Professional Documents
Culture Documents
Abstract—This paper presents a genetic algorithm for resource these problems. Computer evolution and swarm intelligence
scheduling in service discovery to MANETs operating in emer- are examples of what nature has given an inexhaustible font
gency scenarios. The shared resources could be ambulances or of inspiration [2].
support cars. Through an appropriate model for scheduling
resources, we aim to attend the greatest number of victims in The Genetic Algorithms (GAs) consist of a Computer
the affected area. Performance evaluation results on the Network Intelligence technique (CI) of which is fundamental in theories
Simulator NS-3 confirm the effectiveness of genetic algorithm and concepts of genetics and the evolution of species according
for resource scheduling when compared to the A-Star algorithm to the natural selection process and discordance with heredity,
which is simple heuristic method, that prioritizing paths that proposed by Darwin [3]. GAs is a probabilistic search heuristic
seem to be leading closer to the goal.
and recognized for combinatorial problems, and are an excel-
I. I NTRODUCTION lent choice for application types where optimized variables
are very different from each other (e.g., a mixture of integers,
A natural disaster or not is an emergency situation which binary values, and floating points). Its configuration can be
needs the management of efficient rescue teams to attend a modified to include different mutation operators for different
majority of the victims. A search of service providers (e.g. sections of the chromosome.
ambulances, support cars or the fire service) for the victims From the population point of view, by using a representation
should occur quickly after the disaster. by integers, it becomes feasible its use in the scenario proposed
Within this context protocol to discover and resource reser- in this work. It is important to note that even for applications
vation put into place the essential services necessary to en- where other optimization methods may be better (e.g., Simu-
able them to locate the victims and allocate the necessary lated Annealing), a good viable solution search performance
resources (service providers) to attend. However, in emergency can be obtained using GA, provided that by performing several
situations, the communication infrastructure could be damaged experiments with the optimization parameters.
and might be impossible to provide adequate communication The GAs propose to solve complex problems of the real
services for the rescue teams and victims (clients). world where traditional optimization methods that are used
The Mobile Ad Hoc Networks MANETs, are particularly have difficulties to find viable solutions. The main function
attractive for these scenes where there is a need to install consists to explore the space being used in searching for a
a communication Network immediately. MANETs can be population of solutions provided (chromosomes), which are
used in emergency situations to exchange information between evolving by each generation by means of genetic operators
clients and the rescue services (service providers). With this of crossover and mutation. GAs requires a definition of the
information the providers will be able to make decisions to codified problem and possible solutions for the matter in the
see if assistance will be performed [1]. form of chromosomes (individual) thereby evaluating each
The fire and ambulance services are indispensable and are solution according to the function of the evaluation or to the
not always available and satisfactory in many cities, with the (Fitness). Between the basic components of genetic algorithms
proposal of efficient medical assistance in emergencies or after which stand out are: chromosomes (individual or possible
disasters. The swiftness at which teams respond when called solutions), crossover operator, mutation operator, selection
and to provide necessary assistance to the emergency scene operator and aptitude function [4].
are essential for for the success of the service. The number of victims in an emergency situation could
The complexity to manage resources in relation to the increase considerably according to the severity of the catastro-
amount of assistance requires the use of adequate algorithms phe. Therefore, characterizing the scheduling of the resources
of scaling necessary resources. at the scene targets a permutation problem.
The growing complexity of global problems has motivated Besides the number of victims, a scheduling algorithm in
computer scientists to search for efficient methods to resolve this context, should attended the maximum time required
1046
aspects related to the detection of the service in emergency The discovery message will be transmitted to the network,
situations have not been considered in this paper. hop-to-hop, through a diffusion mechanism of the broadcast
Gadallah and Serhani [17] have presented a protocol of type and all nodes can act as a relay, passing the information.
the discovery of services for MANETs distributed in disaster The service discovery protocol must be preconfigured on the
areas. The providers of the service send reports to a central devices of victims and providers. The message will travel
who then select the providers by the clients. The protocol within a radius Ri delimited by the service discovery engine.
allows to locate services based on the requirements specified The service selection mechanism automatically selects the re-
of the client, namely the proximity of the service provider and sponses from the best providers, while forwarding the message
the severity of the emergency. on the network and forwards it to the victims. Because each
The article proposed by [18], presents a distributed ar- provider keeps the information about the resource it makes
chitecture of discovery and service sharing in MANETs. available stored, soon any node outside the radius that receives
The proposal consists to integrate a protocol of proactive the discovery message will discard it.
