Ubiquitous Computing and Communication Journal

(ISSN 1992-8424)
A Framework for Adaptive Best Path
Determination in Real World Traffic
Mohammad A. M. Asmaran, Sulieman Bani-Ahmed
IT Department, A-Balqa Applied University
Prince Abdullah Bin Ghazi Faculty Of Science and IT, Al-Balqa Applied University, Al-Salt 19117, Jordan
[email protected], [email protected]
Abstract Route planners that identify the best step-by-step driving directions between two locations have rapidly become a part of
our everyday life. However, the output quality of route planners depends critically on the current status of roads in order to calculate
the best (shortest, fastest, or cheapest) route from a starting point to a destination. Maintaining accurate and updated statistics about
the status of roads has been made possible due to the integration of navigational technologies. These technologies are the global
positioning system (GPS), database technologies such as geographic information system (GIS) and communication technology. This
paper is to propose an efficient self learning mechanism or framework to help route planners find out best routes based on the current
status of roads in addition to the static road lengths. This helps customizable best route identification. With the proposed framework,
it is possible, for instance, to choose the best route that the minimal time in addition to minimal fuel consumption, which is a very
critical factor due to the significant increase in fuel prices in the last few years. An advantage of the proposed framework is that is
does not require any extra streets equipments. Further, the proposed framework promises a system that requires minimal
administration.
Keywords Route planners; Best path algorithms; GPS; Traffic tracking; Shortest Path.

I. INTRODUCTION
Many real life applications relying on finding the best way
to reach specific destination point in a GIS map. This implies
using a best path algorithm that can find out a best path
according to certain measures [1]. Visualizing the GIS map
as a graph, these measures are usually referred to as weights
of the edges of the path, where an edge is the link between
two linked points in the GIS map. These weights actually
control guide the way that best path algorithm can run and
identify the best route. The weights represent the lengths of
the edges and, thus, the best route can be thought of as the
shortest path [2].
Depending on the criteria represented by the weights of
the edges, the best path can be viewed as the one with the
shortest overall travelling time or the one with shortest path
length. Other optimization aspects are also possible, such as
the total fuel cost. The best route identified may vary
depending on the obstacles that vehicle could face in its way
to the destination. These obstacles could be jams, accidents,
traffic lights, or maintenance operations in streets. In some
cases, these obstacles would make the shortest path in terms
of length to consume a lot of time and/or more fuel cost
when compared to following a longer alternative route [3].
Consequently, the weight of the path or the edge should be
computed according to more factors such as jams that are
one of the most affecting factors in term of time. Another
factor is the number of traffic lights in roads and the stopping
times of traffic lights. In practice, traffic lights are

UbiCC Journal, Volume 6: Issue 2

categorized in to four categories according to the way they

are controlled [4]:
A. Fixed time traffic lights
In this case traffic lights are controlled by a timer that
indicates when traffic light should allow vehicles to move or
stop. This doesn't mean that a given traffic light assign the
same open/close times for each direction on the same cross
of streets. Main streets usually assigned longer time period in
green lights and shorter time period in red lights because of
the fact that they contain more vehicles moving on. Of
course the opposite is true regarding sub-streets. In some
cases this opening/closing time may vary depending on the
time of the day in order to handle peak hours. Such type of
traffic light relies on some statistical measures that are used
by traffic agencies to assign proper opening/closing times for
each traffic light. This type of traffics is very effective in
terms of time measurements when used by best path
algorithms because that time of traffic is predictable and
deterministic [4].
B. Intelligent traffic lights
In this case traffic lights are controlled by an intelligent
control system that controls the traffic according to some
facts read from sensors mounted within the traffic. Sensors
provide some information to the control system in order to
control traffic time. For example, sensors can provide the
control system with the number of vehicles that currently
stopping in each street lane, which is an indicator of the size

