A color-theory-based energy efficient routing algorithm for mobile wireless sensor networks

(1)

A color-theory-based energy eﬃcient routing algorithm

for mobile wireless sensor networks

q

Tai-Jung Chang, Kuochen Wang

*

_{, Yi-Ling Hsieh}

Department of Computer Science, National Chiao Tung University, Hsinchu 300, Taiwan, ROC Received 25 January 2007; received in revised form 31 July 2007; accepted 11 October 2007

Available online 22 October 2007 Responsible Editor: X.S. Shen

Abstract

Wireless sensor networks (WSNs) with nodes spreading in a target area have abilities of sensing, computing, and com-munication. Since the GPS device is expensive, we used a small number of fixed anchor nodes that are aware of their loca-tions to help estimate the localoca-tions of sensor nodes in WSNs. To efficiently route sensed data to the destination (the server), identifying the location of each sensor node can be of great help. We adopted a range-free color-theory based dynamic localization (CDL) [Shen-Hai Shee, Kuochen Wang, I.L. Hsieh, Color-theory-based dynamic localization in mobile wire-less sensor networks, in: Proceedings of Workshop on Wirewire-less, Ad Hoc, Sensor Networks, August 2005] approach, to help identify the location of each sensor node. Since sensor nodes are battery-powered, we propose an efficient color-theory-based energy efficient routing (CEER) algorithm to prolong the life time of each sensor node. The uniqueness of our approach is that by comparing the associated RGB values among neighboring nodes, we can efficiently choose a better routing path with energy awareness. Besides, the CEER has no topology hole problem. Simulation results have shown that our CEER algorithm can save up to 50–60% energy than ESDSR [Mohammed Tarique, Kemal E. Tepe, Mohammad Naserian, Energy saving dynamic source routing for ad hoc wireless networks, in: Proceedings of Modeling and Optimi-zation in Mobile, Ad Hoc, and Wireless Networks, April 2005, pp. 305–310] in mobile wireless sensor networks. In addi-tion, the latency per packet of CEER is 50% less than that of ESDSR.

Keywords: Color-theory-based; Energy saving; Localization; Mobile wireless sensor network; Routing algorithm

1. Introduction

With the advance of recent wireless and VLSI technologies, small and low-cost sensor nodes have become feasible in our daily life. These sensor nodes sense data in the environment surrounding them and transmit the sensed data to the sink (or the ser-ver). The way to transmit sensed data to the server

q

The insightful comments of the reviewers help to improve the quality of the paper. This work was supported by the National Science Council under Grant NSC94-2213-E-009-043 and NCTU EECS-MediaTek Research Center under Grant Q583.

* _{Corresponding author.}

E-mail address:[email protected](K. Wang).

Computer Networks 52 (2008) 531–541

(2)

aﬀects the power usage of each node deeply. Typi-cally, wireless sensor networks (WSNs) contain hundreds or thousands of sensor nodes and these sensor nodes have the ability to communicate with either one another or directly to the destination node (sink). Using a larger number of sensor nodes allows for sensing over wider geographical regions with greater accuracy. Sensor nodes may use up their energy so that they become unavailable in a WSN during communications or measurements. With more and more sensor nodes becoming unavailable, the WSN may be separated into several sub-networks or become sparse, which is not desir-able. Therefore, energy saving is an important issues in the WSN.

It is a great challenge for routing in a WSN due to the following reasons. First, since it is not easy to grasp the whole network topology, it is hard to find a routing path. Secondly, sensor nodes are tightly constrained in terms of energy, processing, and stor-age capacities. Thus, they require effective resource management policies, especially efficient energy management, to increase the overall lifetime of a WSN.

The proposed clustered-based color-theory-based energy eﬃcient routing (CEER) algorithm is based on a range-free color-theory-based dynamic locali-zation algorithm, CDL[1], in which the location of a sensor node is represented as a set of RGB val-ues. With known RGB values for each sensor node, we can ﬁnd out the most possible position of a node by looking up the location database in the server. To keep track of a sensor node’s loca-tion, frequently updating the RGB values of each sensor node and delivering the update to the server is necessary. However, if battery-powered nodes frequently update and report their positions, they may consume energy quickly and also waste band-width. The CEER selects those cluster members that are closer to the anchor than itself as next possible hops by comparing their RGB values. Among the selected cluster members, the sensor node with the highest energy level is chosen as the next hop.

