PAPER An Efficient TOA-Based Localization Scheme Based on BS Selection in Wireless Sensor Networks

Transcription

1 2560 IEICE TRANS. COMMUN., VOL.E97 B, NO.11 NOVEMBER 2014 PAPER An Efficient TOA-Based Localization Scheme Based on BS Selection in Wireless Sensor Networks Seungryeol GO a), Nonmember and Jong-Wha CHONG b), Member SUMMARY In this paper, we present an efficient time-of-arrival (TOA)-based localization method for wireless sensor networks. The goal of a localization system is to accurately estimate the geographic location of a wireless device. In real wireless sensor networks, accurately estimating mobile device location is difficult because of the presence of various errors. Therefore, localization methods have been studied in recent years. In indoor environments, the accuracy of wireless localization systems is affected by non-line-of-sight (NLOS) errors. The presence of NLOS errors degrades the performance of wireless localization systems. In order to effectivelyestimate the location of the mobile device, NLOS errors should be recognized and mitigated in indoor environments. In the TOA-based ranging method, the distance between the two wireless devices can be computed by multiplying a signal s propagation delay time by the speed of light. TOA-based localization measures the distance between the mobile station (MS) and three or more base stations (BSs). However, each of the NLOS errors of the measured distance between the i-th BS and the MS is different due to dissimilar obstacles in the direct signal path between the two nodes. In order to accurately estimate the location in a TOA-based localization system, an optimized localization algorithm that selects three measured distances with fewer NLOS errors is necessary. We present an efficient TOA-based localization scheme that combines three selected BSs in wireless sensor networks. This localization scheme yields improved localization performance in wireless sensor networks. In this paper, performance tests are performed, and the simulation results are verified through comparisons between various localization methods and the proposed method. As a result, proposed localization scheme using BS selection achieves remarkably better localization performance than the conventional methods. This is verified by experiments in real environments, and demonstrates a performance analysis in NLOS environments. By using BS selection, we will show an efficient and effective TOA-based localization scheme in wireless sensor networks. key words: wireless localization, time-of-arrival (TOA), location estimation, non-line-of-sight (NLOS), BS selection 1. Introduction In the era of information and computing, the location awareness techniques of wireless devices have received a great deal of interest in many wireless systems such as cellular networks, wireless local area systems, and wireless sensor networks. For example, location awareness-based services which are utilized for the emergency 911 (E-911) services, smart phones, hospitals, shopping malls and welfare facilities, have demanded accurate location information in order to improve ubiquitous smart device environments. In recent years, wireless sensor networks have been widely used in Manuscript received January 21, Manuscript revised June 24, The authors are with the Department of Electronics and Computer Engineering, Hanyang University, Seoul, Rep. of Korea. a) milkyface@hanyang.ac.kr b) jchong@hanyang.ac.kr (Corresponding author) DOI: /transcom.E97.B.2560 real- time location systems (RTLSs), such as to locate the people inside a building, goods inside a warehouse, and cars inside parking lots, where global positioning system (GPS) signals are not available in indoor environments. Currently, for precise location estimation in wireless sensor networks, the most utilized localization techniques are the time-ofarrival (TOA), time-difference-of-arrival (TDOA), angle-of arrival (AOA) and received-signal-strength-indicator (RSSI) methods in [1]. Three or more base stations (BSs) measure the TOA-based ranging result of the signal propagation from the target device, called the mobile station (MS), in TOA-based localization. The distance data of TOA-based ranging can be computed by using a propagation delay time between the MS and each BS. In the TOA-based location estimation method, the same point of the intersection of TOA circles with the measured distance data gives the location of the MS. However, due to errors such as clock offset [2], unequal reply time of ranging [3], and non-line-of-sight (NLOS) conditions [4] in TOA-based localization, the TOA circles do not accurately estimate the same point of intersection as shown in [5]. Among distance errors, NLOS errors greatly influence the performance of localization systems. Therefore, in