814

Ubiquitous Computing and Communication Journal

(ISSN 1992-8424)
of the jam on a certain street lane. Another approach is to
use image processing algorithms to find out the jam size on a
certain traffic light. In this type of traffic lights, time can not
be predicted even if the system is connected to main traffic
control system because it depends on real time data that can't
be predicted before [4].
C. Manual traffic lights
In this type of traffic lights, traffic lights are controlled by
human operator who is sitting in a control room and watches
real traffic using cameras attached on the traffic lights. These
cameras allow him/her to determine which street(s) to be
allowed to go and for how long. The final goal is to reduce
traffic jams. This type of traffic lights is very effective in real
time traffic system because of the intelligence of people but
its main problem is that no one can expect or determine the
time of traffic light [4].
D. Multi-Control Traffic light:
In this type of traffic lights traffic agency operator would
gain the control according to some circumstances in order to
solve some huge jams or emergency situations [4].
Actually traffic agencies could use all the above types in
the same city or country which could lead to inaccurate time
calculations and predictions. Furthermore, accidents,
maintenance operations, or other obstacles would consume
unpredictable time intervals. Predicting these time intervals
is a very critical input to route planners as we said earlier.
Real time measurement of these time intervals requires using
sensors, cameras, or humans. These traditional techniques
are not usually available in every street or situation and they
may have expensive installation and maintenance costs [3].

g. cameras would provide unclear pictures in case of rain or

dust storms).
Many algorithms are proposed to forecast traffic volumes.
The major concept used in these algorithms is to use
historical data statistics in the same conditions (i.e. same
dates, times, and weather circumstance) in order to predict
what will traffic look like in future [8, 9]. The main problem
of such algorithms is that they may fail in unpredicted
situations such as accidents. Using statistical data provides
clear general view according to same conditions in previous
situations. But if some accidental situation occurs, statistics
results will be very far from actual traffic situation. In
addition to the above drawbacks, this approach implies the
administration overhead of collecting data in order to
determine accurate statistical results.
In this paper, short-term intelligent statistical view of
moving vehicles is used for efficient current edge weight
computation. The proposed framework promises more
accurate and customizable route determination. This is
achieved through using GPS systems mounted to vehicles
currently moving in a given road to report statistics that will
be later used for edge weight computation. This way, the
current experience of vehicles is used for computing up-todate edge weights. The vehicles themselves are used as
agents to improve the performance of the route identification
algorithm for the benefit of all the vehicles in the traffic
system.

II. RELATED WORK

Many algorithms are proposed to manage streets
congestions to provide efficient traffic system [5]. These
algorithms such as RFID tags approach [6]. This approach
evaluates congestion level using some sensors to evaluate
street congestion using speed facts and car counting
according to RFID sensors that sense the RFID tag attached
to each car. This approach is used to control traffic lights in
order to minimize the number of cars moving in a specific
street and, thus, minimize the overall waiting time.

Another approach has been proposed to predict travel time

interval and congestions using ANT Colony algorithms [10].
In [10] edge probabilities are represented by the time the
vehicles spend in order to go to the destination. ANT
algorithm provides efficient way to decrease time; however,
it calculates travel time using sensors attached in every path.
This implies higher cost and maintenance overheads.
Moreover, calculating edge weights is a blind process and,
thus, fails to take into consideration irregular vehicles
movements if used with real world traffic. Another use of
ANT Colony based algorithms is available in [11] which
provides more efficient way to take into consideration the
weather conditions. Weather conditions and past experience
both provide more accurate expectations. Similar to the work
in [11], accidental situations are not considered.

Another approach proposed in [7] to measure congestion

using image processing techniques. This approach is used to
control traffic lights by measuring congestions using cameras
that capture images. These images are feed to an algorithm to
be analysed in order to get car counts that are considered to
be valid measure of the congestion level. These algorithms
involve installing equipments in the streets in order to collect
data from specific locations. This approach increases the
overhead of maintenance and administration. Furthermore,
required equipments are subject to errors according to some
inefficient working conditions such as weather conditions (e.

