Content uploaded by Hesham A. Hefny
Author content
All content in this area was uploaded by Hesham A. Hefny on Jul 12, 2015
Content may be subject to copyright.
A.E. Hassanien et al. (Eds.): AMLTA 2014, CCIS 488, pp. 424–435, 2014.
© Springer International Publishing Switzerland 2014
Improving the Performance of TAntNet-2
Using Scout Behavior
Ayman M. Ghazy and Hesham A. Hefny
Department of Computer & Information Sciences,
Institute of Statistical Studies and Research, Cairo University, Giza, Egypt
aymghazy@yahoo.com, hehefny@ieee.org
Abstract. Dynamic routing algorithms play an important role in road traffic
routing to avoid congestion and to direct vehicles to better routes. TAntNet-2
algorithm presented a modified version of AntNet algorithm to dynamic traffic
routing of road network. TAntNet-2 uses the pre-known information about the
expected good travel time between sources and destinations for road traffic
networks. Good travel time is used as a threshold value to fast direct the
algorithm to good route, conserve on the discovered good route and remove un-
needed computations. This paper presents a modified version of the TAntNet-2
routing algorithm that employs a behavior inspired from bee behavior when
foraging for nectar. The new algorithm tries to avoid the effects of ants that take
long route during searching for a good route. The modified algorithm introduc-
es a new technique for launching ants according the quality of the discovered
solution. The presented algorithm uses forward scout instead of forward ant and
uses two forward scouts for each backward ant, in case of failing the first scout
in finding accepted good route. The experimental results show high perfor-
mance for the modified TAntNet-2 compared with TAntNet and TAntNet-2.
Keywords: Swarm Intelligence, Road networks, Dynamic traffic routing,
AntNet, TAntNet-2, Forward ant, Forward scout, Backward ant, Check ant, bee
behavior, bad route.
1 Introduction
Ant routing algorithms is one of the most promising swarm intelligence (SI) method-
ologies that are capable of finding near optimal solutions at low computational cost.
Ant routing algorithms have been studied in many researches [1-7]. AntNet is a dis-
tributed agent based routing algorithm inspired by the behavior of natural ants [8].
Since its first appearance in 1998, AntNet algorithm has attracted many researchers to
adopt it in both of data communication networks and road traffic networks.
On data networks, it has been shown that under varying traffic loads, AntNet algo-
rithm is amenable to the associated changes and it shows better performance than that of
Dijkstra’s shortest path algorithm [9]. Several enhancements have been made to the
AntNet algorithm. Baran and Sosa [10] proposed to initialize the routing table at each
node in the network. The proposed initialization reflects previous knowledge about
Improving the Performance of TAntNet-2 Using Scout Behavior 425
network topology rather than the presumption of uniform probabilities distribution giv-
en in original AntNet algorithm. Tekiner et al. [11] produced a version of the AntNet
algorithm that improved the throughput and the average delay. In addition, their algo-
rithm utilized the ant/packet ratio to limit the number of used ants. A new type of help-
ing ants has been introduced in [12] to increase cooperation among neighboring nodes,
thereby reducing AntNet algorithm’s convergence time. A study for a computation of
the pheromone values in AntNet has been given in [13]. Radwan et al. [14] proposed an
adapted AntNet protocol with blocking–expanding ring search and local retransmission
technique for routing of Mobile ad hoc network (MANET). Sharma et al. [15] showed
that load balancing is successfully fulfilled for ant based techniques [15].
On road traffic networks, An Ant Based Control (ABC) algorithm has been ap-
plied in [2] for routing of road traffic through a city. In [3] a modification of Ant
Based Control (ABC) and AntNet has been presented for routing vehicle drivers using
historically-based traffic information. Claes and Holvoet [4] proposed a cooperative
ACO algorithm for finding routes based on a cooperative pheromone among ants.
