An Indoor Visible Light Communication Positioning System Using a RF Carrier Allocation Technique

Transcription

1 134 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 1, JANUARY 1, 2013 An Indoor Visible Light Communication Positioning System Using a RF Carrier Allocation Technique Hyun-Seung Kim, Deok-Rae Kim, Se-Hoon Yang, Yong-Hwan Son, and Sang-Kook Han, Member, IEEE Abstract We propose a new indoor positioning system utilizing visible light communication. Intensity modulation/direct detection and carrier allocation methods are utilized in the proposed system. Simultaneous three channel transmissions were applied to calculate the receiver s position. The characteristics of the proposed positioning system were investigated through simulation based on the experimental results, and the feasibility of the proposed system was verified by experimentation. The experimental result shows that the average error of estimated positions is reduced to 2.4 cm using adjustment process by normalizing method, which is compared with cm without adjustment process. Index Terms Carrier allocation, direct detection, intensity modulation, positioning, visible light communication (VLC). I. INTRODUCTION L IGHT EMITTING DIODE (LED) lighting conserves energy, contains no mercury, and provides additional brightness with an extended lifetime compared to traditional lighting. The potential dual function of LED light (wireless communication and lighting) in the context of visible light communication (VLC) is an attractive research topic. Supporting research for such an application has been reported in the literature [1] [7]. VLC can be utilized in intelligent transportation systems, machine-to-machine communication, and indoor wireless communication. Since the global positioning system (GPS) does not work inside buildings, the following localization sensing systems have been developed for indoor positioning: RADAR, Cricket location-support, active badge, and BAT [8] [11]. These systems utilize RF, ultrasound, and infrared signals. However, the use of LEDs for lighting is expected to reduce the cost of installing a positioning system based on VLC. Additionally, the absence of electromagnetic interference in addition to features such as high capacity, high security, and compatibility with wireless communication make LEDs particularly attractive. Positioning based on VLC can be used as an indoor navigation system for location tracking, finding objects, controlling the movement of a Manuscript received October 04, 2011; revised August 03, 2012; accepted October 10, Date of publication October 22, 2012; date of current version December 28, This work was supported by the Broadcasting-Communications R&D Program of the Korea Communications Commission/Korea Communications Agency under Grant The authors are with the Department of Electrical and Electronic Engineering, Yonsei University, Seoul, , Korea ( skhan@yonsei.ac.kr). Color versions of one or more of the figures in this paper are available online at Digital Object Identifier /JLT Fig. 1. An example application of a positioning system based on VLC. Receiver 1 guides the destination. Receiver 2 indicates where the object is. robot, and so on. Fig. 1 shows example applications of a positioning system based on VLC. It was assumed that the database containing the building map and objects was built and programmed in a receiver, or the information was downloaded from the LEDs; The LEDs can be connected with networks via power line communication [12]. The receiver obtains information about its location via signals received from the LEDs. The receiver can then provide the user with a route to a specified place based on the received position information. A three-dimensional space coordinate system using image sensors for a positioning system based on VLC has been suggested [13]. The system can measure the receiver s position and direction in indoor and outdoor environments. However, it requires additional image processing techniques for the wireless communication and the data rate depends on the image processing speed. The intensity modulation and direct detection (IM/DD) method is utilized in this paper to implement a positioning system based on VLC. The IM/DD method is usually utilized in VLC due to the simple implementation of a low cost receiver, the possibility of modulation formats, and high data rate compared with VLC based on image processing. Variation in the intensity of light emitted from a LED can be converted to current variation by a photo-detector (PD); thus, the modulated signal can be recovered at a receiver. AwhiteLEDwith yellow phosphor excited by blue light has been utilized in VLC. VLC research has primarily focused on a single channel system in which an identical signal is modulated in every cell. On-off keying (OOK) and pulse-position modulation (PPM) have been used as modulation methods for a single channel system [14]. Different position information should be transmitted from the respective LEDs. When illuminated LEDs are densely arranged, inter-cell interference may be a serious problem in the single channel system, which uses the same modulation frequency band. In order to overcome the inter-cell interference, a carrier allocation (CA)-VLC system, in which different radio frequency (RF) carriers are adopted for signal modulation among cells, is proposed. A phase-shift keying (PSK) modulation method is useful because the average optical power /$ IEEE