In this paper, our proposal provides a complete solution in

order to use vehicle equipments (i.e. Global Positioning
System (GPS) Navigator). Such devices are already provided
in most of modern cars. Moreover, our proposal provides an
efficient way to consider fuel consumption in computing
edge weights. The fuel consumption factor is very important
as fuel prices increase almost every day. In our approach,
fuel consumption is not only a function of time and distance.
Rather, the type of street (be it highway or not) and the total
number of vehicle stops both affects the fuel consumption
especially in hybrid cars.

UbiCC Journal, Volume 6: Issue 2

815

Ubiquitous Computing and Communication Journal

(ISSN 1992-8424)
Another algorithm is introduced in [12] to generate cost
function using topological and historical data in order to get
best cost estimation. This algorithm assumes that statistical
data is available for various day time periods in addition to
streets topology. The performance of this algorithm is
sensitive to real time changes such as accidental situations
and is negatively affected by the lack of statistics in some
regions and streets. In this algorithm no human interaction is
needed except in defining very basic information that is
available in vehicle specification manuals and streets
connections which is obvious to be available for such
algorithms.
III. PROPOSED ALGORITHM
Computing the expected travelling time of a car in a given
road can't be computed in a very accurate way. However,
estimating this time is a very important factor in weighting
edges of the path. This paper proposes a solution in order to
find more accurate measures of the weights that could
properly guide best path algorithms in determining the best
path.
The proposed solution depends on measuring weights
according to distance, obstacles, and traffic situations. In
order to do that, obstacles would be represented as a weight
value that indicates how much the obstacle would affect the
edge weight. This value is computed according to the time.
In our proposal, travelling time is considered as an
indictor to the presence of the obstacles in the street or edge
as follows: if the moving vehicles that have already entered a
specific road are wasting a lot of time in moving through that
road, the corresponding edge in the graph is assigned a
higher weight. Travelling time of a road is highly affected by
the number and severity of obstacles in that road.
Consequently, edges with longer travelling times are to be
assigned higher weights.
We propose the following approach in order to
dynamically compute the current travelling time of roads.
Travelling time is divided into two components, namely; the
"waiting time" and "walking time" components. Waiting
time is stopping time of the vehicle. This is the total time
during which that vehicle is not moving. Practically, not
moving means that the vehicle speed is near to zero or
below a certain threshold. In our implementation, this
threshold is assumed to be 1 km/h. This threshold depends
on multiple factor and may be changed, some of which is the
nature of the country or the city that this algorithm is applied
to. The second type of time is the walking time which is the
total time during which the vehicle is moving (its speed is
above the given threshold).
A. Edge Weight Computations
Each edge has three weight values to be computed
according to the following equations:

UbiCC Journal, Volume 6: Issue 2

1) Length Weight: This weight is computed according to

the speed low so that length = (Average Speed of the Vehicle)
X (Walking Time).
2) Cost Weight: This weight is computed according to
manufacturer specifications of the vehicle according to the
following equation:
Cost = (length weight) X (Car Fuel Consumption Rate Per
Unit of Length) + (Waiting Time) X (Car Fuel Consumption
Per Stopping Unit of Time).
The above Two factors (Car Fuel Consumption Rate Per
Unit of Length and Car Fuel Consumption Per Stopping Unit
of Time) are manufacturer specifications specified in the car
specifications manual. Car Fuel Consumption Rate Per Unit
of Length has three values according to the type of street that
vehicle is moving through. These values differ for Highways,
Semi Highways, and none Highways [13, 14, 15].
3) Time Weight: This weight is computed according to the
total time that vehicle elapsed in order to go through the edge
which is Total Time = Walking Time + Waiting Time.
So, each edge is assigned three weights (Length, Cost, and
Time). If the user of the best path system needs to go through
shortest length path, best path algorithm will use length
weight for each edge. If the intended path is the lowest cost
path, cost weight will be considered, and the same in case of
lowest time where time weight will be considered. So,
weight of the edge will vary according to user need in order
to match his requirement to get best solution.
B. How to measure edge time
According to previous description regarding traffic lights
systems, it is clear that time can't be predicted, and
emergency obstacles in streets such as accidents,
maintenance workshops, etc. are not predictable either. So,
the best way is to measure the current travelling time of
individual roads. Time is measured by using two components
as we said earlier: one to measure stopping time and another
one to measure walking time. Notice that this time is
measured for the vehicles that are currently in the road. This
means that computed time values are very late for the
vehicles already in the road. However, these values will be
used by other vehicles welling to go through the same edge.
The proposed algorithm is designed to use data collected
from currently moving vehicles to weight edges and provide
guidance to any vehicle going to pass a given edge at any
certain moment.
C. Speed Measurement:
Speed can be measured using normal speed equation by
finding the difference between two vehicle locations using
GPS system and divide it by time period that it spent in order
to move between them. In fact, proposed solution considers
the use of special tracking equipments that can send car

816

Ubiquitous Computing and Communication Journal

(ISSN 1992-8424)
speed in addition to the GPS location to the server [16].
Server should compute the average speed according to the
number of reads along edge.

Table IV contains average statistical data computed every

period of time which is assumed to be in minutes (e. g. 10
minutes).

D. Algorithm Steps:
Step 1: Each vehicle should attach a processing unit that
communicates with server in order to send information such
as speed and GPS position. All map information, mainly
edges information are stored on the server as shown in Table
I.

Step4: Edge weights are computed according to Table IV

values using average stopping and walking times in addition
to the average speed. Table II data are used to calculate Cost
Weight in addition to the values computed in Table IV.

TABLE I. EDGE REPRESENTATION DATABASE TABLE

Edge ID

Highway Level

Edge ID is a unique identifier of the edge and Highway

Level is value that describes type of the edge whether it is
Highway, Semi Highway, or None Highway. Highway Level
column is used to help calculates Cost Weight.
Moreover, vehicle information should be stored such as its
fuel consumption rates as shown in Table II.
TABLE II.
Vehicle
ID

EDGE REPRESENTATION DATABASE TABLE

Highway
Semi Highway None Highway
Stopping
Fuel
Fuel
Fuel
Fuel
Consumption Consumption Consumption Consumption
Rate
Rate
Rate
Rate

Using Table II system will be able to compute fuel cost of

the edge as described before in cost weight equations where
Vehicle ID is a unique identifier of the vehicle. Consumption
rates are found in the specifications manual of the vehicle or
they can be known by experience.
Step 2: Server collects data from vehicles and registers this
information with calculated stopping and walking times
depending on the vehicle speed as described before. All this
information is stored as shown in Table III.
TABLE III.
Edge
ID

Vehicle
ID

Stop
Time

EDGE PASSING VEHICLES VALUES

Walking
Time

Average
Speed

Time

No Of
Familiar
Vehicles

As shown in Table III each vehicle passes an edge will has

a record that contains its measured statistics in addition to the
time in which this record is stored.
Step 3: Statistical data are computed every predefined
period of time and based on the Table III data and stored as
shown in the Table IV.
TABLE IV.
Edge
ID

E. Average Calculation Details:

As seen from above steps, average values should be
calculated every period of time and they are calculated based
on the values registered during last interval period only. But
in real life traffic, some vehicles will be odd in term of their
walking or stopping because of the nature of their journey or
the special conditions they have met. So some errors will
arise if their values are considered in average calculations.
These values should be ignored. In order to ignore them,
algorithm defines threshold values that must not be exceeded
between time and speed average values calculations.
Consider values as shown in Table V.

