Omotayo and wirelessradios are becoming extremely small and

Omotayo G. Adewumi, Karim Djouani, Anish M. KurienF’SATI/Department of Electrical EngineeringTshwane University of TechnologyPretoria, South Africa{adewumiog, djouanik, kurienam}—Research has revealed that the correlation betweendistance and RSSI (Received Signal Strength Indication) values isthe key of ranging and localization technologies in wireless sensornetworks (WSNs). In this paper, an RSSI model that estimatesthe distance between sensor nodes in WSNs is presented. Theperformance of this model is evaluated and analyzed in a realsystem deployment in an indoor and outdoor environment byperforming an empirical measurement using Crossbow IRISwireless sensor motes. Our result shows that there is less error indistance estimation in an outdoor environment compared toindoor environment. The results of these evaluations wouldcontribute towards obtaining accurate locations of wirelesssensor nodes.Index Terms—WSN, RSSI, Localization, Distance.I. INTRODUCTIONResearch on Wireless Sensor Networks (WSNs) have attracteda lot of attention in this current time, and this attention is risingbecause WSNs promise to be an enabling technology of thefuture owing to the fact that processors, sensors and wirelessradios are becoming extremely small and inexpensive. In thenear future, the world we live in will be populated by objectsthat are globally networked such that physical environmentsare enriched by computational power. A WSN is a networkconsisting of a large number of wireless radio nodes equippedwith sensing devices and are densely distributed for specificapplications. Each node is equipped with a transceiver tocommunicate with another node within its communicationradio range 1.Sensed information in many applications of WSN onlybecomes useful when it is accompanied by the location of thearea and accurate distances of where such information is beensensed. Hence, sensor nodes need to know the distancebetween one another in order to calculate their positions 2,3. Currently, there are many techniques for determiningdistance between sensor nodes. Physically, the Time of Signal(Acoustic or RF) Arrival (TOA) calculates the distance by useof signal propagation velocity and propagation time, Angle ofArrival (AOA) is measured by getting the signal directionsend by the adjacent node through the combination of arrayantenna and multiple receivers, while Received SignalStrength Indicator (RSSI) measures received power byreceiving node, calculates propagation loss and transformpropagation loss to distance by theoretical or empirical signalpath loss model. RSSI model is considered in this paperbecause the use of it does not require any additional hardwarebut simply a radio transceiver compared to other range basedmodels 4, 5, 6, 7.In this paper, a model which can accurately describe therelationship between the RSSI values and distance in aspecific environment (indoor or outdoor) is evaluated. TheRSSI signal propagation model is currently of three types;Free Space propagation model, Two-ray ground Model andLog Normal Shadowing Model (LNSM) 8, 9, 10, 11.The first two models have special requirements for theapplication environment while the third model which isconsidered in this paper is a more general signal propagationmodel.The remainder of this paper is organized as follows; inSection 2, the RSSI model is defined together with atheoretical and empirical description. In section 3, theexperimental setups and a brief description of the crossbowmotes is explained while in section 4, the experimentalanalysis together with the implementation is defined. Theexperimentation results are presented and explained in section5. Finally in section 6, conclusions with a short summary ofthe results and future research directions are provided.II. RSSI-DISTANCE MEASUREMENT MODELRSSI, which is a standard feature in most wireless radios,has attracted a lot of attention in recent literature. RSSI isdefined as the voltage measured by a receiver’s received signalstrength indicator (RSSI) circuit. Often, it is equivalentlyreported as measured power, i.e., the squared magnitude of thesignal strength. It eliminates the need for additional hardwarein small wireless devices and exhibits favorable properties withrespect to power consumption, size and cost. Given a model ofradio signal propagation in a building or other environment,RSSI can be used to estimate the distance from a transmitter toa receiver in other to estimate the positions of the sensor nodes.However, this approach requires detailed models of radiofrequency (RF) propagation and does not account for variationsin receiver