An Enhanced Framework for Improving Spatio-Temporal Queries for Global Positioning Systems

International Journal of Research in Computer Science
eISSN 2249-8265 Volume 3 Issue 2 (2013) pp. 11-21
www.ijorcs.org, A Unit of White Globe Publications
doi: 10.7815/ijorcs. 32.2013.061

AN ENHANCED FRAMEWORK FOR IMPROVING SPATIO-
TEMPORAL QUERIES FOR GLOBAL POSITIONING
SYSTEMS
Ishraq Al-Fataftah1, Jalal Atoum2
12
Princess Sumaya University for Technology, Amman, JORDAN
Email: 1ishraqabd85@hotmail.com, 2atoum@psut.edu.jo

Abstract: To efficiently process continuous spatio- or as an added value by mobile operators and service
temporal queries, we need to efficiently and effectively providers. Social networking tools such as Facebook,
handle large number of moving objects and continuous Twitter, Foursquare and Google+ also have helped in
updates on these queries. In this paper, we propose a spreading LBS and in helping users to recognize the
framework that employs a new indexing algorithm that benefit gained from such services when compared to
is built on top of SQL Server 2008 and avoid the traditional ones.
overhead related to R-Tree indexing. To answer range
queries, we utilize dynamic materialized view concept LBS has been developed from traditional
to efficiently handle update queries. We propose an navigation and tracking systems to advanced social
adaptive safe region to reduce communication costs services, safety and security services, payment and
between the client and the server and to minimize billing, advertisements, gaming and entertainment
position update load. Caching of results was utilized to services. Navigation systems nowadays not only can
enhance the overall performance of the framework. To show you the route from place A to place B but also
handle concurrent spatio-temporal queries, we utilize will walk you through the desired route turn by turn
publish/subscribe paradigm to group similar queries showing the exact distance and time left to reach the
and efficiently process these requests. Experiments exact place. Novel search capabilities such as finding
show that the overall proposed framework the location of your friends, nearest gas station and
performance was able to outperform R-Tree index and finding public libraries in the route you are driving in.
produce promising and satisfactory results. Modified search options based on user locations have
minimized the search domain to provide more precise
Keywords: Enhanced Framework, GPS, Spatio- and useful information to users upon request.
Temporal Queries, SQL Server.
Moreover, context aware advertisements tailored
I. INTRODUCTION based on user location, speed, direction or past history
are gaining more success since these advertisements
The ascendance of mobile devices and advances in are becoming more useful and helpful for users [6].
wireless technology nowadays has layered the
These are some examples on how LBS can facilitate
landscape for the spread of services provided based on
and ease our lives. Statistics are showing that 35%
the geo-location of users. Whereas location based
from mobiles around the world are smart phones, 47%
services are not new, the early adaptors in the 90’s
from all phones will be smart phones by 2015, 95%
would have no clue on how important location data
from smart phones users are looking for location data,
would be and the impact it imposes in our daily life.
LBS users reached 63 million in 2009 and expected to
Location based services (LBS) are services that
grow in 2015 and reach about 468 million users [9].
provide geographic information and geo-processing
Such statistics indicate the increased awareness toward
power to mobile users based on their current locations
LBS.
or to stationary (non-moving) users according to the
stationary object/mobile of their interest [5]. Although studies are showing that smart phone
users are not yet reaching the critical mass neither
The emergence of positioning technologies those using LBS, the importance of data in the digital
especially global positioning systems (GPS) have world we are living in is increasing and turning into
revolutionized the mobile market and had a great the number one commodity in this world. Data
impact in enhancing location based services. The level provided in terms of LBS are continuously changing in
of accuracy produced by GPS technology has time and location and the need to provide the right data
encouraged different players in the mobile market to in the right time and location is being critical for such
invest in GPS capabilities whether as built-in services. For instance, if searching for the nearest
technology in smart phones by mobile manufacturers hospital in times of emergency are not delivered in

