Solving Sensor Void Problem in Uncontrolled Mobile Sensor

Networks

Development of Mobile Sensor Database Systems

Soo Seok Park

, Chong Sok Lim

and Soon J. Hyun

Department of Computer Science, KAIST, 373-1 Guseong-dong, Yuseong-gu, Daejeon 305-701, Republic of Korea

Department of Information and Communication Engineering, KAIST, Yuseong-gu, Daejeon 305-701, Republic of Korea

Keywords: Sensor Database System, Mobile Sensor Networks, Query Language, Database.

Abstract: Due to the random movement of sensor nodes in uncontrolled mobile sensor networks, the distribution of

sensor nodes may not be uniform over the area of network coverage. Querying a target space in which

mobile sensors are unavailable at the time of the data request will lead to no return of the result. In this

paper, this problem is referred to as sensor void problem, which leads to difficulties in data management and

application design activities. We address the sensor void problem and offer solutions with a location-aware

time-constrained query processing strategy. The key idea of the proposed strategy is to allow mobile sensor

nodes to keep track of time validity and target location matching checks for the execution of the queries

received from the base station. Programmers can specify a certain degree of time sensitivity in their queries

according to their applications’ semantics. We implemented the proposed mobile query processing scheme

into our previously developed sensor database system. A performance evaluation shows how the proposed

query processing strategy effectively handles the sensor void problem with various parameters of an

uncontrolled mobile sensor network.

1 INTRODUCTION

With the rapid technical advancements in sensor

hardware designs and wireless communications, a

variety of smart sensing functions are increasingly

equipped virtually in all types of objects, including

smart phones, vehicles, wearable devices, unmanned

flying objects, and others. A wireless sensor network

(WSN) is composed of a large number of sensor

nodes to monitor the physical world. Based on the

mobility of the sensor node, there are two types of

WSNs, the stationary sensor network and the mobile

sensor network (MSN). Application services using

the former include disaster management, precision

agriculture, health care and traffic management.

Recently, a variety of new application services, such

as land, ocean and air exploration and monitoring;

automobile applications; habitant monitoring; and a

wide range of other scenarios have been developed

(De Zoysa et al., 2007; Nittel et al., 2007). They

demand data collection from mobile sensor nodes

(i.e., sensors on moving objects) and form a new

class of WSN known as a mobile sensor network

(MSN).

There are a few different types of mobility in MSNs

(Di Francesco et al., 2011). We focus on the type of

MSN known as an uncontrolled mobile sensor

network, which is composed of randomly moving

mobile sensor nodes. This type is illustrated in

Figure 1. In this type of mobile sensor network,

sensors are typically attached to autonomous objects,

such as vehicles, unmanned flying objects, or smart

phones. A few recent studies tried to leverage these

types of sensor networks to support environmental

monitoring and urban monitoring applications

(Abdelzaher et al., 2007; Campbell et al., 2006).

The typical problem in uncontrolled MSNs is,

due to their random movement, that there may exist

some target regions (e.g., grid cells) which do not

have any sensor node to execute the given data

collection query at the time of the application

request. We refer to this situation as sensor voidance

in a region. From time to time, any cell may become

a sensor void region in an uncontrolled MSN, as

sensor nodes move randomly. In other words,

although a cell region may be a sensor void at the

time of querying, some sensor nodes may come into

Seok Park S., Sok Lim C. and J. Hyun S..

Solving Sensor Void Problem in Uncontrolled Mobile Sensor Networks - Development of Mobile Sensor Database Systems.

DOI: 10.5220/0005228400470055

In Proceedings of the 4th International Conference on Sensor Networks (SENSORNETS-2015), pages 47-55

ISBN: 978-989-758-086-4

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

the target region and thus execute the associated

query.

Figure 1: Mobile sensor network with randomly moving

sensor nodes.