Yousefi and Zamani in [6] proposed an optimal routing method for car navigation
system based on a combination between Divide and Conquer method and Ant Colony
algorithm. According to their proposed method, road network is divided into small
areas. Then the learning operation is done in these small areas. Then different learnt
paths are combined together to make the complete paths. This method causes traffic
load balance over the road network. A version of the AntNet algorithm has been ap-
plied in [16] to improve traveling time over a road traffic network with the ability to
divert traffic from congested routes. In [17] a city based parking routing system
(CBPRS) that used Ant based routing has been proposed. Kammoun et al. in [18]
introduced an adaptive vehicle guidance system instigated from the ant behavior.
Their system allows adjusting the route choice according to the real-time changes in
the road network, such as new congestions and jams. In [19] an Ant Colony Optimi-
zation combined with link travel time prediction has been applied to find routes. The
proposed algorithm takes into account link travel time prediction, which can reduce
the travel time. Ghazy et al. [20] proposed a threshold based AntNet algorithm (called
TAntNet) for dynamic traffic routing of road networks, which used the pre-known
information about good travel times among different nodes as a threshold value.
In the last decade, many researches were directed their efforts to produce hybrid
algorithms that combine features from ants and bees behavior [21, 22]. Rahmatizadeh
et al. [23] proposed an Ant-Bee Routing algorithm, which inspired from the behavior
of both ant and bee to solve the routing problem. The algorithm is based on the
AntNet algorithm and enhanced via using bee agents, it use forward agent inspired
from ant and backward agent inspired from bee [23]. Pankajavalli et al. [24] present-
ed and implemented an algorithm based on ant and bee behavior called BADSR for
Routing in mobile ad-hoc network. The algorithm aimed to integrate the best of ant
colony optimization (ACO) and bee colony optimization (BCO), the algorithm uses
forward ant agents to collect data and backward bee agents to update the links state,
the bee agent update data based on checking a threshold. Simulation results represent-
ed better result for the BADSR algorithm in terms of reliability and energy consump-
tion [24]. Kanimozhi Suguna et al. [25] showed an algorithm for on demand ad-hoc
routing algorithm, which is based on the foraging behavior of Ant colony optimiza-
426 A.M. Ghazy and H.A. Hefny
tion and bee colony optimization. The proposed algorithm uses bee agents to collect
data about the neighborhood of the node, and uses forward ant agents to update the
pheromone state of the links. The results showed that the proposed algorithm has the
potential to become an appropriate routing strategy for mobile ad-hoc networks [25].
In this paper, a new modified version of the TAntNet-2 algorithm is proposed for
dynamic routing of road traffic networks, where a performance of the algorithm is
enhanced by avoiding the bad effect of forward ants that take a bad route. The new
modified algorithm uses a threshold to measure the quality of the solution that found
by forward ant. When the result of measuring represents bad solution, the algorithm
ignores the solution of first forward ant and retransmits another forward ant.
For the purpose of this paper, the standard TAntnet algorithm is presented in
Section 2. While, the proposed modified version of the algorithm is introduced in Sec-
tion 3. The simulation experiment is given in section 4. Section 5 concludes the paper.
2 Threshold Based AntNet-2 algorithm
TAntNet algorithm was proposed by Ghazy et.al. [20]. TAntNet is a modified version
of AntNet algorithm for traffic routing of road network. The main idea of TAntNet
algorithm is to get benefit of the pre known information about the good travel time
between a source and a destination. And use this good travel times as threshold val-
ues. TAntNet used a new type of ants called “check ants”. Check ants are responsible
of periodically checking the discovered good route whether it is still good or not.
