Teodorovic, Selmic, Edara
-
Upload
georgehany -
Category
Documents
-
view
218 -
download
0
Transcript of Teodorovic, Selmic, Edara
![Page 1: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/1.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 1/11
Teodorović D., Šelmić M., Edara P.
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE PLACEMENT OF TRAFFIC SENSORS
ON HIGHWAYS
1
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE
PLACEMENT OF TRAFFIC SENSORS ON HIGHWAYS
Teodorović Dušan
Šelmić MilicaUniversity of Belgrade
Faculty of Transport and Traffic Engineering
Vojvode Stepe 305, Belgrade, Serbia
[email protected], [email protected]
Edara Praveen
University of Missouri Columbia
Department of Civil and Environmental Engineering
Columbia MO 65211, U.S.A.
ABSTRACT
In this paper, a model is developed to optimally place traffic detectors on freeways. The
proposed model tries to minimize the error in travel time estimation, while taking into account
the constraints of available capital and maintenance funding. A new metaheuristic, the Bee
Colony Optimization (BCO), is used to solve the formulated problem. The proposed BCO
algorithm, inspired by bees' behavior in the nature, was tested on a real-world freeway
segment in Virginia, U.S.A. The obtained results are very competitive when compared with
the results of Genetic Algorithms achieved in a previous study.
1 INTRODUCTION
Point detectors (or sensors) are deployed on roadways to collect traffic data including
volume, occupancy, and speed. The data is used by Traffic Management Centers in cities to
manage traffic and incidents and provide information to motorists about current conditions.
The spacing of detectors on freeways has a major impact on the travel time estimates derived
from the reported speeds. There exists a tradeoff between detector spacing and travel time
estimate accuracy. As detectors become more closely spaced, the data obtained from them
more closely resemble continuous data available from probes. This additional accuracy also
comes with much higher capital and ongoing costs, as all detectors require regular
maintenance to continue to report good data. Transportation agencies are therefore seeking a
method to indicate the most appropriate locations for detector deployment such that the travel
time estimate error is minimized, within the constraints of available capital and maintenance
funding.
In order to obtain accurate travel time estimates, detectors are to be located so as to
effectively sample the traffic conditions. There is limited past research on how to locate point
detectors for effective sampling. The main purpose of this paper is to propose a methodology
to discover the optimal locations of a finite set of point detectors on a freeway corridor. The
proposed model tries to minimize the error in travel time estimation, while taking into account
the constraints of available capital and maintenance funding. The performed numerical
![Page 2: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/2.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 2/11
Teodorović D., Šelmić M., Edara P.
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE PLACEMENT OF TRAFFIC SENSORS
ON HIGHWAYS
2
experiments make obvious the usefulness of the newly developed methodology. The model
formulation is solved using a new metaheuristic algorithm, the Bee Colony Optimization
(BCO). The proposed algorithm is inspired by bees' behavior in the nature. We propose an
artificial system composed of a number of precisely defined elements (or agents) and then
perform a multi-agent simulation to solve the problem. Behavioral rules are defined for artificial bees and their interactions are simulated. The artificial bees perform only activities
defined by the model and a corresponding computer program.
The paper is organized as follows. The problem of optimal placement of point detectors
is described in Section 2. The formulation of the travel time estimation problem is discussed
in Section 3. Section 4 describes the BCO technique. The implementation details of the BCO
for given problem are in Section 5, while Section 6 contains test results. The paper is
concluded in Section 7.
2 PLACEMENT OF POINT DETECTORS ON FREEWAYS
The problem of the placement of point detectors within a roadway network belongs tothe field of location theory. Many papers have been published during the past four decades
dealing with the problem of locating facilities on the network [1],[2]. “The term LocationAnalysis refers to the modeling, formulation, and solution of a class of problems that can best
be described as sitting f acilities in some given space” [2]. Location analysis tries to find
answers to the following questions: a) What should be the total number of facilities in the
network? b) Where should the facilities be located? c) What is the best allocation of clients to
the facilities?
Point detectors are the infrastructure facilities that we are interested in this paper. With
respect to the application of Operations Research/Artificial Intelligence techniques to study
the facility location problems in transportation, the following application areas have been
studied: (1) detectors for O-D estimation [3], (2) Automatic Vehicle Identification (AVI)readers for travel time estimation [5], and (3) detectors for minimizing the travel time
variance and social costs [4].
