the</strong> data-packet, if<br />
necessary) to <strong>the</strong> actuator for updating <strong>the</strong> correlation tree.<br />
(ii) Static actuators and mobile sensors: In mobile-sensor configuration, <strong>the</strong><br />
path to <strong>the</strong> attached-actuator can be no more valid at any instant in time due to <strong>the</strong><br />
mobility pattern <strong>of</strong> <strong>the</strong> nodes, and can also be <strong>the</strong> result <strong>of</strong> an accidental-failure.<br />
The nodes that have no more defined paths sends a one-hop broadcast. If a reply<br />
is received, <strong>the</strong> same procedure will follow as for <strong>the</strong> static case. O<strong>the</strong>rwise <strong>the</strong><br />
one-hop neighbors send a path-search probe to <strong>the</strong>ir one hop neighbors, and <strong>the</strong><br />
procedures can terminate after acquiring a path to <strong>the</strong> actuator. In-fact, if this is <strong>the</strong><br />
case, <strong>the</strong> neighboring nodes that don’t have a defined path has already sent a pathupdate<br />
request after time-out in most practical scenarios. All <strong>the</strong> nodes have to<br />
au<strong>the</strong>nticate <strong>the</strong> validity <strong>of</strong> <strong>the</strong>ir paths after a certain defined interval T o which can<br />
be a trade-<strong>of</strong>f between <strong>the</strong> mobility model and <strong>the</strong> path-update procedure. In <strong>the</strong><br />
mobile-sensor case <strong>the</strong> constant dissemination <strong>of</strong> energy to keep <strong>the</strong> paths updated<br />
can be a direct mapping to <strong>the</strong> applications’ requirements. The dry analysis for<br />
<strong>the</strong> proposed scheme intuitively satisfies <strong>the</strong> efficient consumption <strong>of</strong> energy and<br />
meeting <strong>the</strong> required constraints for real-time traffic. In contrast <strong>the</strong> total energy<br />
consumed by <strong>the</strong> ADP to keep <strong>the</strong> paths updated is less than <strong>the</strong> total energy consumed<br />
by <strong>the</strong> sensors using any <strong>of</strong> <strong>the</strong> coordination mechanisms proposed earlier<br />
for <strong>the</strong> reason that <strong>the</strong>y were application dependent. With our ADP, we can simply<br />
maintain a trade-<strong>of</strong>f between a predefined latency and sensor energy consumption<br />
by managing <strong>the</strong> interval T o , which is mobility dependent and hence making this<br />
an implementation issue. The alternate-path obtained is also a loop-free optimal<br />
path to <strong>the</strong> actuator node. The Actors coordinate among each o<strong>the</strong>r to keep <strong>the</strong><br />
updated networking paradigm for effective actuation process.<br />
3 Experimental analysis <strong>of</strong> <strong>the</strong> ADP<br />
We finally implemented <strong>the</strong> ADP in ns-2 [21] as an application layer protocol. We<br />
have considered a wide range <strong>of</strong> network topologies and test-validated <strong>the</strong> performance<br />
<strong>of</strong> ADP with all considered topologies. Due to <strong>the</strong> posed space limitation,<br />
we present our results for a complex random network (static and mobile). We<br />
consider a complex random network configuration with 100 nodes (5% actuators),<br />
where both sensor and actuator nodes are static. We have also evaluated <strong>the</strong> average<br />
number <strong>of</strong> sensors per actuator (local cluster) for complex random network<br />
topology. The local clusters have an equilibrium in terms <strong>of</strong> number <strong>of</strong> sensors,<br />
and <strong>the</strong> calculated mean turns out to be 19 sensors per actuator, which proves that<br />
<strong>the</strong> paths chosen by <strong>the</strong> sensors to <strong>the</strong>ir respective actuators are <strong>the</strong> optimal paths.<br />
The CDF for <strong>the</strong> delay distribution shows that over 65% <strong>of</strong> <strong>the</strong> nodes lie in <strong>the</strong><br />