www.ijorcs.org

12 Ishraq Al-Fataftah, Jalal Atoum
near real time and with high degree of accuracy will advance by calculating the velocity of moving objects,
put lives into danger and turn these services into an data can be pre-computed [11].
unreliable and an inefficient facilities. This type of
queries are called spatio-temporal queries in which PLACE, Pervasive Location-Aware Computing
objects change their location and shape overtime and Environment is a framework for processing both
data can be described in both space dimension and moving queries and moving objects in addition to
time dimension [4]. Any delay in these queries will keeping track of the history of these objects [1] and
yield an invalid or obsolete result. [3]. What distinguishes PLACE from the other works
discussed above, is its ability to process different types
The contribution of this paper is as follows: of continuous spatio-temporal queries, the use of
incremental updates to evaluate new results, scalability
− We propose a framework that better utilizes to support concurrent execution of queries through
location based services resources in highly dynamic sharing the execution of similar queries by grouping
environments and supports dynamic queries for them based on certain criteria and no processing is to
moving objects. be performed on the client side. Yaun and others [12]
− We propose an algorithm for spatial indexing that had suggested a framework for handling past and
enhances the performance and response time of present/future position of objects in the disk storage
spatio-temporal queries. and the memory separately. By using a PCFI and PCFI
− We utilize safe region concept by proposing an + index structure for tracking moving objects, past and
adaptive safe region that helps in minimizing present/future queries are processed efficiently in disk
communication between client and server side. and memory.
− We provide a comprehensive set of experiments
Mokbel and others [13] had proposed an algorithm
that show the performance of the overall framework
called SINA: Scalable Incremental Hashed based
and how it is more efficient than just using R-tree
algorithm to process concurrent spatio-temporal
indexing in SQL Server 2008.
queries. The purpose of this algorithm is to efficiently
The rest of the paper is organized as follows: process a large number of concurrent queries using an
Section 2 highlights main related work for handling incremental evaluation. Scalability was achieved by
processing of spatio-temporal queries. Section 3 using a shared execution paradigm by a spatial join
demonstrates a quick overview about spatio-temporal between moving objects and moving queries.
queries. Section 4 proposes a new framework for
handling range queries. Section 5 provides an III. SPATIO-TEMPORAL QUERIES
extensive set of experiments to assess the performance
Spatio-temporal queries are queries in which
of the proposed framework. Finally, section 6
objects change their location and shape overtime in
concludes the paper and presents directions for future
which data can be described in both space dimension
work.
and time dimension [1,3,4]. These queries are a
II. RELATED WORK variation from spatial queries and temporal queries.
Spatial queries are queries that are initiated based on
Spatio-temporal databases have been the subject of the position or location of objects, while temporal
various researches. For instance the queries are queries initiated in temporal databases in
CHOROCHRONOS is a European research project that which we are interested only in the current state of the
studied spatio-temporal databases. Advances in several object in which we only store the current state of
areas have been accomplished including query objects not past states.
processing, data modeling, indexing and architecture
for STDBMSs [7]. The main spatial queries that fall under spatio-
temporal queries are:
A lot of research work had focused on one aspect of
these queries such as the work that had focused on A. Window Queries (Range)
moving queries over stationary objects, other work had These types of queries retrieve data that intersects a
focused on stationary range queries [8,10]. In addition, certain window or region. R-Tree is an example on this
these researches had concentrated on the validity of type of queries as shown in Figure 1. In this case, we
query results when being processed, meaning that are interested only in objects that reside within the
query results are considered valid only if it is within a boundaries of the query window; any object that is
certain region such as valid region concept [8] or outside this window will not be retrieved. Continuous
within a certain period of time such as valid time [10]. range queries are queries that remain active over a
Moreover, to overcome the problem of processing period of time and are continuously evaluated to
capabilities of clients, caching techniques have been retrieve accurate answers (Stojanovic and Dordevic-
developed either in the server side or client side. If the Kajan, 2003) [5].
expected location of moving objects is known in

www.ijorcs.org

An Enhanced Framework for Improving Spatio-Temporal Queries for Global Positioning Systems 13