In this paper, we address the sensor void

problem and offer a solution with a location-aware

time-constrained query processing strategy. The

proposed technique aims to maximize the

probability of data acquisition from the target region

for a given query. The key idea of the proposed

strategy is that it allows mobile sensor nodes to

check and keep track of the time validity and target

location matching for the execution of the queries

received from the base station. Programmers can

specify a certain degree of time sensitivity in their

queries according to their applications’ semantics.

The benefits are two-fold. First, this would give

applications a greater probability of a data return

from the designated target cell region by minimizing

the rate of data no-returns. Second, by avoiding

possibly unnecessary query executions and the

collection of incorrect sensor data, mobile sensor

nodes may reduce their battery consumption levels.

In our location-aware time-constrained query

processing strategy, the overall procedure can be

briefly described in phases, as follows. First, the

base station broadcasts the application’s queries to

all mobile sensor nodes in the sensor network.

Second, sensor nodes store them into a query table if

the current location does not match the target region

of the query. Third, mobile sensor nodes check the

time validity and target matching for all of the

queries stored in its query table. This comparing

procedure is triggered in the sensor node whenever

the node moves into a new cell region. Then, the

node in the target region will finally fire a matching

query and transmit the sensor data to the base

station.

The remainder of this paper is organized as

follows. We introduce the sensor database system

and its major related works in Section 2. In Section

3, we propose a location-aware time-constrained

query processing technique to manage the sensor

void problem of an uncontrolled mobile sensor

network. A performance evaluation of our work is

given in Section 4, and the conclusion of this paper

is given in Section 5.

2 DATA MANAGEMENT IN

MOBILE SENSOR NETWORKS

In WSNs, database management is an important

issue when coping with the excessive and yet mostly

useless amount of streaming data and the energy

drain of battery-operated sensor nodes. In sensor

databases, sensor data are collected by querying the

target sensors and the sensors of the target region

within the sensor network. There have been many

sensor database systems designed with various query

languages and query processing strategies for a wide

range of sensor network applications services

(Diallo et al., 2013; Madden et al., 2004).

Recently, database management in mobile sensor

networks (MSNs) has attracted a considerable

amount of attention which has motivated a wide

range of new applications (De Zoysa et al., 2007;

Nittel et al., 2007). The mobility of sensor nodes is a

major issue in MSNs (Zeinalipour-yazti and

Chrysanthis, 2009). Valkanas et al. (2001) designed

declarative queries for an in-network data analysis,

which can be efficiently optimized to allow for the

automatic deployment of executable code for a

mobile setting. They presented a watchdog as a

future direction; this is a software module that tracks

sensor network changes. Andreou et al. (2011)

proposed SenseSwarm, a framework for the

acquisition and storage of spatio-temporal events in

MSNs. SenseSwarm detects physical phenomena

using a swarm of sensor nodes that are dynamically

organized in the perimeter and core nodes. The two

sensor database systems above assume sensor nodes

with controlled mobility or sensor nodes with low

mobility. In contrast to the above mobile sensor

database systems, our work looks into MSNs in

which mobile sensors have fully random mobility

and where application queries seek sensor data from

any target region.

A few recent studies leveraged human-carried or

vehicle-mounted mobile sensor networks using

short/mid-range radios (e.g., ZigBee, Wi-Fi,

SENSORNETS2015-4thInternationalConferenceonSensorNetworks

Bluetooth) for the purpose of data collection to

support a wide range of new applications

(Abdelzaher et al., 2007; Campbell et al., 2006).

Database management in uncontrolled MSNs will

frequently encounter the absence of sensor nodes in

the target area at the time of the application’s query

request. This sensor void problem results in a no-

return (or empty return) of data to the application.

The sensor void regions are never predetermined

and may exist randomly over modeled cell regions.

Thus, if some sensor nodes near a sensor void cell

region hold requested query for a certain period of

time, some of them may happen to enter the sensor

void target cell region and hence execute the query.