When running TAntNet, it was noticed that the good route between a source and a
destination may disappear after some amount of time of running ants over the net-
work. The reason was the bad effect of the sub path update on the discovered good
route. To overcome this problem, TAntNet-2 was suggested in ([26], [27]) to prevent
the sub path updates for the already discovered good routes. Figure 1 illustrates the
pseudo code of The TAntNet-2 algorithm ([26], [27]):
Algorithm. Threshold-based AntNet (TAntNet-2)
/* Main loop */
FOR each (Node s) /*Concurrent activity*/
t=current time
WHILE ! " # /* T is the total experiment time */
Set d := Select destination node;
Set Tsd = 0 /* Tsd travel time from s to d */
IF (Gd = yes)
Launch Check Ant (s, d); /* From s to d*/
ELSE
Launch Forward Ant (s, d); /* From s to d*/
IF (Tsd<=T_GoodSd)
Set Gd = yes
END IF
END IF
END WHILE
END FOR
Fig. 1. The TAntNet-2 Algorithm
Improving the Performance of TAntNet-2 Using Scout Behavior 427
CHECK ANT ( source node: s , destination node: d)
Tsd = 0
WHILE (current_node destination_node)
Select next node using routing table
(node with highest probability)
Set travel_time= travel time from current node to
next_node
Set Tsd = Tsd + travel_time;
Set current_node = next_node;
END WHILE
IF (Tsd>T_GoodSd)
Set Gd = No
END IF
END CHECK ANT
Forward Ant ( source node: s , destination node: d)
WHILE (current_node destination_node)
Select next node using routing table
Push on stack(next_node, travel_time);
Set current_node = next_node;
END WHILE
Launch backward ant
Die
END Forward Ant
Backward Ant ( source node: s , destination node: d)
WHILE (current node ≠ source node) do
Choose next node by popping the stack
Update the traffic model
Update the routing table as follows:
IF (Tsd<=T_GoodSd)
$%&'( )
$*&'( +!!,!!!!!-. / 0!, . 1 23
/* where: h is the node “come from”, k is the
current node, NK is the set of neighbors nodes,
45 is the destination or sub path destination */
ELSE if (Gsd'= No)
$%&'( $%&'6 78) 9 $%&':
/* where r is the reinforcement value*/
END IF
END WHILE
END Backward Ant
Fig. 1. (continued)
428 A.M. Ghazy and H.A. Hefny
3 The Improved TAntNet-2 Algorithm
The Behavior of Bee during collecting the nectar is an attractive behavior. The em-
ployed forager bee memorizes the location of food source to exploiting it. After the
foraging bee loads a portion of nectar from the food source, it returns to the hive and
save the nectar in the food area. After that, the bee enters to the decision making pro-
cess which includes the decision if the nectar amount decreased to a low level or ex-
hausted, in this case it abandons the food source ([28], [29])
This paper uses the previous idea to enhance the performance of TAntNet algo-
rithm. In TAntNet algorithm, forward ant explores a path between a source and desti-
nation. Because of the probabilistic selection of route, forward ant can take a bad
path. The new modified algorithm tries to treat this bad effect by using an idea in-
spired from the bee foraging behavior, when bee takes a decision of completing forag-
ing depend on the quantity level of nectar. In the modified TAntNet-2, we will use
forward scout instead of forward ant. Forward scout does not launch backward ant
immediately after it finishes its trip, but after forward scout finishes its trip it will
enter in a decision step depend on the quality of the discovered route, to determine
whether to launch backward ant or abandons the forward scout and retransmit another
forward scout to search for another solution. The second forward scout will acts the
same as forward ant, it will launch backward ant after finishing of its trip.
After forward scout finished its trip and before launching the corresponding back-
ward ant, the modified algorithm checks the quality of the discovered route. Quality is
checked compared by the mean value in the local traffic statistics table of the source
node. Formula (1) represents the formula that determined the accepted forward ant.
;<= "!> ?= (1)
Where:
;<=: is the total travel time of the discovered route by the forward ant that
launched from s to d.
: weighs the threshold level.
?= : mean of the trip times of ants that launch from node s to node d
The first forward scout with at most total travel time less than or equal µ will be
accepted, otherwise the algorithm will ignore this first forward scout and second for-
ward scout will be launched. Second scout will be accepted whatever its travel time.