The remainder of this paper is organized as fol-lows. Section 2 brieﬂy introduces the CDL algo-rithm that is the basis of our routing algoalgo-rithm. Section 3 introduces ﬁve related approaches and explains how they work. The network model and the proposed CEER algorithm are detailed in Sec-tion 4. Section 5 evaluates and compares our approach with a classical approach. Simulation

results demonstrate the merits of our algorithm. Finally, concluding remarks are given in Section 6. 2. Preliminary of CDL

In this section, we introduce the color-theory-based dynamic localization (CDL) algorithm [1], which is the basis of our proposed routing algo-rithm. This centralized localization algorithm is based on the color theory to perform positioning in mobile wireless sensor networks. It builds a loca-tion database in the server, which maps a set of RGB values to a geographic position. And the dis-tance measurements between sensor nodes are based on the DV-Hop[3]. After receiving an anchor’ RGB values, a sensor node converts the RGB values into HSV values, using an algorithm RGBtoHSV [16]. Note that HSV stands for Hue, Saturation, and Value[17]. Based on color theory, only the lightness of color fades out with the increasing of propagat-ing distances. That is, the V of HSV of an anchor, which is corresponding to the lightness, is decreased in proportion to the distance from the node to the anchor. With the new HSV values, the adjusted RGB values for the node can be obtained by using another algorithm, HSVtoRGB[16]. The node then calculates its own RGB values by averaging these adjusted RGB values, corresponding to the anchors. The node then sends its RGB values to the server so that the server can ﬁnd its most probable location by looking up the location database.

2.1. The information delivery of anchors[1]

In this section, we introduce some notations that are deﬁned in CDL:

• Davgis the average hop distance, which is based

on the DV-Hop[3].

• hijis the hop count between nodes i and j.

• Dikrepresents the hop distance from anchor k to

node i: Dik= Davg· hik.

Each node i maintains an entry of (Rik, Gik, Bik)

and Dik, where k represents the kth anchor.

• (Hik, Sik, Vik) is a set of HSV values of anchor k,

converted from (Rik, Gik, Bik) by the ith node.

Range represents the maximum distance that a color can propagate.

• (Rk, Gk, Bk) is a set of RGB values of anchor k.

The RGB RGB values of anchors are ﬁrst assigned randomly from 0 to 1. After a sensor node

(3)

i obtains each anchor k’s RGB values and hop count (hik), the RGB values are then converted to

HSV values by Eq.(1):

ðHk; Sk; VkÞ ¼ RGBtoHSV ðRk; Gk; BkÞ: ð1Þ

With hik, Dik can be computed. The updated HSV

values corresponding to node i of anchor k are cal-culated by Eq.(2): Hik¼ Hk; Sik ¼ Sk; Vik ¼ 1 Dik Range Vk: ð2Þ The RGB values of node i corresponding to anchor k are then calculated:

ðRik; Gik; BikÞ ¼ HSVtoRGBðHik; Sik; VikÞ: ð3Þ

The RGB values of node i are the mean of the RGB values corresponding to n anchors, and are com-puted as follows: ðRi; Gi; BiÞ ¼ 1 n Xn k¼1 ðRik; Gik; BikÞ; ð4Þ

where n is the number of anchors that node i re-ceived their RGB values.

Fig. 1 illustrates how node i obtains its RGB values.

2.2. The establishment of location database[1]

A location database is established when the ser-ver obtains the RGB values and locations of all anchors. The mechanism is based on the theorem of the mixture of different colors. With the RGB values of all anchors, the RGB values of all loca-tions can be computed by exploiting the ideas of color propagation and the mixture of different col-ors. Note that our localization approach is a kind of localization fingerprinting approaches[18]. Based on the similar steps of how a sensor node obtains its RGB values, as described in Section 2.1, one can create a location database. The only difference is Euclidean distance instead of hop-count distance is used. Firstly, the Euclidean distance between each location i and anchor k is derived:

dik ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðxi xkÞ 2 þ ðyi ykÞ 2 q ; ð5Þ

where (xi, yi) is the coordinate of location i, and

(xk, yk) is the location of anchor k. First of all, we

have to calculate the HSV values of each location i corresponding to anchor k: ðHk; Sk; VkÞ ¼ RGBtoHSVðRk; Gk; BkÞ; ð6Þ Hik ¼ Hk; Sik¼ Sk; Vik ¼ 1 dik Range Vk: ð7Þ