No existing works conducted an investigation of this

subject by assuming that some mobile (or fixed)

sensor nodes exist in all cell regions (Diallo et al.,

2013; Madden et al., 2004).

3 LOCATION-AWARE

TIME-CONSTRAINED QUERY

IN SNQL

3.1 Time Window Operation

The situation of sensor voidance in uncontrolled

MSNs is a prevalent problem due to the random

mobility of the mobile sensor nodes. It will result in

no return of the sensor data. In an effort to manage

this problem, we designed query operations and

processing techniques, known collectively as a

location-aware time-constrained query processing

strategy, as an extension to our previous

development of a sensor network query language

and processing system (Changbai et al., 2008; Lim

et al., 2014). Below, we briefly explain how this

strategy manages query processing tasks which take

place in a sensor void target region.

The application specifies the time validity in

terms of a time window in its sensor database query,

and the query remains valid at sensor nodes during

the specified time. Even if no sensor nodes are

currently available at the time of the application’s

query request in the target region (i.e., sensor

voidance), the query will be given a chance to be

fired when any mobile node enters the target region

and transmits the sensor data back to the application.

To do this, the query processor will use both the

location information of the mobile nodes and the

time specification facility of the query language

according to the query processing procedure.

Figure 2 shows the EBNF format of the location-

aware time-constrained query expression added to

our previous SNQL construct. The application

specifies the types of sensor data to collect in the

SELECT clause (1); is specifies the information of

the target region in the FROM clause (2); and it sets

several conditions of the sensor data with a WHERE

clause (3).

Figure 2: Time-window query in SNQL.

The target region can be defined as any shape

using the spatial specifications and sophisticated

spatial operations in SNQL (Lim et al., 2014). The

spatial operator in SNQL is based on openGIS

(Herring, 2011), which describes a common

architecture for simple feature geometry. SNQL

supports spatial assignment facilities for specifying

irregularly shaped regions by name, and it provides

spatial operators such as the union, intersection and

minus operators to extend the expressiveness of a

query language.

The COLLECTION VALID FOR clause

specifies the time validity of a query in terms of the

time window (4). A time window can have one or

more time intervals, and a time interval is expressed

by the lower-time bound and upper-time bound (5).

In our present implementation of the location-aware

time-constrained query processing strategy, we

employ only the data type yyyy (year), MM

(month), dd (day), hh (hour) and mm (minutes) (6).

The NOW parameter denotes the time the query is

issued. Other temporal expressions and operations

can be utilized for better support of time window

operations, such as after, before, intersect,

difference, and union (Mkaouar et al., 2011). The

design of temporal operations is beyond the scope of

this paper.

<time-constrained-query> ::=

SELECT

<select-attribute-list>

(1)

FROM

(2)

[WHERE

](3)

[WITHIN <percentage> | CASE WHEN … THEN … ]

COLLECTION VALID FOR <time-window> (4)

<time-window> ::=

<time-interval> | <time-window> <time-interval>

<time-interval> ::= (5)

<lower-time-bound> | NOW,

<upper-time-bound> | NOW

<lower-time-bound> ::= <time>

<upper-time-bound> ::= <time>

<time> ::= <yyyy-MM-dd hh:mm> (6)

SolvingSensorVoidProbleminUncontrolledMobileSensorNetworks-DevelopmentofMobileSensorDatabaseSystems

3.2 Query Processing Technique

In this section, we introduce our query processing

scheme, which efficiently handles the sensor void

problem in uncontrolled MSNs. We describe the

procedure of the proposed location-aware time-

constrained query processing technique from the

query dissemination to the data collection stages.