With carefully placed detectors that are well maintained, travel time estimates can be
derived with an acceptable level of accuracy from point detection, under incident-free travel
conditions. In this paper we develop a method to indicate the most appropriate locations for
detector deployment such that the travel time estimate error is minimized, within the
budgetary constraints. The proposed approach is applicable to both regions without any (or
limited) current detector deployment and regions that currently have dense deployment. The
developed method is primarily intended for use at a planning level, to assist in determining
where to deploy detectors in an area that currently has few or no detectors, or in determining
which detectors need to be (or those that need not be) regularly maintained to obtain goodtravel time estimates in areas with dense detector deployment.
3 TRAVEL TIME ESTIMATION MODEL
Travel time data can be collected by driving probe vehicles equipped with Global
Positioning System (GPS) devices. Probe vehicles are driven on all identified freeway
sections in the study regions. A GPS device is installed in the vehicle and then a driver drives
this vehicle according to the “flow of traffic” throughout the study region. While the vehicle
is running, the GPS device automatically logs latitude/longitude points and times. In order to
obtain reliable travel times, multiple probe vehicles are usually deployed for data collection
with a headway of 5 minutes between subsequent vehicles.
![Page 3: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/3.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 3/11
Teodorović D., Šelmić M., Edara P.
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE PLACEMENT OF TRAFFIC SENSORS
ON HIGHWAYS
3
We define two notions of travel time for a freeway section: Ground Truth Travel Time
(GTTT) and Estimated Travel Time (ETT). Computation of GTTT is simple. For each probe
vehicle, the entering and exiting times for the freeway section are compared to obtain the
GTTT specific to that vehicle. The same procedure is repeated for each probe vehicle on that
section during a specified time period. ETT is calculated indirectly. Travel time for the wholefreeway section will be estimated from the travel times of constituent detector „ zones of
influence‟ travel times. The zone of influence of a detector can be defined as half the distance
upstream and downstream to the neighboring detector.
Since actual detector data are not available in this scenario, travel time for each zone of
influence is estimated from the speed data collected at the detector location using the GPS
vehicle (as a surrogate for real detector data). A key assumption in this calculation is that the
speed measured at the point detector is approximately equal to the average speed for the entire
zone of influence. Obviously, the greater the length of the zone of influence the greater the
potential for differences in speeds across the zone. The speed at a detector location is
calculated as follows: For each detector location, a number of GPS points around the detector
are defined. Speeds reported by the probe vehicle at all these points within the vicinity of thedetector location are averaged to obtain the average speed specific to that probe vehicle at that
detector location. Finally, the speeds are averaged over multiple probe vehicles as well. The
length of the zone of influence (ZOI) is divided by this average speed to obtain the travel time
value. ETT for the entire freeway section is then obtained by adding the travel time estimates
for the all constituent zones of influence.
Detector 1 Detector ( )i-1 Detector n
ZOI 1
Freeway Section
ZOI 2 ZOInZOI i-1 ZOI i ZOI i-1
Detector (i) Detector (i+1)
x i
L
Figure 1: Notations used in the Formulation
We introduce the following notation:
n - Number of detectors on the freeway section (=number of zones of influence); i -
Index for the i-th detector; xi - Position of the i-th detector, measured from the origin of
the freeway section (decision variable in the objective function); L - Length of the
freeway section; K - Total number of probe vehicle runs; ZOI i - Length of Zone of
influence of the i-th detector ( Ln
i ZOI i =∑
1=); V i - Speed reported by the i-th detector; TTi -
Travel time for ZOI i (V i
ZOI iTT i = ); ETT - Estimated travel time for the freeway section
( ∑
1==
n
iTT i ETT ); GTTT - Ground truth travel time for the freeway section; ε - Estimation
Error placement of point detectors on freeways.
The estimation error equals:
![Page 4: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/4.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 4/11
Teodorović D., Šelmić M., Edara P.
