VAST: Volume Adaptive Searching Technique for Optimized Routing ...

10 downloads 0 Views 176KB Size Report
Feb 22, 2011 - for Optimized Routing in Mobile Ad-hoc. Networks. Pragati Dahiya. Research Scholar. USICT, Guru Gobind Singh Indraprastha University.
VAST: Volume Adaptive Searching Technique for Optimized Routing in Mobile Ad-hoc Networks Pragati Dahiya

Rahul Johari

Research Scholar USICT, Guru Gobind Singh Indraprastha University Sector-16C Dwarka, New Delhi, India [email protected]

Assistant Professor USICT, Guru Gobind Singh Indraprastha University Sector-16C Dwarka, New Delhi, India [email protected]

packets, routing protocol are classified in three types: deterministic, enforced and opportunistic approach[6,7]. Routing algorithm for delay tolerant network based on opportunistic approach is proposed in[6].

Abstract— In this paper we propose a searching technique termed as VAST to determine an optimal path from source node to destination node in densely deployed mobile ad-hoc network. We have compared the proposed VAST algorithm with metaheuristic algorithms GRASP, Semi-greedy and Tabu search, in terms of the routing cost and algorithm execution time. The comparison results shows that the proposed VAST algorithm outperforms the other algorithm and that it is suitable for adapting the routing optimization problem.

An extension of the basic GRASP procedure, which they call reactive GRASP, in which the parameter Į is self-tuned and its value is periodically modified according with the quality of the solution obtained recently[3]. Semi-greedy heuristic approach is a multistart approach based on greedy randomized construction, but without local search [3][8]. We have taken three maximum iteration for multi-start metaheuristic approaches GRASP, Reactive GRASP and Semi-greedy heuristic approach. We have used Tabu search in local search phase of GRASP and Reactive GRASP. In GRASP, we have taken Į= 0.2. In Reactive GRASP, we have taken a set of three values of Į (0.200,0.225,0.250).

Keywords- MANET(Mobile Ad-hoc Network), GRASP (Greedy Randomised Adaptive Search Procedure), Reactive GRASP, Tabu search, Semi-greedy heuristic approach.

I.

Introduction

In MANET topology changes frequently as it comprises of mobile nodes which are free to change their location. Since connections of nodes in network changes with time, optimized routing is one of the basic requirement in MANET to deliver the packet timely from source node to destination node before the routing path fails. Optimization in routing is delivery of packets from source to destination in minimum time with minimum routing cost.

The proposed VAST algorithm has two phases: construction and local search. The construction phase builds initial feasible solution of the problem. Local search phase is used to find local minima in neighborhood of initial solution generated from construction phase. Best result from local search is kept as final solution. It involves division of total searching space volume into eight 3D quadrants. It is based on adapting the searching space volume in construction phase until initial route from source node to destination node is established.

Tabu search is a dynamic neighborhood local searching technique[1]. Tabu search algorithm is proposed for optimized routing in MANET which performs better than other metaheuristic simulated annealing and genetic algorithm as in [2]. GRASP is a multi-start metaheuristic for combinatorial problems in which each iteration consists basically of two phases: construction and local search [3]. GRASP construction phase has a parameter Į, its value alters the greediness in the construction phase on the basis of cost and is used to form Restricted Candidate list (RCL). GRASP involves randomization by selecting nodes randomly from the RCL in forming initial solution of construction phase.

II. Vast Algorithm 1) Select the source node=S and destination node=D 2) Source node identifies its quadrant 3) Construction phase: Start: Step1: Searching space volume is the 3D quadrant volume in which source node lies, if initial route is established from source node to destination node in this space then stop otherwise go to step2.