3.2.1 Network Setup

Query routing and data aggregation methods in

stationary WSNs are typically found based on

various types of routing trees. In MSNs, on the other

hand, such routing trees can scarcely be constructed

and maintained due to the mobility of the sensor

nodes, such that the network topology is very likely

to be transient (Zeinalipour-yazti and Chrysanthis,

2009). The routing in MSNs can be modeled in

consideration of the scalability and energy

efficiency. The two-tier network model is known to

be feasible for both scalability and energy efficiency

(Gupta and Younis, 2003). The main theme of a

two-tier network model is clustering. In a cluster,

there is a cluster head and a number of member

nodes. The cluster head is responsible for data

collection and aggregation from other member

nodes. Cluster heads communicate with each other

for data aggregation and to route the sensor data to

the base station. Our network model uses the two-

tier architecture and divides the entire sensor

network into grid cell regions as uniquely

identifiable query targets. Our network model is

described below.

• All the sensor nodes are equal and have mobility,

and they move randomly within the sensor

network space  . They have maximum

transmission range (MTR) which can directly

communicate with other neighboring sensor

node.

• Mobile sensor nodes are location-aware using

GPS or some localization mechanisms (Bulusu et

al., 2000). The location data will be evaluated

against the target data of the query whenever the

node enters a new cell region of grid.

• Sensor nodes are aware of the entire cell regions

of the network space , given as 2



×2



; where

k=log



(). This information is stored as meta-

data.

• The base station is stationary and its MTR covers

the overall sensor network space .

• A target cell region of a query is equal to one cell

region of the grid.

Different types of mobility can significantly

impact the query processing phases in MSNs (Di

Francesco et al., 2011; Zeinalipour-yazti and

Chrysanthis, 2009). In general, the mobility is

modeled in three categories: the deterministic,

partially deterministic, and fully random categories.

Deterministic mobility implies that the movement of

sensor nodes is predefined or can be perfectly

scheduled. The partially deterministic mobility

implies that the mobility patterns of mobile nodes

can be predicted by analyzing the traces of sensor

nodes. In the random mobility model, on the other

hand, the mobility of sensor nodes is entirely

unpredictable and is independent of the previous

movement and, thus, the sensor void situation is

most likely to arise. Modeling our uncontrolled

MSNs, we assume a random mobility model, in that

the sensor nodes randomly select the direction,

speed and travel time (Johnson and Maltz, 1996).

The density of sensor nodes is an important

parameter for the sensor void problem. The sparser

they are, the more probable the sensor voidance

problem becomes. We presume that high degrees of

sparseness prevail in most real-world applications

such that applications will encounter a high

probability of the sensor void problem. We

conducted an experiment to learn their co-relations,

as outlined in Section 4. It should be noted that the

density will not be even among the cell regions due

to the random movement of sensor nodes and that

the difference in the density among the cell regions

may change from time to time.

Figure 3: Sensor network topology model.

Our network model uses a two-tiered

architecture, as shown in Figure 3, and each cell

region may have a head node or none, and each may

have zero or several member nodes. A cell region

head (RH) is responsible for the collection and

aggregation of sensor data from the cell region

members (RM). We employ the scheme for head

selection introduced in earlier work (Nasser et al.,

2012). In that the mobility factor (1) is introduced to

select an RH, each node keeps track of its mobility

Region member

Base station

Region head

SENSORNETS2015-4thInternationalConferenceonSensorNetworks

and records the number of movements it has made as

well as the amount of residual energy it has. A node

may change its cell region as a result of movement

and join a new cell region as an RM if there is an

RH in the new cell region. If the cell region is

empty, the node becomes an RH in the new cell

region. The number of times a node has moved over

cell regions and the number of location changes

along with the remaining energy are used to

calculate the mobility factor (M.F) of the mobile

nodes, as given below.

.  =















. 