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE PLACEMENT OF TRAFFIC SENSORS
ON HIGHWAYS
4
GTTT n
iTT iGTTT ETT ε -∑
1==-= (1)
GTTT n
i V i ZOI iε -∑
1== (2)
We can express ZOI i in terms of the decision variable xi as follows:
ni for
x x
L
ni for x x
i for x x
ZOI
nn
iii
2
1,...,3,22
12
1
11
21
(3)
The relative error r equals:
GTTT r
(4)
Cumulative relative error CRE equals:
∑1=
=
K
k GTTT k
ε
k εCRE (5)
where:
k - index for the speed profile (i.e., GPS travel time run)
k - travel time estimation error for the k -th profile
GTTTk - ground truth travel time for the k -th profile.
Upon substituting relations (1), (2), (3), and (4) into relation (5), the Cumulative relative
error CRE equals:
∑
1=
∑
1-
2=-
)2
1-+nx
(-L
+2
1--1++
2 1
2+1
= K
k GTTT k
n
t GTTT k
nk V
n x
V tk
xt xt
V k
x x
εCRE
(6)
where, V ik is the speed reported by the i-th detector for the k -th profile.
The mathematical formulation of the problem of optimal placement of point detectors
on freeways reads:
Minimize
![Page 5: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/5.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 5/11
Teodorović D., Šelmić M., Edara P.
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE PLACEMENT OF TRAFFIC SENSORS
ON HIGHWAYS
5
∑
1=
∑
1-
2=-
)2
1-+nx(-L
+2
1--1++
2 1
2+1 K
k GTTT k
n
t GTTT k
nk V
n x
V tk
xt xt
V k
x x
(7)
subject to:
ni L xi ,...,2,1=≤ (8)
ni L xi ,...,2,1=≤ (9)
In Eq. (7), the numerators of the first three terms are linear combination of two decision
variables. The denominator has the speed term V ik , which is a function of the distance xi. This
means that the objective function is non-linear. The problem (7)-(9) is combinatorial by its
nature.
We divide freeway section into discrete segments. This means that the detectors can be
deployed only at the mid points of these discrete segments. Taking into account thecombinatorial nature of the considered problem, we decided to solve the problem using a
metaheuristic.
4 BEE COLONY OPTIMIZATION
The Bee Colony Optimization (BCO) is a meta-heuristic for solving combinatorial
optimization problems. The BCO algorithm belongs to the class of stochastic swarm
optimization methods. The proposed algorithm is inspired by the foraging habits of bees in the
nature. The communication systems between individual insects contribute to the configuration
of the ‘‘collective intelligence” of the social insect colonies. Swarm intelligence [6] is the part
of Artificial intelligence based on studying actions of individuals in various decentralizedsystems.
The BCO is inspired by bees' behavior in the nature. The basic idea behind the BCO is
to create the multi agent system (colony of artificial bees) capable to successfully solve
difficult combinatorial optimization problems. The artificial bee colony behaves partially
alike, and partially differently from bee colonies in nature.
a. Bees in the Nature In spite of the existence of a large number of different social insect species, and
variation in their behavioral patterns, it is possible to describe individual insects‟ as capable of performing a variety of complex tasks. The best example is the collection and processing of
nectar, the practice of which is highly organized. Each bee decides to reach the nectar source by following a nestmate who has already discovered a patch of flowers. Each hive has a so-called dance floor area on which the bees that have discovered nectar sources dance, in that
way trying to convince their nestmates to follow them. If a bee decides to leave the hive to get
nectar, it follows one of the bee dancers to one of the nectar areas. Upon arrival, the foraging
bee takes a load of nectar and returns to the hive relinquishing the nectar to a food store. After
it relinquishes the food, the bee can (a) abandon the food source and become again
uncommitted follower, (b) continue to forage at the food source without recruiting the
nestmates, or (c) dance and thus recruit the nestmates before the return to the food source. The
bee opts for one of the above alternatives with a certain probability. Within the dance area, the
bee dancers “advertise” different food sources.
![Page 6: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/6.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 6/11
![Page 7: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/7.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 7/11
Teodorović D., Šelmić M., Edara P.
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE PLACEMENT OF TRAFFIC SENSORS
ON HIGHWAYS
7
10. Output the best solution found.
5 THE BCO APPROACH
In this paper, we propose the BCO heuristic algorithm tailored for problem of trafficsensors placement. As of the authors‟ knowledge this is the first implementation of Swarm
intelligence to a given problem.