Accepting of second forward scout is return to avoid the stuck of the algorithm when
critical changes occur in the traffic situation. The pseudo code for the Modified
TAntNet-2 algorithm is illustrated in Figure 2. The lines of codes appear in bold font
represent the new modifications compared with the TAntNet-2 algorithm. Figure 2
shows the main loop and the forward scout procedure of the modified algorithm while
the procedures of Check Ant and Backward Ant will be the same as them of the
TAntNet-2 algorithm shown in Figure1.
The new enhancement in the algorithm can be seen as adding a scouting process
before launching of backward ant. The scouting process includes the following tasks:
Improving the Performance of TAntNet-2 Using Scout Behavior 429
! Sending first scout to search for route between specific source and specific destina-
tion.
! Test the quality of the discovered route by the first scout.
If the route is accepted [i.e. total travel time of the route is less or equal
than the threshold ( µ)], the algorithm will launch the backward ant.
Otherwise if the route is not accepted [i.e. total travel time of the route is
higher than the threshold ( µ)], the algorithm will launch second scout to
search for another route, and then launch backward ant.
The Proposed Modified TAntNet-2 Algorithm
/* Main loop */
FOR each (Node s) /*Concurrent activity*/
t=current time
WHILE @! " A /* T is the total experiment time */
Set
d
:= Select destination node;
Set
Tsd
= 0
/*
T
sd
travel time from s to d */
IF (
Gd
= yes)
Launch Check Ant (s, d); /* From s to d*/
ELSE
Launch Forward Scout (s, d); /* From s to d*/
IF (
!BCD E!> FD!
)
Die (Forward Scout);
/* Die of First Forward Scout From s to d*/
Launch Forward Scout(s,d);
/*second Forward Scout From s to d*/
END IF
IF (
Tsd
<=
T_GoodSd
)
Set
Gd
= yes
END IF
Launch Backward Ant (d, s)
Die (Forward Scout); /* Die of Second Forward Scout*/
END IF
END WHILE
END FOR
Forward Scout (source node: s, destination node: d)
WHILE (current_node destination_node)
Select next node using routing table
Push on stack(next_node, travel_time);
Set current_node = next_node;
END WHILE
END Forward Scout
Fig. 2. The Modified TAntNet-2 Algorithm
430 A.M. Ghazy and H.A. Hefny
The appearance of µ in AntNet and TAntNet-2 algorithms arises in three main
places (In the procedure of backward ant only) as follows:
! The first appearance is for computing µ itself.
! The second appearance of µ is during computing the pheromone values of the rout-
ing tables.
! The third appearance of µ is arising as a threshold to determine the degree of
goodness of the sub path update to determine if the algorithm performs sub path
update or not.
But in the new enhanced algorithm the appearance of µ is increased to be four
times, the fourth one arises as a part of the decision making during the scouting pro-
cess, this using of µ here represents as a part of the threshold formula; which deter-
mines the acceptance of the solution returned by the forward scout or re launches a
new forward scout to search for a different solution.
The increasing of using µ reflects an enhancement in the learning process with the
new proposed algorithm. Which consequently increases the intelligence of the algo-
rithm. Table 1 represents a comparison between the three types of algorithm AntNet,
TAntNet-2 and the new enhanced algorithm.
Table 1. Comparison between the three types of algorithm
AntNet TAntNet-2 The Modified TAntNet-2
Number of Using of µ (Three Times) (Three Times) (Four Times)
Threshold using
* Sub path update is
performed for only a
solution with a speci-
fied degree of good-
ness.
* Sub path update is
performed for only a
solution with a specified
degree of goodness.
* In case of found good
solutions (Launching
check ant instead of
forward-backward ants)
* Sub path update is performed
for only a solution with a speci-
fied degree of goodness.
