Routing-cost and Ant Routing Algorithm in Wireless Sensor ...€¦ · Routing algorithm consumes less energy, but it finds a route blindly, so that a lot of uncertainty is brought

Routing-cost and Ant Routing Algorithm in Wireless Sensor Networks

1Chen Fengchao, 2 Li Ronglin *1, Corresponding Author School of Electronic and Information Engineering, South China University

of Technology, Guangdong, Guangzhou, China, [email protected] 2, School of Electronic and Information Engineering, South China University of Technology,

Guangdong, Guangzhou, China, [email protected]

Abstract Due to the hardware characteristic of sensor nodes, energy consumption and balance are major

problems in routing algorithm in wireless sensor networks (WSNs). The networks should choose a route which not only has fewer hop count, but also makes the round of the nodes with less residual energy. In this paper, we first investigate the energy characteristic of the route and propose a routing-cost model. Then we compare the performance of routing-cost constructing method with other constructing methods on the platform of the ant routing algorithm in WSNs. The simulation results show that compared with other algorithms, the routing algorithm based on routing-cost has a better performance of energy property and node lifetime.

Keywords: Wireless Sensor Networks, Routing Algorithm, Routing-cost, Energy Balance 1. Introduction

Wireless sensor networks (WSNs) consist of a great many of low-power and low-cost sensor nodes deployed in a designated region. Each node collects information within its sensing range, and then transmits data in a multiple-hop way. Finally, the information collected by all nodes is forwarded to the base station, which is also called the sink node.

Under most circumstances, the sensor nodes are powered by batteries that are not rechargeable. Therefore the protocols and design of WSNs are focused on energy consumption and energy balance.

The improvement of routing algorithm in WSNs has a significant impact on prolonging network lifetime. In some cases, the sensor nodes are deployed randomly or non-uniformly in the monitoring region, and in other cases, the initial energy of each node is different. Therefore, one of the most important problems of the routing algorithm is that how to make full use of the node energy in these non-uniform random cases. On one hand, it would be best to choose a route of fewer hops so that the energy consumption is lower; on the other hand, sometimes it is better to choose a longer route to avoid the sensor nodes with less residual energy.

In this paper, we investigate the energy characteristic of the route in WSNs, and then we propose a routing-cost based ant routing algorithm. Finally the performances of different algorithms are evaluated. The paper is organized as follows. In section 2, related works are outlined. In section 3, we analyze the routing-cost and use it to construct the heuristic factor in the ant routing algorithm. Section 4 shows the performances of different algorithms. Finally a conclusion is drawn in section 5. 2. Related Works

The routing protocol is an important issue since traditional IP routing protocols and wireless routing protocol in ad hoc networks may not work well in WSNs. So far, there are many routing algorithms proposed [1-4]. The most well-known flat routing algorithms include flooding algorithm, Sensor Protocols for Information via Negotiation (SPIN) [1], Directed Diffusion (DD) [2], Rumor Routing [3], and so on. The algorithms which construct a routing table by a flooding method directly or indirectly are robust, such as flooding algorithm, SPIN and DD, but they consume much energy. The Rumor Routing algorithm consumes less energy, but it finds a route blindly, so that a lot of uncertainty is brought. In Fuzzy logic based Adaptive Routing (FAR) algorithm [5], some routing protocols are integrated in each sensor, and the networks can choose one of the protocols according to the environment.

Routing-cost and Ant Routing Algorithm in Wireless Sensor Networks Chen Fengchao, Li Ronglin

International Journal of Digital Content Technology and its Applications(JDCTA) Volume5,Number12,December 2011 doi:10.4156/jdcta.vol5.issue12.8

58

Interference-Minimized Multipath Routing (I2MR) [6] uses a correlation factor, which describes the interferences between two routes, to evaluate the quality of the route. The networks can choose a route with small correlation factor to make the load balance, and the energy consumption is reduced. However, the advantage of the algorithm is not obvious in the networks with less congestion.

Reliable and Energy Efficient Protocol (REEP) [7] setups a routing by flooding a request information from the sink, and it maintains an energy threshold value at each sensor node to achieve energy balance. However, the energy consumption of flooding is still high.