At the beginning of a process, we assume that all bees are in the hive. The hive is an
artificial location, it is not connected to potential sensors location. We allow every artificial
bee to fly out from the hive and to generate NC constructive moves. After that, every bee
returns to the hive. Bees exchange information about the quality of the partial solutions
generated.
a. Constructive moves in forward pass
Authors define the probability ( pi) of choosing a certain alternative based on its utility
(U i) to a user. The probability that specific bee chooses node to be sensor location is:
∑
1=
= K
k k U
iU i p , i=1,2,…,n (10)
where U i is utility of the i-th node and K is the number of “free” nodes (not previously
chosen).
The utility of having a detector at any particular location depends on several factors that
affect travel time estimates. Factors such as the presence of a natural bottleneck at thatlocation (e.g., a lane reduction) that leads to recurring congestion during the peak traffic
periods, level of traffic volumes, etc, can be used to determine the utilities. In this paper, all
potential detector locations are assumed to have equal utilities. However, the proposed model
is still applicable when the utilities of locations differ from each other.
Using Eq. (10) and a random number generator, we assign detector locations to bees.
b. Bee’s partial solutions comparison mechanism
All bees return to the hive after generating the partial solutions. All these solutions are
then evaluated by all bees. Every generated partial solution is characterized by the travel time
estimation error. In this paper, we choose as criteria for comparison, the maximum travel time
error over all travel time runs. We denote by E b the maximum travel time error over all traveltime runs in the case of the partial solution created by the b-th bee. Further, it is normalized as
follows:
[ ] BbbO E E
b E E bO ,1=1,0∈,
min- max
-max= (11)
where:
b E - travel time error value created by the i-th bee
max E , min E - maximum/minimum travel time error value over all partial solutions
generated so far
![Page 8: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/8.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 8/11
Teodorović D., Šelmić M., Edara P.
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE PLACEMENT OF TRAFFIC SENSORS
ON HIGHWAYS
8
The probability that the b-th bee (at the beginning of the new forward pass) is loyal to
its previously discovered partial solution is expressed as follows:
Bbu
bOO
eub p ,...,2,1=,
-max
=1+ (12)
where u is the ordinary number of the forward pass.
In relation (12), the better the generated partial solution (higher Ob value), the higher the
probability that the bee will be loyal to this discovered partial solution. Also, the greater the
ordinary number (u) of the forward pass, the higher the influence of the already discovered
partial solution. In other words, at the beginning of the search process bees are “more brave”to search the solution space. The more forward passes they make, the bees have less courage
to explore the solution space. The more we are approaching the end of the search process, the
more focused the bees are on the already discovered solutions.
Using relation (12) and a random number generator, every artificial bee decides to become an uncommitted follower, or to continue flight along the already known path.
c. Recruiting Process In the case when at the beginning of a new stage bee does not want to expand
previously generated partial solution, the bee will go to the dancing area and will follow
another bee. Within the dance area the bee-dancers (recruiters) “advertise” different par tial
solutions. We have assumed in this paper that the probability the recruiter b‟s partial solutionwill be chosen by any uncommitted bee equals:
∑
1=
= R
k Ok
Obb p , b=1,2 ,…, R (13)
where Ok is objective function value of the k-th advertised solution and R - the number of
recruiters.
Using relation (13) and a random number generator, every uncommitted follower join
one bee dancer (recruiter). Recruiters fly together with a recruted nestmates in the next
forward pass along the path discovered by the recruiter. At the end of this path all bees are
free to independently search the solution space and generate the next iteration constructivemoves.
6 APPLICATION OF BCO TO A CASE STUDY
The problem of optimal placement of point detectors was studied in the case of
Northern Virginia region in Virginia. The data related to probe vehicle runs are taken from the
previous work of Edara et al. [13]. In Northern Virginia, an approximately 11-mile section of
I-66 from exit 43 to exit 57 in both directions was studied. The I-66 corridor is one of the
busiest freeways in Virginia with over 90,000 vehicles per day traveling to and from
Washington DC. Three GPS equipped probe vehicles were driven on three week days during
both the morning and evening peak periods in both the east and west directions. Vehiclesdeparted at 5 minute headways. This resulted in a total of 37 travel time runs during the
![Page 9: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/9.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 9/11
Teodorović D., Šelmić M., Edara P.