NLOS environments, various aspects for localization method with applications to communication systems have been investigated in the literature in [6]. In indoor environments, due to the presence of various obstacles such as walls, people, pillars and steel frames, the received signal between devices is influenced by a combination of lineof-sight (LOS), multipath signals, and weakened signals of navigating obstacles. In this paper, in NLOS environments, we present an efficient TOA-based localization scheme using a BS selection algorithm in wireless sensor networks. The goal of this paper is to improve the performance of location estimation using more BSs in wireless sensor networks. Through the BS selection algorithm, the accuracy of localization is enhanced in localization systems of NLOS environments using more BSs. Simulation results are presented to demonstrate the improved performance due to the BS selection scheme in wireless sensor networks. This paper is organized as follows: In Sect. 2, we will describe the problem of NLOS errors for location estimation in wireless sensor networks. Section 3 will discuss the wireless location estimation techniques, such as the TOAbased ranging and localization methods, in detail. Then we will explain the mathematical modeling of signal measurements, and estimation of MS location using the trilateration Copyright c 2014 The Institute of Electronics, Information and Communication Engineers

2 GO and CHONG: AN EFFICIENT TOA-BASED LOCALIZATION SCHEME 2561 method will be briefly reviewed. Section 4 describes the related works for the localization method. The proposed TOA-based localization scheme using BS selection in wireless sensor networks will be presented concretely in Sect. 5. The performance analysis of the proposed scheme is conducted in Sect. 6 via simulations, and Sect. 7 provides the conclusions. 2. NLOS-Error Issue in Wireless Sensor Networks In wireless sensor networks, the problem of NLOS errors exists. For example, Fig. 1 illustrates signal propagation among more BSs and the MS in indoor environments. As illustrated by the signal propagation for TOA-based ranging, an NLOS error results from an obstacle (e.g. wall, pillar, people, steel frame and humidity) in the direct signal path, the non-signal of the barrier and the reflection signal of multipath environments, as shown in Fig. 1. It is difficult to accurately detect the time of arrival of a signal that is affected by NLOS errors. When NLOS errors occur, due to the high speed of the wireless signal in the TOA-based method, a 1-meter error in the distance can be caused by a 3-ns error in the time of arrival of the signal. As such, the NLOS error bias is the extra data of the measured distance, which always has a non-negative value. The NLOS error can be delineated as a deterministic error, a uniform distributed error, a Gaussian distributed error, or an exponentially distributed error as in [7], [8]. The NLOS error bias b i is a non-negative bias introduced due to the blockage of direct path of signal propagation given by and the concept of the TOA-based wireless location estimation. In Sect. 3.1, the ranging-method using the time of arrival of signal propagation is described specifically. Section 3.2 facilitates the design of TOA-based measurement, and a system model of a TOA-based scenario is presented for NLOS environments. 3.1 TOA-Based Ranging Protocol The signal propagation time between two devices can be calculated using a ranging protocol such as one-way ranging (OWR) [9], two-way ranging (TWR) [10], or symmetric double side two-way ranging (SDS-TWR) [3]. The ranging protocol in Fig. 2 briefly diagrams the system in a scenario between two devices. In the ranging protocol, we can estimate a time of arrival of the signal for the TOA-based localization system. As describe in Fig. 2(a), in OWR, the time of arrival of the signal t i at the i-th device is calculated by T round. In the TWR protocol of Fig. 2(b), the time of arrival of the signal t i at the i-th device is described as t i,twr = 1 2 (T round T reply ), (2) where T round denotes the round trip time at the i-th device, T reply is the reply time of the i-th device, and t p is the propagation delay time. By using two representative ranging protocols, we obtain the time of the arrival of signal propagation b i = { 0, if i-th BS is LOS error i, if i-th BS is NLOS. (1) The unit of b i has a little meter value, error i, if signal propagation between the i-thbsandthemswerenlosenvironments. 3. System Model in Wireless Location Estimation In this section, we briefly introduce the mathematical modeling of the TOA-based localization in NLOS environments Fig. 1 Signal propagation in a localization system. Fig. 2 Ranging protocol: (a) One-way ranging (b) Two-way ranging.