* In case of found good
solution
(Launching check ant instead
of forward-backward ants)
* In case of Bad solution
(when forward ant returned
with a bad solution retransmit
another forward ant)
The expected se-
quence of launching
different type of ants
* Forward-Backward * Forward-Backward
* Check
* Forward-Backward
* Check
*Forward-Forward-Backward
4 Experiment
A simulation is used to test and compare the performance of the modified TAntnet-2,
TAntNet-2 and the original AntNet algorithms. The used network has 16 nodes with
the topology shown in Figure 3. The objective is to get best routes between the source
node 1 and any other node in the network over a certain period of time.
Improving the Performance of TAntNet-2 Using Scout Behavior 431
Fig. 3. The topology used for a network with 16 nodes
The simulation runs to test the original AntNet, TAntNet-2 and the modified
TAntNet-2 algorithms. The modified TAntNet-2 is tested for different parameter (
= 0.5; = 1; = 1.5; = 2). The simulation experiment starts by continuously launch-
ing forward (or check) ants from the source node 1 to any arbitrary node. The time of
each simulated experiment is set to 20 minutes. The experiment is repeated 40 times
for the original AntNet, TAntNet-2 and the modified TAntNet-2 (with =0.5; = 1;
=1.5; =2) algorithms on the same processing unit with completely new generated
data at each run.
The simulation experiments show the following results:
• The modified TAntNet-2 (with =2) allows increase in the number of launched
ants compared with the original AntNet and TAntNet-2 algorithms. This increase
is, even, accelerated further with modified TAntNet-2 (with =1; 1.5). The modi-
fied TAntNet-2 (with =0.5; 1.5; 2) allows a reduction in average travel time. The
reduction is increased further under the modified TAntNet-2 (with =1) as shown
in Table 2.
Table 2. Number of launched ants and the average ants travel time over the simulation period
Average ±Standard deviation
Average No. of ants Average travel time
Algorithm Name Value Percentage of
Increase comparing
with AntNet Alg.
Value Percentage of de-
crease comparing
with AntNet Alg.
AntNet 2058.35±71.31 33.84±2.02
TAntNet-2 2683.92±424.76 23.31 % 32.11±2 5.11 %
The Modified TAntNet-2,
Min-threshold= 0.5 µ 2177.22±156.92 5.46 % 31.64±2.1 6.50 %
The Modified TAntNet-2,
Min-threshold= µ 3040.8±168.37 32.31 % 27.39±1.93 19.06 %
The Modified TAntNet-2,
Min-threshold= 1.5 µ 3063.98±178.78 32.82 % 28.27±2.02 16.46 %
The Modified TAntNet-2,
Min-threshold= 2 µ 2895.78±336.74 28.92 % 29.61±2.1 12.5 %
432 A.M. Ghazy and H.A. Hefny
The increasing in the number of ants reflects a decreasing in computational
complexity, which return to avoiding the ants the takes bad route and in most cases
these ants passes many nodes and the corresponding Backward ant takes a lot of
computations.
Related t-test is used to show the significance of the new enhancement. A one-
tailed t-test in the positive direction is used with degrees of freedom equal to 39, the
tabulated is set to 0.05, so the value tcrit equal to +1.69.
Related t-test is applied to the experimental results of AntNet against TAntNet-2
algorithm, AntNet against the Modified TAntNet-2 ( =1) algorithm and TAntNet-2
against the Modified TAntNet-2 (with =1).
The related t-test analysis applied on the performance index of average travel time
over the simulation period, indicates significant decrease in the three cases as
(T(experimental results) > 1.69) as illustrated in Table 3.
Table 3. Related t-test between Average Travel Time over the Simulation Period
AntNet with TAntNet-2 AntNet with The Modified
TAntNet-2, Min-threshold= µ
TAntNet-2 with The Modified
TAntNet-2, Min-threshold= µ
17.32*1 61.54.11 * 44.38 *
At each simulation minute, the average travelling time to all network nodes for The
Modified TAntNet-2 (with =0.5; 1.5) were less than that of the original AntNet and
TAntNet-2. The reduction is increased further under the Modified TAntNet-2 (with
=1) as shown in Table 4 and Figure 4.