2 KIM et al.: INDOOR VISIBLE LIGHT COMMUNICATION POSITIONING SYSTEM 135 Fig. 3. The proposed positioning system based on CA-VLC and IM/DD. Fig. 2. Illustration of RF carrier modulation (a) and the CA-VLC system used to reduce inter-cell interference (b). remains constant regardless of the application of random signals, e.g., it prevents LED flickering. Since PSK modulation exhibits signal variation from negative to positive values, a DC bias should be applied as shown in Fig. 2(a), as it can prevent signal clipping upon modulating the LED. The use of different frequencies among adjacent cells allows one to distinguish signals being emitted from adjacent cells, thus reducing the inter-cell interference (Fig. 2(b)). In this paper, a positioning system utilizing IM/DD based on CA-VLC is suggested. The CA method was used to mitigate inter-cell interference. The characteristics of the proposed positioning system were investigated by simulation. The modeling of a VLC link in a line-of-sight environment was performed based on measurements. Finally, the proposed system was verified by experimental demonstration. II. OPERATIONAL PRINCIPLE The trilateration method is utilized in the proposed system, and at least three signals are required to calculate the receiver (Rx) coordinates [15]. In order to apply VLC to positioning, signals emitted from the transmitters (Txs) of adjacent cells should be separated at the Rx without inter-cell interference. Thus, CA-VLC, in which different RF carriers,,and are adopted for signal modulation corresponding to Tx1, Tx2, and Tx3, respectively, is used in the proposed positioning system [16]. The Txs transmit respective position coordinates to estimate the receiver position. The Txs were located as arrayed equilateral triangle shapes in order to investigate the basic positioning of the unit cell as shown in Fig. 3. The cell size, CS,is defined as the distance among the Txs. The positioning process was performed as described below. Distances between the transmitters and a receiver were estimated by measuring the received RF power. The signals transmitted from adjacent cells were separated by the CA method. The estimated distances were adjusted using the proposed normalization method because the received RF powers can vary due to radiation directivity and the incidence angle. The Rx position was calculated via a trilateration method using the compensated distances. The channel DC gain of wireless infrared communication was modeled [17]. The receiver optical power in a line-of-sight environment can be expressed as where is the average transmitted optical power, is the distance between the transmitter and receiver, and and are radiation and incidence angles with respect to the transmitter and receiver, respectively. Transmitter radiant intensity and effective signal collection area are given by where the order is related to,whichisthetransmitter semi-angle (at half power), by is the detector physical area, is the gain of the optical filter, and is the concentrator gain. Since is related to both the radiation pattern of the transmitter and the incidence angle of the receiver, the distance between the Txs and detection node (DN), which is the receiver position and can be expressed as coordinates, cannot be accurately estimated by using the received optical power only. Nevertheless, in the proposed system, the distances estimated based on the received optical power from Tx are determined by (1) and can be expressed as The radiation directivity and the incidence angle are not considered and is the optical power constant related to the radiation intensity of the LED, the concentrator gain, the optical filter gain, the physical area of the detector in the normal radiation environment of Tx and the normal incidence of light into the Rx. It may be inefficient to adjust the estimated distance based on measurement of the incidence angle, the radiation pattern of Tx, and the gain profile of the Rx with respect to the incidence angle. The current of the photo detector,, at the Rx is generated in proportion to the received optical power with a certain sensitivity. Consequently, (4) can be converted to (1) (2) (3) (4) (5)

3 136 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 1, JANUARY 1, 2013 where is the RF power received from Tx and is the RF power constant related to the optical-to-electrical conversion efficiency and optical power constant given the following relation:. In the proposed system, estimated distances are determined by (5) using received RF powers regardless of inaccurate distance estimation caused by the optical power attenuation due to the radiation and incidence angles of the Txs and Rx. The radiation pattern and the incident angle are not considered in (5) and real distances are not exactly matchedtothe value obtained using both of the RF powers received from Txs and (5). The value of the estimated distance using (5) can be larger compared with the real distance because the RF power reduced by the angle effects induces the extended distance between transmitter and receiver. Thus, the estimated distance is required to be adjusted by reducing the distance. The adjusted distance,, can be expressed as (6) (7) where is a weighting factor related to both the normalizing factor,, and the normalizing constant,, which is an arbitrary positive constant. The degree of the distance adjustment can be controlled by the value of. When the estimated distance is well reflecting the real distance, the value of is expected as 0. In this case, the adjusted distance is equal to estimated distance. In other words, as approaches 0, the adjusted distance becomes the approximate value of the estimated distance. On the other hand, when the estimated distance is considerably distorted by the angle effects, the value of becomes close to 1. In aspect of positioning using trilateration method, as the adjusted distance becomes close to real distance, the better positioning performance can be obtained. Thus, the improvement of positioning performance can be achieved by adjusting the estimated distance similarly to the real distance value using (6). The determination of the optimal value will be introduced in Section VI. The DN can be calculated using a trilateration method. The method is to solve the problem of determining the intersection of distances between the positions of the Txs and the DN. The equations for trilateration