3 2562 IEICE TRANS. COMMUN., VOL.E97 B, NO.11 NOVEMBER 2014 accurate location of the MS using a distance equation expressed as a non-linear expression by employing trilateration for TOA-based localization. In Sect. 4.2, we briefly explain non-linear least squares (NLS) which is a well-known localization technique for minimizing NLOS errors. We then describe the weighted least squares (WLS) method, which is an advanced localization method of NLS. Section 4.3 depicts the centroid method using the intersection coordinates of TOA-circles. 4.1 Matrix Interpretation for a Non-linear Model in TOA Fig. 3 TOA-based localization in NLOS signal propagation. in order to use TOA-based localization. 3.2 TOA-Based Localization To evaluate the TOA-based localization scenario [5] in Fig. 3, the location of the MS and the location of the fixed i-th BS are expressed as x = [x, y] T and x i = [x i, y i ] T,respectively. The TOA-based measured distance ˆd i from the i-th BS can be calculated by multiplying the TOA signal by the speed of light, and is modeled as ˆd i = c t i = d i + b i + n i, i = 1, 2,...,N, (3) where c is the speed of light, t i is the time of arrival of the signal at the i-th BS using TOA-based ranging as described in Sect. 3.1, d i is the true distance using the TOA method between the MS and the i-th BS, b i is the NLOS bias error, which has a non-negative value as described in Sect. 2, n i is an estimation error of signal propagation using TOA measurements, and N is the number of units using the BS in the localization system. The distribution of n i can be divided into n NLOSi and n LOSi whether an NLOS error exists or not. The statistical distribution of n LOSi can be N (0, σ 2 LOSi ), and the statistical distribution of n NLOSi can be N (0, σ 2 NLOSi ). The noiseless true distance d i is calculated by equation trilateration as follow: d i = x x i = (x x i ) 2 + (y y i ) 2, (4) where represents the Euclidean norm operation over a vector of distance between the target and the i-th BS. 4. Wireless Location Estimation Techniques In this section, we will review a variety of location estimation methods for wireless localization. In Sect. 4.1, we introduce mathematical modeling, which can determine the To solve the wireless localization, we consider the nonlinear expressions of the TOA circle, which is written using the distance Eq. (4) and is modeled as (x x i ) 2 + (y y i ) 2 = di 2, i = 1,...,N, (5) between the fixed i-th BS and the MS in noiseless environments called LOS environments. However, in real environments, considering the presence of noise and NLOS errors, the TOA-circle equation is modeled as (x x i ) 2 + (y y i ) 2 = ˆd i 2, i = 1,...,N, (6) using Eq. (3) in Sect We can calculate the location of the MS using the intersection of Eq. (5). For this reason, to determine the MS s location, we have to convert a nonlinear expression into a linear expression. After some matrix representation and linearization, we can write the matrix as in [11], [12] where Aθ = 1 p, (7) 2 A = θ = x 1 y x 2 y x N y N 0.5 x y, p x 2 + y 2 1 =, x y2 1 d2 1 x y2 2 d2 2. x 2 N + y2 N d2 N. (8) Finally, calculating θ, which has the MS s location value x, we can compute the matrix representation (7) as follows: ˆθ = 1 2 (AT A) 1 A T p, (9) where T is the transpose function of the matrix, and 1 indicates the inverse function of the matrix. The matrix solution process, such as in Eqs. (7), (8), and (9), is called the linear least squares method, as in [13]. However, in NLOS environments, the intersection of the TOA circles using Eq. (6) does not arrive at the same intersection. In Fig. 4, we briefly depict the transformation process of TOA-based localization from LOS environments to

4 GO and CHONG: AN EFFICIENT TOA-BASED LOCALIZATION SCHEME 2563 Fig. 4 Transformation process of TOA-based localization: (a) LOS environments, (b) NLOS environments. NLOS environments. Because of the presence of NLOS errors, the same intersection for the MS s estimated location will be relocated from the true MS location. As a result, in Fig. 4(b), an overlapping area has occurred in response to the location estimation error. It is difficult to find the accurately estimated location of the MS insides the overlapping area. 4.2 Non-linear Least Squares Technique The NLS technique [14] is a localization method for the estimation of the MS s location in wireless sensor networks, and can be calculated from ˆx NLS = arg min x = arg min x N ( ˆd i x x i ) 2 i=1 N (e i ) 2, (10) i=1 where x is the coordinate value of the MS s location, x i is the location of the i-th BS, ˆd i is the measured distance, and e