GRASP-BSMA solves fast the constrained minimum spanning tree(CSMT) problem in multicast routing keeping acceptable tree cost[4]. Hybrid simulated evolution-Tabu Search has better performance than the simulated evolution algorithm for optimized routing in computer network[5]. In delay tolerant network that supports delayed transmission of

c 978-1-4799-2572-8/14/$31.00 2014 IEEE

Step2: Searching space is as in step1 in addition of first quadrant volume, if initial route is established

1

from source node to destination node in this space then stop otherwise go to step3. Step3: Searching space is as in step2 in addition of second quadrant volume, if initial route is established from source node to destination node in this space then stop otherwise go to step4. Step4: Searching space is as in step3 in addition of third quadrant volume, if initial route is established from source node to destination node in this space then stop otherwise go to step5 Step5: Searching space is as in step4 in addition of fourth quadrant volume, if initial route is established from source node to destination node in this space then stop otherwise go to step6. Step6: Searching space is as in step5 in addition of fifth quadrant volume, if initial route is established from source node to destination node in this space then stop otherwise go to step7. Step7: Searching space is as in step6 in addition of sixth quadrant volume, if initial route is established from source node to destination node in this space then stop otherwise go to step8. Step8: Searching space is as in step7 in addition of seventh quadrant volume, if initial route is established from source node to destination node in this space then stop otherwise go to step9. Step9: Searching space is as in step8 in addition of eigth quadrant volume, if initial route is established from source node to destination node in this space then stop otherwise go to step10. Step10: Searching space is total volume of searching space. Route is established from source node to destination node. Stop 4) Local search phase : Tabu search is used in this phase.

transmission range. Nodes lying in transmission range of a node are called here the neighboring nodes of that node. By using vector equation of calculating distance between two locations in Cartesian coordinate system determine and store the neighboring nodes and their distances from a node in global memory that can be accessed by all the nodes. 2) Nodes quadrant function: This function is used to identify quadrant in which a node lies and stores it in global memory that can be accessed by all the nodes. 3) Generating Initial solution function: The selected source node is made the first element of the initial solution. The nodes that are present in initial solution are considered as Tabu nodes. These nodes cannot be inserted again in the initial solution, this is to avoid the routing loop in initial solution Destination node is searched in space restricted feasible neighboring nodes of the inserted node which are the feasible nodes that are present in searching space of each step of the construction phase. If destination node is present in space restricted feasible neighboring node then it is selected and inserted in initial solution and function stops, otherwise a node other than the Tabu nodes is chosen randomly from space restricted feasible neighboring nodes and inserted at next place in the initial solution. In Fig.1. S= Source node, D=Destination node and In= nth Intermediate node S

I1

I2











In

D

Fig.1.Initial solution

4) Remove node move function: This function works on initial solution to improve it. In initial solution, node at ith location searches in its neighboring node memory whether the node at (i+2)th location is there. If it is present then( i+1)th node is deleted and (i+2)th node is made the node next to ith node. Resulting solution is stored in Tabu list. Same procedure is repeated on improved solution obtained by previous step to further improve it. Fig.2.shows working of remove node move function.

5) Selection of the best result. III. Functional Blocks Nodes in transmission range of a node are determined by using Neighboring nodes searching function. Source node and destination nodes are selected. Initial solution of construction phase is generated using Generating Initial solution function. Remove node move function and Replace node move function are local search phase functions of Tabu search that are used to improve the initial solution. All improved solutions are stored in Tabu list. Cost function calculates the cost of all the paths stored in Tabu list and returns the path having minimum cost. Node location change function is used to induce mobility in the network. Scan function is called after each function to check the change in location of nodes. 1) Neighboring nodes searching function: Node location is given by 3D Cartesian coordinates. Each node has a

2

Fig.2. Working of Remove node move function

5) Replace node move function: This function also works on initial solution to find better solution .This function replaces an intermediate node at ith location in between source and destination node in initial solution from the node that is present in neighboring node of the (i-1)th node taking in account this replacing node is not present in tabu nodes. This replacing node now present at ith location searches nodes preceding it, in its neighboring node memory, if any of them is present then that node is made next node to the replacing node. Solution obtained by above procedure is compared with solutions already stored in Tabu list. If it is found different

2014 IEEE International Advance Computing Conference (IACC)

TABLE I

from all solutions stored in Tabu list then it is inserted in tabu list. Fig.3.shows working of replace node move function

FIGURES DESCRIPTION

FIGURE Fig.4.

Fig.5. Fig.6. Fig.3.Working of Replace node move function

Fig.7.