>0 (1)





is the total number of region changes, 



the total number of moves made during t seconds,

and 



is the remaining energy. Each node keeps a

record of its mobility factor. A lower value of the

mobility factor indicates that the node is less mobile

and more of a candidate to become the RH. In

contrast, a higher value depicts frequent movements

and renders a node inconsistent as a RH. The region

head selection procedure starts at each node by

broadcasting its M.F. This broadcast is intended for

the members of the same region and is discarded by

others. Initially, each node keeps its own M.F as the

region head M.F. Once a broadcast is received, the

node compares the region head M.F with the one

received. If the received value is lower than the

current value, the region head M.F and region head

identifier are appropriately updated. At the end of

the broadcast phase, each node has knowledge of the

node with the lowest mobility factor; hence, the

node is considered as the region head. The lowest

value of the M.F ensures that the node will serve as

the region head for a longer duration. Therefore, the

number of procedures for the head selection process

can be minimized.

In our network model, there are three different

sensor node states as shown in Figure 4: idle,

sensing, and discovery. A state transition is triggered

by a cell region change of a sensor node or by the

execution of a query. When an RM leaves its current

cell region, it changes its status to discovery and

broadcasts a discovery message to the RH of the

new cell region. If it receives an acknowledgement

from the existing RH of the new cell region, it sends

its reference information, including its node

identification number and network information to

the RH. On the other hand, if the incoming sensor

node does not receive any acknowledgement of the

RH discovery message, this indicates that there is no

RH in the new cell region. Thus, it must take the role

of the RH of the new cell region. When an RH

leaves its current cell region, it broadcasts a beacon

to notify its members of its status. The RM after

receiving the broadcast changes its status to

discovery and elects a new RM following the same

procedure defined in the ‘region head selection’

procedure described above. The RH is in charge of

the collection and aggregation of the sensor data

from the RM. The sensor node is in sensing status

during the query execution phase and changes its

status to the idle status after finishing the execution.

Figure 4: State transitions of a mobile sensor node.

3.2.2 Query Processing Technique

Based on the uncontrolled MSN topology and head

election strategies given in Section 3.2.1, we

propose a location-aware time-constrained query

processing technique. The proposed query

processing strategy is composed of three phases: (1)

query dissemination, (2) query execution, and (3)

data return. The key idea is to give the mobile sensor

nodes the capability of holding queries for certain

durations that are specified in the time window of

the queries. As mentioned earlier, the time window

can be programmed by an application designer using

the SNQL query language construct for MSNs, as

described in Section 3.1.

1) Query dissemination: For a query request

from an application, the base station broadcasts the

query to the MSNs. We assume that the maximum

transmission range (MTR) of the base station is wide

enough to cover the entire sensor network space. All

of the sensors in the network receive the transmitted

query and check its target specification and the

validity of the time in the time window specification.

If the target cell region of a query matches the

current location of the sensor node, the RH

immediately executes the query and aggregates the

sensor data from the RMs. Otherwise, sensor node

stores it into its query table (QT) for later

evaluations. Figure 5 shows an example description

of a QT.

Idle

Discovery

(Head selection)

Sensing

(Query processing)

Cell region change

SolvingSensorVoidProbleminUncontrolledMobileSensorNetworks-DevelopmentofMobileSensorDatabaseSystems

Figure 5: Example of a query table (QT).

The stored queries in the QT are evaluated for

possible execution in the matching target cell region

at some time. By doing so, the probability of the

sensor void problem (i.e., the data no-return rate)

can be greatly diminished. The mobile sensor node

checks the queries in the QT when it enters another

cell region.

2) Query execution: When a sensor node enters a

new cell region, it checks whether there is an RH in

the new region. If there is an RH, the sensor node

registers itself to the RH. If not, the sensor node by

itself serves as the RH in the new region and checks

the time validity of the query by comparing its time

window specification and the current time. As the

result, all of the timed-out queries will be deleted

from the QT, after which target region matching is

conducted by comparing the target data (i.e., the

region ID in the query table) and the current location

in the cell region just entered. When the time

validity and region matching are found to be true,

the query is executed for data collection. Figure 6

shows how the aforementioned query processing

procedure is implemented.

Figure 6: Execution of queries in the QT.

