Localization in Wireless Sensor Networks using Three Masters

Localization in Wireless Sensor Networks using Three Masters Mayuresh M. Patil, Umesh Shaha,

SPA"

U. B. Desai, S. N. Merchant. Lab

Department of Electrical Engineering Indian Institute Of Technology, Bombay Powai, Mumbai - 400 076 Email IDS: {mayuresh,umeshs,ubdesai,merchant}@ee.iitb.ac.in Abstract In wireless sensor networks, sensor node localization is an important topic because sensor nodes are randomly scattered in the region of interest and they get connected into network on their o m . Localization information is needed in routing of data, reducing energy consumption and in location dependent data acquisition. We propose a localization scheme plus a medium access protocol that (a) reducm power consump tion, (b) does not require high power beacons, (c) provides better localization accuracy and (d) helps in reducing collisions. The proposed localization scheme is based on received signal strength by the three masters. Moreover, it is a distributed algorithm.

-

Indez Term Wireless sensor network, Localization, Three beacons method, MAC layer.

I. I n t r o d u c t i o n A sensor network consists of number of sensor nodes scattered in the region of interest to acquire some physical data. The sensor node should have the ability of sensing, processing and communicating. The sensor network may be thought of the as an ad-hoc network, but with two main differences; sensor network is a power constraint network and has an excessively large number of nodes. Multi-hop routing protocol must be designed with power constraint as the main cost factor. The MAC layer protocol should also be designed for low power consumption. Besides these localization is a very important issue in sensor networks. In the sensor network the location of the node is unknown in almost all applications. Also the nodes may be mobile. Thus localization is critical. In this paper we propose a new localization algorithm based on three masters (three beacons)along with a medium access protocol. The key advantages of the proposed method over other methods are: (a) does not require high power beacons, (b) provides better localization accuracy and (c) helps in reducing collisions. We validate the prosed scheme through extensive simulations.

A . Need OfLocalization Localization algorithms find the location of the sensor nodes relatively or absolutely. There are following uses of localizai

0-7803-8964-61051520.00 02005 lEEE

tion. Localization enables the efflcient routing: A typical sensor network has large number of nodes wihich communicate at very short distance (a few meters). The data sensed by a node has to be communicated to the central unit (or a sink) through several other nodes. Thus multi-hop routing is a must. In order to implelemnt multi-hop routing it is necessary that nodes are of thier localitly, anmely, they know their relative postion with respect to their neighbours. Consequently localization becomes very important. Localization provides the power saving: Suppose we have deployed the sensor network for pollution monitoring. Now the neighbor sensor nodes will have data which will not be dramatically different from each other. Thus to save power it makes sense to combine the data from neighboring nodes and then communicate the combined, reduced data set, thereby conserving power (since communication takes lot more power than local processing). In order to do this local data fusion, we wilt need the location information. Once again locdization becomes vital. Localization assists in the applications like t a r g e t tracking: In this application, we typically need to determine the range, speed and the direction of the target. Sensors are deployed in region which sense the sign& from the moving target, using which the range, speed and direction of the target me monitored. In order to calculate the global orientation of the target it is necessary that we know the location of the sensor nodes. Once again localization becomes necessary. Localization useful in locating the source of the data: In many applications, an event based sensor network is used. Here, the nodes are normally in sleep mode and when an event occurs (say sudden vibrations take place) the nodes are awakened. The nodes than sense and transmit the data. Such data requires a location stamp and therefore localization becomes necessary. From the above examples we see that localization is indeed a necessity for sensor networks. Moreover, localization algorithms vary depending on applications. Localization algorithms for a sensor network with stationary nodes and a sensor network with mobile nodes will be different. next we briefly summarize some of the leading localization algorithms.

384

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 3, 2008 at 00:42 from IEEE Xplore. Restrictions apply.

A75

IS

A19

19 0

0

.U3

0

AI 48

0

0 M

AI1

a

U

Fig. 1. One Master Method Sensor Environment

Fig. 2. Two Master Method Sensor Environment

B. Localization Methods

This method of localization is accurate but consumes a lot of power. It requires high power beacons and a lot of power is wasted in transmitting the relative distances to the master.

We can classify the localization schemes in two categories; range based and range free. The range based model basically uses the channel characteristics to find the distance. The range free localization technique does not use the channel model for calculating the location. In the range based method, there are three main schemes for localization: (i) based on received signal strength indication (RSSI), (ii) based on time of arrival (TOA), and (iii) based on angle of arrival (AOA). The RSSI method is widely used; i t exploits beacons and using the received power from the beacons finds the distance. In the next section we first review existing RSSI based methods and then present OUT new algorithm based on

FtSSI. €3.1 One Master Method Far Localization

This method based on RSSI uses one master node that knows its position, and using this node finds the relative distances and then the location of the other nodes. This technique uses the three different operations to locate all the sensors. The major steps in this algorithm are: 1. Single sensor identification. 2. Imprecise sensor location map definition. 3. Sensor location map refinement. First a beacon initiates the localization algorithm and transmits the signal. Nearby sensors will receive this signal and then they send an acknowledgment. The master then gives a unique ID to each of these nodes. The nodes calculate the distance based on receive power and transmit this information to the master. The master terms these nodes as sons. Next, the sot29 of master will find out their sons by foIlowing the same procedure. This procedure continues until the complete network i s covered. The disadvantage of single master method is that the relative distances first comes to the master from each node and then the master dcuiates the locations of the nodes, and then transmits the whole information back to nodes. The globat master keeps track of all the relative distances. The master thus has the location of all the nodes.