In Delay Guaranteed Routing and MAC(DGRAM) [8], the sensor nodes are deployed uniformly in the sensing area, and the node should find the next hop from the inner tier in the right time slot until the data reach the sink. In the algorithm, the routing has a definite direction to the sink node. However, each node should know its position with respect to the sink, and the energy balance cannot be achieved in WSNs with non-uniform node distribution.

In Energy Balanced Dynamic Routing Protocol based on Probability (EBDRP-P) [9], next hop is chosen in accordance to the ratio of its residual energy to the transmitting distance. This algorithm improves the energy consumption and network lifetime compared with DD algorithm. But unfortunately, it can only achieve local balance of energy consumption but not the global balance.

Due to the characteristics of distributed computing and self-organization, Ant Colony Optimization (ACO) fits the dynamic routing in WSNs. Different from the algorithms mentioned above, ACO can find the optimal path without a flooding process, and the algorithm can avoid searching a route blindly with the help of pheromone. The AntNet algorithm, which is constructed by the traffic load and communication delay in the networks, performs well in wired networks [10]. There are also ant based routing protocols proposed for mobile ad hoc networks (MANETs) [11, 12]. These algorithms pay much more attention to the adaptive and dynamic nature of the networks and QoS problem. Some ant routing algorithms for WSNs had also been proposed. Zhang improved the AntNet algorithm and proposed three ant-routing algorithms [13]. The heuristic factor in ACO was constructed by the distance between sensors and the sink node. Latency, energy consumption and energy efficiency were chosen to be performance metrics. However, the energy balance and the network lifetime were not considered in this paper. Guo used the energy of neighboring nodes and the distance between sensors and the sink node to describe heuristic value, he also used the distance to determine the quantity of pheromone laid on paths [14]. However, using the information of neighboring nodes can only achieve local balance of energy consumption too. Camilo used the average and minimum energy of the path to improve the amount of pheromone trail dropped by the backward ant [15]. It also reduced the communication load of the ants to decrease the energy consumption. The author considered the energy information of all sensor nodes on the path and reflects the routing-cost more objectively than the aforementioned algorithms. This paper used average energy, standard deviation and energy efficiency to evaluate the performance of the algorithm. Sun solved the path optimization problem by a combination of ant colony algorithm (ACA) and genetic algorithm (GA) [16]. GA was introduced to the iterative process of ACA; thus it could improve the convergence speed. The path fitness function was defined by the path length, energy consumption of the path, and the network energy equilibrium consumption. However, the complexity of the algorithm is increased, and the energy equilibrium consumption, which is defined by the length of the chromosomes, is not fit for the situation that the energy distribution changes all the time. Wu proposed an energy-aware ant routing algorithm for MANETs; he provided a parameter called linkcost to evaluate the energy metric of the routes [12]. However, a route consists of some links, so linkcost is just part of the routing-cost. 3. Routing-cost and Ant Routing Algorithm

Because the sensors are distributed randomly and the energy distribution in WSNs is not uniform, the routing algorithm has a great impact on energy consumption. A shorter route will consume less energy. But if a node is used to relay data frequently, it will fail faster. The routing algorithm in WSNs should take both the energy consumption minimization and energy balance into account, so that the network lifetime can reach a maximum. Ant routing algorithm is adopted in this paper for its adaptive, dynamic and distributed performance.


59

3.1 Network Model

In this paper, we assume that the sensor nodes are deployed randomly in the sensing region with probability density function ( ),f u v . The data generation rate of each node is constant, and the initial energy of each sensor node is limited, while the sink node has sufficient energy. The communication range of each node, including the sink node, is limited and the same. Each sensor node only knows its neighbors, nodes that are within its radio communication range. In this paper, the routing table in each node maintains the probability information for choosing a neighboring node as the next hop.

For the energy model, we only consider energy consumption for data transmitting and receiving. Let ( )TXE d and RXE denote energy consumption of transmitting and receiving 1-bit data

respectively. The first order radio model [4] is adopted here, and we have

( ) 2TX elec ampE d E de= + (1)

RX elecE E= (2) where d is the transmitting distance, elecE is the energy consumed by the transmitter or the receiver, and ampe is the amplification coefficient.