i is the error value between the MS and the i-th BS, as described by Sect The goal of NLS is to find the coordinate value of the MS s location which estimates the minimized value of the distance error among the MS and all BSs. To supplement the NLS technique, the WLS estimation has been studied in [15], and some weights β i can be utilized featuring the reliability of each measured distance between the MS and the i-th BS. The WLS estimation of x is represented as ˆx WLS = arg min x = arg min x N ( β i ˆd i x x i ) 2 i=1 N β i (e i ) 2, (11) i=1 where β i is the variance of the TOA error in [16], and is expressed as β i = 1,whereσ 2 σ 2 i is denoted by i σ 2 i = E { ( ˆd i d i ) 2}, (12) where E{ }is the expected value operator. 4.3 Centroid Method for Location Estimation of the MS The centroid method for the wireless location estimation is a Fig. 5 Diagram of the 2-D TOA-based localization with intersections. localization method using the center of gravity of the overlapping area of the TOA circles as in [17], [18]. As illustrated in Fig. 5, the TOA-based localization is drawn in 2- dimensions. In Fig. 5, the measured distances among the MS and the three BSs are simply denoted, and then the intersections of the TOA circles are indicated. If the TOA- measurements are LOS or environments without noise, three TOA-circles will intersect to a single point as shown in Fig. 4(a). However, in NLOS environments, the overlapping area is confined within the intersection region of all of the TOA circles (i.e. the region within points A, B, and C, as shown in Fig. 5). It is difficult to accurately estimate the MS s location in the overlapping area, so the centroid method has proposed wireless localization using the center of gravity of the intersection of TOA-circles in [17]. As shown in [19], the MS s location is estimated so that the geometric dilution of precision (GDOP) value is minimized at the center of gravity of the overlapping area. 5. BS Selection Scheme in Wireless Sensor Networks In this section, we will present the localization method of wireless sensor networks using the BS selection scheme. TOA-based localization is fundamentally estimated by using three BSs. However, in the environment for localization, three or more BSs are installed to improve the performance of the wireless localization. Accordingly, the location estimation of the MS is conducted within the range of the signal propagation of the BSs in wireless sensor networks. We propose an efficient TOA-based localization scheme that adds a process of BS selection. We will introduce the process for calculating the overlapping area of the TOA circles in Sect Section 5.2 will introduce the performance comparison of the localization, and we will briefly present the problem of localization without the BS selection scheme. Lastly, in Sect. 5.3, we will describe an efficient localization method, which is the optimized BS selection scheme using the overlapping area obtained as in Sect. 5.1.

5 2564 IEICE TRANS. COMMUN., VOL.E97 B, NO.11 NOVEMBER 2014 Fig. 6 Illustration of assumed relationship between the true overlapping area and the triangular region. 5.1 Computation of Overlapping Area in TOA-Circles As illustrated in Fig. 5, three TOA-measurements for the distances from the BSs to the MS are shown in 2- dimensional space. Two or more circles that have the sum of the radius of two circles longer than the distance between the fixed locations of two BSs will meet at two points called the intersections. The intersection points can be calculated using the information of the measured distances and the location of the i-th BS. The goal is to obtain the solution process of the assumed area of the overlapping area, as shown in Fig. 6, which is calculated using the intersection point. In Sect. 3.2, we expressed the MS s location as x = [x, y] T,the i-th BS s location as x i = [x i, y i ] T and the measured distance as ˆd i. As shown in [20], for example, using the intersection between the circles of BS1 and BS2, the intersection point x = [ x, ỹ] T can be calculated as x = x 1 + ˆd 1 cos θ ỹ = y 1 + ˆd 1 sin θ where θ is or x = x 2 + ˆd 2 cos θ ỹ = y 2 + ˆd 2 sin θ, (13) θ = arctan Y ˆd X 2 ± arccos 1 ˆd D2 2 ˆd 1 D = arctan Y ˆd X 2 ± arccos 2 ˆd D2 (14) 2 ˆd 2 D and where D = X 2 + Y 2, X = x 2 x 1, and Y = y 2 y 1. (15) Also similarly, other intersection points such as between BS2 and BS3, and between BS3 and BS1 can be easily computed as in [20]. By using the calculation of the intersection point, we can estimate the assumed value of the overlapping area, which consists of the intersection of the TOA circles. In