6) Cost function: For all routing solution from source to destination stored in Tabu list, cost is calculated by this function. Cost is calculated by summing the link cost for entire solution which is proportional to the distances between two consecutive nodes. The solution having minimum cost value is the optimized routing path between source and destination. 7) Node location change function: This function is used to support mobility of nodes. During computation of optimized route, it is called random number of times between values zero and four. This function selects randomly the number of nodes in network that will change their location. Change value is selected randomly between one and N (number of nodes in network) for all three coordinates. The coordinates of the randomly selected nodes are incremented or decremented, which is also chosen randomly, by the random change value. 8) Scan function: This function helps to monitor the mobility of nodes. This function is called after every function to test if there is any change in location of any node. If there is change in value of sum of x-coordinate or sum of y-coordinate or sum of z-coordinate of all the nodes present in network, then this means that node have changed its locations. This function then calls the neighboring nodes searching function to determine and store the changes in neighboring node and their distances in global memory.

Fig.8. Fig.9. Fig.10. Fig.11.

Fig12.

Fig.13.

Fig.14.

Fig.15.

Fig.16.

EXPLANATION Circular dots denotes the nodes in network (N=30) which are numbered from 1 to 30, present in volume 300x300x300m3 First Change in location of mobile nodes which is carried out by using Node location change function Shows routing path from source node=1 to destination node=27 using Semi-greedy heuristic approach Shows routing path from source node=1 to destination node=27 using Tabu search algorithm. Shows routing path from source node=1 to destination node=27 using GRASP algorithm. Shows routing path from source node=1 to destination node=27 using Reactive GRASP algorithm. Shows routing path from source node=1 to destination node=27 using VAST algorithm. Comparison of the routing cost of different paths determined by l above stated algorithms for N=30 node network present in volume 300x300x300m3 Comparison of the computation time of different paths determined by above stated algorithms for N=30 node network present in volume 300x300x300m3 Comparison of the routing cost of different paths determined by above stated algorithms for N=20 node network present in volume 250x250x250m3 Comparison of the computation time of different paths determined by above stated algorithms for N=20 node network present in volume 250x250x250m3 Comparison of the routing cost of different paths determined by above stated algorithms for N=15 node network present in volume 200x200x200m3 Comparison of the computation time of different paths determined by above stated algorithms for N=15 node network present in volume 200x200x200m3 27 24

IV. Simulation Results

9

26 30 12

200

29

23

6 17

Z-axis

Simulations are done in MATLAB. We have compared VAST algorithm and other algorithms simulation results for finding routes in 3 different networks having N= 15, N=20 and N=30 nodes in volume of 200x200x200m3, 250x250x250m3 and 300x300x300 m3 respectively. Each node in every network has transmission range of 100m in all directions. At least one other node should lie in transmission range of a node. All the nodes have routing capability.

18 15

25021

3

14 28

15020

8 25 11

100

5

22 16 2

5019 250

13 7

200 150

150 100

Y-axis

200

4

10

100

1 50

50

X-axis

Fig.4. Initial node location (for N=30 network)

2014 IEEE International Advance Computing Conference (IACC)

3

300

300

27

27 24

24 18

250

3

28

150

17

14 25

20

23

8

100

13

50 300

7

19 250

5 16

11

10

0

Y-axis

50

X-axis

0

Y-axis

18

27

11

100

50

6

3 25 14

20

22 2

19

9

200

28 17

16 5

15

21

9

26 30 29

23

18

250

15

Z-axis

150

8

23

4

8 5 16

11 2

13 7

19 250

250

200

200

200 150

200 150 100

50

50 0

1

100

100 50

0

Y-axis

X-axis

50

X-axis

Fig.9. Reactive GRASP algorithm (for N=30 network)

Fig.6. Semi-greedy heuristic approach (for N=30 network)

27

27 300

300

24

24 18

18 21 12

Z-axis

200

11

100

50

6

3 25 14

20

150

16 5 2

8

13

10

250 200 150 100 0

4

8

13

4 1

250

250

200

200 150

150 100

100 50

X-axis

Fig.7. Tabu search algorithm (for N=30 network)

2

10

50 0

16 5

22

19

100 50

Y-axis

11

28 17

7

1

150

6

3 25 14

0 300

200

9

26 30 29

23

20

50

4

250

150

100

7

0 300

12

200

28 17

22

19

15

9

30 26 29

23

21

250

15

Z-axis

250

250

4

10 150

150 100 0

17

22 50 300

1

Y-axis

28 14 25

7

250

3

150 20

13

6

26 30

29

12

100

10

0 300

X-axis

24

21 12

50

300

27

200

150 100

Fig.8. GRASP algorithm (for N=30 network)

