Third International Workshop on Advanced Computational Intelligence August 25-27,201 0 - Suzhou, Jiangsu, China

An Ant System Optimization QoS Routing Algorithm for Wireless Sensor Networks Fengjun Shang, Yin Wang

Abstract-Routing of sensor data has been one of the

challenging areas in wireless sensor network, specially

being in QoS routing. In this paper, a QoS routing

algorithm (APAS) with adaptive parameters based on

Max-Min Ant System(MMAS) Algorithm is presented.

The performance of algorithm is improved by the

adaptive pheromone and evaporation coefficient.

Furthermore, to save energy consumption, the beamed

broadcasting is introduced, the node of energy changed

only need announce the necessary neighbors. The

simulation for APAS shows that the improved algorithm

can improve search QoS routing and extend network

lifetime.

I. INTRODUCTION

D ecent advances in miniaturization and low-power design .Rb.ave led to active research in large-scale, highly

distributed systems of small-size, wireless unattended sensors. Over the last few years, the design of sensor networks has gained increasing importance due to their potential for some civil and military applications such as combat field surveillance, security and disaster management. On the military side, applications of sensor networks are numerous[l). Routing of sensor data has been one of the challenging areas in wireless sensor network research. It usually involves multi-hop communications and has been studied as part of the network layer problems. Despite the similarity between sensor and mobile ad-hoc networks, routing approaches for ad-hoc networks proved not to be suitable to sensors networks. This is due to different routing requirements for ad-hoc and sensor networks in several aspects. For instance, communication in sensor networks is from multiple sources to a single sink, which is not the case in ad-hoc networks. Moreover, there is a major energy resource constraint for the sensor nodes. As a consequence, many new algorithms have been proposed for the problem of routing data in sensor networks. However, the development of video and imaging sensors requires the consideration of quality of service (QoS) in sensor networks, which magnifies the

Manuscript received April 9, 2010. This work was supported in part by the Chongqing Natural Science Foundation under Grant No. 2009882081 and the Project Sponsored by the Scientific Research Foundation for the Returned Overseas Chinese Scholars, State Education Ministry.

Fengjun Shang is with the College of Computer Science and Technology, Chongqing University of Posts and Telecommunications, Chongqing 4000�5, China (corresponding author to provide phone: +86-13752871911; e-maIl: shangtj@cqupt.edu.cn).

978-1-4244-6337-4/10/$26.00 @2010 IEEE 339

difficulties associated with the energy efficiency and awareness.

While many mechanisms have been proposed for routing QoS constrained real-time multimedia data in wire-based networks, they cannot be directly applied to wireless sensor networks due to the limited resources, such as bandwidth and energy that a sensor node has.

On the other hand, a number of protocols have been proposed for QoS routing in wireless ad-hoc networks taking the dynamic nature of the network into account.

Akyildiz survey the state of the art in algorithms, protocols, and hardware for wireless multimedia sensor networks. Architectures for Wireless Multimedia Sensor Networks (WMSNs) are explored, along with their advantages and drawbacks. Currently off-the-shelf hardware as well as available research prototypes for WMSNs are listed and classified. Existing solutions and open research issues at the application, transport, network, link, and physical layers of the communication protocol stack are investigated, along .

d . . . [2) with possible cross-layer synergIes an optlmizations .

Spadoni uses adaptable mobile agents to carry out data collection, fusion and delivery of information. The key features of these agents are their novel dynamic addressing scheme and their flexibility to behavior and structure changes that can guarantee the application required quality of service in harsh conditions, such as in fire situations[3). In Ref. 4, it presents a query-based routing protocol for a

WSN that provides different levels of Quality of Service (QoS): energy-efficiency, reliability, low latency and fault-tolerance-under different application scenarios. The algorithm has low computational complexity but can dynamically guarantee different QoS support depending on the requirement of the applications.

Suriyachai presents the implementation of a new WSN MAC protocol that is able to give deterministic bounds for message transfer delay and reliability. This implementation shows that a deterministic MAC protocol with reasonable energy consumption patterns is practical[5).

Chen proposes a novel mechanism to fmd multiple-paths between one sink and multiple-sources with the consideration of reducing collision occurred at nodes that are receiving and forwarding packets on behalf of the source nodes. Simulation results are provided to show the potential and effectiveness of our solution[6). In Ref. 7, aiming at three basic services, abnormal event

alarm, information query and stream query service, the paper abstracts a QoS muting model for the multimedia sensor networks. Moreover, based on the traditional ant-based algorithm, we propose an ant-colony optimization based

service aware muting(ASAR). The ASAR chooses the suitable paths to satisfy with the diverse QoS requirements from different kinds of service, thus maximizes network utilization and improves network performance. Finally, extensive simulation using NS2 is conducted to verify the effectiveness of our solution. Compared to the traditional ant-based algorithm, our ASAR algorithm has better convergence and significantly provides better QoS for multiple types of services in the multimedia sensor networks.