Fig. 6, it is assumed that the overlapping area is a triangular region using the intersection points, because it is difficult to find an overlapping area of the TOA circles in Fig. 6(a). In Fig. 5, the area of the triangle of the assumed overlapping area can be calculated with the cross product in vector space (or matrix) in Appendix A as follows: S = 1 (B A) (C A), (16) 2 Fig. 7 Performance comparison of localization without BS selection scheme at a location (7.9, 6) in a gymnasium. where A = [ x 12,ỹ 12 ] T, B = [ x 23,ỹ 23 ] T,andC = [ x 31,ỹ 31 ] T are intersection points of the TOA circles among the fixed i-th BSs for i = 1, 2 and Localization Problem without a BS Selection Scheme TOA based localization is fundamentally used by three BSs in the location estimation system. In wireless sensor networks, three or more BSs are installed to improve the localization performance. However, to accurately estimate the MS s location, it is necessary to provide the optimal localization method using the BS selection scheme as in [21], [22]. For example, in our gymnasium experiment in Fig. 14, we were able to identify the localization performance without a BS selection scheme, and the results of the performance are shown in Fig. 7. Figure 7 compares four different experiment results where the number of BSs changes from three fixed BSs to six fixed BSs. In Fig. 7, the estimation error of the MS means the difference between the coordinate of true MS s location as x and the coordinate of estimated MS s location as ˆx using localization method in Sect. 4, is expressed by Estimation Error = x ˆx. (17) For example, as shown in Fig. 7, the estimation error of the MS is assumed to indicate less result of the location estimation than 4 meter at three BSs. In localization experiment using four BSs, the estimation error of the MS is estimated as result of the location estimation of more 8 meters. However, in other localization experiments using five and six BSs, the estimation error of the MS is more degraded than localization using four BSs. As a result, we can see that a scenario with more BSs in wireless sensor networks does not surely provide better performance than that provided by less BSs. Generally speaking, accuracy of location estimation does not certainly provide better performance in localization with more BSs, because it is difficult to judge the NLOS or LOS environments between the i-th BS and the MS. Therefore, it is necessary to utilize localization scheme with three se-

6 GO and CHONG: AN EFFICIENT TOA-BASED LOCALIZATION SCHEME 2565 lected BSs, which eliminates the effect to NLOS errors in wireless sensor networks. Then, we can obtain minimum estimation error of MS with the localization system using three BSs with less NLOS errors. Therefore, to enhance the location estimation, we will propose an efficientlocalization system that adds a BS selection scheme in wireless sensor networks. 5.3 BS Selection Scheme Using an Overlapping Area In Sect. 5.1, the key idea is to find the area of the triangle that has the relationship of the MS s location with NLOS errors. That is, if the area of triangle S, which was composed of the three intersections of the TOA circles, were estimated to have a zero value, the NLOS error of the TOA-based localization would become minimized, which approaches the MS s true location. By using the area of the triangle that is implied by the intersection points, we present an efficient localization method in a localization system that has more than three BSs. To estimate an efficient localization using BS selection in wireless sensor networks, the measured distance between the i-th BS and the MS is calculated by TOA measurements. Then, the area of the triangle is calculated by a combination of three BSs. In the next step, the area of the triangle made by the TOA measurements is computed using the combination that incorporates the i-th BS, and is denoted by S (k), where k is the number of combinations of three BSs from the total BSs. Using the triangle of the area, we can estimate the MS s location by using the minimizing step, which is expressed as arg min S (k). (18) k It would be remarkable that (18) can obtain minimum estimation error of MS using combination scheme of three BSs with less NLOS errors. The process of finding k that minimizes the area of the triangle denotes the BS selection scheme, and Fig. 8 briefly shows the flow chart of the localization process using the BS selection scheme. In Table 1, we introduce the algorithm using the proposed BS selection scheme, and the BS selection scheme can apply a general localization method