Figure 1: Window query using R-Tree

B. Nearest neighbor queries (NN): F. Future queries:

This type of queries retrieve objects that are near These queries are interested in predicting the future
the current location of the mobile device, as shown in location of the objects. Extra information such as the
Figure 2. In such scenario, we can cache the result of velocity of moving objects should be stored along with
this query at the client side as long as the mobile the location of each object.
device did not change its location. But mobile devices This paper handles only NOW queries in Range
are moving all the time, to overcome this issue the spatio-temporal queries.
concept of valid region was introduced. When a
mobile device issues NN query, the result returned IV. THE FRAMEWORK
include with it the region in which this result is
considered valid and the mobile determines whether to The proposed framework is built around a
issue new request or use the valid region by publish/subscribe paradigm in which moving objects
determining whether it is still in this region or it and queries are mapped together through a notification
moved outside [11]. system/engine. Figure 3 demonstrates the infra-
structure for the framework proposed. The reason for
choosing this approach is that many suggested
solutions for handling the large number of concurrent
queries is through sharing where similar queries are
grouped together based on rules extracted from each
query and same answer are returned for different query
owners such in [13, 14 ,1]. Special engines were
developed to carry out these operations. The
publish/subscribe paradigm simplifies this concept in
which similar queries are treated as subscriptions,
these subscriptions share the same interests, once a
Figure 2: KNN Query data is available (in our case upon requesting of data),
the notification engine is responsible for delivering the
C. Reverse nearest neighbor queries: results to all subscribers. The main component of the
framework is the Notification system/engine that is
It is a variation from NN queries in which it works
responsible for collecting location data from publishers
in the reverse order. These types of queries are
and subscribers, handling and processing continuous
interested in finding all objects in which q is their
queries and returning results to the subscribers.
nearest neighbor.
D. Historical queries:
These queries are interested in past location of
objects. This requires the system to store the location
of objects in different period of time. Usually, when
the object changes its location, a record is moved to
the history database and calculated from that database.
E. Now queries:
These queries are interested in the current location
of objects. This requires the system to maintain a
record for the latest position of all objects.
Figure 3: Main Framework Infrastructure

www.ijorcs.org

Let us consider an example in which we are also is built over circular regions. Moreover, research
interested in retrieving all restaurants within one papers for using the shape of safe region proved that
kilometer radius. In this scenario, mobile devices as using circular safe region has more advantages over
moving objects (when driving a car or walking) are using rectangular safe region [15]. Figure 4 depicts
interested in searching for static objects (Restaurants) the circular safe region. As long a moving object (O) is
within a range of 1K radius. This type of spatio- within the safe region (O1-Ok), no query will be issued
temporal queries is of type moving queries over static although the location of the object is changing. Once
objects in which we have built our framework around. an object is outside the safe region (Ok+1), an update
The client application installed on the mobile device location query would be issued to retrieve an updated
will provide the client with a graphical user interface set of results.
(GUI) that will display a geographical map such as
Google maps. An indicator of the client current
location will be displayed on the map that will fit the
display resolution of the mobile device. The
application will request the service providing the client
spatial coordinates obtained from the GPS embedded
within the mobile device.
Publishers in this framework are static objects we
are interested in providing their data so that others can
search for them. For instance, restaurants in our
example are publishers which register their location
and type of service provided to the notification system.
When a mobile device moves, it obtains its location
through GPS system embedded within the device and
transmits its location through a wireless
communication system to the notification engine. The
efficiency of the system is highly dependent on the
Figure 4: Circular Safe Region
wireless communication network that affects the
transmission and reception of information. We will The caching manager is responsible for caching the
disregard any implication of the wireless query result for each query region to process any query
communication network in this framework. The issued by other objects that share the same query
publishers would provide information about their region and type of service. The notification engine is
services and location information through traditional responsible for handling concurrent queries by
means of registration. grouping similar objects that share the same safe
region and type of service (Subscription model).
The framework works in the following way:
As we have mentioned earlier, the notification
1. When a mobile device issues a request for a service, engine is the core part of the system where all retrieval
a subscription ticket is issued to identify the mobile and creation operations are taken place. The design
object and its current location. decision taken for searching and retrieving spatial data
2. A circular safe region is constructed around the is done through the use of a spatial indexing structure
moving mobile device to minimize the number of over the publisher table. Our testing on actual data
location updates issued by the mobile device. showed a significant variation in the query response
3. The notification engine will process the query based time when data retrieved from tables is directly
on the location provided. compared to data retrieved with the presence of a
4. The result will be published back to the mobile spatial index which was an R-Tree index. Therefore,
device. we have proposed an algorithm that partition the
5. A cached version of the results will be stored to spatial space using a grid into equally spaced
answer any query issued and have the same query partitions, each partition is covered with a circular
results. shape to construct the structure of the index as shown
The main idea of using a safe region is that as long in Figure 5. For the uncovered areas in the intersection
as the mobile device is within the safe region, there is point between the four squares we use a filter circle to
no need to continuously issue update to the mobile include these points in the index. Each circle will
device since the result will be the same. The reason for cover a number of points that lies on its circumference
using a circular safe region is that the underlying and inside the circle; these points will be used as
Euclidian space in this framework is partitioned as parent and child nodes for constructing the index tree.
circular regions and the indexing structure proposed