router Destination-Sequenced Distance Vector (DSDV) with a In the discovery message, the client node sends the follow-
protocol of discovery service. The protocol DSDV Maintains ing information: an identification of the node, an identifica-
information used by the user in the routing table. This infor- tion of the request, the sequence number of the request, its
mation is used to discovery the service location according to geographical coordinates, the maximum time for requesting
the type needed. Also nodes are being used for information the request (Δtmax ), The service sought, and the number of
like geographic position, network topology and the type of providers wanted [1] and [13].
service. The architecture of the protocol incorporates: service After receiving the responses from the apt providers, the
discovery, a register of the service and service summoned. client sends a service invocation message to the selected
In general, service discovery protocols implement the ser- provider(s). These, in turn, should send a service confirmation
vice invocation phase, as occurs with protocols LADP [13], message to the client. Finally, the client sends a confirmation
BAHSD proposed by [14], and used by authors Gadallah and acknowledgment message to the provider.
Serhani [17] and Park [18]. Each confirmation acknowledgment message received is
However, protocols for discovery services existing in the intended to notify the client that the provider will physically
literature were evaluated in the work of [Kniess, 2011] and move to the location where the resource necessary. The algo-
[13], where the author verified that the protocols do not rithm GAS acts in the phase of the invocation of the protocol
provide mechanisms to optimize the use of the providers service LADP.
in relation to the victims (people) in the affected area. To The entity responsible for the scheduling is the provider
address this limitation of service discovery protocols for the node. The service request is performed by the client through
target scenario the algorithm proposed in this article has been the discovery message. After the provider is selected as
developed. it establishes the protocol selection mechanism LADP, the
The discussion presented about the protocols [1], [13], [14], client sends an invocation message to the provider that upon
[17] and [18], enforce the necessity of a new mechanism receiving it initiates a timer to wait for invocation messages
of innovation which contemplates the peculiarities of the from other clients.
target scenario of this work. In the target scenario the service The Figure 1 illustrates the function of the scheduling
providers should dislocate to a place when asked and necessary algorithm of resources GAS from when the summons message
within a time limit and attend a maximum of victims possible. is sent to the client (1, 2, 3 and 4) requesting for an ambulance
In view of the constraints of the protocols of the detection (provider P) to attend in a maximum time of 10 min. After
service operating to attend the scheduling of resources to provider P receives the first message of invocation from a
the target scenario proposal, this article offers a solution in client (e.g., client 1), the provider initiates a timer ΔAG . This
the form of a genetic algorithm , observing the geographical timer represents the waiting time of the node provider for
location of the client, the dislocation time of the local provider the invocation messages. For invocation messages received
and the maximum attendance time determined by the client the node provider sends the client a message confirming the
to decrease the dislocation time of the provider and also to invocation in unicast mode and waits for the acknowledgment
increase the number attended. to be confirmed.
When the timer ΔAG ends, an adjacent matrix is generated
III. G ENETIC A LGORITHMS FOR S CHEDULING THE containing information related to the time of dislocation of
R ESOURCES - GAS the provider to each of the soliciting clients and also to
This section presents genetic algorithms for scheduling their pair/partner a carrier, and a vector containing the time
resources operating in emergency scenarios, the detail of the stipulated by the client for the service (windowing). The
problem (Section III-A) and the developed GA (Section III-B). Figures 2(a) and 2(b) brings a conceptual illustration of the
adjacencies and the windowing.
A. Detail of the Problem Next, it is shown how the GAS algorithm organize sequence
In protocol LAPD, the victims begin the service discovery of service in order to aiming maximize the number of clients
process through a mobile device interconnected with MANET. served within the defined time window. Having received the
1047
response of GAS, the provider sends a message notifying the then go to the next point until all paths are complete, without
attendance of the clients and starts travelling towards the first repeating the edges and without returning to the initial position
client. of the provider. The Equation (4) determines where it not
possible for a node to go by itself. The Equation (5) represents
confirmation of the shift xij , thus xij = 1, if the provider goes
to dislocate from node i to node j, or xij = 0, if the contrary
where to happen.
In the Equation (6) adjusts itself a time window with the
time to attend for each point of the defined path, N = (i, j).
With this, for each client node i defines a time window [tmaxi ],
being tmaxi a vector contains the attendance time for each
client. This way a provider cannot arrive at the location after
tmaxi of each victim.