Considering the energy hole problem in WSNs, to achieve the subbalanced energy depletion, the deployment of sensor nodes must follow a geometric distribution. It means that when the common ratio 2q > , the percentage of the neighboring nodes of the sink node must be more than 50% [17]. However, in most of the time, these conditions cannot be fulfilled and the absolute global energy balance cannot be achieved. Therefore, "energy balance" discussed in this paper means the energy balance of the nodes which have approximate distance or hop count away from the sink, such as the nodes which are 2-hop or 3-hop away from the sink. 3.2 Energy Characteristic of the Route

We assume that the networks select a route of h -hop to transmit data from source to destination and each node in this route consumes energy to forward message. Therefore, the cost of the route to forward the message should be determined by the energy information of h nodes on the route. The energy consumption characteristic of a route is shown in Figure 1.

Figure 1. Energy consumption characteristic of a route

According to the characteristic of WSNs, we provide some definitions to compute the routing-cost. Value of node m : Value of node m is defined as the reciprocal of the residual energy of the node, that is


60

1m

m

VE

= (3)

where m ( 0,1, ,m h= L ) is the node of the m th hop on the route. When 0m = , it means the source node; when m h= , it means the sink node. Because we assume that the sink node has sufficient energy, the energy consumption of sink node is neglected. mV is value of node m , and mE is the residual energy of node m . In WSNs, the less energy left, the larger value the node has. In other words, as time passes by, the value of the node increases.

1-bit energy consumption of data forwarding: Applying (1) and (2), when node m forwards 1-bit data, its energy consumption md can be

expressed as

( )( )

2

2

, 02 , 1, 2, , 1

TX m elec amp mm

RX TX m elec amp m

E d E d mE E d E d m h

ed

eì = + =ï= í + = + = -ïî L

(4)

where md is the transmitting distance of node m . As shown in Figure 1, the source node ( 0m = ) transmits data, while other relay nodes ( 1,2, , 1m h= -L ) receive and then transmit data. The energy consumption of sink node is neglected.

Node-cost of data forwarding: Node-cost mw is defined as the difference between the node value mV after forwarding 1-bit

data and that before forwarding 1-bit date. So it can be formulated as

( )

1 1 mm

m m m m m mE E E Ed

wd d

= - =- -

(5)

(5) shows that the node with less energy will cost more value to forward data. Routing-cost of data forwarding: Routing-cost g is the sum of all the node-cost on the route, that is

( )

1 1

0 0

h hm

mm m m m mE E

dg w

d

- -

= =

= =-å å (6)

Because energy consumption of data forwarding is often far less than the residual energy,

that is m mEd << , (6) can be simplified as

1

20

hm

m mEd

g-

=

»å (7)

From (3), (4), (5) and (7), we can get that the definition of routing-cost is consistent with the

energy consumption characteristic of WSNs. A. The larger the distance md is, the more the route must cost to forward the data. B. The cost of data forwarding increases as a route of more hops is chosen. C. The route with less energy cost more when it forwards data. D. The nodes with less residual energy on the route have much more impact on the routing-

cost. Routing-cost can reflect the route characteristic objectively, so according to (7), we can

choose a route with less routing-cost.


61

3.3 Ant Routing Algorithm

Considering the advantages of ACO described in section 2, we evaluate the performance of a routing-cost based routing algorithm on the platform of ACO.

The basic ant colony algorithm can be simply described by two rules, that is, the node transition rule and the updating rule [18]. In this paper we design our routing algorithm based on AntNet [10].

Node transition rule defines the probability to select next hop for the forward ant as

[ ] [ ], if

0, otherwise

ki

ij ij kik

il ilijl N

j Np

a b

a b

t h

t hÎ

ì é ù é ùë û ë ûï Îï= íïïî

å (8)

where k

ijp is the probability with which ant k transits from node i to node j , ilt ( kil NÎ ) is

pheromone value on the link ( ),i l , and ilh is heuristic factor. kiN is the set of neighboring nodes

which are not visited. a and b are parameters that control the relative importance of ilt and ilh . When a node i receives a backward ant from its neighboring node j , it uses pheromone

updating rule to update its routing table. As the pheromone on the link ( ),i j increases, the pheromone on other links evaporates in the following manner:

( )1 , , ,

il ilil k

il il i

l jl N l j

t r tt

t rt+ - =ì

= í- Î ¹î

(9)

where r ( 0 1r< < ) is an evaporation coefficient.

