MULTI-ROBOT SENSOR RELOCATION TO ENHANCE CONNECTIVITY IN A WSN

International Journal of Ad hoc, Sensor & Ubiquitous Computing (IJASUC) Vol.7, No.2, April 2016
DOI : 10.5121/ijasuc.2016.7201 1
MULTI-ROBOT SENSOR RELOCATION TO
ENHANCE CONNECTIVITY IN A WSN
Sahla Masmoudi Mnif and Leila Azouz Saidane
National School of Computer Science, University of Manouba, Manouba, Tunisia
ABSTRACT
Ensuring connectivity in Wireless Sensor Networks (WSN) is a challenging issue, especially in hazardous
areas (like battlefield). Many applications of WSN require an important level of connectivity in the network
to detect a given event (like detection Intrusion) and forward it to the ”sink” node in order to alert users.
For these risky areas the deterministic deployment is not usually guaranteed and the network is composed
by a set of disconnected Islands. We present in our work two strategies to relocate sensors in order to
improve the connectivity using mobile Robots. These two solutions are called Multi-Robot Island-based
Relocation (MRIBR) and Multi-Robot Grid-Based Island-based Relocation (MRGIR). Through several
simulations, we show that MRGIR outperforms MRIBR. Our study can be used especially to make a trade-
off between the number of deployed sensors and the numbers of the used mobile robots, according to the
quality needed for the application.
KEYWORDS
Sensor, Wireless Sensor Network, Mobile Robot, Redundant, Connectivity, Relocation
1. INTRODUCTION
Wireless Sensor Network (WSN) applications have been suggested for many applications,
especially in risky and hazardous regions. The main objective of these applications is to detect
any abnormal event in the area of interest. One relevant application is detection intrusion in a
battlefield and coast and border surveillance. We remind that a WSN is composed of a set of
small entities called “sensors,” each sensor is characterized by a limited energy and is constrained
in its computation and communication resources. When a sensor detects a phenomenon, in the
given area, the deployed sensors have to communicate together in order to forward this event to
the “sink” node. The sink node will alert the user by the detected event. An illustrative
application is military applications especially for detection intrusion in battlefield or frontier
control. The deployment in these risky areas cannot be assisted by human, and a random
deployment is necessary. Random initial deployment leads in most of cases, to a portioned
network. In each network partition redundant nodes can appear. However the connectivity must
be usually guaranteed to survey correctly the controlled zone. We propose to use mobile robots to
enhance the connectivity in the network using redundant nodes. The main role of the mobile
robots is to discover the network topology and to connect the disconnected parts of the network
simultaneously. In addition, robots must coordinate their movement and cooperate during
operation.

2
We propose in this work two strategies for redundant sensors relocation using mobile robots in
order to improve the connectivity in a wireless sensor network. The main difference between the
proposed strategies is the functioning and the movement adopted by mobile robots to achieve
their mission.
2. RELATED WORK
In recent years, sensor relocation has been a challenging matter that was studied by many
researchers. Several solutions have been proposed to solve the redeployment issue. One relevant
solution was to provide motion capability to all sensors. This way, the sensors can move and
relocate themselves in order to adjust the topology and achieve the connectivity and/or the
coverage. The sensors must synchronize their movement to enhance the network topology.
Among the proposed solutions, we mention particularly the cascade motion which is detailed in
[2]: instead of moving directly to the target, the sensor nodes adopt a cascade movement which
means that the nearest node to the target point will move there, and the location of nearest node is
replaced by moving another sensor and so on. Virtual Forces Aspect has been also proposed as a
solution for sensor relocation. In this way, deployed sensors communicate together and compute
their new locations in order to ensure connectivity and/or coverage. Then these sensors exerce a
repulsive or an attractive force to move to their estimated locations. This strategy was studied and
presented in [18]. The mobility of nodes is very efficient and improves the network topology, but
it requires an important energy consumption which causes the node depletion and decreases the
network lifetime. Other solutions consist of the use of fixed sensor nodes and the network is
assisted by some ”actors” like mobile robots. Some studies, proposed to use the robot to carry
data between disconnected sensors so that robot collects the detected event from nodes and then
delivers these information for the other nodes. This approach is presented by [19]. In this way the
event is delayed and a latency time is introduced which can be considered as a shortcoming for
critical applications.
Another set of related works include algorithms using DATA MULES [14][15], which are
wireless devices integrated on mobile entities (animals, vehicles). A DATA MULE is a data
collector; it picks up data from nodes and relays it to other nodes. So that, data would not be
relayed on long routes and the network lifetime is increased In other proposed solutions, the
actors are advanced mobile sensors that exploit the redundant nodes and relocate them to achieve
better connectivity and /or coverage trying to preserve the network lifetime as long as possible. In
recent years many researches have been interested by sensor relocation to enhance the
connectivity in the WSN.
3. PROPOSED SOLUTION
In our work, we consider particularly a WSN over a risky or unreachable region of interest. We
suppose that the WSN is containing a large number of static sensors which are deployed in the
considered region in a random manner (eg, dropped from an airplane). This deployment leads
generally to the partition of the network, so that the connectivity between nodes is not
guaranteed. We suppose that the position of the sink node is fixed and known. Each sensor has a
communication range noted by rc and a sensing range noted by rs.
We note that, all deployed sensors have the same connectivity range and the same sensing range.
Upon deployment, the position of each sensor is fixed and known by the use of a localization