The main objective will be reached and a reduction of the
cost of dislocation and minimizing the total travel time and
increasing the number of clients served.
Fig. 1. Illustration of the problem
B. GAS Project
The mathematical formalization for the problem of schedul- The problem to be solved is to make the GA, from a
ing of resources based on TSPTW require a definition of list of adjacencies of time of displacement as shown in the
a group of fundamental parameters: N = {1,2,...,n} which Figure 2(a), generate a queue of attendance, having as priority
represent a group with n nodes on the plan. The cost associated the time of displacement and the time of attendance (Δtmax )
with travelling time between the nodes i, j ⊂ N is given specified in the detection service message sent by the client.
for dij , and, ending in, xij , represents the number of times The codification of the individual and given by the permu-
the connection between the node i and node j (designated by tation of a finite group and their elements, in the case of the
decision variables). clients. The codification defines a path at the origin of the
The mathematical model of calculation of the function destination in which the time of dislocation for the attendance
objective is represented in Equation (1), which is consistent will be measured. Once the initial solution is formed randomly
to adjust travelling time to the service provider which goes it is necessary to calculate the aptitude of each individual case,
through distinct point once only. The variable p consists of representing its fitness.
the number of penalties that will occur for each possible In the case if this example the cost is associated with the
solution in the population, in which your feasibility is verified, path, in the mode of the lower the cost the best solution will be
in other words, valid. The variable m represents the value of found. The Equation (1) generates an evaluation grade starting
the sanction. The Equations (2) and (6) are restrictions of the from the matrix of time of dislocation generated by Network
objective function. Simulator 3 (NS3) [19].
n
n Each individual is considered a possible path. For the formu-
min dij xij + p.m (1) lation of the function some restrictions have been considered,
i=1 j=1 for example, verifying the viability if the provider can attend
Subject to restrictions: the next client or not, if the time is exceeded (Δtmax ), and if
n the provider continues to the next attendance or not. It will be
xij = 1 ∀i ∈ N (2) considered a valid solution if the result that makes it possible
i=1
to attend all clients within the maximum time of attendance.
n In cases where it does not result in a solution which caters to
xij = 1 ∀i ∈ N (3) all clients, the one that offers the best solution is chosen.
j=1 In the objective function (Equation 1) weights and measures
n penalties are given according to the requirements of atten-
xij = 0 ∀i ∈ N (4) dance. The variable p in the objective function is consistent
i,j∈S
to the number of penalties. In case there is no valid solution
which composes a way, the same is judged invalid. As a result,
xij ∈ {0, 1} ∀i, j ∈ N, i = j (5) receiving a high score you will be penalized thus decreasing
n
their chances of being selected in the next generation, adopting
di + dj ≤ ti ∀i, j ∈ N (6) the following premise: the higher the score the worse the path.
i,j∈S The individual which represents the route of the provider
Equations (2) and (3) establish restrictions for which the for attendance of clients is composed of a ranking between
provider does once only for each node apploicant and should solicited clients.
1048
The Figure 2(b) presents an example of a possible solution not seen in the original version could occur.
according of matrix to time dislocation (Figura 2(a)). In the Consider in Figure 1 when a client sends a message of
Matrix demonstrates a trajectory containing 4 (four) solicited acknowledgment confirming the service for the provider. How-
clients from a window of 10 minutes. ever, assume the message was lost in the network and the
To attend each client, the provider P took a total of 8.15 provider did not receive the message acknowledging confir-
minutes, thus defined as fitness. In other words it was possible mation within the time period ΔAG . In this case, the provider
to attend all the clients within the solicited time. The Figure will send a message (NACK recognizing) the client, commu-
2(b) represents a valid solution given to everyone who was nicating that the message of acknowledgment of confirmation
attended within the specified time (Δtmax ) by the client. That was not received. Before starting on route to the first client,
is, there were no penalties. the provider should try another two times to communicate with
the client.
If the provider receive acknowledgment of these clients
confirmation before starting the displacement, the provider will
include again the client in the GAS calculation so as to define
the sequence of attendance. It is worth to highlight when the
execution of GAS started after the set time ΔAG expired.
If a failure occurs, as mentioned above, considering that the
provider receives the message acknowledging confirmation of
the client when on route, soon, was defined in GAS that the
(a) provider will continue service in the direction of the first client.