Average
Stop
Time

STATISTICS TABLE

Average
Walking
Time

Average
Speed

UbiCC Journal, Volume 6: Issue 2

Calculation
Time

TABLE V.
Edge
ID

1
1
1
1

SAMPLE STATISTICAL COMPUTATIONS

Vehicle
ID

1
2
3
4

Stop
Time

20
18
23
200

Walking
Time

100
98
105
300

Average
Speed

43
45
40
15

All values in the table look like to be close to each other

except values determined for vehicle number 4. It is clear
that this vehicle has none logical values comparing to other
vehicles values. So, its values shouldn't be considered in
average calculations. In order to do that, threshold value is
defined for every column as the following example:
Stop Time Threshold Value = 10
Walking Time Threshold Value = 15
Average Speed Threshold Value = 10
Values differences should not exceed above threshold
values for every column (i.e. Stop Time, Walking Time, and
Average Speed).
When new vehicle values are received, they are checked
across all interval vehicles. If the absolute value of the
difference doesn't exceed threshold value. No of Familiar
Vehicles column should be incremented by one for the
corresponding vehicle and new vehicle.
Initially "No Of Familiar Vehicles" should be one. At the
end, table should contain each vehicle value and the number
of familiar values related to it in the same interval. Table VI,
VII, VIII, and IX show steps calculation of the previous
example:

817

Ubiquitous Computing and Communication Journal

(ISSN 1992-8424)
TABLE VI.

FIRST VEHICLE VALUES DURING INTERVAL

Edge
ID

Vehicle
ID

Stop
Time

Walking
Time

Average
Speed

20

100

43

TABLE VII.
Edge
ID

SECOND VEHICLE VALUES DURING INTERVAL

Vehicle
ID

Stop
Time

Walking
Time

Average
Speed

20
(diff.
is 2)

100
(diff.. is
2)

43
(diff. is 2)

18

98

45

TABLE VIII.
Edge
ID

Vehicle
ID

No Of
Familiar
Vehicles
2

THIRD VEHICLE VALUES DURING INTERVAL

Stop
Time
20
(diff. is
3)
18
(diff. is
5)
23

TABLE IX.

No Of
Familiar
Vehicles
1

Walking
Time

Average
Speed

100
(diff. is
5)
98
(diff. is
7)
105

43
(diff. is
3)
45
(diff. is
5)
40

No Of
Familiar
Vehicles
3

FOURTH VEHICLE VALUES DURING INTERVAL

Edge
ID

Vehicle
ID

Stop
Time

Walking
Time

Average
Speed

100
(diff. is
200)

43
(diff. is
28)

20
(diff.
is
180)
18
(diff.
is
182)
23
(diff.
is
177)
200

98
(diff. is
202)

45
(diff. is
30)

105
(diff. is
195)

40
(diff. is
25)

300

15

No Of
Familiar
Vehicles
3

As shown in the above example vehicle 4 is neglected

because of its irregular values.
Once the interval end is reached, average is calculated
according to the vehicle having the maximum "No Of
Familiar Vehicles" column value. This is done by finding the
familiar vehicle values related to it and dividing their
summations by "No Of Familiar Vehicles" column value.
Table X shows above example results.