(4)

The RGB values of location i corresponding to an-chor k can be derived by Eq.(8):

ðRik; Gik; BikÞ ¼ HSVtoRGBðHik; Sik; VikÞ: ð8Þ

Then the RGB values of location i can be calculated by averaging all RGB values of location i corre-sponding to N anchors. ðRi; Gi; BiÞ ¼ 1 N XN k¼1 ðRik; Gik; BikÞ; ð9Þ

where N is the number of anchors.

In this way, the location for each sensor node can be constructed in the location database by maintain-ing the coordinate (xi, yi) and the RGB values

(Ri, Gi, Bi) at each location i. Then, the location of

a sensor node can be acquired by looking up the location database based on the derived RGB values. 2.3. Mobility[1]

When a mobile node arrives at a new location, it sends an anchor information request to neighbor nodes. If a neighbor node has the anchors’ RGB values, it transmits packets that contains the RGB values of each anchor and the hop count from the anchor to the node. After receiving the packets from neighbor nodes, node i compares and calculates the Dik to the kth anchor and selects the smallest Dik.

With the RGB values and Dikto all anchors, node

i can update its RGB values using Eqs. (1)–(4). The new RGB values are then transmitted to the server and the position of node i will be updated in the location database.

3. Related work

3.1. Existing routing protocols

In this section, we review several routing proto-cols in WSNs, which can be categorized into ﬂat rout-ing, hierarchical routrout-ing, location-based routrout-ing, and source routing. Flat routing, which is a kind of data-centric routing, is that when a node queries for data in its communication range, neighbor nodes which have the data will transmit the data to that node. Example ﬂat routing protocols include SPIN

[4] and Direct Diﬀusion [11]. Hierarchical routing builds a hierarchical topology in the region. It per-forms data aggregation and fusion in cluster heads in order to decrease the number of transmitted messages to the base station. Example hierarchical

routing protocols include TTDD [5], MECH [12], and PEGASIS[13]. As to the location-based routing, sensor nodes are addressed by their locations. We can use the direction of the destination to forward data to the nearest neighborhood recursively until reaching the destination. Example location-based

routing protocols include GEAR [6] and SPAN

[14]. Source routing is a routing technique in which the source node determines the complete sequence of nodes to forward the packet. The source node explicitly lists this route in the packet’s header. It identiﬁes each forwarding ‘‘hop’’ by the address or node’s ID of the next node to transmit the packet on its way to the destination node. Example source routing protocols include DSR [7], ESDSR[2]and MSR [15]. In the following, we review a classical routing protocol from each category.

3.1.1. Flat routing – sensor protocols for information via negotiation (SPIN)[4]

SPIN disseminates all the information at each node to neighbor nodes in the network. A source node ﬁrst advertises its data to neighbor nodes. If a node within the radio range wants the data adver-tised by the source node, it sends a reply packet to the source node. The source node will send data to the request node which queried for the data. This is illustrated inFig. 2. The disadvantage of SPIN is

B B REQ B A B A DATA ADV ADV ADV ADV ADV B A B A DATA DATA DATA REQ REQ REQ

a

b

c

d

e

f

A ADV A

Fig. 2. Flat routing – SPIN. Node A starts by advertising its data to node B (a). Node B responds by sending a request to node A (b). After receiving the requested data (c), node B then sends out advertisements to its neighbors (d), who in turn send requests back to B (e and f)[4].

(5)

that it cannot guarantee the delivery of interested data. If a node is interested in the data that is far away from the source node and nodes between source and destination nodes are not interested in that data, data will not be delivered to the destina-tion node at all.