Fig. 4. The average travel time at each minute for all network nodes
* Means significant at = 0.05
Improving the Performance of TAntNet-2 Using Scout Behavior 433
Table 4. Average Travel Time at Each Minute
Minute AntNet TAntNet-2
The Modified
TAntNet-2,
Min-threshold=
0.5µ
The Modified
TAntNet-2, Min-
threshold= µ
The Modified
TAntNet-2,
Min-threshold=
1.5µ
The Modified
TAntNet-2,
Min-threshold=
2 µ
1 40.68±4.88 39.05±6.27 37.08±6 32.08±4.45 32.62±4.88 34.02±5.2
2 33.6±4.46 31.51±5.22 32.41±6.21 27.69±4.87 28.32±4.85 29.74±5.36
3 33.98±4.73 31.46±6.16 31.13±5.6 26.8±4.44 27.67±4.47 28.63±4.9
4 33.36±3.94 31.34±5.59 31.06±6.46 26.94±4.91 27.74±4.63 28.95±5.33
5 33.35±4.28 31.29±5.85 31.03±6.35 26.18±4.57 27.25±4.85 28.53±5.33
6 35.14±4.4 34.13±6.13 33.47±5.11 29.58±4.97 29.71±4.75 31.08±4.8
7 32.61±4.29 32.54±5.75 32.04±5.43 27.97±4.77 28.85±4.8 29.91±4.76
8 33.74±4.17 32.02±6.12 32.42±6.13 27.42±4.9 28.57±4.93 29.64±5.22
9 32.47±4.67 32.26±5.73 31.1±5.78 27.37±4.67 28.8±5 29.82±5.17
10 33.95±4.19 31.81±5.64 31.34±5.75 27.38±4.65 28.51±4.87 29.68±4.93
11 34.37±3.32 34.69±4.26 34.21±4.47 30.19±4.23 30.79±4.01 32.21±4.07
12 34.78±3.52 33.55±4.42 34.36±4.32 28.56±3.89 29.74±3.9 31.08±4.54
13 35.02±4.35 33.68±4.63 33.09±4.91 28.14±3.7 29.26±4.44 30.79±4.6
14 35.37±3.91 33.01±4.46 32.48±4.42 27.9±4 29.53±4.34 30.59±4.53
15 34.77±3.73 33.38±4.56 32.32±4.22 27.69±3.82 29.18±3.79 30.79±4.57
16 32.83±3.97 31.02±4.08 30.15±3.94 27.07±4.34 27.38±3.6 29±3.57
17 30.69±3.79 29.76±4.51 28.73±4.75 25.09±3.64 25.95±3.92 27.59±3.87
18 31.94±3.65 28.95±4.5 28.04±4.77 24.73±3.81 25.44±3.73 26.89±3.97
19 31.56±3.52 28.53±4.68 27.98±4.53 24.53±3.85 24.98±3.91 26.27±4
20 32.74±3.87 28.36±4.88 28.42±5.1 24.58±3.84 25.05±3.93 25.94±4.06
5 Conclusion and Future Works
In this paper, a modified version of the TAntNet-2 algorithm is presented to be ap-
plied to dynamic traffic routing of road networks. The new algorithm inspires a new
feature from bee foraging behavior, to enhance the performance of TAntNet-2 algo-
rithm. The new algorithm performs a scouting process before launching of the back-
ward ants. The scouting process uses a threshold to determine the accepted solution.