in the proposed positioning system are expressed as follows: where,and are the coordinates of Tx, and the position data are sent from Tx. The estimated position can be calculated using two linear equations, which can be simply obtained by subtracting the second and third equations from the first in (8), and the two linear equations in the condition of can be expressed in matrix form as follows: (8) (9) Fig. 4. The experimental setup for RF power, EVM, and illumination intensity measurements. The equation can be solved using the linear least square method given by [18] because the distances between Txs and DN are approximately estimated by measuring the received RF powers. III. ANALYSIS OF BASIC TRANSMISSION CHARACTERISTICS Fig. 4 shows the experimental setup for the measurements of transmission performance and illumination intensity. The Txs consisted of 2 2 white LED arrays. A constant bias voltage and a QPSK signal were applied to the respective Tx by adopting bias-t. A 0.25-M symbol per second (s/s) QPSK signal with 2 MHz, 2.5 MHz, and 3 MHz carriers were generated by an arbitrary wave generator (AWG) and a vector signal generator (VSG), and then applied to Tx1, Tx2, and Tx3, respectively. Transmitted signals were recovered using a PD, trans-impedance amplifier (TIA), and low noise amplifier (LNA). The TIA converts the variation in photocurrent to voltage. Subsequently, those signals were analyzed by both a vector signal analyzer (VSA) and an RF spectrum analyzer (RFSA). A. Effects of Radiation and Incidence Angles Before implementing the proposed positioning system as theoretically described in Section II,theeffectsofTxsandRxdirectivity were analyzed because the RF powers received after transmission varied as a function of radiation and incidence angles. The illumination intensity according to the variation in distance between the Txs and Rx in the y direction was measured using an illuminometer while keeping the radiation and incidence angles at 0. The received optical powers had similar attenuation tendencies as shown in Fig. 5(a). The illumination intensity was measured for Tx1 as the radiation angle was varied, while the distance between Tx1 and the illuminometer and the incidence angle were maintained as 60 cm and 0,respectively. Upon increasing the radiation angle from 0 to 90, the received optical power decreased as shown in Fig. 5(b). The radiation patterns for the other Txs were similar to that of Tx1. Fig. 6 shows the illumination intensities according to the Rx movement variation in the x-direction as illustrated in Fig. 4. In this case, the received optical powers decreased because of the effects of the radiation and incidence angles. For example, when the incidence and radiation angles were normal, the illumination intensities were 120 Lux at 80 cm as shown in Fig. 5(a). However,at50cminthexdirection,whichresultedinadistance of 78 cm between the Txs and the Rx, the illumination intensities were approximately 90 Lux for each Txs as illustrated in

4 KIM et al.: INDOOR VISIBLE LIGHT COMMUNICATION POSITIONING SYSTEM 137 Fig. 5. Variation in illumination intensity as a function of the y-directional movement of Rx (a) and the radiation directivity of Tx1 (b). Fig. 7. Measured RF spectra of signals received from respective Txs and received RF power, definition. Fig. 6. Variation in illumination intensity according to the x-directional distance of Rx and the real distance between Tx and Rx. Fig. 5(a). Thus, we propose adjusting the distance using the normalization method expressed in (6). The proposed method can be conducted using software and does not require calculation of the radiation and incidence angles in order to compensate for the reduced signal powers. B. Distance Estimation by Received RF Signal Power The distance between the Txs and Rx is estimated by referring to the RF powers received under normal radiation and incidence as expressed in (5). Since the two signals resulting from the AWG had less power than that from VSG, two LNAs were utilized to equalize the signal powers. Fig. 7 shows the measured RF spectra of received signals at 55 cm in the x-direction using the RFSA. The signals of both Tx1 and Tx2 were distorted and broadened compared with the Tx3 signal because of the increased non-linearity due to additional LNA use. The received RF power,,issimplydefinedasthe peak signal power obtained from Tx (Fig. 7). The received RF powers were measured as the y-directional distance for the Rx was varied as a reference for estimating distances. Fig. 8 shows the measured and theoretical results. The transmission performances of the Txs had similar tendencies as the transmission distance was varied. The experimental results coincided with the theoretical results obtained using (5) when the RF power constant,,wasset to [cm mw]. The theoretical result was utilized to estimate the distance between the Txs and Rx by measuring the Fig. 8. The relationships between received RF power and transmission distance determined using both measured and theoretical results. received RF powers using the RFSA. For example, if the received RF power of Tx1 is dbm, the estimated distance between Tx1 and Rx, will be cm. C. Signal Performance Variation due to Transmission Distance Variation The coordinates of LED should be provided for position calculation at the Rx, thus deriving the estimated position using received RF powers, ;theestimated position can be determined with (9). To investigate the transmission performance, EVM (error vector magnitude) values were measured as the y-directional distance between the Txs and the Rx were varied and the results were compared to those detailed in Section III.B. The received signal power decreased as the transmission distance increased, and consequently, the signal-to-noise-ratio (SNR) was also degraded. Fig. 9 shows the measured EVM values and the constellation of Tx3 at a specific distance using the VSA. The degradation tendencies of the EVM value were slightly different among the Txs. This might be a result of the variations in the modulation efficiency, signal quality, and frequency response due to the use of different generators and RF carriers. We considered the received RF powers to be equal