3.1.2. Hierarchical routing – two tier data dissemination (TTDD)[5]

An approach that provides data delivery to mul-tiple mobile sinks is called TTDD, is illustrated in

Fig. 3, which is two tier data dissemination. Each data source proactively builds a grid structure that is used to disseminate data to the mobile sinks by assuming that sensor nodes are stationary and loca-tion aware. The sink chosen by the server may leave its old position to another position. Once this occurs, sensor nodes surrounding the sink process the signal sent by the sink and one of them becomes the new source node to generate data reports. 3.1.3. Location based routing – geographic and energy aware routing (GEAR)[6]

This algorithm, as illustrated inFig. 4, discusses the use of geographic information while disseminat-ing queries to appropriate regions since data queries often include geographic attributes. Each node in GEAR keeps an estimated cost and a learning cost of reaching the destination through its neighbors. The estimated cost is a combination of residual energy and distance to the destination. The learning cost is a reﬁnement of the estimated cost that

accounts for routing around holes in the network. A hole occurs when a node does not have any closer neighbor to the target region than itself. Two phases are used in this algorithm, forwarding packets towards the target region and forwarding packets within the region. In the ﬁrst phase, when receiving a packet, a node checks its neighbors to see if one neighbor node is closer to the target region than itself. In the second phase, if the packet reaches the target region, it can be diﬀused in that region

by using recursive geographic forwarding or

restricted ﬂooding.

3.1.4. Source routing – dynamic source routing

(DSR)[7] and ESDSR[2]

The DSR routing[7]is based on the source rout-ing. It uses each node’s cache to store information about the routing path which will be maintained by each sensor node if the routing path is broken or not. Two operations are used to build a routing path, route discovery and routing maintenance. Route discovery is the mechanism by which a source node finds a route to the destination. When a source wants to send information to the destination, it searches its own cache to find a routing path. If the source node can not find a routing path to the destination, it starts to perform route discovery. First of all, the source node initiates a local broad-cast to start the routing. If the node which receives the packet is the route destination, it will route back to the source node. If it is not the destination, it also broadcasts a packet to its neighbors until reaching the destination. The source will choose the shortest path – the smallest hop-count path – to route to the destination.

The route maintenance operation is the mecha-nism which a node is able to detect any changes in

Node Source Sink

B

S

Fig. 3. Hierarchical routing – TTDD: one source B and one sink S[5].

Node Data packet Target region

Fig. 4. Location based routing – GEAR. Recursive geographic forwarding: data disseminate to four sub-regions recursively until regions with only one node are left[6].

(6)

the network topology. While a node is unavailable, its one hop neighbor nodes around the routing path should route back to the source such that every node’s cache can update the route information.

ESDSR[2]is an energy saving DSR protocol. It changes the way of choosing a routing path from the shortest path in DSR to the maximum expected life of the path. Once a source node setups a route discovery, the return packet of each route path will contain the minimum expected battery life. The source node chooses a routing path with the maxi-mum expected battery life among all routing paths. Each node also has its power table which contains its neighbor’s power state information so that it will choose a more energy eﬃcient path to the destination.

3.2. Comparison of different routing protocols In terms of the following metrics: energy cost, mobility, position awareness, and localization, the above mentioned approaches are compared, as shown inTable 1. The proposed color-theory-based energy efficient routing (CEER) is also included in the table, which will be described in Section4. First, the metric of energy cost indicates the average energy cost of the protocol for a packet sending from the source to the destination. CEER and ESDSR have the lowest energy cost. These two algorithms will be further evaluated quantitatively in our simulations later. Mobility indicates whether nodes in the WSN are fixed or mobile. DSR, ESDSR, and the proposed CEER can support mobility. The position awareness indicates that if a protocol needs each node to know its position, e.g., using GPS. TTDD and GEAR require this information to route data to the destination. The

last metric, localization, indicates that the proposed CEER can perform localization because the source node uses RGB values to assist routing [1].

4. Color-theory-based energy eﬃcient routing algorithm

In this section, we propose a color-theory-based energy efficient routing (CEER) algorithm for WSNs, which is based on the CDL[1]. The network model is first described. The routing process can be organized into three phases: setup phase, data dis-semination phase, and refinement phase, which will be described next.

4.1. Network model

The network model for our routing protocol is described as follows:

• A server is used for building a location database, localization, and collecting sensed data.

• The area is divided into clusters and a cluster head for each cluster is selected by the server. • There are four anchors, which are placed in the

four corners of the area [9]. Anchors collect aggregate data received from cluster heads. • All sensor nodes have a uniform energy at the

beginning.

• Each node wakes up at a ﬁxed rate to check if its RGB values have changed.

• All sensor nodes are mobile.

• CSMA/CA is used to avoid collision of packets.