UbiCC Journal, Volume 6: Issue 2

TABLE X.
Edge
ID

Average Stop
Time

(20+18+23)/3
= 20.3

FINAL STATISTICAL COMPUTATIONS

Average
Walking Time
(100+98+105)/3
= 101

Average Speed
(43+45+40)/3=42.7

F. Complexity
Average calculation process complexity is O(n) where n is
the number of values read during interval.
Insertion process complexity is O(m) where m is the
number of previously inserted values from the beginning of
the interval. [17]
G. Determining "Highway Level" column
"Highway Level" column varies according to interval
average speed so that if it is more than 80 km/h, then
Highway Level is 3. If the interval Average Speed is between
50 km/h and 80km/h, then Highway Level is 2. Otherwise
Highway Level is 1.
Speed ranges depend on the nature of traffic system in the
city or country.
H. Multithreading problems
This algorithm involves multi access operations and
updates at the same time on the same vehicle values. So,
average calculation and insertion process is defined as
atomic functions. Or, it is left to be handled by database
management system locking mechanism if the system is
implemented using database.
I. Measure Enhancements
In order to get more accurate results edges shouldn't
contain any cross or sub turn with another street regardless of
crossing street size.
J. Implementation
Implementation of the proposed algorithm is done using
object oriented to encapsulate vehicle data and functions. A
special function is defined to add records and increase time
index every interval end to initiate next time interval and
other weights calculation functions. As another
implementation, database is used to utilize database
management systems fast access algorithms.
To demonstrate the superiority of the proposed
framework consider figure 1 as an example. Figure 1 shows
two different routes between two points in Salt area in Jordan.
Few statistics will be presented next. Those statistics are
based on real observations. One of the routes is via a street
with a possible average speed of 96 km/h (red colored in
figure 1). This route is longer than another possible route by
6 kilometers with a total length of 11 kilometers (blue
colored in figure 1). The average travel time for the longer
route was 430 seconds. This is because that this route

818

Ubiquitous Computing and Communication Journal

(ISSN 1992-8424)
contains a single traffic light that stops traffic for a while.
Walking time average was 413 seconds but average stop time
was 17 seconds. Based on the above dynamically calculated
statistics; this street (in the longer route) is considered a
highway by our system compared to the other shorter route.
The other street length is 5 kilometers and consists of 3
traffic lights in addition to a congested cross in the middle.
The average speed was 41 km/h and average travel time was
673 seconds. Average walking time was 439 seconds and
average stop time was 234 seconds. Dijkstra algorithm can
be used to find out best path in terms of routes length.
However, with our proposed framework, it is possible to
consider the time as well as fuel cost factors. For instance,
time factor produces the first route as an optimal one
although second one is shorter in terms of length. Moreover,
in case of using a Toyota Camry car, fuel cost of the second
path will be 639 cc (cubic centimeters) of fuel while the cost
of longer route is 792 cc of fuel. This example gives a flavor
of how valuable our framework can be as it is customizable
and dynamic.

solution is the complexity of insertion process and the high

level of process synchronization needed in the algorithm. But
this delay could be over came or neglected because that
system will use last computed averages and weights till new
ones are available so that system won't stop or wait. In order
to achieve very high accuracy, GPS navigation equipment
should be attached to significant number of the vehicles
passing through streets. In fact, most of new model vehicles
have embedded GPS navigation equipments already installed
from their vendors. In future, new research is planed to
minimize time needed to perform calculations and algorithm
steps as a try to reach as much as possible to the optimality.

Implementation is done using manual stop watches used

by volunteer guys who take snap velocity reads and enter
them to an excel sheets on their laptops. These excel sheets
are migrated to the database of special implementation and
applying average calculation process in order to get results.
Calculation time interval length is used to control the
sensitivity of the algorithm to sudden small congestions. In
case of very sensitive algorithm, time interval should be
small interval so that sudden accidental situations are sensed.
In case of low sensitivity to sudden small congestions, time
interval is set to be long so that sudden congestions
participation vehicles are treated as irregular reads in average
calculations so that they are excluded.
IV. RESULTS AND CONCLUSIONS:
The main advantage of our proposal is that it is self
learning and self maintaining. Consequently, the proposed
framework is capable of providing its customizable best
route identification based on multiple possible optimization
factors such as travel time, fuel cost, and distance. Our
framework also implies less overhead of administration
(except the initial step of defining edges and vehicles
information). Moreover, weights can be used by Traffic
Agencies in order to achieve Intelligent Transportation
System that control streets limited speeds and traffic lights to
minimize congestions and derive efficient traffic flow. On
the other hand, traffic agencies do not have to install any
extra equipments or sensors in roads because that moving
vehicles are used as moving agents that provide the server of
travel agencies with the current statistical data regarding the
status of the roads. The availability of such statistics travel
agencies can find out obstacles or accidents especially in
critical situations. The main disadvantage of the proposed

UbiCC Journal, Volume 6: Issue 2

Figure 1.