5 138 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 1, JANUARY 1, 2013 Fig. 9. The measured EVM values with respect to Txs by the vairation of transmission distance; the insets are the constellation of the Tx3 signal. for positioning purposes. The EVM values of 17%, 25%, and 35% corresponded to bit error rates of, and, respectively. IV. SIMULATIONS The proposed positioning system was simulated in order to analyze the system characteristics without measurement error, which may occur during the measurement of received RF power. Simulation is a convenience means of assessing system performance variation according to fluctuations in different parameters such as the radiation pattern of the transmitter, receiver field of view, and normalizing factor. In this section, the design of the VLC link based on experimental results is described and the performance of the proposed system is discussed. A. VLC Link Modeling Using the Lambertian Property, the received RF power under normal conditions, which refers to when only the normal radiation and incidence are considered, can be calculated as (10) using (5), where and are the distance between a transmitter and receiver, and the RF power constant, respectively. In a real VLC link, the received RF power is affected by the radiation and incidence angle. Thus, (10) can be modified as (11) where and are radiation and incidence angles with respect to the transmitter and receiver, respectively. and are the normalized radiation and incidence gains of a transmitter and a receiver with respect to and, respectively. The characteristics of and depend on the LED lens of the transmitter and the concentrator lens of the receiver. In order to model the normalized radiation and incidence gains, the received RF powers were measured using RFSA by rotating the transmitter and receiver. The results are shown in Fig. 10. A 3 db reduction in the received RF power occurred when the transmitter and receiver were rotated approximately 50 and 22, respectively. The Lambertian property of transmitter radiation Fig. 10. The normlized radiation (a) and incidence (b) gain according to and, respectively. has been used previously to model the VLC transmission link in optical wireless communication [2], [17]. Based on the experimental results, and were modeled using a Lambertian property and expressed as (12) (13) where and are related to and, which are the transmitter and receiver semi-angles at half power, and are given by and. with and with are illustrated in Fig. 10. Even though the modeling result using the Lambertian property was not consistent with the experimental results, the modeling was useful for analyzing the system characteristics according to variations in,and. Since the normalized receiver gain exhibited narrow semiangles, the CS was set to 60 cm and height,,wassetto60 cm as shown in Fig. 3. Signal transmissions were achieved at a distance greater than 170 cm under normal conditions, and the narrow coverage for signal detection was thought to be due to the properties of the measured and. We expect that the detection coverage can be expanded by using higher optical power radiated from the LEDs, improved modulation efficiency, and a wider field of view for both the transmitter and receiver. The three Txs were arranged as an equilateral triangle to investigate the basic positioning of the unit cell. The coordinates of the Txs were set to, and. It was assumed that the signals transmitted from the Txs included the position information and were well recovered at the receiver. The RF powers received from the Txs were calculated using (11) with,and [cm mw]. Subsequently, the positioning process was carried out using the (5), (6), and (8).

6 KIM et al.: INDOOR VISIBLE LIGHT COMMUNICATION POSITIONING SYSTEM 139 Fig. 12. Positions estimated utilizing VLC modeling via the Lambertian property with adjustment using nomalizing factors, (a) and (b). Fig. 11. The estimated position, which was calculated using (8), according to the various radiation and incidence semiangles: and, (a) and,(b) and,(c)and and (d). B. Position Estimation As mentioned in Section III.A, since the received RF power was reduced due to the effects of the radiation and incidence angles, the estimated distance was greater than the actual distance. The estimated distance was determined by (5) without considering the angle effects. Initially, the effects of the radiation and incidence angles were investigated as the transmitter and receiver semi-angles were varied. Positioning was achieved using (8) without adjustment. The distances estimated using (5) were substituted in (8). Positioning was conducted at 69 detection nodes within the inner equilateral triangle to investigate positioning performance. The results are shown in Fig. 11. As the transmitter and receiver semi-angles increased, the positioning errors decreased. The Lambertian pattern at a semi-angle of 90 remained approximately constant with respect to the various angles as shown in Fig. 10(a). Thus, the estimated distance is accurate at and. Since the positioning error originated from an inaccurate value of, a reduction in this error can be achieved by adjusting using (6) and (7). The estimated positions as a function of the normalizing factor,,areshowninfig.12.as increased, the estimated position converged at the center of mass of the triangle. The position errors continued to decrease up to, but began to increase again thereafter due to excessive adjustment. The inner and outer position errors were defined as the position error when the estimated position was nearer and farther from the triangle center than the detecting position, respectively. The