www.ijorcs.org


Figure 5: Partitioned Spatial Space Grid
Spatial indexes are considered heavy and impose a holds the geo-location of the current subscribers (area
huge overhead for insertion and deletion operations. of interest), so that only updates to the underlying
As spatio-temporal queries are continuously changing tables that affect this area will be materialized.
their locations, new publishers are added to the system
and old publishers may be deleted from the system, we Caching manager is responsible for caching results
have decided to take advantage of materialized views to a cache table to answer any query that share the
to represent portion of data subscribers interested in same subscription interest and handle out of sync
rather than querying over the original tables that cases. Caching of data is triggered whenever the
maintains a huge amount of data. To overcome the moving object reaches half the range of the interests
maintainability issues for the MV, we have decided to region in order to pre-compute data and provide more
use a dynamic materialized view based on work in precise results to subscribers in a timely manner.
[16], this view is controlled by a control table that Figure 6 summarizes the proposed framework
architecture design.

Figure 6: Proposed Framework Architecture

www.ijorcs.org


A. Proposed Indexing Algorithm sub tree. When the root node is created, it is placed in
the first page. All parents then will be inserted in the
The main algorithm proposed to index the spatial same page until the pages is full. New page will be
data is presented below. In a nutshell, this algorithm created to hold subsequent nodes.
divides the spatial space into equally spaced portions,
each with an overlay circle shape that set the main When a point is involved in an insert, delete or
structure of the algorithm. For each circle, we consider search operation, parent nodes that intersects the point
points that lie on the circle circumference as parent of interest or those lie within its range are retrieved.
nodes and points that lie within the circle as child All children that satisfy the same criteria will be
nodes. The tree will hold parent nodes with children retrieved. The search relies on the location of the node
that lie on the left side of the circle as left sub tree and in the left or right side of the circle in order to filter
children that lie on the right side of the circle as right and minimize nodes to search through.

// Loop through all cells in the spatial space grid and filter circles
Foreach(cell in space grid)
{
// Retrieve only data that intersects the circle cell
Get all data that intersects cell into TempIntersectTable;
Foreach(point in TempIntersectTable)
{
// Select points that lies on circle circumference as parent nodes
IF(point lies on cell circumference)
{
// Parent node lies on the Left side of the cell
IF(point.distance <= cell center)
{
Insert record into ParentTable with LeftChild = ‘L’ AND RightChild =
NULL;
}
// Parent node lies on the Right side of the cell
ELSE
{
Insert record into ParentTable with LeftChild = NULL AND RightChild =
‘R’;
}
}
// Select points that lies inside circle as child node
ELSE
{
// Child node lies on the Left side of the cell
IF(point.distance <= cell center)
{
ChooseParent(‘L’);
Insert record into ChildrenTable with Left = ‘L’ AND Right = NULL;
}
// Child node lies on the Right side of the cell
ELSE
{
ChooseParent(‘R’);
Insert record into ChildrenTable with LeftChild = NULL AND RightChild
= ‘R’;
}
}
}
}
// Select a parent node from the parents table (nodes that lies on circle circumference)
// first node that is not full and lies on left or right side of the circle
// If no such node exists, create a dummy node that will be the parent of all nodes
Function ChooseParent(location)
{
var parentNode;
SELECT parentNode = TOP record FROM ParentTable