4.2. Setup phase

In this phase, we need to choose a sensor node to be the cluster head in each cluster so that data col-lection can be started. First, we use a sensor node’s radio range R such that the grid size is R· R. All sensor nodes in a cluster (grid) are within their clus-ter head’s radio range. Secondly, each anchor ﬂoods its RGB values and average hop distance to each sensor node so that each sensor node can calculate its hop count to the anchor and adjust its RGB val-ues based on the hop count. Then, the server will receive the RGB values from each sensor node. By looking up the location database, the server can cal-culate each sensor node’s location and choose a cluster head which is close to the center of the grid, as shown in Fig. 5. Finally, the server uses the

Table 1

Comparison of six routing protocols, including the proposed CEER protocol Approach Energy cost Mobility Position awareness (e.g., using GPS) Localization

SPIN[4] Mid N/A No No

TTDD[5] Mid Low Yes No

GEAR[6] Mid Low Yes No

DSR[7] Mid Mid No No

ESDSR[2] Low Mid No No

CEER (proposed)

(7)

closest anchors to send information to the selected sensor nodes to let them know that they are cluster heads, as shown inFig. 6.

4.3. Data dissemination phase

After the setup phase, each cluster head will receive its cluster member’s information. If a sensor node’s position changes, its hop counts to the anchors may also change. The node then updates its RGB values and transmits the RGB values as well as collected data (if available) to its cluster head. When a cluster head receives information from its cluster member, it will setup a timer and wait for other possible information from other members until the timer expires. The advantage of setting up a timer is that we can wait for other clus-ter members to see if they have data to send. In this way, communication cost can be reduced.

When a cluster head wants to forward the aggre-gate data toward the server, two steps are

per-formed in CEER. In the ﬁrst step, it selects its one hop neighbors that are closer to a nearby anchor than itself as the next possible hop by comparing their respective RGB values, as shown in Fig. 7. In the second step, among those selected cluster members, the sensor node with the highest energy level is selected as the next hop, as shown in

Fig. 8. The selected node will receive aggregated data and again follows these two steps to transmit the data to its next hop until reaching the server via the anchor.

Our CEER algorithm does not have the topology hole problem that often occurs in location-based routing [6]. The CEER is based on the CDL [1]

and is basically a hop-count-based routing

approach, not a location-based routing approach. We useFig. 9to explain how our routing approach will not encounter the topology hole problem. In

Fig. 9, it shows a topology hole caused by an obsta-cle. For node 2, in a location-based routing

Anchor Cluster head Sensor

Fig. 5. Setup step 1: Choose a cluster head which is close to the center of the grid.

Anchor Cluster head Fig. 6. Setup step 2: The server uses the closest anchors to send information to the selected sensor nodes to let them know that they are cluster heads.

Anchor Cluster head Selected sensor Unselected sensor

Radio range

Fig. 7. Data dissemination step 1. Select one hop neighbor nodes closer to the anchor.

Anchor Cluster head Selected sensor Unselected sensor E=0.6 E=0.7 E=0.3 E=0.5 Radio range

Fig. 8. Data dissemination step 2. Choose the node with the maximum energy level from the selected nodes.

(8)

approach, it will choose node 3 as the next node since node 3 is closer to the destination in terms of Euclidean distance. Thus, it will encounter the topology hole problem. However, in the CEER, node 2 would choose node 1 as the next hop because the color of node 1 is closer to the anchor (destina-tion) than node 3, even if node 3 is geographically closer to the anchor. This is because if a node with a color closer to the color of the destination, it also implies the node has a smaller hop-count distance to the destination based on color theory. Since each sensor node can receive the hop-count information from adjacent sensor nodes, we can always ﬁnd a sensor node with RGB values closer to that of the anchor. Therefore, the CEER will not encounter the topology hole problem when routing data to the anchor.

4.4. Reﬁnement phase

There are several situations that a cluster head may fail to do its job. For example, if a cluster head has low power, it may not live long enough to be a cluster head. On the other hand, if a cluster head moves away from the grid, it may lose contact with some cluster members. In these situations, it is nec-essary to choose another sensor node to become a new cluster head so that the collected information can still be forwarded to the server. In our approach, the server is responsible to choose cluster heads because the server knows the speed, position, and energy level of each node. However, if we change cluster heads too often, it will waste energy and induce extra communication cost to switch the cluster head’s job from one sensor node to another. To save energy, under the following two conditions, the role of a cluster head will not be replaced:

• A cluster head with an energy level higher than half of the original battery capacity.