In the event of the first attendance, the mobile device of the
provider analysis new messages invocation received. In the
case that the node provider verifies the viability of including
a new attendance on route, without prejudicing the window
previously defined, the provider will send a alert message to
the clients already selected and the new clients announcing
the service.
1049
TABLE I
min. And in the second case study, the value of Δtmax is of C ASE 1
maximum 20 min. However, it was randomly assigned values
between 10 and 20 min for client service. For both studies, GAS A-Star
Clients Time Clients Time
the number of clients requests varies between 10 and 30. Clients
Served Served Served Served
The metrics evaluated consist of: the number of clients 10 10.00 ± 0.00 16.81 ± 1.59 10.00 19.21
20 16.00 ± 1.05 19.06 ± 0.73 10.00 13.99
attended with in the time limit and the fitness (in relation to 30 19.70 ± 1.06 19.27 ± 0.64 19.00 19.78
genetic algorithms). The fitness considers the time the provider
takes to arrive at the client. We considered 1,000 generations TABLE II
because we did not know the behavior of a GA in this type C ASE 2
of application. GAS A-Star
Some parameters are not modified during the experiments. Clients Time Clients Time
Clients
These are: (i) the Displacement Time Matrix, which is the Served Served Served Served
10 09.50 ± 0.53 15.76 ± 1.81 08.00 14.88
time it takes to reach each client to the provider leaving and 20 12.80 ± 0.79 16.45 ± 1.22 10.00 13.41
its adjacent pairs; (ii) The Attendance Time (window), which 30 15.40 ± 1.35 16.76 ± 1.21 13.00 13.33
represents the maximum time attendance. This acts on the
problem as a limited space to find the solutions, ensuring that
the valid script does not exceed the time provided by the client; algorithms regarding the amount of customer served and the
(iii) the Genetic Algorithm, in which the number of runs = 20, time of service. In the simulation tests were considered with
the population = 100, the number of generations = 1,000, the 10, 20 and 30 clients for service. For Case Study 1, it was
value of the penalty m = 100, the method to change the order considered the homogeneous service time equal to 20 minutes
of the genes = 1% (SwapMutator) and, the Partial Matched for all clients. In the Case Study 2 a time of Heterogeneous
Crossover - PMX = 90%. service that varies between 10 and 20 min.
For the purpose of comparison the performance and results In Case 1 (Table I), the GAS algorithm obtained the best
obtained from GAS, algorithm was compared with the results attendance average when compared to the A-Star algorithm.
from the algorithm A-Star, being a search algorithm in graphs When 10 clients requested the appeal, the GAS reached the
that finds a path between a given source node and a given average travel time of 16.81 min to serve all clients. In this
destination node. same scenario, A-Star also served 10 clients, however, it spent
The algorithm A-Star, is known for its performance and 19.21 min for commuting.
accuracy and works well when making the search path for As the number of clients increases, the GAS algorithm
optimization problems. Therefore, it makes the algorithm pointed to better performance. For example, with 30 clients
a good candidate for optimal path optimization problems. the GAS serviced on average 19.70 clients within the time of
The characteristics of this algorithm motivated us to explore 19.27 min. The A-Star served close to 19 clients, but in a time
its operation in the optimization problems for the traveling of 19.78 min.
salesman with time windows [22]. The results presented in Case 2 (Table II), show the effect of
A-Star using min spanning tree heuristic chooses the current GAS on different treatment times. The results obtained confirm
best path base on function compound: the effectiveness of the GAS algorithm in the optimization of
the displacement time and in the maximization of the number
f(n)=g(n)+h(n) (7) of attendances.
f (n) ≤ ti (8) In a scenario with 10 clients, the GAS served, on average
9.5 clients in 15.76 min. On the other hand, A-Star served 8
The Function 7 is represented per g(n), where is the cur- clients with an average time of 14.88 min.
rently computed cost and h(n) an estimation of the remaining For the scenario with 20 clients, it was observed that the
cost. Only the most promising node (the one with the smallest GAS serviced on average 12.80 clients in 16.45 min and the
f(n) respecting the client time window ti (Function 8)) at the A-Star served only 10 with the time of 13.41 min.
moment will be expanded. In the scenario with 30 clients, the GAS served an average
The programming language used is Java on a hardware of 15.40 clients in 16.76 min, A-Star served only 13 clients
feature of the computer used in the simulations comprise a in 13.33 min.