3
technique (eg GPS). In our proposed solution, mobile robots are used to relocate redundant nodes
in order to enhance the network connectivity and as a consequence the coverage will be
improved. A mobile robot is a mobile platform which has a large amount of energy and can be
recharged when needed. The robot is also characterized by an important computation capability
and has a communication range noted by Rc. Each mobile robot handles a number of fixed
sensors that can be used to heal connectivity holes. Each robot knows the frontiers of the
controlled area.
After a random deployment over the WSN, the total connectivity between all sensor nodes is not
guaranteed. The resulted network is partitioned. When an event is detected, it may be not
delivered to the sink node. The formed network is composed by a set of disconnected islands,
we call island every set of connected sensor nodes. The Island containing the sink node is called
“MainIsland”. To ensure the connectivity in the WSN, we should connect islands to the
MainIsland. Generally, each Island contains redundant nodes.
3.1. Redundant sensors Identification
To identify redundant nodes, we use hexagonal pavement of the region, sensors belonging to two
adjacent cells are able to communicate. A sensor is said to be redundant if its cell (perception
zone) is covered by other cells.
3.2. Identification of Island-Head
For each Island in the network, an Island-Head is elected. The main role of the Island-Head is to
collect information concerning all sensor nodes belonging to the Island (position of each sensor,
position of redundant nodes, number of redundant nodes, and number of deployed nodes…). To
elect an Island-Head we define criterion described by this equation
where Eres and Emax represent respectively the residual energy and the maximum level of energy
for a given node. Nbn and Nbnodes refer to the number of neighbours of a sensor node and the
numbers of nodes in a given Island. The node with the highest value of f is elected as an Island-
Head. Meaning that, node with the highest amount of energy and having the largest number of
one hop neighbours is chosen as an Island-Head. In case of multiple candidates, the node with
higher Identity is elected. A backup Island-head is chosen to replace Island-Head in case of its
depletion. When the Island-Head is elected, it collects the positions of redundant nodes and these
nodes go to the passive mode (sleeping mode) to save energy of the whole network.
In our solution, we propose to use some robots to relocate redundant sensors and connect Islands
to the MainIsalnd. For this purpose, used robots should communicate together and synchronise
their movements and functions. We propose two strategies for robots to improve the connectivity
in WSN. We call the first one “Multi-Robot Island-Based Relocation” and we note it by
“MRIBR”, in which robots moves is based on Islands. The second strategy is called “Multi-
Robot Grid-Based Island-Based Relocation and is noted by “MRGIR”, with this strategy, in
addition to the island-based model, the controlled area is portioned according to a grid.