The threshold uses the historical data saved in the local traffic statistics table. Scout-
ing process use retransmits of new scout, in case of rejected first scout. The new algo-
rithm works on preventing the bad effect of bad forward ant. Also the new enhance-
ment decreases the processing time that used by backward ant which corresponds to
forward ant that takes bad route and passes many nodes. Experimental results show
high performance for the modified TAntNet-2 compared with TAntNet and TAntNet-
2. Among different values of for threshold of the modified TAntNet-2, =1 repre-
sents the best value.
We will work in the future on extend the simulated experiments that compared the
modified TAntNet-2 with AntNet and TAntNet-2 algorithms and using the statistics
to test and analyze the performance of modified TAntNet-2. Also we will work on test
the modified algorithm on a larger network.
434 A.M. Ghazy and H.A. Hefny
References
1. Kassabalidis, I., El-Sharkawi, M.A., Marks, R.J., Arabshahi, P., Gray, A.: Adaptive-SDR:
adaptive swarm-based distributed routing. In: Proceedings of the International Joint Con-
ference on Neural Networks, Honolulu, HI, vol. 1, pp. 351–355 (2002)
2. Kroon, R., Rothkrantz, L.: Dynamic vehicle routing using an ABC-algorithm. In: Trans-
portation and Telecommunication in the 3rd Millennium, Prague, pp. 26–33 (2003)
3. Suson, A.: Dynamic routing using ant-based control. Master thesis, Faculty of Electrical
Engineering, Mathematics and Computer Science, Delft University of Technology (2010)
4. Claes, R., Holvoet, T.: Cooperative ant colony optimization in traffic route calculations. In:
Demazeau, Y., Müller, J.P., Rodríguez, J.M.C., Pérez, J.B. (eds.) Advances on PAAMS.
AISC, vol. 155, pp. 23–34. Springer, Heidelberg (2012)
5. Shah, S., Bhaya, A., Kothari, R., Chandra, S.: Ants find the shortest path: a mathematical
proof. Swarm Intelligence 7(1), 43–62 (2013)
6. Yousefi, P., Zamani, R.: The Optimal Routing of Cars in the Car Navigation System by
Taking the Combination of Divide and Conquer Method and Ant Colony Algorithm into
Consideration. International Journal of Machine Learning and Computing 3, 44–48 (2013)
7. Jabbarpour, M.R., Malakooti, H., Noor, R.M., Anuar, N.B., Khamis, N.: Ant colony opti-
misation for vehicle traffic systems: applications and challenges. International Journal of
Bio-Inspired Computation 6(1), 32–56 (2014)
8. Di Caro, G., Dorigo, M.: AntNet: distributed stigmergetic control for communications
networks. J. Articial. Intell. Res. (JAIR) 9, 317–365 (1998)
9. Dhillon, S.S., Van Mieghem, P.: Performance analysis of the AntNet algorithm. Computer
Networks 51, 2104–2125 (2007)