• A cluster head with an energy level lower than half of the original battery capacity but higher than the average energy level.

If a cluster head with low energy needs to be replaced, the server collects information of the sen-sor nodes in the corresponding grid. Then the server sorts each node’s speed from low to high and chooses a node if its energy level is higher than half of the original battery capacity or higher than the average energy level. When a sensor moves away from its cluster head, it broadcasts a packet to its neighbor nodes to ask for a new cluster head and hop counts. If there are more than two cluster heads for the sensor node to choose, the cluster head with RGB values closer to itself will be chosen.

5. Simulation results and discussion

In this section, we compare the proposed CEER with the ESDSR[2]by measuring their total energy consumption and latency per packet with respect to various numbers of sensor nodes. Since our routing algorithm is based on the CDL[1], which was imple-mented in C++. We also used C++ to construct the simulation environment.

5.1. Simulation model

Our simulation model follows that of CDL [1]

since our routing algorithm is based on the CDL.

All nodes were randomly placed in a 500 m·

500 m area and a modiﬁed random waypoint models [8] were used to simulate the mobility characteristic of each sensor node. In addition, a sensor node selects a moving destination and a velocity randomly. After reaching the destination, the sensor node pauses for a period of time (pause time). The node speed is uniformly distributed within [Vmin, Vmax]. The simulation parameters are

shown in Table 2. We used the same energy

consumption model as that of ESDSR [2]. The

energy consumption per data-packet of size D bytes over a given link can be modeled as

EðD; PtÞ ¼ K1PtDþ K2;

where Pt is the minimum transmit power required

for successful reception, and the typical values of constant K1and K2in IEEE 802.11 MAC

(9)

ment at 2 Mbps are 4 ls per byte and 42 lJ, respectively.

The two performance parameters are deﬁned as follows:

• Total energy consumption (mJ): summary of all sensor nodes’ energy consumption after 50 time slots (time units).

• Latency per packet (s): average time spent to send a packet with aggregate data, such as RGB val-ues and collected data, from a sensor node to the server.

In [9], it has shown that the location estimated error is minimized when anchors are placed at the corners. So we put four anchors (A1–A4) at the four

corner of the area, as shown inFig. 10, where S rep-resents a sensor, a solid line reprep-resents a direct link with one hop, and a dotted line represents an indi-rect link with more than one hop. In addition,

Fig. 11 shows the impact of seed (anchor) density on location estimated error [1]. Note that anchor density is the average number of anchors in one hop transmission range [1]. This ﬁgure shows that the more anchors (more speciﬁcally, the higher

anchor density), the better position accuracy (the lower estimated error).

Fig. 10 is also used to illustrate the diﬀerence between CEER using RGB and a revised CEER using (absolute distance, hop count). In Fig. 10, if sensor node S1intends to send data to anchor A3,

for a revised CEER using (absolute distance, hop count), it will select either S2 or S3 as the next

hop, as both sensor nodes have the same hop count to A3. However, for CEER using RGB, it will

always select S3as the next hop since S3’s color is

closer to A3’s color than S2’s color. This is because

CEER using RGB has better location accuracy than CEER using (absolute distance, hop count), as shown in [1]. That is, CEER using RGB always chooses a path with a shorter actual distance. There-fore, the direct beneﬁts of using RGB are to have lower total energy consumption and lower latency per packet in CEER.

5.2. Comparison with ESDSR[2]

Kim et al.[10]mentioned that a well dissemina-tion routing protocol should have three characteris-tics: energy efficiency, self-configuration, and scalability. By simulations, we have shown that our algorithm is better than ESDSR in terms of energy efficiency. Our algorithm has the characteris-tic of self-configuration in terms of cluster head selection, as described in Section4.4. As to the sca-lability, our algorithm is more scalable than ESDSR when the area is bigger, which will be described at the end of this section.