4
4. MULTI-ROBOT ISLAND-BASED RELOCATION
In this solution, we use mobile robots to redeploy sensors in order to improve the connectivity in
WSN. Upon random deployment, the islands are formed and the Island-Heads are elected as
described in the previous section. We remind that, mobile robots have no idea on the network
topology. Nevertheless, each robot is aware by the position of the “sink node”. Robots should
communicate to connect islands to the “MainIsland”.
To achieve their goal, each mobile robot can be in one of the three states:
• Free: the robot walks in the controlled area with a random manner.
• Busy: the robot places and relocates sensors in order to connect an island to the
“MainIsland”.
• Topology Discovery: Periodically, the robot stops (after a distance of 2* Rc and sends a
Hello-Robot Message. Each sensor receiving a ”Hello-Robot”, forwards this message to
its ”Island-Head” and the ”Island-Head” replies with ”Island information” message
containing all information concerning this island (position of nodes, positions of
redundant sensors, sensors identities, number of redundant nodes...).
Each robot starts its work with Topology Discovery state, after sending Hello-Robot message, the
robot waits for a given period.
o If the robot does not receive any reply, it continues its walk in a random direction and
stays in Topology Discovery state.
o If the robot receives an ”Island-information”,
• If the encountered Island is the “Mainland”, the robot memorizes all information
concerning the “MainIsland” and stays in the Topology Discovery state.
• If the encountered Island is an ordinary Island, the robot computes the position of
the nearest node of the ”sink” and then it calculates the number of needed sensors
to connect the island to the “MainIsland”.
• If this requested number of sensors is available on the robot, the robots changed
its state to busy state. It then relocates them (the nodes will be relocated
according to hexagonal pavement).
• If this requested number of sensors is not available on the robot, it continues its
walk in a random manner.
To identify a robot we use an Identity number (ID-Robot) which is assigned to each robot to
make distinction between all robots. Each robot maintains a sequence number for each visited
Island which will be increased after every new visit to an Island.

5
When a robot places sensors to connect an Island to the “MainIsland”, the robot should verify if
the locations of the sensor is still empty by sending “Hello-Robot” Messages.
After connecting a given Island to the “MainIsland”, all sensors belonging to this Island must
mark their membership to the “MainIsland”.
When a mobile robot encounters redundant sensor nodes, it picks up these sensors unless its
maximum capacity is not reached. We note also that when a mobile robot encounters another one,
it can ask the encountered robot, if needed and possible, to give it some redundant nodes.
5. MULTI-ROBOT GRID-BASED ISLAND-BASED RELOCATION
In this solution, we maintain the concept of Islands and our goal is also to connect all the Islands
to the “MainIsland”. We propose also to divide the controlled area vertically to zones as
described in Figure 1.
Each robot is placed on a zone and knows the frontiers of its zone. Each mobile robot is
responsible of its zone and is not allowed to leaves its zone.
The identification of redundant nodes is made in the same manner like in MRIBR technique and
the placement of nodes in the controlled area is made according to hexagonal pavement.
Figure 1: Network Model
When the sink node belongs to only one zone and when the position of the sink node is known by
all mobile robots, we define Reference Points (RP) which are fictive points replacing sink node in
each zone as described in Figure 1.
Each robot in its zone will function as described in the above section but by replacing the sink
node by RP. Hence a mobile robot will discover its zone and tries to connect discovered Islands
to RP Points. The mobile robot chooses its nearest RP to connect Islands.

6
For the intermediate zone the robot should also connect its two RP, if they are not already
connected.
Reference Points can be used as points of collect (collection of fixed senors); a mobile robot can
place sensors in these points when the robot has a large number of redundant sensors exceeding
its needs (greater than a predefined threshold).
6. PERFORMANCE EVALUATION
Our proposed solution is implemented under NS2 simulator using MannaSim packages. Several
simulations were established with different scenarios. For all simulations we use a large number
of deployed sensors to ensure full connectivity and coverage over the network. The sensors are
initially deployed randomly through a square ROI, we set rc = rs = 25m and Rc = 60m. We set the
dimension of the Region of Interest to 600*600. The number of deployed sensor nodes is set to
150 sensors in the first time. In a second step, we will vary the number of deployed sensors from
50 to 450 sensors.
To evaluate and compare our proposed strategies MRIBR and MRGIR, we fix a set of metrics
like connectivity time, the average travelled distance by a robot and the connectivity rate.
6.1. Connectivity Time
We note by CT this criterion. Connectivity Time is defined as the needed time to ensure
connectivity in the network. This criterion should be minimized. We present the connectivity
Time according to the number of mobile robots as described in figure 3 and then we present this
connectivity according to the numbers of sensors as described in Figure 2.
Figure 2: Connectivity Time/Numbers of sensors