10. Baran, B., Sosa, R.: AntNet routing algorithm for data networks based on mobile agents.
Inteligencia Artificial, Revista Iberoamericana de Inteligencia Artificial 12, 75–84 (2001)
11. Tekiner, F., Ghassemlooy, F.Z., Al-khayatt, S.: The AntNet Routing Algorithm - Improved
Version. In: Proceedings of the International Symposium on Communication Systems
Networks and Digital Signal Processing (CSNDSP), Newcastle, UK, pp. 22–28 (July
2004)
12. Soltani, A., Akbarzadeh, T.M.-R., Naghibzadeh, M.: Helping ants for adaptive network
routing. Journal of the Franklin Institute 343(4), 389–403 (2006)
13. Gupta, A.K., Sadawarti, H., Verma, A.K.: Computation of Pheromone Values in AntNet
Algorithm. International Journal of Computer Network & Information Security 4(9),
47–54 (2012)
14. Radwan, A., Mahmoud, T., Houssein, E.: AntNet-RSLR: a proposed ant routing protocol
for MANETs. In: Proceedings of the First Saudi International Electronics, Communica-
tions and Electronics Conference (SIECPC 2011), April 23-26, pp. 1–6 (2011)
15. Sharma, A.K.: Simulation of Route Optimization with load balancing Using AntNet Sys-
tem. IOSR Journal of Computer Engineering (IOSR-JCE) 11(1), 1–7 (2013)
16. Tatomir, B., Rothkrantz, L.: Dynamic traffic routing using Ant based control. In: IEEE In-
ternational Conference on Systems, Man and Cybernetics (SMC 2004) on Impacts of
Emerging Cybernetics and Human-machine Systems, vol. 4, pp. 3970–3975 (October
2004)
17. Boehlé, J., Rothkrantz, L., van Wezel, M.: CBPRS: a city based parking and routing sys-
tem. Technical report ERS-2008-029-LIS, Erasmus Research Institute of Management,
ERIM, University Rotterdam (2008)
Improving the Performance of TAntNet-2 Using Scout Behavior 435
18. Kammoun, H.M., Kallel, I., Adel, M.A.: An adaptive vehicle guidance system instigated
from ant colony behavior. In: 2010 IEEE International Conference on Systems Man and
Cybernetics (SMC), pp. 2948–2955. IEEE (2010)
19. Claes, R., Holvoet, T.: Ant colony optimization applied to route planning using link travel
time predictions. In: 2011 IEEE International Symposium on Parallel and Distributed Pro-
cessing Workshops and Phd Forum (IPDPSW), pp. 358–365. IEEE (2011)
20. Ghazy, A.M., El-Licy, F., Hefny, H.A.: Threshold based AntNet algorithm for dynamic
traffic routing of road networks. Egyptian Informatics Journal 13(2), 111–121 (2012)
21. Kashefikia, M., Nematbakhsh, N., Moghadam, R.A.: Multiple Ant-Bee colony optimiza-
tion for load balancing in packet switched networks. International Journal of Computer
Networks & Communications 3(5), 107–117 (2011)
22. Raghavendran, C.V., Satish, G.N., Varma, P.S.: Intelligent Routing Techniques for Mobile
Ad hoc Networks using Swarm Intelligence. International Journal of Intelligent Systems
and Applications (IJISA) 5(1), 81–89 (2013)
23. Rahmatizadeh, S., Shah-Hosseini, H., Torkaman, H.: The Ant-Bee Routing Algorithm: A
New Agent Based Nature-Inspired Routing Algorithm. Journal of Applied Sciences 9(5),
983–987 (2009)
24. Pankajavalli, P.B., Arumugam, N.: BADSR: An Enhanced Dynamic Source Routing Algo-
rithm for MANETs Based on Ant and Bee Colony Optimization. European Journal of Sci-
entific Research 53(4), 576–581 (2011)
25. Kanimozhi Suguna, S., Uma Maheswari, S.: Bee - Ant Colony Optimized Routing for Ma-
nets. European Journal of Scientific Research 74(3), 364–369 (2012)
26. Ghazy, A.: Enhancement of dynamic routing using ant based control algorithm. Master
thesis, Institute of Statistical Studies and Research, Cairo University (2011)
27. Ghazy, A.: Ants Guide You to Good Route: Dynamic Traffic Routing of Road Network
using Threshold Based AntNet. Lap Lambert Academic Publishing (2012)
28. Baykasoglu, A., Ozbakir, L., Tapkan, P.: Artificial bee colony algorithm and its applica-
tion to generalized assignment problem. In: Chan, F.T.S., Tiwari, M.K. (eds.) Swarm Intel-
ligence. Focus on Ant and Particle Swarm Optimization, pp. 113–144. ITech Education
and Publishing, Vienna (2007)
29. Akbari, R., Mohammadi, A., Ziarati, K.: A novel bee swarm optimization algorithm for
numerical function optimization. Communications in Nonlinear Science and Numerical
Simulation 15(10), 3142–3155 (2010)