outer and inner position errors are shown in Figs. 11(a) and 12(b), respectively. The inner position error cannot exceed the distance between the edge detecting position and the center of the triangle. However, the estimated position outside the triangle tended to diverge from the triangle. Thus, should be determined given that the estimated positions should be inside triangle according to the various possibilities for and. Fig. 13. The experimental setup for the proposed positioning system. V. THREE CHANNEL TRANSMISSION EXPERIMENT Fig. 13 shows the experimental setup and a picture of the proposed positioning system. and were both set to 60 cm. Thus, the coordinates of the Txs were set to,and. As in the previous experiments described in Section III, 0.25-M s/s QPSK signals with 2 MHz, 2.5 MHz, and 3 MHz carriers were generated by the AWG and VSG. The signals were simultaneously applied to Tx1, Tx2, and Tx3. Both the EVM values and the received RF powers of the signals were measured at 18 detection nodes in the inner equilateral triangle to investigate the positioning performance of the proposed system. The results are shown in Figs. 14 and 15. Fluorescent light lamps were turned off during the measurements. We recorded the values of received RF powers by observing RF spectra in the RFSA with the receiver fixed at given detection nodes. The received RF powers were used to estimate the distance between the Txs and Rx. The RF spectra at several detection nodes are illustrated in Fig. 14(d). The estimated distances, s, were calculated based on the received RF powers. For example, considering detection node 1, the received RF powers of,and dbm were converted to 70.47, 198.6, and cm, respectively, using (5) with a of [cm mw] as illustrated in Fig. 8. The EVM performance was degraded with increasing transmission distance as illustrated in Figs. 14(a), (b), and (c). The effects intensified due to the radiation and incidence angles. For example, the real distance between Tx2 and Rx was approximately cm at detection node 1. The distance corresponded to a received RF power of dbm under normal radiation and incidence (Fig. 8). However, the received RF power was dbm at detection node 1, which corresponded to an estimated distance of

7 140 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 1, JANUARY 1, 2013 Fig. 16. The results of positions estimated experimentally and via VLC link modeling using exponential properties before adjustment (a) and after adjustment (b). Fig. 14. The measured EVM values of Tx1 (a), Tx2 (b), and Tx3 (c) signals and the measured RF spectra after simultanesous transmission at detection nodes 4, 10, and 12 (d). Fig. 17. The normalized gains by the variation of angle with respect to experimental results, Lambertian property, and Exponential method. model of and in order to fit these values to the experimental results. and were modeled using an exponential property and were expressed as (14) (15) Fig. 15. Received RF powers at detection nodes through the simulation and measurement cm. Thus, the effects cause not only SNR deterioration, but also the inaccurate estimation when calculating the distances between the Txs and Rx. We propose an adjusting the estimated distance value by adopting a weighting factor,, to mitigate the effects of the reduced RF powers as expressed in (6) and (7). The values were adjusted with both an of and a of , and the adjusted distances, s, were determined using (6) and (7). Finally, positions were estimated by substituting s into (8), and the results are shown in Fig. 16. The fact that the value of from the simulation described in Section IV was different might be due to measurement error and the mismatch of both the radiation pattern and the receiver field of view pattern between the experimental results and the link modeling using the Lambertian property. Usually, a lens is packaged with a LED; thus, the radiation pattern is affected by the lens shape, the internal refractive index of the lens, and the LED arrangement. The influence of the lens also applies to receiver efficiency because it causes variation in the incidence angle at the receiver. Thus, link modeling using the Lambertian property may be inaccurate. We suggest general where and are the slope constants with respect to and, respectively, and and are related to and by and, respectively. It is expected that and are related to the lens shape and the internal refractive index. Both with and with are illustrated in Fig. 17. The received RF power decreased further in simultaneous transmission. The RF power received from Tx1 was approximately dbm under normal conditions (Fig. 8) and dbm at detection node 1 (Fig. 15). The result was induced from the fact that the PD was saturated by the intense optical power during simultaneous transmission of three channels as the PD had a limited response to the received optical power [19], [20]. Therefore, the channel gain of was assumed in the VLC modeling of three channel transmission. The received RF powers in VLC modeling of the exponential property are shown in Fig. 15. Fig. 16 shows the results of positioning with VLC link modeling using the exponential property. The adjustment parameters were an of and a of The measurement error of the received RF power affected the estimated distance more seriously in the lower RF power region. The measurement error of a 2.25-dB power reduction resulted in estimated distance errors of 20 cm and 7 cm at dbm and dbm, respectively. Thus, positioning using unadjusted measured values tended to yield larger positioning errors in the region of the triangle sides than in the triangle center as compared to positioning using VLC link modeling with the exponential property. The adjustment method using