24

250

1 50

Fig.5. Nodes location, after first random call of function “Node location change function” (for N=30 network)

300

200 100

100

250

4

10 150

150

50

13 7

200

200 1

100

Z-axis

2 19 250

250

4

150

17

22

2

200

28 14 25

22 50 300

6

26 30

3

20

5 16

11

29

12

150

8

100

9

200 Z-axis

Z-axis

23

6

30 26

29

12

15

21

9

200

18

250

15

21

Y-axis

50 0

0

Fig.10. VAST algorithm (for N=30 network)

2014 IEEE International Advance Computing Conference (IACC)

X-axis

0.7

600

0.6

500

0.5

Computation time

Cost

400

300

200

0.4

0.3

0.2

0.1

100

0

0

SemiGreedy

Tabu search

GRASP Approach

Reactive GRASP

SemiGreedy

Tabusearch

VAST

Fig.11. Routing cost comparison result of different algorithms for N=30 nodes network.

GRASP Approach

Reactive GRASP

VAST

Fig.14.Comparison result of computation time of different algorithms for N=20 nodes network

1

450 0.9

400

0.7

350

0.6

300

0.5

250 Cost

Computation time

0.8

0.4

200 0.3

150 0.2

100

0.1

0

SemiGreedy

Tabusearch

GRASP Approach

Reactive GRASP

50

VAST

Fig.12.Comparison result of computation time of different algorithms for N=30 nodes network

0

SemiGreedy

Tabu search

GRASP Approach

Reactive GRASP

VAST

Fig.15.Routing cost comparison result of different algorithms for N=15 nodes network. 400 0.25

350

0.2

300

Computation time

Cost

250

200

150

0.15

0.1

100 0.05

50

0

SemiGreedy

Tabu search

GRASP Approach

Reactive GRASP

VAST

Fig.13. Routing cost comparison result of different algorithms for N=20 nodes network

0

SemiGreedy

Tabusearch

GRASP Approach

Reactive GRASP

VAST

Fig.16.Comparison result of computation time of different algorithms for N=15 nodes network

2014 IEEE International Advance Computing Conference (IACC)

5

CONCLUSION Simulation results shows that VAST algorithm performs best in comparison with other algorithm for all tested instances in terms of routing cost and computation cost. ACKNOWLEDGMENT . We express our sincere thanks and gratitude to the administration of Guru Gobind Singh Indraprastha University for providing the academic environment to pursue our research activity. REFERENCES [1] Fred Glover ,Manuel Laguna, "Tabu Search" ACM Digital Library, Modern heuristic technique for combinatorial problems, Pages 70150, 1993. [2] Kill-Woong Jang, A tabu search algorithm for routing Optimization in mobile ad-hoc networks, published Online:22 february 2011 open access at Springerlink.com.

6

[3] Mauricio G.C. Resende and Celso C. Ribeiro, "Greedy Randomized Adaptive Search Procedures", To appear in the State of Art Handbook in Metaheuristics, F.Glover and G.Kochenberger, eds, Kulwer, 2002. [4] Gang Feng, "GRASP-BSMA: A Fast Algorithm for Delay Constrained Multicast Routing", ICC IEEE 2012 Next-Generation Networking Symposium. [5] Salman A.Khan, "A Simulated Evolution-Tabu Search Hybrid Metaheuristic for Routing in Computer Network", IEEE 2007 Congress on evolutionary computation (CEC 2007). [6] T.Spyropoulos, R.N.B. Rais, T.Tureletti, K.Obraczka & "Routing for distruption tolerant networks: taxonomy and design." In wireless networks (pp. 1-22) 2010. [7] W. Pedrycz and A.Vasilakos, "Coputational intellignce in telecommunications networks" USA: CRC Press (2001). [8] J.P.Hart and A.W.Shogan," Semi-greedy heuristic: An empirical study", Operations Research Letters volume 6 Number3, july 1987.

2014 IEEE International Advance Computing Conference (IACC)