7
Figure 3: Connectivity Time/Numbers of robots
Figure 2 and figure 3 show that the curves have the same shape. We remark that, the connectivity
time decreases when the number of sensors and the number of robots increases. This is explained
by when the number of robots increases, the tasks are realized by more than one robot
minimizing the total connectivity time. We remark also that from a given number of robots (4
robots in figure 3), the connectivity time starts to stabilize. We remark also that MRGIR
outperforms MRIBR in term of connectivity Time.
6.2. Average travelled distance
The Average Travelled Distance is the average distance travelled by a mobile robot after
performing the algorithm. It is giving by the following equation
Total Travelled distance by all robots
Number of robots
Figure 4 shows the average travelled distance by a mobile robot according to the number of
sensors. We remark that the travelled distance decreases when the number of deployed sensors
increases. This can be explained by the fact that the number of holes decreases when the number
of deployed sensors increases. Figure 4 shows also that MRGIR outperforms MRIBR.

8
Figure 4 : Average Travelled Distance/Numbers of sensors
Figure 5 : Average Travelled Distance/Numbers of robots
Figure 5 shows the average travelled distance by a mobile robot according to the number of used
mobile. We remark that the travelled distance decreases when the number of used. But this
numbers starts to be static from a certain threshold.. Figure 5 shows also that MRGIR
outperforms MRIBR.
6.3. Connectivity Rate
The connectivity rate (CR) is he average of connected sensors in the network, this metric should
be maximized to enhance the performance of the tested algorithm. CR is giving by the following
equation:

9
Number of connected sensors to the MainIsland
Number of deployed sensors
We present the connectivity rate according to the number of mobile robots as described in figure
5 and then we present this connectivity according to the numbers of sensors as described in
Figure 6.
Figure 5: Connectivity Rate/Numbers of robots
Figure 6: Connectivity Rate/Numbers of sensors

10
We remark that the connectivity rate increases when the number of robots and the number of
sensors increases too. We note also from these figures that MRGIR outperforms MRIBR in terms
of connectivity rate.
6.4 Results Interpretation
Based on the three fixed criteria we show that MRGIR outperforms MRIBR, this can be
explained by the importance of the use of hybrid solution mixing the Island concept with the
Grid-based partition of the controlled area. For MRGIR strategy, the controlled area is divided
into rectangular zones, each zone is handled by one mobile robot. This robot is responsible by
connecting all deployed nodes in its zone. This functioning reduce the responsibility of a mobile
robot comparing to MRIBR strategy in which the movement of robot is not limited to a given
zone and is more random.
Our study can be used especially to analyse the trade-off between the number of robots to use and
the quality of connectivity. Hence, according to the application we can use our configuration.
For example, the needed rate of connectivity for a precision agriculture application is not the
same for a military application in which the total connectivity must be guaranteed.
From this study, we can also determine the optimal number of mobile robots to achieve a given
connectivity time or rate.
Hence through this study we can fix the characteristics of our application like the needed
connectivity time, the minimum connectivity rate and then we can determine the optimal number
of deployed sensors and the optimal number of used robots.
7. CONCLUSION
Ensuring and maintaining connectivity in WSN is a challenging issue especially for WSN in
Hazardous and risky areas like military applications (battlefield survey, intrusion detection). In
this paper, we deal with a WSN deployed with a random manner. This deployment leads to
partition of the networks and a set of disconnected Islands is formed and redundant sensors are
identified.
We propose to use many mobile robots in order to relocate redundant sensors and connect all
nodes in the network. We present two techniques in our work, the first one is an Island-based
technique (MRIBR) and the second one is Grid-based Island-Based strategy (MRGIR), here we
used the concept of islands and we divided into virtual grid, the movement of each robot is made
considering this grid. We show through extensive simulation that MRGIR outperforms MRIBR
in term of connectivity time, connectivity rate and the average travelled distance, this result can
be explained by the fact that in MRGIR strategy, the controlled area is divided in a grid and each
robot is responsible by one zone.
The movement of each robot is limited in one grid. Nevertheless, in MRIBR strategy, the
movement of robot are more random. We notice that our study can be used especially to make a
trade-off between the number of deployed sensors and the numbers of the used mobile robots for
each kind of application. We can also fix the number of needed robots for a given connectivity