www.ijorcs.org

WHERE record is not full AND
(LeftChild = location OR RightChild = location);
// If such node exists, then return parentNode
IF(parentNode !=null)
{
Return parentNode;
}
// If no such node exists, create dummy node
ELSE
{
var dummyParentNode = sequential number AND LeftChild=’L’ AND RightChild=’R’;
Return dummyParentNode;
}
}

B. Dynamic Materialized View within Area (A1) and requested all data within range
(R), this information will be located in the control
For retrieving only portion of data that each table. When the query is processed, the MV will return
subscriber is interested in and lies within the specified only locations that lie within R range or distance from
range, a dynamic materialized view will be created. A1 instead of searching the entire publishers table. The
The materialized view will be created over the indexed script below demonstrates how to create dynamic
publishers table and will retrieve data that only exists materialized view.
within the control table. For instance, if a subscriber is

// Create Control Table
CREATE TABLE [dbo].[Control](
[ObjectID] [int] IDENTITY(1,1) NOT NULL,
[Time] [datetime] NOT NULL,
[GLocation] [geography] NOT NULL

// Create Dynamic Materialized View
CREATE VIEW V_SpatialView WITH SCHEMABINDING AS
SELECT P.locId,
P.city,
P.country,
P.postalCode,
P.region
FROM dbo.Publishers P
WHERE EXISTS
(SELECT 1
FROM dbo.[Control] ctrl
WHERE ctrl.GLocation.STDistance(P.GEOG) <=@Range)

3. Continuous maintenance operations as a result of
The control table used in this design is of type
introducing new publishers or removing existing
Range Control table. This table supports range
ones are efficiently materialized resulting in
predicates for range or point queries such as
enhanced maintenance and re-indexing operations.
calculating distance from the point (STDistance),
determine within a point within specific area C. Caching Manager
(STWithin), retrieving all points that intersects a
When a query is made against the notification
certain point (STIntersects), etc.
engine, the caching manager check the cache table for
Using dynamic materialized views inherits the an answer for this query, if the cache is empty
following advantages: indicating this is the first request by the object, the
cache manager will answer the query from the
1. Only data that satisfies the client request will be dynamic indexed materialized view. A record of the
created as an actual table stored in the database to new object will be inserted in the control table and the
enhance query processing and data retrieval. query is retrieved from the MV. After the answer is
2. Spatial Index will be created against the MV and returned to the object, the caching manager populates
will be efficiently utilized against partial set of data the cache table with the answer of the issued query.
instead of entire spatial table. This will ensure that subsequent requests made from
the subscribers to the same service and same safe