C. Two Master Method For Localization This method uses two high power masters (beacons), and two masters are placed in the corners of the region of interest as shown in the Figure 2. The unknown node detects the signal strength from the two masters and immediately calculates its location based on circular triangulation. Finally, when the acknowledgmentkfrom all the nodes reaches the beacon, the localization is compIete. This method requires less power than the one master method. The disadvantage is that it requires high power beacons. Each beacon should have enough power to reach the farthest node. Moreover, since one can not get unique location using two beacon nodes by circular triangulation, there will he ambiguity. This is resolved by restricting the beacone node to be at the.corners of the boundry. Our proposed three master method with simple random hack-off requires less power and reduces collisions. 11. Three Master Method for Localization

The method using two beacons requires high power beacons, moreover when the nodes are mobile, localization needs to be done frequently resulting in power wastage by the beacon as well as nodes (due to communication). We propose the three master method, which uses three beacons. The key advantage of the proposed scheme is that, the beacon nodes need not be high power nodes; any node in the network can work as a beacon. There is further saving in power, since the communication overhead (due to collisions) is much less. The proposed algorithm is explicated next.

A. Algorithm Our proposed algorithm assumes that we have three beai con nodes. Note that these are not special nodes. These are having the same capabilities as any other node in the

3 85 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 3, 2008 at 00:42 from IEEE Xplore. Restrictions apply.

neighbours. This way the command is received by all the nodes in the network in a multi-hop fashion. A.1 Circular Triangulation

*I9

0 N O

0 AI1

Ht

151

0

0

AIS

0

O

A16

The actual location is found from the received signal strength which gives the distance information. The distance from the three masters is found and then using the circular triangulation, sensor node can find its location. Distance calculation using received signal strength is given by

110

a

M

A13

0

U

AIS

A16

0

where

P, - Received power, W

Pi - Transmitted power, W

Fig. 3. Three Masters Method Sensor Network Environment

A,, network. The only requirement for the beacons is that, they know their location and can communicate with one another directly. The basic algorithmic steps are: 1. One of the beacon nodes (always the same, e.g node Bl in Figure 3) transmits the localization packet. This packet

contains transmitted power, its ID and its location. 2. This packet is received by other beacon nodes and neighboring unlocalized nodes in the network. The other two beacon nodes (82, B3) do not transmit at same time as B1. They back-off by random time interval and transmit the localization packet. 3. Some nodes in the neighborhood of beacons will receive the packet from all the three beacons. These nodes now can calculate their distances from beacons by measuring the received signal strength. Then these nodes will calculate their location using circular trdngulation (discussed in next section). These nodes can work as beacon nodes €or the next iteration. 4. Now the newly localized nodes will back-off by a random time. and the one whose back-off period expires first will start the next iteration by transmitting a localization packet . 5. Now steps 2 to 4 are repeated till each node in the network is localized. In short, the nodes in the neighborhood of beacons get localized and in turn act as beacon nodes. Note that for getting localized, the node must receive the localization packet from a t least three beacons for circular triangulation. If the node receives the localization packet from more than three beacons then the three with highest received signal strength are selected. In a sensor network the nodes can be mobile. Thus l o d i m tion needs t o be done perodically. Also some particular node may wish to relocalize itself (may be because of its mobility or reduced received signal strength). The relocalization can be initiated by any node in the network (including beacon node) by issuing relocalize command. Generally, perodic relocalization is initiated by beacon node (Bl). When the relocalize command is issued, it is received by their neighbours. Now all these neighbors will broadcast this command to its

- Effective aperture of antenna, m2

Act - Effective aperture of antenna, m2 r - Distance, m X - Wavelength, m If the transmitter employs an isotropic source, (1) then,

Pr Aw P, = 4nr2

All constants can be grouped into a single constant K

In general, the received power can b e expressed as

PTj= k 4 "

(4)

where k is a constant which takes into account carrier €requency and transmitted power, d j is the distance between the sensor and the j f h beacon, j = I , 2, and 3, and cy is the attenuation exponent. Thus, if we know k and a then we can estimates the distance form the beacons using

(5) A fiensor node will get the signal from the three beacons. The following system of equations

(s(i)- b Z l ( i ) ) 2

( z ( i )- b Z Z ( i ) ) 2

+ (g(i) - b y 1 ( 0 ) 2 = &i)

+ (v(i)- by2(i))z

= d#)

(z(2) - b z 2 ( ( i ) ) 2 -t (y(l) - by2(i))2 =

}

4(i)

(6)

is solved for x(.)and y(.). Now, the new sensor nodes whose location is found can be made beacons for the next iteration. The procedure is repeated by the new beacon nodes till the location of all the nodes is found.