as in Fig Performance Analysis In this section, to verify the proposed localization method using the BS selection scheme, we use the chirp spread spectrum (CSS)-based wireless sensor network system developed in [23], [24]. The experiments were implemented in real indoor environments. We performed the simulation tests using six BSs in NLOS environments with interference and multipath. Then, we conducted localization experiments from several MS locations in experimental conditions. The basic idea of the proposed method for NLOS environments incorporates a three BS selection algorithm among the many installed BSs. These measured results Fig. 8 Flow chart of the localization process using BS selection. Table 1 An efficient localization scheme algorithm. Fig. 9 Localization process: (a) General Localization (b) Proposed Localization. ranging between the i-thbsandthemsaresenttoaserver computer, and then the server computer estimates the MS s location using the localization methods. The simulation results of the proposed method with the BS selection scheme show more accurate estimation results than the conventional methods such as the NLS, WLS, and Centroid methods. 6.1 Gymnasium with NLOS Environments The evaluation for the localization is conducted in a meters gymnasium with the presence of NLOS errors as shown in Figs. 11 and 12. For the experimental test, we installed six BSs at locations with the coordinate values [1.5,

7 2566 IEICE TRANS. COMMUN., VOL.E97 B, NO.11 NOVEMBER 2014 Fig. 10 Server computer and CSS-based device for performance evaluation. Fig. 11 Experimental environments (Gymnasium). Fig. 13 Comparison of distributions of shot pattern at a location (27.8, 15) in a gymnasium. Fig. 12 Locations of sensor devices (Gymnasium). Table 2 Estimation error of the NLS and proposed method. 1.1] T, [1.5, 20.5] T, [34.7, 20.5] T, [34.7, 1.1] T, [19.1, 1.1] T, and [19.1, 20.5] T, as shown in Fig. 12. The MS s location for the localization simulation included locations for several different experiments with coordinate values of [7.9, 6] T, [19.1, 10.5] T, and [27.8, 15] T, as shown in Fig. 13. In order to compare the estimation error of the accurate MS location, we compute the MS s estimated location by using the localization methods. Tables 2, 3, and 4 show the simulation results of the performance analysis of the localization system by comparing various conventional methods with the proposed localization scheme. To verify the effectiveness of localization system, the number of 100 shot patterns for the location estimation is drawn in overlap as shown in Fig. 13. Comparing the shot pattern for NLS estimation [14], WLS estimation [15], and centroid method [17] with the proposed BS selection scheme, we can see that the shot pattern of the MS in localization system with BS selection is more distributed accurately than the others. Table 3 Table 4 Estimation error of the WLS and proposed method. Estimation error of the Centroid and proposed method. For example, Fig. 13 shows shot pattern of experiment results of three different localization methods at a location (27.8, 15) in a gymnasium. The shot patterns of estimating MS location are distributed near the MS location. As

8 GO and CHONG: AN EFFICIENT TOA-BASED LOCALIZATION SCHEME 2567 Table 5 Estimation error of the NLS and proposed method. Table 6 Estimation error of the WLS and proposed method. Fig. 14 Experimental environments (Parking Lot). Table 7 Estimation error of the Centroid and proposed method. Fig. 15 Locations of sensor devices (Parking Lot). shown in Figs. 13(a) and (b), comparing the general localization without BS selection with the proposed localization with BS Selection, the distributions of shot pattern of estimating MS location in proposed localization is more accurate than the general localization scheme without BS selection. Tables 2, 3, and 4 show comparison values at three experiments locations, which is one data in shot patterns of localization experiment results. 