II. APAS ALGORITHM

MMAS[8) is a type of stochastic heuristics optimization algorithms. In this paper, APAS (Adaptive Pheromone Ant System) for QoS Routing in wireless sensor network is proposed base on adaptive pheromone evaporation coefficient. In AP AS, it improves the convergence rate of MMAS by adaptive pheromone evaporation coefficient and increases global search speed. APAS uses the beamed broadcasting[9), the node of energy changed only need to announce the necessary neighbors. The simulation for APAS shows that the improved algorithm can search QoS routes effectively and have longer life. Definition 1 Given undirected graph G = (V, E), where V is vertex aggregate, E is edge aggregate, vertex Vi' V j E V ,

edge e(i, j) E E , T ij is pheromone intensity of edge e(i, j) and its value is corrected by ant, T ij is Heuristicgene.

Definition 2 In graph G, VI is source node of the path

P = (VI' v2" '" vs) and Vs is destination node, named

P(vl' vs). Definition 3 Neighbor aggregate of node A(XA,YA) is

Neighbour = {(Xp ¥;, a1), (X2' Yz, a2), .. · , (Xn' Y",an)} ,where

X,Y is Plane Coordinate, an denotes direction angle from

node A to the nth node.

A. Ant Colony Algorithm

Given n be number of traveling salesman problem (TSP) and m be the total number of ant. In period t, an ant k successively chooses the next node probability is listed as follows from node i to node j. 1 ['ij(t)r .[1Jij(t)]p

p/(t) = L ['is (t)r·[1Jis (t)]P ,j E allowedk (I) g seallawedk O,j � allowedk

a is response factor of Pheromone, which denotes Pheromone effecting on ant motion process. f3 is response factor of heuristics gene, which denotes heuristics information effecting on ant selecting path.

When all ants traverse n nodes, it renews Pheromone. In period t+n, Pheromone level of path (ij) is adjusted using

340

formula (2), where P denotes the pheromone evaporation coefficient, I:l.'ij(t) denotes increment of Pheromone in this

path (ij) 'ij(t + n) = Px ,uet) + I:l.'ij(t) (2)

MMAS avoids the Premature Convergence by setting the upper and lower bounds of Pheromone and only add Pheromone on corresponding arc. It may avoid Pheromone on some path far more than other path so that it makes all ants concentrate one path.

B. Adaptive Pheromone

Pheromone Strength denotes the tracks of ant effects global search optimal solution and convergence speed. APAS proposes adaptive Pheromone is as follows.

PX'ij(n)+I:l.'in), ,in) >'rrt &&UP=o

1.1'y(n), 'y(n) < 'rrt &&UP = 1

'y(n+l)= (3) 1.1'y(n), ,in) > 'rrt &&UP = 1

Px'y(n)+I:l.'y(n), ,in) <'rrt &&UP=o

In an ant search from source node Vo to destination node vs'

with increasing search times, good path PA will accumulate

many pheromone so that pheromone intensity , is risen on

path PA and then this makes the follow-up ants convergence

on path PA• Furthermore, this will lead to reduce search

path PB so that pheromone intensity , is reduced on path

PB and then this will be disadvantageous for search optimal

solution in graph G and nodes quickly waste energy on path PA

'

This algorithm introduces threshold of pheromone intensity

Tml'min < Tmt < Tmax ) . When 'ij < 'mt' 'ij will stop

reducing. This may avoid search local optimal solution in relatively optimal path PA and inferior path PB owning to

pheromone intensity , being different.

When 'ij increases to 'max' indictor variable up=l, else

up=O.

C. Adaptive Evaporation Coefficient

Ants have memory ability, but with the time going, information is lost. Evaporation Coefficient P effects global search ability and convergence speed. APAS proposes adaptive evaporation coefficient is as follows.

P(n+l)=

O.95P(n),P(n) > Pmin &&pp =-1

1.05P(n),P(n) < Pmin &&pp = I (4) 1.05P(n),P(n) > Pmin &&pp = 1 1.15Po'pp = 0

In an ant search from source node Vo to destination node vs' if P is big, the search ability is reduced. In APAS, ftrstly, Evaporation Coefficient reduces O.95P(n) every round up to O.95P(n) < Pmin•

D. Pheromone and Heuristic Factor

In order to satisfy QoS and saving energy in wireless sensor network, Pheromone increment and Heuristic Factor is designed as follows.

E(

.) I

n.. =k._

1

_+ +

m +-

q- (5)

"II} Eo C(i, j) D(i, j) d(j)

1 �:r = (6) I} a.