Core i7 processor, 8 GB RAM memory and a 100 GB SATA The results obtained in the study of Case 2 presented
disk. an inferior performance in relation to the study of Case 1,
attributed this result to the fact that the provider is willing to
B. Computational Results the time of heterogeneous service requested by the clients.
The results described refer to the average of 20 runs Based on the described results, it was observed that the
followed by their standard deviation. The number of clients probabilistic method of the GAS algorithm demonstrated better
served and the service time obtained in minutes are considered. performance when compared to the simpler heuristic method
Tables I and II show the results obtained for Case 1 and used with the A-Star.
Case 2 respectively of the execution of the GAS and A-Star From this experiment, it was found that as the displacement
1050
time between the points increases, the GAS served more clients TSPTW is adequate for a possible solution of the problem.
in relation to A-Star algorithm. Specifically, the GAS algorithm was effective against all the
The results presented by the A-Star algorithm suggest that metrics evaluated, allowing a provider to be able to attend to
the absence of adaptive change in the parameters of the a greater number of victims and in a timely manner.
evaluation function and the decision parameters resulted in a The comparative approach is simple yet feasible for solving
decrease in the performance of the algorithm when compared the problem. The objective is to analyze heuristic methods
to the GAS. applied to MANETs operating in emergency scenarios.
Another objective of this work was to evaluate the process- In the short term we will look at cases of failures where
ing time of both algorithms. The GAS processing time was on the provider receives the acknowledgment message from the
average 0.88 seconds. In contrast, the A-Star averaged 0.055s. client’s confirmation when the provider is already on the move.
However, the GAS served a larger number of clients and within The average time, measurements of the overhead and power
the defined time limit. consumption of the devices involved in the scenario will be
Aiming to reduce service time and increase the number of carried out. It is also thought to evaluate other probabilistic
clients served, the target object of the genetic algorithm for methods (e.g., Simulated Annealing, Ant-Colony) and to com-
proposed resource scheduling, the convergence capacity of the pare them with the genetic algorithm. It is also suggested to
GAS algorithm was evaluated. The performance of the GAS model the GAS in a multi-objective approach considering the
had an equivalent behavior in all case studies and an example minimization of time and the maximization in the time of
can be visualized in the graph of convergence of the Figure 3. service.
R EFERENCES
[1] J. Kniess, O. Loques, and C. V. N. de Albuquerque, “Descoberta de
servio em redes ad hoc mveis,” Ph.D. dissertation, Universidade Federal
Fluminense (UFF), 7 2011.
[2] R. S. Parpinelli and H. S. Lopes, “New inspirations in swarm intel-
ligence: a survey,” International Journal of Bio-Inspired Computation,
vol. 3, no. 1, pp. 1–16, 2011.
[3] B. M. Varghese and R. J. S. Raj, “A survey on variants of genetic
algorithm for scheduling workflow of tasks,” in Science Technology
Engineering and Management (ICONSTEM), Second International Con-
ference on. IEEE, 2016, pp. 489–492.
[4] A. Shukla, H. M. Pandey, and D. Mehrotra, “Comparative review
of selection techniques in genetic algorithm,” in Futuristic Trends on
Computational Analysis and Knowledge Management (ABLAZE), 2015
International Conference on. IEEE, 2015, pp. 515–519.
[5] C. Xiong and W. Qidi, “Formulating the steel scheduling problem as a
TSPTW,” in Intelligent Control and Automation, 2002. Proceedings of
the 4th World Congress on, vol. 3. IEEE, 2002, pp. 1744–1748.
Fig. 3. Fitness - Graphic Convergence [6] A. N. Mian, R. Baldoni, and R. Beraldi, “A survey of service discovery
protocols in multihop mobile ad hoc networks,” IEEE Pervasive Com-
puting, vol. 8, no. 1, pp. 66–74, 2009.
The x-axis represents the fitness and the y-axis and the [7] D. Chakraborty, A. Joshi, Y. Yesha, and T. Finin, “Toward distributed
number of generations. The average of the 20 executions service discovery in pervasive computing environments,” IEEE Trans-
actions on Mobile Computing, vol. 5, pp. 97–112, 2006.
achieved in each generation is illustrated in terms of the cost
[8] S. Helal, N. Desai, V. Verma, and C. Lee, “Konark - a service discovery
of travel time to clients. and delivery protocol for ad-hoc networks,” Wireless Communications
Analyzing the convergence, we can see the convergence of and Networking, WCNC 2003. IEEE, pp. 2107–2113 vol.3, 2003.
algorithm GAS there occurred the same before the maximum [9] M. Klein, B. Konig-Ries, and P. Obreiter, “Lanes – a lightweight overlay
for service discovery in mobile ad hoc networks,” Universit at Karlsruhe,
was reached of the generations available. This result indicates Faculty of Informatics, Tech. Rep. 2003/6, 2003.
that a solution can be reached before the performance of all [10] V. Lenders, M. May, and B. Plattner, “Service discovery in mobile ad hoc
the generations. networks: A field theoretic approach,” Pervasive Mob. Comput., vol. 1,
pp. 343–370, September 2005.