6.2 Parking Lot with NLOS Environments In a meter parking lot with the presence of NLOS errors, as shown in Figs. 14 and 15, the verification for the localization was performed. Since a steel frame, car, wall, drain, and ventilator exist, the parking lot is a more realistic environment than the gymnasium. In addition, the base stations of a telecommunications company were installed in the underground parking lot in which the experiment was conducted. Because of that, our wireless devices, which are based on CSS [23], [24], can experience interference from 2.4 GHz devices [25]. Furthermore, LOS environments are not assured due to obstacles such as walls and steel frames. To evaluate the localization comparison, we installed the BSs at locations with the coordinate values [10.52, 11.5] T, [10.52, 34.6] T, [90.04, 34.6] T, [90.04, 11.5] T, [56.98, 11.5] T, and [56.98, 34.6] T,andtheMSis installed in locations with the coordinate values of [30.4, 19.7] T, [56, 19.7] T, and [89.8, 19.7] T asshowninfig.14. The simulation results are shown in Tables 5, 6, and 7. The performance analysis proves that the proposed method has better localization performance than the other localization Fig. 16 Comparison of distributions of shot pattern at a location (56, 19.7) in the underground parking lot. methods in a parking lot. In Fig. 16, Shot pattern of the results of localization experiment is showed at a location (56, 19.7) in the under-

9 2568 IEICE TRANS. COMMUN., VOL.E97 B, NO.11 NOVEMBER 2014 ground parking lot. To compare the accuracy of estimating MS location, the number of 100 shot patterns is drawn in localization system as shown in Fig. 16. To evaluate an accuracy of the localization system, Figs. 16(a) and (b) are represented as localization without BS selection and with BS selection, respectively. In Fig. 16, the results of shot patterns of the proposed method are more accurately distributed near the MS location than general localization methods. Tables 5, 6, and 7 fill in tables, which is estimation error between general method and proposed method in one data of shot patterns of localization experiment results. 7. Conclusion The presence of NLOS errors is an interesting issue in wireless localization because accurately estimating MS location is difficult in NLOS environments. As a result, a variety of localization methods have been investigated. In this paper, an efficient localization method using BS selection based on TOA measurements is proposed for wireless sensor networks. In wireless sensor networks, three and more BSs are installed for the localization system. As we select three optimal BSs, optimal in terms of NLOS errors, the performance of the location estimation is improved. This BS selection scheme has been implemented to optimize localization in NLOS environments. The proposed technique seeks to accurately estimate the MS s location by using the BS selection scheme before estimating the MS s location using a method that has been proposed in wireless location estimation. To verify the proposed localization scheme, we performed localization experiments using a CSS-based wireless sensor network. The TOA-based localization system is demonstrated by a performance analysis in NLOS environments. The performance shows that the localization estimation error could be minimized in wireless sensor networks. Acknowledgments This work was supported by the National Research Foundation of Korea (NRF) Grant funded by the Korean Government(MOE) (NRF-2013R1A1A ). References [1] Kupper and Axel, Location-based services: Fundamentals and operation, Wiley, New York, ISBN: [2] Y. Jiang and V.C.M. Leung, An asymmetric double sided twoway ranging for crystal offset, International Symposium on Signals, Systems and Electronics, ISSSE 07., pp , July [3] L.J. Xing, L. Zhiwei, and F.C.P. Shin, Symmetric double side two way ranging with unequal reply time, IEEE 66th Vehicular Technology Conference 2007., VTC-2007, pp , Sept [4] Y.T. Chan, W.Y. Tsui, H.C. So, and P.C. Ching, Time-of-arrival based localization under NLOS conditions, IEEE Trans. Veh. Technol., vol.55, no.1, pp.17 24, Jan [5] I. Guvenc and C.C. Chong, A survey on TOA based wireless localization and NLOS mitigation techniques, IEEE Communications Surveys & Tutorials, vol.11, no.3, pp , 3rd Quarter [6] S. Gezici, Z. Tian, G.B. Giannakis, H. Kobayashi, A.F. Molisch, H.V. Poor, and Z. Sahinoglu, Localization via ultra-wideband radios: A look at positioning aspects for future sensor networks, IEEE Signal Process. Mag., vol.22, no.4, pp.70 84, July [7] L. Cong and W. Zhuang, Non-line-of-sight error mitigation in TDOA mobile location, IEEE Global Telecommunications Conference GLOBECOM 01., vol.1, pp , [8] C. Ma, R. Klukas, and G. Lachapelle, A nonline-of-sight errormitigation method for TOA measurements, IEEE Trans. Veh. Technol., vol.56, no.2, pp , March [9] H.S. Ahn, H. Hur, and W.S. Choi, One-way ranging technique for CSS-based indoor localization, 6th IEEE International Conference on Industrial Informatics, INDIN 2008., pp , July [10] H. Kim, Double-sided two-way ranging algorithm to reduce ranging time, IEEE Commun. Lett., vol.13, no 7, pp , July [11] K.W. Cheung, H.C. So, W.K. Ma, and Y.T. Chan, Least squares algorithms for time-of-arrival-based