11
rate and a given connectivity time depending on the type and the specifities of the needed
application.
REFERENCES
[1] R.Kershner, The number of circles covering a set, American Journal of Mathematics, p665-671,1939.
[2] T.La Porta, G.Wang, G.Cao and W.Wang, Sensor Relocation In Mobile Sensor Networks, Infocom
2005.
[3] F.Wang, M.Thai, D.Du, on the construction of 2-connected virtual backbone in wireless networks,
IEEE Transactions on Wireless Communications 2009.
[4] A.Abbasi, M.Younis and K.Akkaya, Movement-assisted connectivity restoration in wireless sensor
and actuator networks. IEEE Transactions on Parallel Distributed Systems 2009.
[5] G.Wang, G.Cao and T.Laporta, A Bidding Protocol for Deploying Mobile Sensors, The 11th IEEE
International Conference on Network Protocols (ICNP) 2003.
[6] G.Wang, G.Cao and T.Laporta, Movement-assisted sensor deployment, Infocom 2004.
[7] G.Fletcher, X.Li, A.Nayak and I.Stojmenovic, Carrier-based sensor deployment by a robot team,
IEEE SECON, 2010.
[8] Y.Mei, C.Xian, S.Das, Y.C.Hu, Y.H.Lu, sensor replacement using mobilerobots, computer
communication 30(13) 2007.
[9] D.Xuan, Z.Yun, X.Bai, S.Kumar and Ten H.Lai, Deploying Wireless sensors to achieve both
coverage and connectivity, Mobile Ad Hoc Networking and Computing, 2006.
[10] X.Li and N.Santoro, ZONER: a zone-based sensor relocation protcol for mobile sensor networks,
IEEE WLN, 2006.
[11] X.Li and N.Santoro and I.Stojmenovic, Mesh-based sensor relocation for coverage maintenance in
mobile sensor networks, UIC 2007.
[12] A. Howard, M. J. Mataric and G. S. Sukhatme, Mobile Sensor Networks Deployment Using Potential
Fields: A Distributed, Scalable Solution to the Area Coverage Problem, the 6th International
Symposium on Distributed Autonomous Robotics Systems, 2002.
[13] E.Egea-Lpez, J.Vales-Alonso, A.S.Martnez-Sala, P.Pavon-Mario, J. Garca-Haro, Simulation Tools
for Wireless Sensor Networks, Summer Simulation Multiconference - SPECTS, 2005.
[14] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam and E. Cayirci, A Survey On Sensor Networks, IEEE
communications Magazine, 2002.
[15] T.Watteyne,Using Existing Network Simulators for Power-Aware Self- Organizing Wireless Sensor
Network Protocols, INRIA, 2006
[16] G.Chalhoub, Rseaux de capteurs sans fil, Clermont Universit,2009. [17] A.Gallais, J.Carle,
D.Simplot-Ryl , La k couverture de surface dans les rseaux de capteurs, AlgoTel, 2007.

12
[18] X.Wang, Sh.Wang, D.Bi, Virtual force-directed particle swarm optimization for dynamic deployment
in wireless sensor networks, ICIC, 2007.
[19] W.Zhao, M.Ammar, and E.Zegura, A message Ferrying Approach for data deliveryin sparse mobile
adhoc networks, Mobihe3uoc 2004.

MULTI-ROBOT SENSOR RELOCATION TO ENHANCE CONNECTIVITY IN A WSN

More Related Content

What's hot (18)

Similar to MULTI-ROBOT SENSOR RELOCATION TO ENHANCE CONNECTIVITY IN A WSN (20)

Recently uploaded (20)

MULTI-ROBOT SENSOR RELOCATION TO ENHANCE CONNECTIVITY IN A WSN