The ant finds the optimal route in accordance with the above rules. ilt and ilh are two important variables in ACO, which determine the performance of the algorithm. In this paper we get ilt by the same rule as in AntNet, which is better for a quick exploitation of new discovered paths.

ijh is often constructed in accordance with the route characteristic. For example, ijh is constructed by the distance of the nodes in the basic ACO, and it is constructed by the length of the data queue in AntNet. In WSNs, we construct the heuristic factor ijh by routing-cost, that is

11

1 20

1 11

20

jj

j

ll

ki l

ki

RhmR

mj mij Rh

l ml N R

ml N m

E

E

dg

hg d

--

-=

- --

Î=Î

æ öç ÷ç ÷è ø= =æ öç ÷è ø

å

åå å

(10)

where lR is the optimal route from neighboring node l to sink, and lg is routing-cost of lR . lR

md and lR

mE are the 1-bit energy consumption and residual energy on the m th hop of lR , respectively. lh is the hop count of lR . The hop count in routing-cost determines that the ant will move in the direction of the sink, and the energy information in routing-cost determines that the ant will choose a route with more residual energy. (8) and (10) shows that the route with less routing-cost can be selected with a larger probability. Therefore, the ant routing can avoid flooding searching and blind searching.

Based on AntNet, the ant routing algorithm can be described as follows. A. To balance the exploration ability, all the probabilities and pheromones of the neighboring nodes

are initialized to be equal, respectively. B. At regular intervals, a forward ant is generated by the source node to find a path to the sink.


62

C. When a sensor node receives a forward ant, it will detect whether it has received the same ant before to avoid a cycle. Then the node uses the probability in the routing table to select the next node.

D. When the forward ant reaches the sink, it dies and a backward ant is generated and returns to the source node in the opposite direction of the forward ant. The routing-cost parameter in the backward ant is initialized to be 0 at the sink.

E. When a sensor node receives a backward ant, the pheromone is updated according to (9), and the routing-cost information in the backward ant is updated as follows.

2m

mEd

g g¬ + (11)

where md and mE are the 1-bit energy consumption and residual energy of the sensor node.

According to (11), in the sink node, we have 0g = ; in the node which 1m = , we have

1 12 2

1 1

0 h h

h hE Ed d

g - -

- -

= + = ; in the node which 2m = , we have 1

1 22 2 2

21 2

hh h m

m hh h mE E Ed d d

g-

- -

= -- -

= + = å ; the rest

can be deduced by analogy. Here the backward ant carries only one parameter and (7) can be achieved.

The heuristic factor is updated according to (10), and the sensor uses (8) to update the probability in the routing table.

If the sensor is the source node, the backward ant dies, and this round of searching finishes; if not, the sensor transmits the ant to the next node until it returns to the source node. 3.4 Updating of Routing Table

The routing table in each node includes the probability information for choosing a neighboring node as the next hop. The updating of routing table consists of three processes: first, the probability of each neighboring node is initialized to be the same value. Second, the routing table updates according to (8) when the node receives a backward ant. Third, some new nodes may enter the networks or some old nodes may fail, which is different from wired networks in AntNet. At regular intervals, the node first detects the changing of its neighboring nodes, and then creates or deletes the corresponding links. At the same time, the routing table in the node will update the probability information.

In the third updating process, let N denote the set of neighboring nodes which work after detection, and let M denote the set of neighboring nodes which keep working before and after detection. Therefore, ( )N M- is the set of the new neighboring nodes. The updating of probability information follows three principles. First, the total probabilities of M and ( )N M- are in proportion to the size of these two sets; so the probability ratio between M and

( )N M- is ( ):M N M- . Second, after updating, the probability distribution in M should

keep the same as before. Third, the probability of the new nodes in ( )N M- is set equally. Therefore, the updating rule of the third updating process can be described as

( )

,

1 ,

ij

ill Mij

pMj M

N pp

j N MN

Î

ì× Îï

ï¢ = íï

Î -ïî

å (12)


63

where ijp is the probability before updating, and ijp ¢ is the probability after updating. From

(12) we notice that the total probabilities in M and ( )N M- are MN

and N M

N-

,

respectively. 4. Performance Evaluation

In this section, we evaluate the routing algorithms in WSNs. In the simulation, the sensor nodes are deployed randomly in a square of 600×600 m2.