Ecost{P)+b.C{P)+c.D{P)

In Formula (5), where Eo is the initial energy, E(j) is next

hop node j energy; C(i, j) is lost packet ratio from node

i to node j; D(i, j) is delay from node i to node j ;

d (j) is distance from next hop node to base station. k ,

1, m , q

is weight factor. In Formula (6), where

Ecost(P) is the average energy on path P; C(P) is the

average value of lost packet ratio on path P; D( P) is delay

on path P. a, b, c is weight factor. Formula (5) and (6) consider delay, packet loss rate and

energy consumption, so this may satisfy QoS requirement and extend network lifetime.

E. Beamed Broadcasting

Given a graph G, the optimal path P is searched so that it makes QoS value be maximum. Formula (7) is QoS function of path P. The Q(P) is great, the QoS of path P is great.

Q(P)= '.

1

+k ._

1

_+1 ._

1

_ (7) 1

0 Ecost(P) 0 C(P) 0 D(P)

Where Ecost(P) is energy consumption of path P; C(P) is

packet loss rate, D(P) is delay; jo ' ko' 1

0

is

corresponding weight factor.

F. QoS Function

In order to quicken search speed, this paper limits search area by search angle[101. It may avoid too many search ants to quicken search speed. Search angle is () aiming at base station. After ants search from source node V

0

to destination node

Vs ' next hop node is only selected from

Neighbour = {(XI' �,al)' (X2' 1';, a2), ... , (Xn' Y", an)} within search angle ().

341

Because it introduces search angle, in ending this round, some nodes need broadcast the rest energy to Neighbour

by beamed broadcasting and other nodes don't need broadcast. Beamed broadcasting formula is list as follows.

B a P- r

360 (8)

Where r is broadcast radius, a is the transmission decay coefficient. APAS can save energy M by using beamed broadcasting, where M is as follows

G. AP AS Algorithm

M = (1-

360 - B )ra 360

Algorithm 1 is CA TM Pseudo-code as follows. 1. At the beginning, Nc=l;

(9)

2. m forward ants start from source node S, use equation (1) to search the next node; 3. While EA(t+n) *" EA(t) do

4. node A broadcasts E A (t + n) to Neighbourneed;

5. End while 6. When m forward ants reach base station, calculate the QoS

value of each ant with equation (7),ant with the highest value is the best ant,

7. Send backward ant to update pheromone of path which the best ant has passed by equation (2) and (6);

8 .1f Tij<TmtAND up=1 OR Tij>TmtAND up=1 then

9. Tij(Nc+l)=1.1TyCNc)

10. Else 11. Tij(Nc+l) = Px Tij(Nc) +/).Tij(Nc)

12. End If 13. If pp=-1 14. P(Nc+l)=0

.95p(Nc)

15. Else If pp=1 16. P(Nc+l)=1

.05p(Nc)

17. Else If pp=O 18. P(Nc+l)=1.15Po

19. End If 20. EndIf 21. End If 22. If Nc<Nmax 23. Nc=Nc+1, return to 2 24. Else If 25. End

III. SIMULATIONS AND ANALYSIS We select a scenario to simulate our algorithm using

MA TLAB and the parameter set is shown in Table 1. Table I SIMULA nON PARAMETERS

Parameter Value Network coverage (0,0)�(300,300) Base station location (0,0) N 100 Initial energy 3J Data packet size 20000 bits The number of round 200

jo 2

ko 1

10 2

Eelec 5xlO-8 8ft 10-11 8mp 1.3 x 10-15

do 70

Cinit 0.05

Formula (10) is the radio hardware energy dissipation. Both the free space (d2 power loss) and the multi-path fading (d4

power loss) channel models are used in the model, depending on the distance between the transmitter and receiver. Transmission ( Erx ) and receiving costs ( E R.x) are calculated as follows:

_ {lEelee + 1& Jsd2, d < do Erx(/,d) - 4 lEelee + I&mpd ,d > do

Where d is the distance between the transmitter and the receiver.

(10)

In Formula (11), Cinit is the packet loss rate, R\ is the

radius of noise source 1. When node i is within R\, lengt� can be computed, etc. C(.) ",._, O.2*(� -length,) O.2*(Rz -lengthJ O.2*(� -lengt!;) ( 11 ) I =urut+ + I

� Rz �

'900 Optimal ,em

QoS Value '700 -=: .. .. .1

'600

'500

'400

..... .

1 ... . �:1s 1 '-0�����40��ro�OO��,�00�'���14�0�'ro��,oo��

Number of Rounds Figure 1 Comparison the optimal QoS between APAS and

MMAS

342

Figure 1 shows relation between QoS value and number of round. From figure 1, with time passing, APAS is great MMAS, because pheromone strength is great in MMAS so that ants can aggregate the good path and omit search other path and then this leads to resolve local optimal roots, but APAS may search the global optimal roots by Adaptive Pheromone and Evaporation Coefficient. The optimal QoS in APAS is 1.2 times as much as that one in MMAS. The optimal QoS in MMAS is 1.8 times as much as that one in APAS.