8 KIM et al.: INDOOR VISIBLE LIGHT COMMUNICATION POSITIONING SYSTEM 141 (6) reduced the average position errors from and cm to 0.4 and 2.4 cm in simulation and experimental results, respectively. These results are compared to other electromagnetic wave-based positioning systems as follows; In RADAR system, it was reported that the propagation modeling method provides a resolution of 1.86 m for the 25th percentile [8]. The target resolution of Cricket location-support and active badge systems is room-sized granularity within a few square feet, so as to distinguish portions of rooms [9], [10]. It was presented that the 95% of readings are within 9 cm of true positions using measurements in BAT system [11]. The resolution of a few meters in most wireless indoor positioning systems using GPS-based, radio frequency identification (RFID), cellular-based, and wireless local area network (WLAN) techniques was reported except for the UWB positioning system with the accuracy of less than one meter [21]. VI. NORMALIZATION PROCESS The variation of system performance as a function of the normalizing factor,, was investigated. The optimal value depends upon the system environment parameters such as,and. These parameters influence the received optical power as they are related both the incidence and radiation angles. In this section, three techniques to determine the optimal value of are suggested: mathematical analysis, calibration, and fixed for transmitter diversity. A. Mathematical Analysis for Determining the Optimal It is assumed that the link property of VLC is known and the Rx knows,and basedonthe information received from the Txs. Thus, the adjusted distance can be calculated using (6), (7), and (11) and rearranged as (16) is the distance under normal conditions estimated by (5); is the accurate estimated distance without the effect of power reduction by the radiation and incidence angles. is an attenuation coefficient, which was 1/3 due to the PD saturation by multiple Txs discussed in the previous section. The desired value of is varied from to because the real distance between the Rx and Tx cannot be shorter than or longer than. For example, in Fig. 3, the distance between position A and Tx3 is, and the radiation and incidence angles are both 0; the distance is and the angles are at position B. Thus, two equations can be obtained from (16) via the boundary condition through the minimum and maximum values of. In the case of minimum and. In the second case,, and both and can be calculated. The solution can be expressed as (17) The solution was applied to the positioning system through VLC link modeling with Lambertian and exponential properties as discussed in Sections IV and V. In Lambertian modeling, when,and, the optimal and were and 60, respectively. In exponential modeling, when,and, the optimal and were and , respectively. B. Calibration Method for Determining the Optimal It is assumed that the system environment is known in order to optimize via mathematical analysis. In the calibration method, we assume that a detection node inside the positioning unit cell is known. The known detection node is defined as a calibration node and expressed as. The Rx is located in the calibration node, and receives signals from the Txs. The Rx calculates the s based on measured RF powers. When the arbitrary is determined to be a positive value, the objective is to identify the estimated position,, which is similar to using (5), (6), (7), and (9). The objective can be formulated as follows: A simple way to solve the above problem is to determine the minimum position error, by substituting all possible values of from0to1in(7).inotherwords,thevaluesof,and1aresubstitutedin(7), the estimated positions are determined by (9), the position errors can be calculated, and finally an optimal can be selected when is minimized. In this algorithm, as the number of substitution increases to find optimal value, the more calculation load is piled up. However, the positioning performance can be improved by comparing position errors densely with the compact step size of.thealgorithm process requires the calculation of (6), (7), and (9) for each value in addition to the comparison process to find the minimum position error where the calculation load becomes heavy due to quadruple of the number of substitution. When, and 90, the simulation results showed, and , respectively. The simulation parameters were,and using the exponential method. C. Determination of a Fixed for Transmitter Diversity The mathematical analysis and calibration method can provide the high resolution of the estimated position. However, since the transmitter s characteristics such as transmitter semi-angle and modulation efficiency may not be equalized for all transmitters, it may be an inefficient manner to store transmitter s characteristics and with respect to all LED cells for mathematical analysis and calibration method. In this section, the positioning performance is investigated through the fixed rather than determination of optimal according to each characteristic of transmitters. The transmitters with different make the received RF power fluctuated even in the same transmission distance. The

9 142 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 1, JANUARY 1, 2013 Fig. 18. The estimated positions with respect to value of (a) and (b) by applying random Tx semi-angle from 40 to 60. concept for determination of is that the variance of the estimated position induced from the fluctuated RF power can be alleviated through the excessive adjustment by using higher value of a fixed without the consideration of various.thedetermination of a fixed starts from the condition in which the estimated positions have inner position errors by excessive adjustment as shown in Fig. 12(b). The primary property of normalizingmethodistoenabletheestimatedpositiontoconverge into the center of triangle by using high value of.inmathematical analysis, a possible lowest RF power given by is considered to determine the optimal because the lowest RF power corresponds to the longest distance which can be existed in a given cell. By