As shown inFig. 12, our approach (CEER) saves more energy than the ESDSR when the number of nodes increases. This is because the CEER uses a cluster head to aggregate data and adopts dynamic path ﬁnding to ﬁnd a neighbor node with maximum energy to route data to a nearby anchor. The

Table 2

Simulation parameters

Parameter Value

Area size 500· 500 m2

Node speed Randomly choose from

[Vmin, Vmax]

Node transmission range (R) 50 m

Pause time 0

Measurement period 50 tu

Update interval 5 tu

Time slot length (time unit) tu

A3 A2 A4 A1 S1 S2 S3 S4 Anchor

Fig. 10. Anchor deployment.

(10)

ESDSR uses local broadcast to ﬁnd a path and chooses the best route with the maximum expected life, but the source node does not aggregate data of its neighbor nodes to send data to the destination.

InFig. 13, the CEER has less latency per packet than the ESDSR because the ESDSR uses the path with the maximum expected life in order to balance the load of the entire sensor network. In contrast, our approach selects a next hop node to deliver data based on the RGB values and the maximum energy levels of neighbor nodes, which is more eﬃcient.

We also simulated our routing algorithm without cluster heads to verify that our routing algorithm also works well without clustering. We used the two data dissemination steps in Section4.3so that each node will route its data to the nearest anchor without relying on a cluster head.Fig. 14shows that our algorithm also works well without cluster heads and saves more energy than the ESDSR.

We also used the same network model as that of ESDSR [2] to validate the correct implementation of the ESDSR and CEER. Forty sensor nodes were randomly placed in an area sized from 200 m· 200 m, 300 m· 300 m, 400 m · 400 m, to 500 m · 500 m, respectively, while keeping the number of sensor nodes (40) and connection (20) constants.

Fig. 15 shows that, in a small area, our approach

did not perform well compared to the ESDSR in terms of number of dead nodes at the end of the simulation. This is because we used a cluster head to forward collected data, even if a node is within the transmission range of an anchor. When the area grows bigger, the routing path is long enough so that the overhead of routing packets to the cluster head can be reduced. Therefore, the CEER is more scalable than the ESDSR when the area grows big-ger, as illustrated inFig. 15.

6. Conclusion

In this paper, we have presented an eﬃcient color-theory-based energy eﬃcient routing (CEER) algorithm based on a color-theory-based dynamic localization (CDL) algorithm. The key idea of the CEER is that it selects those cluster members that are closer to the anchor than itself as next possible hops by comparing their RGB values. Among the selected cluster members, the sensor node with the highest energy level is chosen as the next hop. Sim-ulation results have shown that our routing algo-rithm can save up to 50–60% energy compared to the ESDSR [2]in mobile wireless sensor networks. CEER ESDSR 0 1 2 3 4 5 200 300 400 500 Number of nodes

Latency per packet(sec)

Fig. 13. Latency per packet vs. number of nodes.

CEER ESDSR 0 10 20 30 40 50 200 300 400 500 Number of nodes

Total energy consumption

Fig. 14. Energy consumption vs. number of nodes (without cluster heads). 0 10 20 30 40 50 200 300 400 500 Number of nodes

Total energy consumption

CEER ESDSR

Fig. 12. Total energy consumption vs. number of nodes (with cluster heads). CEER ESDSR 0 2 4 6 8 10 12 40 90 160 250 Area (1000 m2₎

Number of dead nodes

(11)

In addition, the latency per packet of the CEER is 50% less than that of the ESDSR.

References

[1] Shen-Hai Shee, Kuochen Wang, I.L. Hsieh, Color-theory-based dynamic localization in mobile wireless sensor net-works, in: Proceedings of Workshop on Wireless, Ad Hoc, Sensor Networks, August 2005.

[2] Mohammed Tarique, Kemal E. Tepe, Mohammad Naserian, Energy saving dynamic source routing for ad hoc wireless networks, in: Proceedings of Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks, April 2005, pp. 305–310.

[3] J.G. Lin, S.V. Rao, Mobility-enhanced positioning in ad hoc networks, in: Proceedings of the IEEE Wireless Communi-cations and Networks Conference, vol. 3, March 2003, pp. 1832–1837.

[4] W. Heinzelman, J. Kulic, H. Balakrishnan, Adaptive proto-cols for information dissemination in wireless sensor net-works, in: Proceedings of the 5th ACM/IEEE MOBICOM, Seattle, WA, August 1999, pp. 174–185.

[5] F. Ye, et al., A two-tier data dissemination model for large-scale wireless sensor networks, in: Proceedings of the ACM/ IEEE MOBICOM, 2002, pp. 148–159.