sensitivity and orientation 11.Wireless sensors communicate with neighboring sensors.As a result, the RSSI of RF signals can be measured by eachreceiver during normal data communication without presentingadditional bandwidth or energy requirements. RSSImeasurements are relatively inexpensive and simple toimplement in hardware. In distance prediction based methods,RSSI data are directly mapped to distances through a signal978-1-4673-4569-9/13/$31.00 ©2013 IEEE 1534propagation model. In the case of outdoor deployments, wherethere are no obstacles, a single signal propagation model is inall the profiling data required. In the case of indoordeployments, the same also applies, but the signal propagationmodel might be different across different buildings withdifferent furniture and wall arrangements in this case 11,12, 13.Given an underlying signal propagation model, RSSI datacan be mapped to specific distances. This distance is useddirectly in order to perform localization. The accuracy of thedistance prediction based methods depends heavily on theaccuracy of the signal propagation model used for translatingRSSI values to distances 13.The theoretical and empirical model for RSSI are classifiedand explained belowA. Basis of RSSI ModelThe principle of RSSI ranging techniques describes therelationship between the received power, transmitted power ofwireless signals and the distance among wireless sensor nodes.This relation is shown in equation 1. Pr is the received powerof wireless signals. Pt is the transmitted power of wirelesssignals, d is the distance between the sending nodes and thereceiving nodes, n is the transmission factor whose valuesdepend on the environment of propagation (indoor andoutdoor) 12.Pr = Pt × (1/d)n (1)Pt, the transmitted power of nodes is given, the receivedpower in dBm is then given as;Pr(dBm) = A – 10 × n × logd (2)In equation 2, it is clearly shown that the values ofparameter A and parameter n determine the relationshipbetween the strength of received signals and the distance ofsignal transmission 12.B. RSSI Based Ranging Theoretical ModelAs stated in 8, 9, 10, 11, RSSI propagation models inwireless sensor networks include the free space model, two rayground model and log-normal shadowing model. The freespace propagation model is used to predict received signalstrength when the transmitter and receiver have a clearunobstructed line of sight path between them. As with mostlarge scale radio wave propagation models, the free spacemodel predicts that received power decays as a function of theTransmitter-Receiver separation distance. The free spacepower received by a receiver antenna separated from aradiating transmitter antenna by a distance d, and is given bythe Friis free space in equation 3.Pr(d) = (Pt × Gt × Gr × 2) / (4)2 × d2 × L (3)where Pt is the transmitted power, Pr(d) is the receivedpower which is a function of the transmitter to receiverseparation, Gt is the transmitter antenna gain, Gr is the receiverantenna gain, d is the transmitter to receiver separation distancein meters, L is the system loss factor not related to propagation(L is greater or equal to 1) and  is the wavelength in meters.The two ray ground model is the case when the singledirect path between the transmitter and the receiver is the onlyphysical means of propagation of the radio signal. The two-rayground model considers two propagation paths; the directedpath and a ground reflected propagation between thetransmitter and the receiver.Pr(d) = (Pt × Gt × Gr × (ht2 hr2)) / d4 (4)In the two-ray ground propagation model, the receivedpower at distance d is given in the equation (4) above where htis the transmitter antenna height and hr is the receiver antennaheight.Log-normal shadowing model is a general propagationmodel. It is suitable for indoor and outdoor environments. Thismodel provides a number of parameters which can beconfigured according to different environment (indoor andoutdoor) and is shown in equation 5.PL(dB) = PL(d0) + 10 × n × (logd / d0) + X (5)In equation 5, d0 is the near-earth reference distance whichdepends on the experiential value, n is the path-loss indexwhich depends on specific propagation environment and itsvalue will become larger when there are obstacles. X is thezero mean Gaussian random variable. The Log-normal shadowmodel is most suitable for wireless sensor networksapplications and is the main focus of this paper because of itsuniversal nature and the ability of being configured accordingto environments.C. RSSI Based Ranging Empirical ModelReceived signal strength techniques measures the power ofthe signal at the receiver based on the known transmit power.The respective