3) Data return: In contrast to the query

transmission procedure, the data return procedure

relies on a multi-hop fashion due to the limited MTR

of the sensor nodes. Furthermore, the random

mobility of the sensor nodes will cause frequent link

failure problems and the communication void

problems (i.e., a node with no neighbor closer to the

destination than itself) due to possible sensor void

cell regions during the data return processes. Our

data return strategy is based on a stateless

geographic routing protocol for mobile ad-hoc

networks (Karp and Kung, 2000). We designed the

data return procedure, as follows.

In our network model, only the RHs participate

in the routing process for returning sensor data to the

base station. When an RM finishes its query

execution, it sends the sensor data to the RH in the

same cell region. The RH which receives the sensor

data from the RM undertakes aggregation and starts

to send the data to the base station. When an RH as

the data holder returns sensor data to the base

station, it uses a stateless geographic routing

protocol by selecting the next data holder from

among the neighboring RHs which has the highest

positive forwarding cell region hop count (i.e., the

neighboring RH which is closest to the base station

is the next data holder). For a communication void

region in which the data holder itself has the highest

positive forwarding cell region hop count within the

maximum transmission range, it broadcasts a beacon

and holds the sensor data until there is a response

message from a neighboring RH which has a higher

positive forwarding cell region hop count.

At this stage, Figure 7 shows a situation in which

the data holder is in a communication void situation

and is moving out of the current cell region. In this

situation, the data holder (RM) sends the sensor data

to the RM in the new cell region. If the new cell

region is empty, the sensor node becomes an RH in

the new cell region. It then broadcasts a beacon and

holds the sensor data until there is a response

message from a neighboring RH which has a higher

positive forwarding cell region hop count.

Figure 7: Example of a data holder (RM) movement in

communication void situation.

Query ID Region ID Time window

0000150 754

02-09-2014 02:00

02-09-2014 03:15

0004033 367

02-08-2014 04:20

02-09-2014 04:30

.. .

. . . . . .

0000117 019

02-09-2014 02:45

02-09-2014 03:25

Algorithm

Sensor node moves to a new cell region

Check whether there is an ZH in the new cell region

if there is an ZH in the new region then

else

Take the role of an ZH in the new cell region

Check the time window of queries in the QT

if there is a timed-out query then

Delete the query from the QT

end if

Conduct target region matching of queries in the QT

if there is a matching query

(both the time validity and the region matching)

Execute the query

Remove the executed query from the QT

end if

Current sensor data holder

Region member

Base station

Region head

SENSORNETS2015-4thInternationalConferenceonSensorNetworks

4 EVALUATION

In this section, we show the experimental results of

the proposed location-aware time-constrained query

processing technique based on a grid-based two-tier

network model. Our experiment includes a

measurement of the occurrence ratio of the sensor

void problem. We also examined the ratio of

successful query executions in association with the

number of sensor nodes in a sensor network (i.e., the

node density), the velocity of the sensor nodes, and

the amount of validity time in terms of the time

window sizes.

For the experiment with the proposed location-

aware time-constrained query processing strategy,

we used MobiSim to generate a mobility trace

(Mousavi et al., 2007). MobiSim is a java-based

mobility management utility which was mainly

designed to investigate mobility issues in mobile ad-

hoc networks. We used a grid-based two-tier

network model and analyzed the mobility traces of

the sensor nodes.

Table 1: Experiment settings.

Parameters Values

Mobility pattern Random waypoint

Network size 1000 x 1000

Region size 100 x 100

Simulation(s) 10,000

Number of trials 50

Min speed(m/s) 1

Time window interval (s) Simulation dependent

Number of nodes Simulation dependent

Max speed(m/s) Simulation dependent

Table 1 shows the MobiSim experimental

parameters used here. The sensor network size was

set to 1000m * 1000m with a cell region size of

100m * 100m. We assume that the sensors have

random mobility. We set the minimum speed of the

sensor node to 1m/s with a maximum pause time of

100 seconds. In the experiment, we vary the number

of sensor nodes, the maximum speed of the sensor

node and the time window intervals. This result was

produced by averaging 50 experiments for each

parameter variable.