Optimal QoS

Value

Number of Rounds

Figure 2 Comparison the average QoS between APAS and MMAS

From figure 2, in beginning phase, the average QoS value in APAS likes in MMAS, because energy consumption of nodes is little. But after it run 80 rounds, the average QoS in APAS is great, because APAS consider the rest energy of next hop node.

70

ro

50

Number 40 of Dead Nodes 30

�

'0 , -.' 00 � 40 60 80 100 120 140 160 180 2IXl

Number of Rounds

Figure 3 Comparison the number of dead nodes between APAS and MMAS

By observing the number of dead nodes from figure 3, it can be seen that there are no dead nodes in 30 rounds of AP AS. In 30 rounds of MMAS, there are 8 nodes at least, which is 8% of number of total nodes. The number of dead node shows balance network energy consumption. The less there are dead nodes, the better we can do balance network energy. APAS both prolongs lifetime of network and reduces the number of dead nodes. Hence, APAS more efficiency balances the

energy consumption of network compared to the other strategies.

Packet Loss Rate

Delay

20

19

18

17

16

15

14

13 1 1.5 2 2.5

Times 3.5 4.5

Figure 4 Comparison the packet loss rate

x 10" ' . 8 ;;-;-:'-�-�-�-�-�--�-�-. '75 ... . ....•....•. . ..•. .... .

I·--·:::s I 1.7 1.65

1.6 1.55

1.5

1.45

'.4 1L-----:'"'=.5-----::---::"2.':-5 ---=---3::':.5:---�----:4"'=.5-----!5 Times

Figure 5 Comparison the delay

By observing the packet loss rate from figure 4, it can be seen that the packet loss rate in APAS is 1.3 times as much as that one in MMAS. At the same time, in figure 5, it can be seen that the delay in MMAS is 1.2 times as much as that one in APAS.

IV. CONCLUSIONS In this paper, a QoS routing algorithm(APAS) with adaptive

parameters based on Max-Min Ant System(MMAS) Algorithm is presented. The performance of algorithm is improved by the adaptive pheromone and evaporation coefficient. With the beamed broadcasting, the node of energy changed only need to announce the necessary neighbors. The simulation for APAS shows that the improved algorithm can search QoS routes effectively and have longer life. In future research, we will consider NS2 simulation platform using event-driven mechanism to simulate performance of the AP AS algorithm.

343

REFERENCES

[I] H. Wen, C. Lin, and F. Y. Ren, "QoS Architecture in Wireless Sensor Network", Chinese Journal of Computers, vol.32, no.l, pp.432-440, 2009.

[2] L F. Akyildiz, "A survey on wireless multimedia sensor networks", Computer Networks(Elsevier), vo1.51, no.4, pp. 921-960,2007.

[3] Igor M. B. Spadoni, Regina B. Araujo, Cesar Marcondes, "Improving QoS in Wireless Sensor Networks through Adaptable Mobile Agents", in Proc. of the 28th IEEE international conference on Computer

Communications Workshops, Brazil, 2009,pp.361-362. [4] 1. Sen and A. Ukil, "An adaptable and QoS-aware routing protocol for

Wireless Sensor Networks", In Proc of the Wireless VITAE, 2009, pp.767-771.

[5] P. Suriyachai, U. Roedig, and A. Scott, "Implementation of a MAC protocol for QoS support in wireless sensor networks", In Proc of the

IEEE International Conference on Pervasive Computing and

Communications, 2009, pp.I-6. [6] Y. F. Chen and N. Nasser, "Enabling QoS Multipath Routing Protocol

for Wireless Sensor Networks", In Proc of the IEEE International

Conference on Communications, 2008, pp.2421-2425. [7] Y. Sun, H. D. Ma, and L. Liu, "An Ant-Colony Optimization Based

Service Aware Routing Algorithm for Multimedia Sensor Networks", Acta Electronica Sinica, vol.35, no.4, pp. 705-711, 2007.

[8] S. Thomas and H. H. Holger, "Max-Min ant system",Future

Generation Computer Systems,voI.16, no.8, pp.889-914, 2000. [9] Jeffrey E. Wieselthier, Gam D. Nguyen, Anthony Ephremides,

"Energy-Aware Wireless Networking with Directional Antennas: The Case of Session-Based Broadcasting and Multicasting", IEEE

transaction on mobile computing, yoU, no.3, pp.l76-190, 2002. [10] G. S. Yin, G. Yang, and W. Yang, "An Energy-Efficient Routing

Algorithm for Wireless Sensor Networks", In Proc of ICICSE, 2008, pp.181-186.