Our aim in this simulation is to evaluate the performances of different heuristic factor constructing methods on the platform of ant routing algorithm in the networks. We compare the simulation results of the routing-cost based constructing method with the results of some other basic constructing methods used in some proposed ant rouging algorithms. Here give some other basic constructing methods [13-16].

A. Heuristic factor constructed by residual energy of neighboring nodes:

1

1

j

l

ki

R

ij R

l N

EE

h

Î

=å

(13)

where 1

lRE is the residual energy of the neighboring node in lR . B. Heuristic factor constructed by the hop count (or distance):

1

1

ki

jij

ll N

hh

h-

-

Î

=å

(14)

C. Heuristic factor constructed by residual energy of neighboring nodes and hop count

(or distance) together:

1

11

1

j

l

ki

Rj

ij Rl

l N

E hE h

h-

-

Î

=å

(15)

In (8), b is set to be larger than a so that the algorithm can adapt to the dynamic energy

distribution. Here we set 1a = and 3b = . Evaporation coefficient r is set to be a moderate value of 0.5. The parameters elecE and ampe in the energy model (1) and (2) are set to be 50 nJ and 100 pJ/m2, respectively.

As mentioned before, "energy balance" discussed in this paper means the energy balance of the nodes which have approximate distance or hop count from the sink. However, it is difficult to provide the energy properties of all nodes. Moreover, from the energy hole problem we know that the energy properties of the neighboring nodes of the sink, the nodes which are 1-hop away from the sink, are most important [17]. Therefore, we only show the simulation results of these nodes. We do not adopt the average energy and the standard deviation as performance metrics. It is possible that there exist a few nodes which consume energy fast even if the performances of average energy and standard deviation are very good, so it is necessary to provide the energy consumption data of each neighboring node.

A. Simulation 1 200 sensors are deployed randomly in the sensing region. The probability density function ( ),f u v is a uniform function in the region and the sink is deployed at (300,300), but the


64

networks are not homogeneous in different direction. In the simulation, we assume that each sensor has the same data generation rate, and initial energy is 1 J.

In this simulation, we will show the energy consumption of each neighboring node of the sink so that we can clearly observe the energy distribution in this region. The indices of these nodes are 6, 13, 53, 64, 100, 101, 125, 152, 173 and 199.

Figure 2 shows the energy consumption of different heuristic factor constructing methods in WSNs with a uniform random distribution. Each curve represents the energy consumption of the corresponding node, and the dispersion of all curves represents the energy balance property. The networks with more dispersive curves have a worse balance property. Figure 2a) is the result of Eq. (13), and in this case heuristic factor is constructed by residual energy of neighboring nodes, that is the first hop of the route, so the energy distribution can reach local balance, that is, as time passes by, the curves will become more and more dispersive slowly. Moreover, there is another problem: most nodes fail fast because the energy consumption minimization or hop count is not considered in Eq. (13). Figure 2b) is the result of Eq. (14). We can see that the nodes consume energy more slowly than that in Figure 2a). But it will lead to energy imbalance in the networks if we only pursue energy consumption minimization. The difference of the node lifetime is large in Figure 2b). Figure 2c), which is the result of Eq. (15), is a compromise between Eq. (13) and Eq. (14). The balance property and the node lifetime are considerably improved. The difference of the node lifetime in Figure 2c) is much smaller than that in Figure 2b), but n[64] has much energy left when the simulation is finished. Figure 2d) is the simulation result of routing-based ant algorithm which takes into account the information of all nodes in the route. Figure 2d) shows the best performance. We can notice that all the nodes almost fail at the same time and have a long lifetime. Figure 2d) also has a good balance property. At the beginning, when the nodes have much energy, the curves disperse slowly. As the energy decreases, the node with less energy affects the routing more significantly; thus the curves converge and finally all nodes almost fail at the same time.

a) Simulation result of Eq. (13)


65

b) Simulation result of Eq. (14)

c) Simulation result of Eq. (15)

d) Simulation result of Eq. (10) (Routing-cost based ant routing)

Figure 2. Residual energy of neighboring nodes of the sink in Simulation 1

B. Simulation 2 300 sensor nodes are deployed randomly in the sensing region. The probability density

function ( ),f u v is shown in (16) and the distribution is asymmetrical. The sink is deployed at (300,300). In the simulation, we assume that each sensor has the same data generation rate.