www.ijorcs.org

region will be answered directly from the cached properties have a direct impact on the query processing
version of the results. operations including search, indexing operations and
caching. This should be kept in mind since better
Moreover, when the moving object reach half the
physical properties provides better results.
distance of the range specified in the search query, the
caching manager will be triggered to issue an
automatic query that updates the control table with a B. Testing Results
new location, execute the query against the MV and We conducted a number of experiments to measure
populate the cache table with new data. This will the performance of the proposed solution to enhance
ensure that search results will automatically be range queries in spatio-temporal database. We will
executed and available with new and more precise first evaluate the proposed spatial index algorithm we
data, the client will need only to issue the first request have developed in this framework for search
and send an indication that the object has reached half operations.
the way and there is no need for the client to issue
subsequent update requests to maintain data. We ran the test over different sample of data series
(300,000, 1,000,000 and 1,300,3000 records) one
V. PERFORMANCE EVALUATION hundred consecutive times. As shown in Figure 7, the
time taken to execute the queries measured in
In this section, we will present an evaluation milliseconds (ms) shows an acceptable execution time
of the performance of the proposed framework to even for large data size. In addition, the algorithm
assess its efficiency in comparison with R-Tree maintained a stable behavior when executed repeatedly
indexing approach implemented in SQL Server. We overtime; the same results were obtained when we ran
have created a simulation environment for testing the the algorithm more than a hundred times.
proposed framework using SQL Server 2008 R2. We
ran a series of experiments to evaluate each component
of the framework as a single unit.
A. Preliminaries
All experiments were performed on a workstation
with 2.53 GHz Core 2 Duo processor, 4 GB memory
and one 500 GB disk running Windows 7 operating
system. We have used for testing real sample spatial
database for world cities and countries that is
downloaded from (MaxMind).
The performance of queries response time was
measured using the following properties:
Figure 7: Proposed Index CPU Time (Search)
1. Dataset size: the number of objects that lie within a
specific range to determine the speed of indexing To assess the efficiency of our proposed algorithm,
operations and data retrieval and transfer rate. In we have compared its performance against a spatial
addition, the maintenance operations on the table with the application of R-Tree index as shown in
materialized view are highly dependent on data Figure 8. It can be shown that a significant
size. We have considered data sizes within a range improvement have been obtained when compared to a
between 300,000 and 1,300,000 objects. table with no index. The reason is that, applying an
2. Range/Window size: this property has a great index on the table where data stored in a well-defined
impact on the number of objects covered by each data structure significantly improved the data retrieval.
node in the indexing structure and on how much Moreover, storing the data in a tree based on a circular
data can be returned. We have considered window shape where points are classified as parent and child
size of 5 and 10 Kilometers. nodes based on their location on the circumference or
3. Type of query: we have considered Range spatio- inside a filter circle enhanced data storage in physical
temporal queries that are concerned with NOW pages and data retrieval CPU time. Also, it can be
queries. noticed that the performance of the R-Tree is not stable
when executed repeatedly over time in contrast to our
4. Number of concurrent queries: The publish/
proposed algorithm that maintained a stable behavior
subscribe paradigm and caching is used to handle
when executed repeatedly over time. This is due to the
multiple concurrent queries over the same area of
points were grouped based on a spatial geometry that
interest.
enclosed points within a specific region in a well-
Moreover, the client machine physical properties defined tree and sub-tree where no overhead for
including CPU speed, memory size and other handling overlapped regions is necessary.

www.ijorcs.org

view. Materialized views reside in disk where views
are generated on the fly in main memory. The
execution plan of the three scenarios will show that
fully materialized views are shown as single lookup
whereas dynamic materialized view will be shown
with two branches one for the control table and the
other for the underlying table.
Figure 11 depicts the advantage gained for using a
dynamic materialized view over fully materialized
view and regular view when executed several times for
different location requests. Using a control table that
contains the current location of the moving objects and
creating a materialized view that calculates the range
Figure 8: Proposed Index CPU Time query based on the current location of the moving
object produce more enhanced results even when the
We ran a series of experiments to measure the location of the moving object was continuously
performance of insertion and deletion operations using updated.
the proposed indexing algorithm in Figures 9 and 10,
respectively. We can notice that these operations does
not produce any overhead since the resulting tree does
not need to be balanced so no additional computation
is needed. In addition, the simplicity in selecting
parent nodes and creating new sub-trees gave our
indexing algorithm advantage over R-Tree index
especially in the deletion operation.

Figure 11: Search query response time using different views

Figure 12 demonstrates the performance of the
framework from issuing the first request through the
entire process compared with R-Tree index. We can
notice that both methods register the same
performance for the first request then we can notice
how the continuous change in the moving object
location affected the R-Tree index performance where
the response time increased overtime whereas our
proposed framework outperformed the R-Tree index
where we noticed an enhanced response time overtime
Figure 9: CPU Time (Insert) which is the purpose of this thesis.