propagation loss can be calculated. Theoreticaland empirical models are used to translate this loss into adistance estimate. This method has been used mainly for RFsignals. The relationship between RSS values and distances arederived as follows by 15.The received signal power is inversely related to distance as;P  1 / Dn (6)RSS  10 × log(1 / D)n (7)RSS = -10 × n × log(D) + C (8)D is the distance of deployed sensor and n is the path-lossexponent factor. C is considered to be a fixed constant. In morecompact form, RSS can be represented as:RSS = -m × log(D) + C (9)1535In equation 9, m is the slope of linear equation betweenRSS and log(D). The path-loss exponent factor is given as:n = m / 10 (10)III. EXPERIMENTAL SETUPIRIS Motes are the motes used for our experimentation andthey form part of the latest generation of Motes from CrossbowTechnology, XM2110 (2400 MHz to 2483.5 MHz band). Theradio used by the IRIS is an IEEE 802.15.4 compliant RFtransceiver designed for low-power and low-voltage wirelessapplications. It uses Atmel’s AT86RF230 radio that employsO-QPSK (Offset Quadrature Phase Shift Keying) with half sinepulse shaping. The 802.15.4 radio includes a DSSS (DigitalDirect Sequence Spread Spectrum) baseband modem providinga spreading gain of 9 dB and an effective data rate of 250 kbps17.Fig. 1. Figure taken from 17, XM2110 IRIS mote with standard antenna andits respective block diagramTABLE I. This table from 17 summarizes the details regarding IRISXM2110 Crossbow moteMote HardwarePlatformPlatform Types IRIS XM2110MCU ChipTypeProgram Memory(KB)SRAM(KB)AT Mega 12817.37 MHZ, 8Bit1288Sensor BoardInterfaceType10Bit ADCUART51 Pin7, 0V to 3VInput 2RF Transceiver(Radio)ChipRadio Frequency(MHZ)Max Data Rate(Kbits/sec)Antenna ConnectorRF 2302400250MMCXDefault PowerSourceTypeTypical Capacity(mA-hr)AA, 2x2000IV. EXPERIMENTAL ANALYSISThe experiment was carried out in an indoor and outdoorenvironment. The IRIS sensor nodes were programmed beforeperforming any forms of measurements. One end of theUniversal Serial Bus (USB) extension cable is connected to anavailable USB port on the laptop and the other end to thegateway USB connector. Each node was programmed bymounting them on the Gateway. The Gateway used is MIB520Interface board by crossbow. Application software called moteviewwhich acts as an interface between the user and thedeployed wireless sensors was also installed on the laptop. Itprovides the tools to simplify deployment and monitoring. Italso makes it easy to connect to a database to analyze and tograph sensor readings.The indoor experimentation was done in the corridor of thedepartment of electrical engineering building of TshwaneUniversity of Technology. The outdoor experiments were donein an open rugby field of the University where there are noforms of obstructions between the sensor nodes. Only twosensor nodes were used for performing the experimentation,one node act as the transmitter and the other as the receiver.The effects of the antenna orientation of the sensor nodeswere taking into consideration so therefore all the sensor nodesare resting on their bases with antenna pointing verticallyupwards. RSSI measurements are prone to noise andinterference which leads to error in measurements. Hence, inorder to avoid these errors and to provide ideal environment forthe measurements, measurements were taken in such a waythat no other device working in the range of 2.4GHz werepresent in the vicinity of experimental location and it wasensured that the deployed sensor nodes are in line of sight withthe base node and no obstacles were present between them.The RSSI readings were taken at a fixed distances atdifferent time intervals in order to take into consideration theeffect of shadowing due to the fact that RF channels vary withtime.V. EXPERIMENTATION RESULTTo compare the measured results with a simulated model,simulations were performed in MATLAB v.7.10 on a PentiumM processor of 2GHz with 2GB RAM. On the simulationplatform, sensor nodes were randomly deployed according to aspecified connectivity level. The network topology ofrandomly deployed sensor nodes with connectivity level of 5meters is shown in figure 2. The full distance matrix wascalculated between the pair-wise sensor nodes that are in thesame radio range and the log-normal shadowing model wasused to calculate the power received between the sensor nodesthat are in the same range. The relationship between thereceived power and the distance is shown in figure 3.1536Fig. 2. Network Topology Graph of a Randomly Deployed Sensor NodesFig. 