First, we examined the occurrence of a sensor

void situation depending on the number of sensor

nodes in the sensor network. We set a single cell

region as the target region to be queried. The target

region is randomly selected in this experiment. As

shown in Figure 8, as the number of sensor nodes

increases, the ratio of sensor voidance decreases.

This is obvious, because the sensor void situation is

quite proportional to the average density of mobile

sensor nodes, that the size of the cell region and the

size of the network space will show similar ratios.

Figure 8: Sensor voidance ratio on network density.

Second, we examined the average time of sensor

void recovery depending on the size of the time-

window interval. This experiment shows that the

rate of successful query execution depends on the

amount of time validity specified in the time

windows, as shown in Figure 9. For this experiment,

we set the maximum speed of the sensor node to

5m/s and number of sensor nodes to 10. We verified

that a longer validity time led to a higher the rate of

successful query executions, as the time window

interval becomes large and thus the probability of

the inflow of the sensor node in the target region is

increased.

Figure 9: Query execution rate on time window intervals.

Figure 10: Average delay of a query execution on the

maximum velocity of mobile sensor nodes.

100

0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400

Rate of query execution (%)

Time window interval (sec)

SolvingSensorVoidProbleminUncontrolledMobileSensorNetworks-DevelopmentofMobileSensorDatabaseSystems

Finally, we conducted an experiment on how the

maximum velocity of mobile sensor nodes can affect

the recovery rate of the sensor void situation. We

increased the maximum velocity of the sensor node

from one to twenty and checked the average delay in

the query execution. Figure 10 shows that a faster

mobile sensor node leads to a lower average delay

for query executions. This is derived from the fact

that the sensor is more likely to flow into the target

region if the sensor has higher velocity.

5 CONCLUSIONS

In this paper, we propose a sensor database system

for MSNs with uncontrolled mobile sensor nodes. In

contrast to a stationary sensor network, the

availability of sensor nodes in the target region is not

guaranteed in a mobile sensor network, and the

random mobility issue gives rise to difficulties in

sensor database management, which is referred to as

the sensor void problem in this paper. In this paper,

we have proposed a location-aware time-constrained

query processing technique which is highly effective

for handling the sensor void situation of sparse

MSNs with uncontrolled mobile sensor nodes. We

have demonstrated the proposed query processing

procedure in operational phases for query

dissemination, execution, data collection and

aggregation, and data return. Our experiments show

that various mobility parameters are correlated with

the occurrence rates of sensor void situations.

Finally, we plan to develop our query processing

strategy further and implement more functions into

our mobile sensor network database management

system.

ACKNOWLEDGEMENTS

This work was supported by ICT R&D program of

MSIP/IITP. [B0101-14-0334, Development of IoT-

based Trustworthy and Smart Home Community

Framework]

REFERENCES

Abdelzaher, T., Anokwa, Y., Boda, P., Burke, J. A.,

Estrin, D., Guibas, L., Reich, J., 2007. Mobiscopes for

Human Spaces. Pervasive Computing, 6(2), pp.20–29.

Andreou, P., Zeinalipour-Yazti, D., Chrysanthis, P. K.,

Samaras, G., 2011. In-network data acquisition and

replication in mobile sensor networks. Distributed and

Parallel Databases, 29(1-2), pp.87–112.

Bulusu, N., Heidemann, J., Estrin, D., 2000. GPS-less low-

cost outdoor localization for very small devices.

Personal Communications, IEEE, 7(5), 28-34.

Campbell, A. T., Eisenman, S. B., Lane, N. D., Miluzzo,

E., Peterson, R. A., 2006. People-centric urban

sensing. Proceedings of the 2nd annual international

workshop on Wireless internet - WICON ’06, p.18–es.