386 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 3, 2008 at 00:42 from IEEE Xplore. Restrictions apply.

Fig. 4. Typical Range of a Beacon Node 0'

M

1W

150

200

254

3W

3M

400

IS0

Fig. 6. Localization error Vs Iteration for T w o master and Three master method

.

all the newly localized nodes will back-off randomly The node whose back-off period expires first initiates the next iteration by transmitting the localization packet. This packet will be received by other beacons as well as other nodes and the above procedure is repeated. Note that in this method when any beacon listens to other beacon signal it gernerates a random number and then backs off correspondingly. This reduces collisions and retransmissions. Fig. 5. Circular Triangulation

B. Example Consider the sensor network environment as shown in Fig-

ure 3. We have three beacon nodes (black filled square) B1, 3 2 , and B3. The other nodes are scattered randomly in the region. Note, in our we do not need to place the beacon

In short, during localization the beacons transmit the packets in a predecided and organized manner, so collisions will be very small in number. This method of localization is distributed in nature and saves beacon power since we are communicating only with nearby nodes. Also the fault tolerance of the network increases since any node can become a beacon node. 111. Simulation

nodes at the corner of the region. Nevertheless, they should be neighbors of each other. One of the beacon nodes (always the same) say B1 transmits the localization packet. This packet consists of transmitted power, its ID and its location. After beacon B1 transmits the packet, it is received by the surrounding nodes and other beacons B2, B3. Let us say that it is received by nodes A l , A2, A3, A4, B2, and B3. As soon as the beacons B2 and 33 receive the signal from B1, they generate a random number and back-of€proportionally. After back-off period B1 and B2 will transmit the localization packet. Because of random back-off the probability of collisions will be very small. Assume nodes A l , A2, and A3 receive the packets transmitted by the beacons B l , R2 and B3. Then using the received signal strength from B1, B2 and B3; A 1 cdculates its relative distance from E l , B2, and B3. A I localizes itself using circular triangulation as shown in the Figure 5. Similarly A2 and A 3 will find their location. Now

We tested our proposed algorithm by simulating a sensor n e b work in square region of size 100 units x 100 units area and with 100 and 1000 nodes. We assume BPSK modulation and the packet length during localization to be of 2' bytes. We have simulated the two master algorithm and the proposed three master algorithm. We ran the simulation 500 times and calculate localization error. The simulation results axe depicted Figure 6, 7, and 8, As seen from the Figure 6 localization error in case of the two masters method is much higher than the three masters method. Another advantage of the above scheme is that of we do not require the beacons with high power capabilities. Figure 7 shows the power dissipation of sensor nodes and beacons under ideal condition (namely, no retransmissions). From the graph we observe, that beacon power required in the two master method is very

387 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 3, 2008 at 00:42 from IEEE Xplore. Restrictions apply.

500

V. References

Fig. 7.

[l] Yu-Chee Tseng, Shih-Lin Wu, Wen-Hwa Liaa, Chih-Min Chao, "Location Awareness in Ad Hoc Wireless Mobile Network, "IEEE Computer,,Vo1.34,No 6,PP-46-52,June,2001. [2] J. B. Andersen, T. S. Rappaport, and S. Yoshida, "Prcpogation measurement and models for wireless communication channels, " IEEE Comna~nic~tion Mag.: vo1.33,PP. 4249, Jan. 1995. [3] Filippo Mondinelli, and Zsolt M. Kovacs-Vajna, "SelfLocalizing Sensor Network Architectures, " JEEE Zhnsaction On Intmnaentation and Measurenaent,Vol.,53,No.2,April, 2004. [4] P. Bergamo, G.Mazzini, " Localization in Sensor Network with fading and Mobility, "IEEE Thnscation On Coommimication,JuneJ002 [5] Neal Patwari, Alfred 0. Hero, Matt Perkins, N. S. Correal, R. J. O'Dea, " Relative Location Estimation In Wireless Sensor Network, "IEEE h n s a e t i o n s On Signal Processing, vo1.,51,No.8,August,2003.

Power Dissipation in Two Masters algorithm without MAC

layer

Power Dissipation in Three Masters algorithm without MAC layer

Fig. 8.

high. Whereas, in the three master case, the beacon power required is the same as any ordinary node. In our algorithm during localization beacons back-off by random time interval. This reduces retransmissions.

IV. Conclusion The proposed three master method performs much better than two master method and one master method with respect to localization error and power consumption. Also the proposed method does not require high power beacons. The random back-off used during localization reduces the collision and hence the retransmissions. This also reduces the power consumption and localization time for the network.

388 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 3, 2008 at 00:42 from IEEE Xplore. Restrictions apply.