3. Plot of RSSI-Distance VariationAccording to the simulation result shown in the figure (2),the sensor nodes are randomly deployed in an area of 20 metersby 20 meters and the nodes connect to its neighbors in therange of 5 meters. Using LNSM (log Normal shadowingmodel), we could observed in figure (3) the relationshipbetween the RSSI of the sensor nodes and their respectivedistances. The simulation results in figure (3) shows that as thedistance between the two sensor nodes in the same rangeincreases, the received signal strength reduces.A. Experimentation Result in an Indoor EnvironmentThe result of the indoor experimentation is shown in thefigure 4. The RSSI readings were taken from I meter to 10meters in distance. Figure 4 shows the plot of the RSSI valuesin dBm recorded at different time interval of each distance.Fig. 4. Plot of RSSI Values at each Distance in an Indoor EnvironmentFig. 5. Plot of mean of the measured RSSI values in an Indoor EnvironmentTABLE II. This table shows the mean, variance and the standard deviationof the RSSI values in an indoor Environment.Distance(m) Mean VarianceStandardDeviation1 -13.0000 2.7368 1.65432 -17.2500 1.6711 1.29273 -19.6500 3.2921 1.81444 -21.8500 4.3447 2.08445 -24.0000 0.6316 0.79476 -27.2000 5.1158 2.26187 -30.5500 1.5237 1.23448 -32.5000 0.5789 0.76099 -33.7500 0.7237 0.850710 -35.4500 0.4711 0.68630 5 10 15 2024681012141618Network topologyX (meters)Y (meters )12345678 910111213141516171819202122232425262728293031323334353638 373940414243444546474849502 4 6 8 10 12 14 16 18 20-60-55-50-45-40-35-30-25Distance(m)RSSI(dBm)Simulation Result of RSSI-Distance0 5 10 15 20 25 30 35 40-40-35-30-25-20-15-10Time(minutes)RSSI values(dBm)Plot of RSSI Values at Different Distances (Indoor)1m2m3m4m5m6m7m8m9m10m0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-40-35-30-25-20-15-10log of Distance (D in meters)mean Values of RSSI(dBm)Plot of RSSI vs Distance (Indoor)1537The corresponding plots of the mean of the measured RSSIvalues in an indoor environment are shown in the figure 5.B. Experimentation Result in an Outdoor EnvironmentThe outdoor experimentation was done at an open rugbyfield of the institution and the corresponding result is shown infigure 6.Fig. 6. Plot of RSSI Values at each Distance in an Outdoor EnvironmentTABLE III. This table shows the mean, variance and the standarddeviation of the RSSI values in an Outdoor Environment.Distance(m) Mean VarianceStandardDeviation1 -10.5000 0.2778 0.52702 -12.6000 0.2667 0.51643 -13.4000 0.4889 0.69924 -16.7000 1.3444 1.15955 -20.5000 0.5000 0.70716 -22.9000 0.3222 0.56767 -24.8000 0.1778 0.42168 -26.7000 0.2333 0.48309 -28.1000 0.1000 0.316210 -30.5000 0.9444 0.9718Figure 7 shows the corresponding plots of the mean of themeasured RSSI values in an outdoor environment.C. Calibration and Error AnalysisSince the measured RSSI values vary randomly with time,the mean values of the RSSI were considered for calculatingthe distances of the deployed sensor nodes. In order tocorroborate and support our initial theoretical analysis fromwhich we derived distance as an exponential function of RSSI,we conducted an experiment for generating the calibrationequation relating RSSI to distance.Fig. 7. Plot of mean of the measured RSSI values in an Outdoor EnvironmentThe plot in figure 8 and figure 9 shows the data used forcalibration. It was obtained with RSSI(dBm) along the Y-axisand log10(Distance) along the X-axis. The calibration equationand curve were fitted with MATLAB’s basic fitting tool whichshows a linear curve between RSSI and log10(Distance.)The calibration equation and curve shown in figure 8 is forthe indoor RSSI measurement. It gives the relationshipbetween RSSI(dBm) and the actual distance and is derivedbelow.y = -23x – 10 (11)Where y represents RSSI(dBm) and x represent log10(D),where D is the distance.D = 10((y +10) / -23) (12)The corresponding calibration equation for an outdoorenvironment which is show in figure 9 is also derived as;y = -21x – 6.9 (13)D = 10((y +6.9) / -21) (14)We used this calibration equation in equation 11 andequation 13 for computing distances for different known pointson our test-bed (Indoor and outdoor).From equation (12), the path-loss exponent factor has beenfound to be 2.3 for indoor environment and from equation 14the path-loss exponent for an outdoor environment is 2.1.We observed that the path-loss exponent of the indoorenvironment is larger than the outdoor due to the fact that thereare different forms of obstacles involving in the indoorenvironment compared to the outdoor environment.Table 4 shows the estimated distance for the indoor andoutdoor which were obtained using the equation 12 andequation 14 and their respective distance estimated error. Thedistance error is the difference between the actual distance andthe estimated distance.0 2 4 6 8 10 12 14 16 18-35-30-25-20-15-10Time(minutes)RSSI values(dBm)Plot of RSSI Values at Different Distances (Outdoor)1m2m3m4m5m6m7m8m9m10m0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-35-30-25-20-15-10log of Distance (D in meters)mean Values of RSSI(dBm)Plot of RSSI vs Distance (Outdoor)1538Fig. 8. Calibration curve of the linear regression based on the mean of themeasured RSSI values in an Indoor EnvironmentFig. 