66

Here we set the initial energy 1.3 J for n[199], 1.5 J for n[173], 1.8 J for n[152], 2 J for n[125] and 1 J for other nodes.

( )6 4110 10 , 0 , 600

, 324 540, others

u u vf u v

- -ì ´ + ´ < £ï= íïî

(16)

The indices of the neighboring nodes of the sink are 6, 13, 53, 64, 100, 101, 125, 152, 173,

199, 266 and 270. In this simulation, we show the good performance of different algorithms in WSNs with a non-uniform distribution and different initial energy setting.

Figure 3 shows the energy consumption of different heuristic factor constructing methods in Simulation 2. We can see that the simulation results are similar to that in Figure 2. Figure 3a) is the result of Eq. (13). The energy distribution reaches local balance and the node lifetime is very short. In Figure 3b), the energy curves and node lifetime are very dispersive, just like Figure 2b). Figure 3c) shows the result of Eq. (15), which is a compromise between Eq. (13) and Eq. (14). Compared with Figure 3a) and Figure 3b), the energy balance is considerably improved in this constructing method, but the networks cannot make full use of the energy of n[64]. Figure 3d) shows the simulation result of the routing-cost based ant routing algorithm. We can see that the energy balance property is the best of all. At the beginning, the nodes with more initial energy, such as n[125] and n[152], consume energy faster than others. As time passes by, the curves converge and the difference of the node lifetime is slight.

a) Simulation result of Eq. (13)

b) Simulation result of Eq. (14)


67

c) Simulation result of Eq. (15)

d) Simulation result of Eq. (10) (Routing-cost based ant routing)

Figure 3. Residual energy of neighboring nodes of the sink in Simulation 2 In Figure 2 and Figure 3, we notice that most constructing methods cannot make full use of the

energy of n[64], while the routing-cost based method can deal with this problem better. 5. Conclusions

In this paper, we study the energy characteristic of the route and propose a routing-cost based ant routing algorithm. Then we compare the routing algorithms of different constructing methods on the platform of ACO. We believe that the routing-cost constructing method takes into account the energy information of each node and hop count information on the route, which is more rational than the algorithms constructed by the residual energy of the neighboring node or hop count. Therefore, the routing-cost ant routing algorithm has a good energy property and node lifetime property.

The research in this paper focuses on the constructing method of heuristic factor in ACO. However we have not discussed other problems of ACO, such as the convergence problem and the generation rate of the forward ant, which affect the energy consumption of the networks greatly and need further investigation. 6. Acknowledge

This work was supported by the National Nation Science Foundation of China (60871061), the Guangdong Province Natural Science Foundation (8151064101000085), and the Specialized Research Fund for the Doctoral Program of Higher Education (20080561).


68

7. Reference [1] Wendi Rabiner Heinzelman, Joanna Kulik, Hari Balakrishnan, "Adaptive protocols for information

dissemination in wireless sensor networks", In Proceedings of the 5th Annual ACM/IEEE International Conference on Mobile Computing and Networking, pp. 174-185, 1999.

[2] Chalermek Intanagonwiwat, Ramesh Govindan, Deborah Estrin, "Directed diffusion: a scalable and robust communication paradigm for sensor networks", In Proceedings of the 6th Annual International Conference on Mobile Computing and Networking, pp. 56-67, 2000.

[3] David Braginsky, Deborah Estrin "Rumor routing algorthim for sensor networks", In Proceedings of the 1st ACM International Workshop on Wireless Sensor Networks and Applications, pp. 22-31, 2002.