Figure 10: CPU Time (Delete)
Figure 12: Proposed Framework CPU Time
To demonstrate the effectiveness of using dynamic Finally, we have measured the I/O statistics for
materialized view, we have compared the behavior of both the R-Tree index and our proposed framework.
the dynamic MV against a fully MV and a regular

www.ijorcs.org

Table 1 demonstrates I/O statistics obtained from both
operations including average and total I/O operations
(average and total number of IO operations), average
and total logical reads (average and total number of
pages read from data cache), average and total logical
writes (average and total number of pages written to
cache data), average physical reads (average number of
pages read from disk) and total physical writes (total
number of pages written to disk). The proposed
algorithm outperformed the R-Tree index in terms of
I/O operations as indicated in Figure 13 as well.
Figure 13: Proposed framework and R-Tree I/O operations.

Table 1: I/O operations for R-Tree Index and the Proposed Framework

Total Avg Total Avg Total
Total Avg Logical
Count Avg I/Os Logical Logical Logical Physical Physical
I/Os Reads
Reads Writes Writes Reads Reads

R-Tree Index 100 342 34291 342 34291 0 0 0 0
Proposed
100 258 14165 256 14163 2 2 0 0
Framework

VI. CONCLUSION AND FUTURE WORK Due to limited resources and battery life of moving
objects, we intend in future work to move the
This paper had addressed the challenge of handling
calculation of the half distance in a safe region from
range queries for moving queries over static objects in
the moving object to the server side. By incorporating
location based services. We proposed a framework that
future queries with this approach, our framework will
addressed the need to efficiently and effectively
replace pull queries to push queries where our
answer continuous queries for concurrent users at the
framework will be able to predict the future location of
same time. Concurrent query execution is handled by
the moving object based on speed and velocity and
utilizing the publish/subscribe paradigm. Moreover, an
push new results to the client automatically. Finally, in
adaptive safe region is introduced to reduce the
term of practice, we had implemented our framework
overhead of issuing continuous updates to the location
and spatial index structure in SQL Server as temp
of the moving object. Efficient spatial index over
tables in order to simulate the real behavior of spatial
dynamic materialized views is proposed to reduce the
index. One future work would be to implement our
response time for evaluating the continuous queries. A
index structure as part of the DBMS especially in an
caching manager is proposed to handle update requests
open source DBMS such as MySQL.
from the safe region and answer all subsequent
requests from concurrent users. This approach had
VII. REFERENCES
outperformed the R-Tree index when executed several
times over time in terms of CPU performance time and [1] Mohamed F. Mokbel, Xiaopeng Xiong, Moustafa A.
I/O operations which had enhanced the overall Hammad, Walid G. Aref، “Continuous Query
Processing of Spatio-temporal Data Streams in
response time for spatio-temporal queries.
PLACE”, Geoinformatica V.9, pp.343-365, 2005. doi:
There is a number of promising directions we can 10.1007/s10707-005-4576-7
extend for future work based on research work [2] Z.W. Yuan, M. Cheng, H.S. Kim, H.Y. Bae and J. W.
presented in this paper. For the proposed indexing Ge, “Querying Spatial-temporal Data in Location-Based
structure, it was built on the concept of range queries Services”, 2005.
that is directly dependent on the window/range size of [3] Mohamed F. Mokbel, “Continuous Query Processing in
the circular regions built on top of the spatial space. Spatio-temporal Databases”, Springerlink, p 364-367,
One direction would be testing this index structure on 2005. doi: 10.1007/978-3-540-30192-9_10
nearest neighbor queries and work on enhancing its [4] Mohamed F. Mokbel, Walid G. Aref, Susanne E.
structure if needed. Other direction would be to work Hambrusch, Sunil Prabhakar, “Towards Scalable
on history and future queries since the current Locationaware Services: Requirements and Research
framework is developed to handle now queries. The Issues”, GIS '03 Proceedings of the 11th ACM
promising thing in this approach is that the framework international symposium on Advances in geographic
information systems, pp.110-117, 2003. doi:
is easily extendable to support these queries since the
10.1145/956676.956691
publish/subscribe paradigm can incorporate any type
of queries.