the same way, we can consider the possible lowest RF power with respect to various transmitters which have different characteristics of. Normally, the received RF power is considerably affected by the transmitter semi-angle. In first, positioning was evaluated considering the with the smallest transmitter semi-angle for decision of a fixed. As the transmitter semi-angle decreases at given value, the estimated position has the tendency of diverging toward the outside of triangle; thus, for positioning optimization, the higher value is required with respect to the transmitter with relatively narrow semi-angle. For example, the optimal value of and corresponding to the semi-angle of 40 and 70 can be calculated by mathematical analysis. Although the value of is optimal with respect to the transmitter with the semi-angle of 40, it can excessively adjust the estimated position with respect to transmitters with the semi-angle higher than 40. Basically, the LED coordinates should be provided in order to calculate receiver position by trilateration. Therefore, the deployment structure of LED cell can be obtained by the coordinate information of LED cells. The fixed value with respect to the transmitter with the minimum semi-angle can be calculated by the mathematical analysis. This value is not optimal for the other transmitter with the higher semi-angle and causes the excessive adjustment. However, the value has tendency of confining the estimated position to the inside of triangle for the transmitters having a semi-angle larger than 40. The estimated positions were evaluated by simulation in the environment of Fig. 13. Both the CS and height were 60 cm. The of (14) with was used for all transmitters. The attenuation coefficient was fixed as 1/3 with respect to all LED cells. The transmitter semi-angle was randomly chosen and varied between 40 and 60. The estimated positions at 18 detection nodes were evaluated by using measurements Fig. 19. Probability distribution of average position error with respect to various by applying random Tx semi-angle from 40 to 60. in order to consider LED cells and the transmitters with different semi-angles. The constellations of estimated positions were depicted in Fig. 18 with respect to fixed values of (a) and (b) which were calculated by the (17) for the semi-angle of 70 and 40.As value increased, the deviation of estimated position by the fluctuated RF power decreased. This result was induced from the fact that the fluctuation of estimated distance by the random variance of received RF power was alleviated by the adjustment process using (6). We measured average position errors with respect to various in order to evaluate the respective position accuracy for LED cells at given value as shown in Fig. 19. It can be seen that 100% of average estimated position were within cm with the fixed value of corresponding to the semi-angle of 40. In case of value higher than , although the deviations of average position error were relatively smaller than those of other values, the estimated positions were moved to the center of the mass of triangle which resulted in less accurate estimated positions. In case of value smaller than , the deviation increased due to the growth of outer position errors as the value was reduced. Including the semi-angle variation between 40 and 60, the additional random variation within % of the received RF power was applied for the consideration of the high optical power effects and various transmitter characteristics such as and. The constellations of estimated positions with respect to various LED cell structures were illustrated in Fig. 20 in the conditions of a fixed corresponding to the semi-angle of 40, the semi-angle variation between 40 and 60, and received RF power variation of %. In mathematical analysis, since the lowest RF power corresponding to the longest distance is considered, the in (17) should be the largest value among the lengths of triangle sides which can be calculated by using LED coordinates. For example, swere, 86, and 89.4 in case 2, 3, and 4, respectively. The fixed values corresponding to semi-angle of

10 KIM et al.: INDOOR VISIBLE LIGHT COMMUNICATION POSITIONING SYSTEM 143 VII. CONCLUSION We have proposed an indoor positioning system adopting VLC based on IM/DD in a line-of-sight environment, which was investigated by simulations and experiments. Even though the CA method was utilized to mitigate inter-cell interference in this paper, it is expected that the proposed algorithm can be utilized with other channel separation methods including wavelength division multiplexing by a RGB-LED and optical filters, time division multiplexing, and code division multiplexing. Although it was inaccurate when estimating distance based on the effects of radiation directivity and the incidence angle, we were able to reduce the positioning error using the proposed adjustment method. We achieved the positioning accuracy within a few centimeters using the adjustment method with optimal and a fixed through simulation and experiment. The proposed indoor positioning scheme would be useful for various indoor location-based applications. Fig. 20. The estimated positions with respect to various LED cell structures by applying random power fluctuation and Tx semi-angle. Fig. 21. Probability distribution of average position error with respect to various LED cell structures by applying random power fluctuation and Tx semiangle. 