average-delay (according to our findings) range as shown in Figure 4 along with<br />
<strong>the</strong> mean number <strong>of</strong> nodes per cluster. We have also shown, in Figure 5, <strong>the</strong> mean<br />
path length as a function <strong>of</strong> network size. The mean path length, which is related<br />
5
to <strong>the</strong> end-to-end latency for both cases: Fixed number <strong>of</strong> actuators, and secondly<br />
actuators increasing by 5% with network size. However, <strong>the</strong> increase is more gradual<br />
with <strong>the</strong> increasing number <strong>of</strong> actuator nodes, which would be a more practical<br />
assumption for <strong>the</strong> SANETs.<br />
90<br />
80<br />
70<br />
Number <strong>of</strong> Sensors<br />
60<br />
50<br />
40<br />
30<br />
20<br />
100 Nodes in <strong>the</strong> Network & 5% <strong>Actuator</strong>s<br />
10<br />
0<br />
0.5 1 1.5 2 2.5 3 3.5 4<br />
Delay (in ms)<br />
Figure 4: CDF for Attachment-Delay for Static Complex Network Configuration<br />
8<br />
7<br />
Fixed Number <strong>of</strong> Actors<br />
Actors increasing by 5%<br />
6<br />
Mean Path Length<br />
5<br />
4<br />
3<br />
2<br />
1<br />
0<br />
0 50 100 150 200 250<br />
Nodes in <strong>the</strong> Network<br />
Figure 5: Mean Path Length for Static Complex Network Configuration<br />
In Figure 6, we have obtained <strong>the</strong> average local cluster densities as a function<br />
<strong>of</strong> time for both static and mobile SANETs. The cluster average shows <strong>the</strong> optimal<br />
equilibrium obtained in terms <strong>of</strong> efficient energy utilization inside each local<br />
cluster. This phenomenon leads <strong>the</strong> network toward minimal total network energyconsumption,<br />
if cost functions are well driven. In analyzing <strong>the</strong> performance <strong>of</strong><br />
6
ADP, <strong>the</strong> interesting measurement <strong>of</strong> a pre-promised delay for any type <strong>of</strong> traffic<br />
comes from <strong>the</strong> increased optimal energy consumption, which was usually not <strong>the</strong><br />
case for previous work.<br />
25<br />
Mobile-Case, path updates and recovery<br />
Static-Case, no updates required<br />
20<br />
Number <strong>of</strong> Sensors<br />
15<br />
10<br />
5<br />
0<br />
0 5 10 15 20 25 30 35 40 45 50<br />
Time (in sec.)<br />
Figure 6: Average Cluster Size over time for Static and Mobile SANETs<br />
Similarly in <strong>the</strong> dynamic node configuration, we have again considered a random<br />
complex network with 100 nodes (5% actuators), where <strong>the</strong> sensors are mobile<br />
and actuators have defined positions. We haven’t thought <strong>of</strong> any particular mobile<br />
application <strong>of</strong> SANETs, so we focus our major goal to-wards optimizing <strong>the</strong> global<br />
energy-consumption while maintaining an optimal delay for stern real-time traffic.<br />
We have applied a random way-point mobility model in our ns-simulations and analyze<br />
<strong>the</strong> performance <strong>of</strong> ADP in terms <strong>of</strong> energy estimates for maintaining paths<br />
(to-wards <strong>the</strong> optimal-actuator) in order to keep a promised QoS for time-stringent<br />
applications. The proposed ADP has also achieved a good balance toward energyconsumption<br />
while keeping a low end-to-end delay (optimal number <strong>of</strong> ’messageexchange’<br />
to keep <strong>the</strong> paths updated). The mobility <strong>of</strong> <strong>the</strong> sensors is kept slow<br />
(mean speed = 1-m/s) and <strong>the</strong> duration <strong>of</strong> stay at one place is kept more (pause<br />
time = 20s) for <strong>the</strong> appropriate sensing <strong>of</strong> <strong>the</strong> deployed environment. The Cluster<br />
densities for two different mobility patterns vs. time are shown in <strong>the</strong> Figure 7 and<br />
8 respectively. In both cases, <strong>the</strong> clusters tend to acquire an equilibrium in <strong>the</strong>ir<br />