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE PLACEMENT OF TRAFFIC SENSORS
ON HIGHWAYS
9
morning peak and 40 runs during the evening peak. Currently, detectors are deployed at
approximately one-half mile spacing resulting in a total of about 20 detectors over the study
segment.
One of the main purposes of developing proposed methodology was to generate tradeoff
plots between the travel time error and the number of detectors which would give the optimal placement of detectors for different levels of available funding. Tradeoff plots were generated
by varying the actual number of detectors (q) from 2 to 19 in increments of 1. The BCO
algorithm parameters were set to the following values: the total number of bees B engaged in
the search process was equal to 10; the number of moves, NC, during one forward pass was
equal to 1; the number of iterations I within one run was equal to 100. All the tests were
performed on a AMD Sempron (tm) Processor with 1.60 GHz and 512 MB of RAM under
Windows OS. Results of the BCO runs are shown in Figure 2. For a given number of
detectors, the obtained optimal placement would result in a travel time estimation error for
each travel time run. The maximum of these errors versus the detector deployment is plotted
in Figure 2. Edara et al. [13] used Genetic Algorithms (GA) to discover the optimal detector
deployment. The maximum error versus the detector deployment obtained by the Geneticalgorithms (GA) is also plotted in Figure 2.
Figure 2: Maximum Travel Time Estimation Error Plot (BCO vs GA)
As it can be seen from the plot, the maximum error value is high when only a few
detectors are deployed; however, as the deployment increases the error value decreases. After
reaching a certain level of deployment any further increase in the number of detectors may not
decrease the error.
Table 1: Optimal detectors locations Number
of Detectors
Deployed
Maximum TT Estimation Error (Minutes)
RecommendedPlacement of Detectors
2 6.7827 13 19
3 5.7032 3 13 19
4 4.4583 8 12 18 20
5 3.9362 8 11 17 18 20
6 2.7712 5 8 10 12 18 20
7 2.9008 3 5 10 11 12 18 20
8 2.5474 2 3 9 11 15 17 19 20
9 2.1731 1 3 7 8 10 14 17 19 20
10 1.9631 2 3 5 9 12 14 17 18 19 20
11 1.9433 1 3 4 7 10 13 15 16 17 19 20
12 1.6342 1 4 6 7 10 11 12 14 17 18 19 20
13 1.4200 1 4 6 7 9 10 11 12 14 17 18 19 20
14 1.2423 1 2 4 7 8 9 10 12 13 14 17 18 19 20
0
1.7
3.3
5
6.7
8.3
10
2 4 6 8 10 12 14 16 18 20
GA Solutions
BCO Solutions
Number of Detectors
M a x i m
u m T
T E s t i m a t i o n E r r o r ( m i n u t e s )
![Page 10: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/10.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 10/11
Teodorović D., Šelmić M., Edara P.
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE PLACEMENT OF TRAFFIC SENSORS
ON HIGHWAYS
10
15 1.2584 1 2 3 4 5 7 10 12 14 15 16 17 18 19 20
16 1.2959 1 2 3 4 5 6 7 10 12 14 15 16 17 18 19 20
17 1.3334 1 2 3 4 6 7 9 10 12 13 14 15 16 17 18 19 2018 1.3621 1 2 3 4 5 6 7 9 10 12 13 14 15 16 17 18 19 20
19 1.7793 1 2 3 4 5 6 7 8 9 10 12 13 14 15 16 17 18 19 20
From Figure 2 it can be inferred that deploying 14 detectors would result in the least
maximum TT error (~ 1.24 minutes) and an acceptable error distribution. We can further
conclude that the 20 detectors currently deployed are more than is needed to provide
reasonably accurate travel time estimates. We show in Table 1 the optimal detector locations
obtained by the BCO algorithm for different numbers of detectors along with the
corresponding maximum travel time errors.
7 CONSLUSIONS
The model developed could be used for planning purposes. In a very short time it is
possible to generate a large number of different solutions for detectors placement, and observe performances (numbers of detectors employed, and travel time estimation error) achieved for
each one. In this way, state and local departments of transportation can decide where to add
new detectors. The proposed model also suggests which of the existing detectors should go on
with getting maintenance, given agencies‟ financial resource constraints.In this paper, the Bee Colony Optimization was used to solve the problem of optimally
locating traffic detectors on freeways. The BCO belongs to the class of constructive methods.