9. Calibration curve of the linear regression based on the mean of themeasured RSSI values in an Outdoor EnvironmentThe mean distance error for the indoor environment is0.9753meters and for the outdoor environment is 0.8831metersTABLE IV. This table shows the estimated Distance in an indoor andoutdoor environment with their respective estimated errors in meters.ActualDistIndoorEst. Dist.OutdoorEst. Dist.IndoorErrorOutdoor Error1 1.3503 1.4840 0.3503 0.48402 2.0664 1.8682 0.0664 0.13183 2.6276 2.0395 0.3724 0.96054 3.2750 2.9286 0.7250 1.07145 4.0616 4.4424 0.9384 0.55766 5.5953 5.7797 0.4047 0.22037 7.8249 7.1184 0.8249 0.11848 9.5118 8.7671 1.5118 0.76719 10.7798 10.2217 1.7798 1.221710 12.7797 13.2987 2.7797 3.2987VI. CONCLUSIONIn this paper, the performance evaluation of RSSI basedmodel in an indoor and outdoor environment was presented.The study was conducted by using empirical measurements toestimate the distance. Based on the result that was obtainedwhich resulted in distance estimation error of 0.9753meters inan indoor environment and 0.8831 for outdoor environment,we concluded that the results is very encouraging and could beused for approximation algorithms under ideal condition toperform localization in WSN. Currently, research work is inprogress in developing an accurate algorithm to performlocalization of wireless sensor nodes in a given environment.REFERENCES1 I.F. Akyildiz, W. Su, Y. Sankarasubramamiam, E. Cayirci,”Wireless Sensor Networks: A Survey,” Elsivier Science B.VJournal of Computer Networks, 393-422 (2002).2 Jamal, N.A., Ahmed E.K, “Routing Techniques in WirelessSensor Networks: A Survey,” IEE Wireless CommunicationMagazine (2005).3 Adel, G.A.E., Hussein, A.E., Salwa, E.R., Magdy, M.I, “AnEnergy Aware WSN Geographic Routing Protocol,” UniversalJournal of Computer Science and Engineering Technology, 105-111 (2010).4 Amitangshu, P, “Localization Algorithm in Wireless SensorNetworks: Current Approaches and Future Challenges,”Network Protocol and Algorithm, 1943-3581 (2010).5 Dong, B., Mahdy, A.M, “Underwater Wireless SensorNetworks: Efficient Schemes using SemiDefiniteProgramming,” International Journal on Advances in Networksand Services, 186-195 (2010)6 He, T., Huang, C., Blum, B.M., Stankovic, J.A., Abidelzaher, T,”Range-Free Localization Schemes For Large Scale SensorNetworks,” Mobicom’03, 81-95 (2003).7 Mohit, S., Puneet, G., Bijendra, N.J, “Experimental Analysis ofRSSI-Based Location Estimation in Wireless Sensor Networks”.8 Giacomin, J.C., Correta, L.N.A., Heimfarth, T., Pereira, G.M.,Silva, V.F., De Santa, J.L.P, “Radio Channel Model of WirelessSensor Networks Operating in 2.4GHZ ISM Band”.9 Willis, S., Kikkert, C.J, “Radio Propagation Model for LongRange Wireless Sensor Networks,” IEEE, (2007)10 Sommer, C., Dressler, F, “Using the Right Two-Ray Model? AMeasurement Based Evaluation of PHY Models in VANETS”.11 Benkic, K., Malajner, M., Planinsic, P., Cucej, Z, “Using RSSIValue for Distance Estimation in Wireless Sensor NetworksBased on Zigbee”.12 Xu, J., Liu, W., Lang, F., Zhang, Y., Wang, C, “DistanceMeasurement Model Based on RSSI in WSN” Wireless SensorNetwork, 606-611 (2010).13 Tadakamadla, S, “Indoor Local Positioning System for Zigbee,Based on RSSI,” An Msc Thesis Report, Mid SwedenUniversity, (2006).14 Malajner, M., Benkic, K., Planinsic, P., Cucej, Z, “TheAccuracy of Propagation Models for Distance MeasurementBetween WSN Nodes,” IEEE, (2009)15 Kumar, P., Reddy, L., Varma, S, “Distance Measurement andError Estimation Scheme for RSSI Based Localization inWireless Sensor Networks,” IEEE, (2009)16 Awad, A., Frunzke, T., Dressler, F, “Adaptive DistanceEstimation and Localization in WSN using RSSI Measures”.17 Crossbow.: MPR-MIB Users’ Manual: Revision A, (2007).0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-40-35-30-25-20-15-10log of Distance (D in meters)mean Values of RSSI(dBm)Plot of RSSI vs Distance (Indoor)y = – 23*x – 10RSSI ValuesLinear Best Fit0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-35-30-25-20-15-10-5log of Distance (D in meters)mean Values of RSSI(dBm)Plot of RSSI vs Distance (Outdoor)y = – 21*x – 6.9RSSI ValuesLinear Best Fit1539Powered by TCPDF (