sensor density based on <strong>the</strong> mobility pattern. Similarly <strong>the</strong> average cluster size<br />
over time for both cases is shown in Figure 9.<br />
We computed matrices for energy consumption and delays for mobile SANETS.<br />
Figure 11 shows <strong>the</strong> increased energy consumption for updating <strong>the</strong> paths that are<br />
lost due to <strong>the</strong> mobility in <strong>the</strong> network as opposed to energy consumed for static<br />
networks. The update as explained before comprises <strong>of</strong> only one ’Hello’ message<br />
exchange between any two sensors that are part <strong>of</strong> <strong>the</strong> multi-hop path to-wards <strong>the</strong><br />
optimal actuators. It is worth mentioning that <strong>the</strong> delay has been considerably low,<br />
7
40<br />
35<br />
30<br />
Actor 1<br />
Actor 2<br />
Actor 3<br />
Actor 4<br />
Actor 5<br />
Number <strong>of</strong> Sensors<br />
25<br />
20<br />
15<br />
10<br />
5<br />
0<br />
0 100 200 300 400 500 600 700 800 900 1000<br />
Time (in sec.)<br />
Figure 7: Cluster Densities avg_speed = 1-m/s, pause_time 20s, Tx_range = 11 m, distance d (between any<br />
two nodes) = 10 m<br />
as shown in Figure 10 even in <strong>the</strong> course <strong>of</strong> mobility due to <strong>the</strong> periodic update<br />
<strong>of</strong> multi-hop paths. The total network energy consumed during <strong>the</strong> lifetime <strong>of</strong> <strong>the</strong><br />
SANETs is considerably low with ADP compared to previous proposed coordination<br />
protocols.<br />
4 Conclusion and Future Work<br />
For sensor-actuator coordination, <strong>the</strong> proposed ADP can acquire a promised QoS<br />
for time stringent traffic at <strong>the</strong> cost <strong>of</strong> optimal energy consumption. For actuatoractuator<br />
coordination, <strong>the</strong> proposed unified framework using AODV can be exploited<br />
by different applications to always select <strong>the</strong> best networking paradigm<br />
that is possibly available according to <strong>the</strong> sensed information and to perform <strong>the</strong><br />
necessary operations in an efficient way. The issues related to node-failures and<br />
mobility are well handled by <strong>the</strong> periodic exchange <strong>of</strong> ’Hello’ messages in all deployment<br />
scenarios. The energy consumption for obtaining new paths in-case <strong>of</strong><br />
failure is also energy efficient and is driven by <strong>the</strong> collaborative nature <strong>of</strong> SANETs.<br />
For <strong>the</strong> moment, we are working on <strong>the</strong> design <strong>of</strong> a delay-energy sensitive<br />
routing protocol to work in collaboration with our ADP, to make <strong>the</strong> architectural<br />
and coordination platform sound for <strong>the</strong> future <strong>of</strong> SANETs. We are also working<br />
on MAC optimization based on <strong>the</strong> proposal for <strong>the</strong> efficient utilization <strong>of</strong> sensorenergy<br />
and to satiate <strong>the</strong> real-time problems experienced so far in <strong>the</strong> SANETs.<br />
References<br />
[1] M. Farukh Munir and F. <strong>Filali</strong>, A Novel Self Organizing Framework for SANETs, to appear in Proc, <strong>of</strong><br />
8
40<br />
35<br />
30<br />
Actor 1<br />
Actor 2<br />
Actor 3<br />
Actor 4<br />
Actor 5<br />
Number <strong>of</strong> Sensors<br />
25<br />
20<br />
15<br />
10<br />
5<br />
0<br />
0 100 200 300 400 500 600 700 800 900 1000<br />
Time (in sec.)<br />
Figure 8: Cluster Densities avg_speed = 2-m/s, pause_time 10s, distance d (between any two nodes) = 10 m<br />
<strong>the</strong> 12th European Wireless Conference (EW’06), April 2006.<br />
[2] V. Paruchuri, A. Durresi, and L. Barolli, “Energy Aware Routing <strong>Protocol</strong> for Heterogeneous WSNs”, in<br />
Proc. <strong>of</strong> DEXA, August 2005.<br />
[3] Q. Fang, J. Gao, and L. J.Guibas, “Locating and Bypassing Routing Holes in Sensor Networks” in proc.<br />