[4] Wendi Rabiner Heinzelman, Anantha Chandrakasan, Hari Balakrishnan, "Energy-efficient communication protocol for wireless microsensor networks", In Proceedings of the 33rd Annual Hawaii International Conference on System Sciences, pp. 908-918, 2000.

[5] Hee Wan Kim, Hee Suk Seo, Sun Ho Hong, Chul Kim, "Modeling of Energy-efficient Applicable Routing Algorithm in WSN", JDCTA: International Journal of Digital Content Technology and its Applications, Vol. 4, No. 5, pp. 13-22, 2010.

[6] Jenn-Yue Teo, Yajun Ha, Chen-Khong Tham, "Interference-Minimized Multipath Routing with Congestion Control in Wireless Sensor Network for High-Rate Streaming", IEEE Transactions on Mobile Computing, vol. 7, no. 9, pp. 1124-1137, 2008.

[7] F. Zabin, S. Misra, I. Woungang, H.F. Rashvand, N.-W. Ma, M. Ahsan Ali, "REEP: data-centric, energy-efficient and reliable routing protocol for wireless sensor networks", IET Communications, vol. 2, no. 8, pp. 995-1008, 2008.

[8] Chilukuri Shanti, Anirudha Sahoo, "DGRAM: A Delay Guaranteed Routing and MAC Protocol for Wireless Sensor Networks", IEEE Transactions on Mobile Computing, vol. 9, no. 10, pp. 1407-1423, 2010.

[9] Ming Zhang, "An Novel Energy Balanced Dynamic Routing Protocol Based on Probability in Wireless Sensor Netwoks", JCIT: Journal of Convergence Information Technology, vol. 6, no. 3, pp. 10-17, 2011.

[10] Gianni Di Caro, Marco Dorigo, "AntNet: Distributed stigmergetic control for communications networks", Journal of Artificial Intelligence Research, vol. 9, no. 1998, pp. 317-365, 1998.

[11] Osama H. Hussein, Tarek N. Saadawi, Myung Jong Lee, "Probability routing algorithm for mobile ad hoc networks' resources management", IEEE Journal on Selected Areas in Communications, vol. 23, no. 12, pp. 2248-2259, 2005.

[12] Zheng-Yu Wu, Han-Tao Song "Ant-based Energy-aware Disjoint Multipath Routing Algorithm for MANETs", The Computer Journal, vol. 53, no. 2, pp. 166-176, 2010.

[13] Ying Zhang, Lukas D. Kuhn, Markus P. J. Fromherz, "Improvements on ant routing for sensor networks", In Proceedings of the 4th International Workshop on Ant Colony Optimization and Swarm Intelligence, pp. 154-165, 2004.

[14] Wenjing Guo, Wei Zhang, Gang Lu, "PEGASIS Protocol in Wireless Sensor Network Based on an Improved Ant Colony Algorithm", In Proceedings of the 2nd International Workshop on Education Technology and Computer Science, pp. 64-67, 2010.

[15] Tiago Camilo, Carlos Carreto, Jorge Sá Silva, Fernando Boavida, "An energy-efficient ant-based routing algorithm for wireless sensor networks", In Proceedings of the 5th International Workshop on Ant Colony Optimization and Swarm Intelligence, pp. 49-59, 2006.

[16] Yang Sun, Jingwen Tian, "WSN path optimization based on fusion of improved ant colony algorithm and genetic algorithm", Journal of Computational Information Systems, vol. 6, no. 5, pp. 1591-1599, 2010.

[17] Xiaobing Wu, Guihai Chen, Sajal K. Das, "Avoiding energy holes in wireless sensor networks with nonuniform node distribution", IEEE Transactions on Parallel and Distributed Systems, vol. 19, no. 5, pp. 710-720, 2008.

[18] Marco Dorigo, Vittorio Maniezzo, Alberto Colorni, "Ant system: Optimization by a colony of cooperating agents", IEEE Transactions on Systems Man and Cybernetics Part B-Cybernetics, vol. 26, no. 1, pp. 29-41, 1996.


69

Documents

Routing-cost and Ant Routing Algorithm in Wireless Sensor ...€¦ · Routing algorithm consumes less energy, but it finds a route blindly, so that a lot of uncertainty is brought