Implementation roads paths

REFERENCES
[1] Dijkstra, E. W. (1959). "A note on two problems in
connexion with graphs", Numerische Mathematlk, vol.
1, pp. 269271, Nov. 1959.
[2] Stuart Russell and Peter Norvig, Artificial Intelligence:
A Modern Approach, 3rd Ed, Prentice Hall, 2009
[3] (2010) FHWA Arterial Management Website, latest
information on traffic signal operations [Online].
Available: http://ops.fhwa.dot.gov/arterial_mgmt/
[4] Wikipedia (2011) Traffic light page on Wikipedia
[Online].
Available:
http://en.wikipedia.org/wiki/Traffic_light
[5] Wikipedia (2011) Congestion pricing page on
Wikipedia
[Online].
Available:
http://en.wikipedia.org/wiki/Congestion_pricing

819

Ubiquitous Computing and Communication Journal

(ISSN 1992-8424)
[6] Khalid A. S. Al-Khateeb, Jaiz A.Y. Johari and Wajdi F.
Al-Khateeb, Dynamic Traffic Light Sequence
Algorithm Using RFID Journal of Computer Science,
vol. 4, pp. 517-524, 2008.
[7] P. Pongpaibool, P. Tangamchit,K. Noodwong, and
Klong Luang NECTEC, "Evaluation of road traffic
congestion using fuzzy techniques" In Proc. TENCON,
2007, paper 10.1109/TENCON.2007.4429119, p. 1
[8] Yan-hong Tang and Bao Xi, "Dynamic forecasting of
traffic volume based on quantificational dynamics: A
nearness perspective", Scientific Research and Essays,
vol. 5, pp. 389-394, Feb. 2010.
[9] Tom Thomas, Wendy Weijermars, and Eric van
Berkum , "Variations in urban traffic volumes", EJTIR,
vol. 8(3), pp. 251-263, September 2008.
[10] Ronald Kroon, "Dynamic vehicle routing using ant
based control", Master of Science thesis Delft
University of Technology, Delft, Nederland, May 2002.
[11] Bogdan Tatomir, Adriana-Camelia Suson and Leon
Rothkrantz, Dynamic Routing and Travel Time
Prediction with Ant Based Control, in Proc. 6th
International Conference ANTS, 2008, paper
10.1007/978-3-540-87527-7, p. 404.
[12] O. Aloquili,A. Elbanna, and A. Al-Azizi , "Automatic
Vehicle Location Tracking System Based on GIS
Environment", IET Softw., vol. 3(4), pp. 255-263, 2009.
[13] Honda (2011) 2011 Honda Accord Sedan Specs.
[Online].
Available:
http://automobiles.honda.com/accordsedan/specifications.aspx
[14] Mercedes-Benz (2011) 2011 Mercedes-Benz C300
Specs.
[Online].
Available:
http://www.mbusa.com/mercedes/vehicles/explore/spec
s/class-C/model-C300W
[15] Chevrolet (2010) 2010 Chevrolet COBALT COUPE
Specs.
[Online].
Available:
http://www.chevrolet.com/cobalt/coupe/features-specs/
[16] (2010) STARTCOMS Systems website. [Online].
Available: http://www.starcomsystems.com/
[17] Thomas H. Cormen, Charles E. Leiserson, Ronald L.
Rivest, and Clifford Stein, Introduction to Algorithms,
3rd Edition,Cambridge, Massachusetts London,
England: The MIT Press, 2009

UbiCC Journal, Volume 6: Issue 2

820