40 in various LED structures were slightly changed by the different and. Fig. 21 shows the probability distribution of average position error using measurements for various LED cell structures. It was observed that 100% of average estimated position was within approximately 7 cm for various LED cell structures. Although the well-equalized transmitters make the proposed system to have better positioning performance, we believe that positioning could be well performed by the fixed for the excessive adjustment despite of the variance of received RF power. REFERENCES [1] Y. Tanaka, S. Haruyama, and M. Nakagawa, Wireless optical transmissions with white colored LED for wireless home links, in Proc. 11th Int. Symp. PIMRC, 2000, pp [2] T. Komine and M. Nakagawa, Fundamental analysis for visible-light communication system using LED lights, IEEE Trans. Consumer Electron., vol. 50, no. 1, pp , [3] K.Lee,H.Park,andJ.R.Barry, Indoorchannelcharacteristicsfor visible light communications, IEEE Commun. Lett., vol. 15, no. 2, pp , [4] T. Komine, J. Lee, S. Haruyama, and M. Nakagawa, Adaptive equalization system for visible light wireless communication utilizing multiple white LED lighting equipment, IEEE Trans. Wireless Commun., vol. 8, no. 6, pp , [5] H. Minh, D. O Brien, G. Faulkner, L. Zeng, K. Lee, D. Jung, Y. Oh, and E. Won, 100-Mb/s NRZ visible light communications using a postequalized white LED, IEEE Photon. Technol. Lett, vol. 21, no. 15, pp , [6] S. Okada, T. Yendo, T. Yamazato, T. Fujii, M. Tanimoto, and Y. Kimura, On-vehicle receiver for distant visible light road-to-vehicle communication, in Proc. IEEE Intelligent Vehicles Symp., [7] H. Chinthaka, N. Premachandra, T. Yendo, M. Tehrani, T. Yamazato, H. Okada, T. Fujii, and M. Tanimoto, High-speed-camera image processing based LED traffic light detection for road-to-vehicle visible light communication, in Proc. IEEE Intelligent Vehicles Symp., [8] P. Bahl and V. Padmanabhan, RADAR: An in-building RF-based user location and tracking system, in Proc. INFOCOM, 2000, vol. 2, pp [9] N. Priyantha, A. Chakraborthy, and H. Balakrishnan, The cricket location-support system, in Proc. Int. Conf. Mobile Comput. Netw., 2000, pp [10] R. Want, A. Hopper, V. Falcao, and J. Gibbons, The active badge location system, ACM Trans. Info. Syst., vol. 10, no. 1, pp , [11]A.Harter,A.Hopper,P.Steggles,A.Ward,andP.Webster, The anatomy of a context aware, in Proc. ACM/IEEE MOBICOM, [12] T. Komine and M. Nakagawa, Integrated system of white LED visible-light communication and power-line communication, IEEE Trans. Consumer Electron., vol. 49, no. 1, pp , [13] M. Yoshino, S. Haruyama, and M. Nakagawa, High-accuracy positioning system using visible LED lights and image sensor, in Proc. IEEE Radio Wireless Symp., [14] H. Sugiyama, S. Haruyama, and M. Nakagawa, Experimental investigation of modulation method for visible-light communications, IEICE Trans. Commun., vol. E89-B, no. 12, pp , [15] D. Manolakis, Efficient solution and performance analysis of 3-D position estimation by trilateration, IEEE Trans. Aerospace and Electron. Syst., vol. 32, no. 4, pp , [16] H. Kim, D. Kim, S. Yang, Y. Son, and S. Han, Mitigation of inter-cell interference utilizing carrier allocation in visible light communication system, IEEE Commun. Lett., vol. 16, no. 4, pp , [17] J. Kahn and J. Barry, Wireless infrared communications, Proc. of the IEEE, vol. 85, no. 2, pp , 997.

11 144 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 1, JANUARY 1, 2013 [18] F. Reichenbach, A. Born, D. Timmermann, and R. Bill, A distributed linear least squares method for precise localization with low complexity in wireless sensor networks, Distributed Computing in Sensor Systems Lecture Notes in Computer Science, vol. 4026, pp , [19] K. J. Williams, R. D. Esman, and M. Dagenais, Effects of high space-charge fields on the response of microwave photodetectors, IEEE Photon. Technol. Lett., vol. 6, no. 5, pp , [20] H. Kim, D. Kim, S. Yang, Y. Son, and S. Han, Analysis of inter-cell interference and crosstalk in carrier allocation based visible light communication, in Proc. OFC/NFOEC, 2012, Paper JTh2A.50. [21] H. Liu, H. Darabi, P. Banerjee, and J. Liu, Survey of wireless indoor positioning techniques and systems, IEEE Trans. Syst., Man, and Cyber., Part C: Appl. Rev., vol. 37, no. 6, pp , Hyun-Seung Kim received the B.S. degrees in electrical and electronic engineering at Yonsei University, Seoul, Korea in He is currently studying in the unified Master s and Doctor s course in electrical and electronic engineering at Yonsei University. His current research interests are visible light communication, radio over fiber system, optical fiber transmission system, optical device, and optical system for communications. Deok-Rae Kim received the B.S. degrees in electronic engineering at Dankook University, Korea in He is currently working as Master s course in electrical and electronic engineering at Yonsei University. His current research interests are visible light communication, optical device, and CAN communication. Se-Hoon Yang received the B.S. degrees in electrical and electronic engineering at Yonsei University, Seoul, Korea in He is currently studying in the unified Master s and Doctor s course in electrical and electronic engineering at Yonsei University. His current research interests are radio over fiber system, visible light communication, and optical system for communications. Yong-Hwan Son received the B.S. and M.S., and Ph.D. degrees in electronic engineering at Hoseo University, Korea in 1999, 2001, and 2008 respectively. He is currently working as a Research Professor in electrical and electronic engineering at Yonsei University. His current research interests are optical device and visible light wireless communication. Sang-Kook Han (M 95) received the B.S. degrees in electronic engineering from Yonsei University, Seoul, Korea, in 1986 and the M.S. and Ph.D. degrees in electrical engineering from the University of Florida, Gainesville, in From 1994 to 1996, he was with the System IC Laboratory, Hyundai Electronics, where he was involved in the development of optical devices for telecommunications. He is currently a Professor with the Department of Electrical and Electronic Engineering, Yonsei University. His current research interests include optical devices/systems for communications, microwave-photonics technologies, and LED based visible light communications.