<strong>of</strong> IEEE Infocom March 2004.<br />
[4] J-H. Chang, and L. Tassiulas, “Maximum Lifetime Routing in WSNs”, IEEE/ACM Transactions on Networking,<br />
Vol. 12, No. 4, August 2004.<br />
[5] Akyildiz, I.F, Kasimoglu, I.H, “sensor and actor networks: research challenges”. Article Ad Hoc Networks,<br />
Vol. 2, No. 4, pp 351-367, October 2004.<br />
[6] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, “WSNs: a survey”. Article Computer Networks,<br />
Vol. 38, No. 4, pp 393-422, March 2002.<br />
[7] T. Melodia, D. Popili, V. C. Gungor and I. F. Akyildiz, “A distributed coordination framework for WSANs:<br />
in Proc. <strong>of</strong> MobiHoc, May 2005.<br />
[8] W. Hu, N. Bulusu, and S. Jha, “A Communication Paradigm for Hybrid SANs”, International Journal <strong>of</strong><br />
Wireless Information Networks, Vol. 12, No. 1, pp 47-59, January 2005.<br />
[9] E. Yoneki, J. Bacon, “A survey <strong>of</strong> WSN technologies: research trends and middleware’s role” Technical<br />
Report, No. 646, September 2005.<br />
[10] V. P. Mhatre, C. Rosenberg, D. K<strong>of</strong>man, R. Mazumdar, N. Shr<strong>of</strong>f, “A Minimum Cost Heterogeneous<br />
Sensor Network with a Lifetime Constraint”, IEEE Transactions on Mobile Computing, Vol. 4, No. 1, pp.<br />
4-15, January 2005.<br />
[11] Xiang Ji, and Hongyuan Zha, “Sensor Positioning in Ad-hoc WSNs Using Multidimensional Scaling”, in<br />
Proc. <strong>of</strong> IEEE Infocom, March 2004.<br />
[12] Yi Shang, and Wheeler Ruml, “Improved MDS-Based Localization”, in Proc. <strong>of</strong> IEEE Infocom March<br />
2004.<br />
[13] G. Sarma, and R. Mazumdar, “Hybrid SNs: A Small World”, in proc. <strong>of</strong> MobiHoc, May 2005.<br />
9
30<br />
av.speed = 2-m/s, pause-time = 10s<br />
av.speed = 1-m/s, pause-time = 20s<br />
25<br />
Number <strong>of</strong> Sensors<br />
20<br />
15<br />
10<br />
5<br />
0<br />
0 5 10 15 20 25 30 35 40 45 50<br />
Time (in sec.)<br />
Figure 9: Average Cluster Size for Mobile SANET Configuration Vs. Time<br />
[14] A. Cerpa, J. L. Wong, M. Potkonjak, and D. Estrin, “Temporal Properties <strong>of</strong> Low Power Wireless Links”,<br />
in Proc. <strong>of</strong> MobiHoc, May 2005.<br />
[15] I. F. Akyildiz, M. C. Vuran and O. B. Akan, ” On Exploiting Spatial and Temporal Correlation in WSNs”,<br />
in Proc. <strong>of</strong> IWWAN, May 2004.<br />
[16] H. Gupta, V. Navda, S. R. Das, and V. Chowdhary, “Efficient Ga<strong>the</strong>ring <strong>of</strong> Correlated Data in Sensor<br />
Networks” in Proc. <strong>of</strong> MobiHoc, May 2005.<br />
[17] G. Xing, C. Lu, Y. Zhang, Q. Huang, and R. Pless, “Minimum Power Configuration in WSNs”, in Proc.<br />
<strong>of</strong> MobiHoc, May 2005.<br />
[18] F. Kuhn, T. Moscibroda, and R. Wattenh<strong>of</strong>er, “Initializing Newly Deployed Ad Hoc and Sensor Networks”,<br />
in Proc. <strong>of</strong> MobiHoc, May 2004.<br />
[19] J. L. Bredin, E. D. Demaine, M. T. Hajiaghayi, and D. Rus, “Deploying Sensor Networks with Guaranteed<br />
Capacity and Fault Tolerance”, in Proc. <strong>of</strong> MobiHoc, May 2005.<br />
[20] B. Liu, P. Brass, O. Dousse, P. Nain, and D. Towsley, “Mobility Improves Coverage <strong>of</strong> Sensor Networks”,<br />
in Proc. <strong>of</strong> MobiHoc, May 2005.<br />
[21] The Network Simulator - ns (version 2.27). http://www.isi.edu/nsnam/ns/.<br />
10
4.5<br />
4<br />
3.5<br />
3<br />
Delay (in ms)<br />
2.5<br />
2<br />
1.5<br />
1<br />
Static-Case, 100 Nodes, 5% <strong>Actuator</strong>s<br />
Mobile-Case, 100 Nodes, 5% <strong>Actuator</strong>s<br />
0.5<br />
0<br />
0 20 40 60 80 100<br />
Sensors in <strong>the</strong> Network<br />
Figure 10: Delay Comparison for Static and Mobile SANET Configuration<br />
0.006<br />
0.0055<br />
Energy Consumption (in joules)<br />
0.005<br />
0.0045<br />
0.004<br />
0.0035<br />
0.003<br />
0.0025<br />
0.002<br />
ADP for Static SANET, no Updates<br />
ADP for Mobile SANET, updates and path-recovery<br />
0.0015<br />
0.001<br />
0 20 40 60 80 100<br />
Sensors in <strong>the</strong> Network<br />
Figure 11: Energy-Consumption for Static and Mobile SANETs<br />
11