www.ijorcs.org

[5] D. Stojanov and S. Dordevic–Kajan, “Modeling and [12] Z. Yuan, M. Cheng, H. Kim, H. Bae and J. Ge,
Querying Mobile Objects in Location–based Services”, Querying Spatial-temporal Data in Location-Based
FACTA UNIVERSITATIS (NI¡S), Ser. Math. Inform. Services, 2005.
18 (2003), 59–80. [13] Mohamed F. Mokbel, Xiaopeing Xiong, Walid G. Aref,
[6] Comarch technology, Technology Review (Comarch “SINA: Scalable Incremental Processing of Continuous
no.2), Poland, 2009. Queries in Spatio-temporal Databases”, In Proceedings
[7] M. Koubaraki, Y. Theodoridis and T. Sellis, “Spatio- of the 2004 ACM SIGMOD international conference on
temporal Databases in the Years Ahead”, Spatio- Management of data (SIGMOD '04), ACM, pp.623-
Temporal Databases, Lecture Notes in Computer 634, 2004. doi: 10.1145/1007568.1007638
Science Volume 2520, pp.345-347, 2003. doi: [14] Xiaopeng Xiong, Mohamed F. Mokbel, Walid G. Aref,
10.1007/978-3-540-45081-8_9 S.E. Hambrusch, S. Prabhakar, “Scalable spatio-
[8] Jun Zhang, Manli Zhu, Dimitris Papadias, Yufei Tao, temporal continuous query processing for location-
Dik Lun Lee, “Location-based Spatial Queries”, In aware services”, IEEE, Proceedings 16th International
Proceedings of the 2003 ACM SIGMOD international Conference on Scientific and Statistical Database
conference on Management of data (SIGMOD '03). Management, pp.317-326, 2004. doi:
ACM, pp.443-454, 2003. doi: 10.1145/872757.872812 10.1109/SSDM.2004.1311223
[9] Blur Marketing (2011, October), http://blur- [15] Shengsheng Wang, Chen Zhang, “A Dynamic Interval
marketing.com/blog/trends-and-statistics-in-location- Based Circular Safe Region Algorithm for Continuous
based-services/ Queries in Moving Objects”, International Journal of
Communications, Network and System Sciences, 2011.
[10] Baihua Zheng, Dik Lun Lee, “Semantic Caching in doi: 10.4236/ijcns.2011.45036
Location-Dependent Query Processing”, Springer,
Advances in Spatial and Temporal Databases, 97-113, [16] Jingren Zhou, Per-åke Larson, Jonathan Goldstein,
2001. doi: 10.1007/3-540-47724-1_6 Luping Ding, “Dynamic Materialized Views”, IEEE,
23rd International Conference on Data Engineering,
[11] Dik Lun Lee, Manli Zhu, Haibo Hu, “When Location- 2007. doi: 10.1109/ICDE.2007.367898.
Based Services Meet Databases”, ACM, Mobile
Information Systems, vol.1, issue 2, pp.81-90, 2005.

How to cite
Ishraq Al-Fataftah, Jalal Atoum, "An Enhanced Framework for Improving Spatio-Temporal Queries for Global Positioning
Systems". International Journal of Research in Computer Science, 3 (2): pp. 11-21, March 2013. doi: 10.7815/ijorcs.
32.2013.061

www.ijorcs.org

An Enhanced Framework for Improving Spatio-Temporal Queries for Global Positioning Systems

Recommended

More Related Content

What's hot (13)

Viewers also liked (9)

Similar to An Enhanced Framework for Improving Spatio-Temporal Queries for Global Positioning Systems (20)

More from IJORCS (20)

Recently uploaded (20)

An Enhanced Framework for Improving Spatio-Temporal Queries for Global Positioning Systems