[6] Y. Yu, D. Estrin, R. Govindan, Geographical and Energy-Aware Routing: A Recursive Data Dissemination Protocol for Wireless Sensor Networks, UCLA Comp. Sci. Dept. Tech. Rep., UCLA-CSD TR-010023, May 2001.

[7] J. Borch, D.B. Johnson, D.A. Maltz, The Dynamic Source Routing Protocol for Mobile Ad Hoc Networks, IETF Internet Draft, draft-ietf-manet-dsr-00.txt, March 1998. [8] J. Yoon, M. Liu, B. Noble, Random waypoint considered

harmful, in: Proceedings of the IEEE INFOCOM. April 2003, pp. 1312–1321.

[9] Hongyi Wu, Chong Wang, Nian-Feng Tzeng, Novel self-conﬁgurable positioning technique for multihop wireless networks, IEEE/ACM Transaction on Networking 13 (3) (2005) 609–621.

[10] Sooyeon Kim, Sang H. Son, John A. Stankovic, Yanghee Choi, Data dissemination over wireless sensor networks, in: IEEE Communication Letters, September 2004, pp. 561–563. [11] C. Intanagonwiwat, R. Govindan, D. Estrin, Direct diﬀu-sion: a scalable and robust communication paradigm for sensor networks, in: Proceedings of the ACM MobiCom, 2000, pp. 56–67.

[12] R.-C. Chang, C.-J. Kuo, An energy eﬃcient routing mech-anism for wireless sensor networks, in: Proceedings of the Advanced Information Networking, and Applications, April 2006.

[13] S. Lindsey, C. Raghavendra, PEGASIS: power-eﬃcient gathering in sensor information systems, Proceedings of the IEEE Aerospace Conference 3 (9–16) (2002) 1125–1130. [14] B. Chen et al., SPAN: an energy-eﬃcient coordination

algorithm for topology maintenance in ad hoc wireless networks, Wireless Networks 8 (5) (2002) 481–494. [15] L. Wang et al., Adaptive multipath source routing in ad hoc

networks, in: Proceedings of the IEEE International

Conference on Computer Communications, 2001, pp. 867–871.

[16] E. Vishnevsky, RGB to HSV & HSV to RGB. <http:// www.cs.rit.edu/~ncs/color/t_convert.html>.

[17] DevX.com, ‘‘Color Models’’. <http://www.devx.com/pro-jectcool/Article/19954/0/page/7>.

[18] A. Taheri, A. Singh, A. Emmanuel, Location ﬁngerprinting on infrastructure 802.11 wireless local area networks (WLANs) using Locus, in: Proceedings of the 29th Annual IEEE International Conference on Local Computer Net-works, November 2004, pp. 676–683.

Tai-Jung Chang received the B.S. degree in Computer and Information Science from the National Chiao Tung Univer-sity, Taiwan, in 2002, and the M.S degree in Computer Science and Engi-neering from the same university in 2006. He is currently a System Engineer in Sunplus Technology Co., Ltd. He served in the army as a second lieutenant artil-lery platoon leader from 2002 to 2004. His research interests include wireless sensor networks and network ﬁle systems for consumer electronic products.

Kuochen Wang received the B.S. degree in control engineering from the National Chiao Tung University, Taiwan, in 1978, and the M.S. and Ph.D. degrees in elec-trical engineering from the University of Arizona in 1986 and 1991, respectively. He is currently a Professor in the Department of Computer Science, National Chiao Tung University and a Acting Director of the Computer and Network Center at the university. From 1980 to 1984, he was a Senior Engineer at the Directorate Gen-eral of Telecommunications in Taiwan. He served in the army as a second lieutenant communication platoon leader from 1978 to 1980. His research interests include wireless (ad hoc/sensor) net-works, mobile computing, and power management for portable wireless/multimedia devices.

Yi-Ling Hsieh received the B.S. and M.S. degrees in computer and information science from the National Chiao Tung University, Taiwan, in 2003 and 2005, respectively. She is currently a Ph.D. student in the Department of Computer Science, National Chiao Tung Univer-sity. Her research interests include wire-less ad hoc/sensor networks, mobile computing, and eﬃcient energy man-agement for portable wireless/multi-media devices.