Changbai, C., Jaehyoung, L., Juyeon, H., Insung, J.,

Minsoo, K., Hyun, S. J., 2008. SNQL: A Query

Language for Sensor Network Databases. In

Proceedings of the 7th WSEAS International

Conference on Telecommunications and Informatics,

pp. 114-119.

De Zoysa, K., Keppitiyagama, C., Seneviratne, G. P.,

Shihan, W. W. A. T., 2007. A public transport system

based sensor network for road surface condition

monitoring. Proceedings of the 2007 workshop on

Networked systems for developing regions - NSDR

’07, p.1.

Di Francesco, M., Das, S.K., Anastasi, G., 2011. Data

Collection in Wireless Sensor Networks with Mobile

Elements. ACM Transactions on Sensor Networks,

8(1), pp.1–31.

Diallo, O., Rodrigues, J. J. P. C., Sene, M., Lloret, J.,

2013. Distributed database management techniques

for wireless sensor networks.

Gupta, G., Younis, M., 2003. Load-balanced clustering of

wireless sensor networks. IEEE International

Conference on Communications, 2003. ICC ’03., 3,

pp.1848–1852.

Herring, E.J.R., 2011. Open Geospatial Consortium Inc .

Status : Corrigendum Category : OpenGIS ®

Implementation Standard OpenGIS ® Implementation

Standard for Geographic information - Simple feature

access - Part 1 : Common architecture.

Johnson, D. B., Maltz, D. A., 1996. Dynamic source

routing in ad hoc wireless networks. In Mobile

computing, pp. 153-181, Springer US.

Karp, B., Kung, H.T., 2000. GPSR : Greedy Perimeter

Stateless Routing for Wireless Networks. In

Proceedings of the 6th annual international conference

on Mobile computing and networking, pp. 243-254.

LIM, C. S., LEEb, J. H., PARK, M., HYUN, S. J., 2014.

Design and Implementation of Spatial Operators and

Energy-efficient Query Processing Strategy in

Wireless Sensor Network Database System.

Madden, S.R., Franklin, M.J., Hellerstein, J.M., 2004.

TinyDB : An Acquisitional Query Processing System

for Sensor Networks

1. , V(212).

Mkaouar, M., Bouaziz, R., Moalla, M., 2011. Querying

and manipulating temporal databases.

Mousavi, S. M., Rabiee, H. R., Moshref, M.,

Dabirmoghaddam, A., 2007. Mobisim: A framework

for simulation of mobility models in mobile ad-hoc

networks. In Wireless and Mobile Computing,

Networking and Communications, 2007. WiMOB

2007, pp. 82-82.

SENSORNETS2015-4thInternationalConferenceonSensorNetworks

Nasser, N., Al-Yatama, A., Saleh, K., 2012. Zone-based

routing protocol with mobility consideration for

wireless sensor networks. Telecommunication

Systems, 52(4), pp.2541–2560.

Nittel, S., Trigoni, N., Ferentinos, K., Neville, F., Nural,

A., Pettigrew, N., 2007. A drift-tolerant model for data

management in ocean sensor networks. Proceedings of

the 6th ACM international workshop on Data

engineering for wireless and mobile access - MobiDE

’07, p.49.

Valkanas, G., Gunopulos, D., Galpin, I., Gray, A. J.,

Fernandes, A. A., 2011. Extending query languages

for in-network query processing. Proceedings of the

10th ACM International Workshop on Data

Engineering for Wireless and Mobile Access -

MobiDE ’11, p.34.

Zeinalipour-Yazti, D., Chrysanthis, P. K., 2009. Mobile

sensor network data management. In Encyclopedia of

Database Systems (pp. 1755-1759), Springer US.

SolvingSensorVoidProbleminUncontrolledMobileSensorNetworks-DevelopmentofMobileSensorDatabaseSystems