mobile location, IEEE Trans. Signal Process., vol.52, no.4, pp , April [12] Y.T. Chan, H.Y.C. Hang, and P.C. Ching, Exact and approximate maximum likelihood localization algorithms, IEEE Trans. Veh. Technol., vol.55, no.1, pp.10 16, Jan [13] I. Guvenc, C.C. Chong, and F. Watanabe, Analysis of a linear leastsquares localization technique in LOS and NLOS environments, IEEE 65th Vehicular Technology Conference, VTC2007- Spring., pp , April [14] J.J. Caffery and G.L. Stuber, Overview of radiolocation in CDMA cellular systems, IEEE Commun. Mag., vol.36, no.4, pp.38 45, April [15] B.T. Sieskul, F. Zheng, and T. Kaiser, Time-of-arrival estimation in path attenuation, IEEE 10th Workshop on Signal Processing Advances in Wireless Communications, SPAWC 09., pp , June [16] B.T. Sieskul, F. Zheng, and T. Kaiser, Time-of-arrival estimation in path attenuation, IEEE 10th Workshop on Signal Processing Advances in Wireless Communications, SPAWC 09., pp , June [17] C.H. Chen, K.T. Feng, C.L. Chen, and P.H. Tseng, Wireless location estimation with the assistance of virtual base stations, IEEE Trans. Veh. Technol., vol.58, no.1, pp , Jan [18] P. Barsocchi, S. Lenzi, S. Chessa, and G. Giunta, A novel approach to indoor RSSI localization by automatic calibration of the wireless propagation model, IEEE 69th Vehicular Technology Conference, VTC Spring 2009., pp.1 5, April [19] N. Levanon, Lowest GDOP in 2-D scenarios, IEE Proc. Radar, Sonar and Navigation, vol.147, no.3, pp.149, 155, June [20] B.H. Lee, H. Hur, and H.S. Ahn, Environmental-adaptive bias calibration in wireless localization, IEEE Commun. Lett., vol.17, no.4, pp , April [21] Y. Ji, S. Biaz, S. Wu, and B. Qi, Optimal sniffers deployment on wireless indoor localization, Proc. 16th International Conference on Computer Communications and Networks, ICCCN 2007., pp , Aug [22] K. Liu, Y. Cui, J. Zhang, and F. Hu, Anchor selection scheme for multi-energy level localization method in wireless sensor networks, Proc. 4th International Conference on Ubiquitous Information Technologies & Applications, ICUT 09., pp.1 4, Dec [23] IEEE Standard a-2007, pp.1 203, Aug [24] ISO/IEC Standard , pp.1 72, March [25] H. Hur and H.S. Ahn, A circuit design for ranging measurement using chirp spread spectrum waveform, IEEE Sensors Journal, vol.10, no.11, pp , Nov [26] D. Zwillinger, Standard Mathematical Tables and For Mulae, CRC Press, New York, ISBN:

10 GO and CHONG: AN EFFICIENT TOA-BASED LOCALIZATION SCHEME 2569 Appendix A.1 Cross Product (or Vector Product) In vector space, a binary operation on two vectors a = [a 1, a 2 ] T and b = [b 1, b 2 ] T is conducted by the cross product. The cross product is presented as [26] A = a b = a b sin θ = a 2 b 2 (a b) = a 1 b 2 a 2 b 1, (A 1) where θ is the angle between vector a and vector b, andthe quantity of A represents the area of the parallelogram determinedbyvectora and vector b. Then the area of triangle S, which has vector a and vector b as boundaries, is denoted as S = 1 A. (A 2) 2 Seungryeol Go received B.S. and M.S. degrees in Electronics and Computer Engineering from Hanyang University, Seoul, Korea, in 2011, and 2013 respectively. Currently, he is working toward a Ph.D. degree with the Department of Electronics and Computer Engineering, Hanyang University. His current research interests include indoor wireless communication SoC design for ranging and positioning, location estimation algorithm, two-way ranging protocols, and timing and frequency synchronization of chirp signals. Jong-Wha Chong received B.S. and M.S. degrees in Electronics Engineering from Hanyang University, Seoul, Korea, in 1975, and 1979, respectively, and a Ph.D. degree in Electronics & Communication Engineering from Waseda University, Japan, in Since 1981, he has been a professor of the Department of Electronics Engineering, Hanyang University. From 1979 to 1980, he was a researcher at the C&C Research Center of Nippon Electronic Company. From 1983 to 1984, he was a visiting researcher at the Korean Institute of Electronics & Technology. In 1986 and 2008, he was a visiting professor at the University of California, Berkeley, USA. He was the chairman of the CAD & VLSI society of the Institute of the Electronic Engineers of Korea in 1993, and the president of the IEEK and of the KIEEE in 2007 and from 2009 to 2010, respectively. He is currently the Chairman of the Fusion SoC Forum. His current research interests are SoC design methodology, including memory centric design and physical design automation of 3D-ICs, indoor wireless communication SoC design for ranging and location, video systems and power IT systems.