The proposed BCO algorithm was tested on a real-world case study in Virginia with travel
time data collected using highly accurate GPS-equipped probe vehicles. The obtained results
were compared with the results of a Genetic Algorithm methodology achieved in a previous
study. The comparison indicated that the BCO results were better than or equal to the GA
results for majority of sensor deployments. The promising results obtained in this work encourage further investigation of the Bee Colony Optimization to similar combinatorial
optimization problems.
REFERENCES
1. C. ReVelle, “A perspective on location science,” Location Science, 5, 1997, pp. 3-13.
2. C. ReVelle, H.A. Eiselt, “Location analysis: A synthesis and survey,” European
Journal of Operational Research, 165, 2005, pp. 1 – 19.
3. H. Yang, J. Zhou, “Optimal Traffic Counting Locations for Origin-DestinationMatrix Estimation,” Transportation Research Part B,32, 2005, pp. 109-126.
4. A. Ehlert, M. Bell, S. Grosso, “The Optimization of Traffic Count Locations in
Road Networks,” Transportation Research B; 40, 2006, pp. 460-479.
5. D. Teodorović, M. Van Aerde, F. Zhu, F. Dion, “Genetic Algorithms Approach to
the Problem of the Automated Vehicle Identification Equipment Locations,“ Journal
of Advanced Transportation, 36, 2002, pp. 1-21.
6. S. Camazine, J. Sneyd, “A model of collective nectar source by honey bees: self-
organization through simple rules,” Journal of Theoretical Biology, 149, 1991, pp.547- 571.
![Page 11: Teodorovic, Selmic, Edara](https://reader030.fdocuments.in/reader030/viewer/2022021118/577cdd1a1a28ab9e78ac3ae3/html5/thumbnails/11.jpg)
7/30/2019 Teodorovic, Selmic, Edara
http://slidepdf.com/reader/full/teodorovic-selmic-edara 11/11
Teodorović D., Šelmić M., Edara P.
BEE COLONY OPTIMIZATION APPROACH TO OPTIMIZE PLACEMENT OF TRAFFIC SENSORS
ON HIGHWAYS
11
7. P. Lučić, D. Teodorović, “Bee system: modeling combinatorial optimizationtransportation engineering problems by swarm intelligence,” in Preprints of the
TRISTAN IV Triennial Symposium on Transportation Analysis. Sao Miguel, Azores
Islands, 2001, pp. 441-445.
8. P. Lučić, D. Teodorović, “Vehicle routing problem with uncertain demand at nodes:the bee system and fuzzy logic approach”, In: Verdegay J.L, (Ed.) Fuzzy Sets based
Heuristics for Optimization,. Physica Verlag: Berlin Heidelberg, 2002, pp. 67-82.
9. P. Lučić, D. Teodorović, “Computing with bees: attacking complex transportationengineering problems,” International Journal on Artificial Intelligence Tools, 12,
2003, pp. 375-394.
10. D. Teodorović, M. Dell‟Orco, “Mitigating traffic congestion: solving the ride-
matching problem by bee colony optimization,” Transportation Planning and
Technology, 31, 2008, pp. 135-152.
11. G. Marković, D. Teodorović, V. Aćimović Raspopović, “Routing and wavelength
assignment in all-optical networks based on the bee colony optimization ,” AI
Communication - The European Journal of Artificial Intelligence, 20, 2007, pp. 273
– 285.
12. T. Davidović, M. Šelmić, D. Teodorović, “Scheduling Independent Tasks: Bee
Colony Optimization Approach,” Proc. Тhe 17th IЕЕЕ Mediterranean Conference on
Control and Automation, MED'09, Thessaloniki, Greece, June 24-26, 2009, pp.
1020-1025.
13. P.Edara, J. Guo, B.L. Smith, C. McGhee, “Optimal placement of point detectors on
Virginia‟s Freeways: Case studies of Northern Virginia and Richmond,” Final Report
VTRC 08-CR3, Virginia Transportation Research Council, Richmond, VA, 2008.