[11] O. Ratsimor, D. Chakraborty, A. Joshi, T. Finin, and Y. Yesha, “Service
V. C ONCLUSIONS discovery in agent-based pervasive computing environments,” Mobile
The use of a genetic algorithm to solve the problem of Networks and Applications, vol. 9, pp. 679–692, December 2004.
[12] A. Varshavsky, B. Reid, and E. de Lara, “A cross-layer approach
scheduling of resources was motivated by the efficiency of to service discovery and selection in MANETs,” IEEE International
GAs in the solution of combinatorial problems. The results Conference on Mobile Adhoc and Sensor Systems Conference, vol. 0, p.
obtained through the NS3 simulator demonstrated that the 466, 2005.
[13] J. Kniess, O. Loques, and C. V. N. de Albuquerque, “Service discovery
heuristic methods are suitable for the scheduling of resources with time constraints in mobile ad hoc networks,” Earth Science
in emergency scenarios. Informatics, vol. 8, no. 2, pp. 439–452, 2015.
The proposed solution, we characterized the scaling of [14] G. Arenella, F. de Santis, and D. Malandrino, “Beeadhocservicedisco-
very: A MANET service discovery algorithm based on bee colonies,” in
resources in the target scenario as a permutation problem, Informatics in Control, Automation and Robotics (ICINCO), 2014 11th
so it was identified in the literature that the meta-heuristic International Conference on, vol. 1. IEEE, 2014, pp. 244–251.
1051
[15] S. Chalotra, S. K. Sehra, and S. S. Sehra, “A systematic review of
applications of bee colony optimization,” in Innovation and Challenges
in Cyber Security (ICICCS-INBUSH), 2016 International Conference
on. IEEE, 2016, pp. 257–260.
[16] H. N. Saha, A. Chattopadhyay, and D. Sarkar, “Review on intelligent
routing in MANET,” in Computing and Communication (IEMCON),
2015 International Conference and Workshop on. IEEE, 2015, pp.
1–6.
[17] M. A. Serhani and Y. Gadallah, “A service discovery protocol for
emergency response operations using mobile ad hoc networks,” in
Telecommunications (AICT), 2010 Sixth Advanced International Con-
ference on. IEEE, 2010, pp. 280–285.
[18] J. C. Park, M. S. Choi, B. J. Lee, K. T. Kim, and H. Y. Youn, “Distributed
semantic service discovery for MANET,” in Ubiquitous Intelligence
and Computing, 2013 IEEE 10th International Conference on and
10th International Conference on Autonomic and Trusted Computing
(UIC/ATC). IEEE, 2013, pp. 515–520.
[19] M. Ikeda, E. Kulla, L. Barolli, M. Takizawa, and R. Miho, “Performance
evaluation of wireless mobile ad-hoc network via NS-3 simulator,” in
Network-Based Information Systems (NBiS), 2011 14th International
Conference on. IEEE, 2011, pp. 135–141.
[20] D. Chenghao, “An improved routing protocol based on gauss-markov
model in ad hoc networks utilizing prediction of link quality,” in Control
Conference (CCC), 2015 34th Chinese. IEEE, 2015, pp. 6507–6511.
[21] N. Kumari, S. K. Gupta, R. Choudhary, and S. L. Agrwal, “New
performance analysis of AODV, DSDV and OLSR routing protocol for
MANET,” in Computing for Sustainable Global Development (INDIA-
Com), 2016 3rd International Conference on. IEEE, 2016, pp. 33–35.
[22] J. Huang, L. Sun, F. Du, H. Wan, and X. Zhao, “Genetic adaptive a-star
approach for ttrain trip profile optimization problems,” in Computational
Intelligence in Vehicles and Transportation Systems (CIVTS), 2014 IEEE
Symposium on. IEEE, 2014, pp. 129–134.
1052