Sage Journals: Discover world-class research

Abstract

The accuracy and efficiency of radio-frequency identification device tag localization is significant to a large number of existing and upcoming radio-frequency identification device applications. In this article, we have introduced a novel method to locate a radio-frequency identification device tag in three-dimensional space. The proposed method uses readers with multiple power levels to locate the passive radio-frequency identification device tags. First, we preprocess the data and estimate the distances between the readers and target tag. Then, we use sphere trilateration method to estimate the location of the radio-frequency identification device tags in three-dimensional space. Finally, we use planar fit method and the method of least squares to reduce the localization error. Both the simulation and testbed experiment results have shown that the proposed method can reach a higher positioning accuracy with less radio-frequency identification device readers and more robust to the environment than other existing methods.

Keywords

Radio-frequency identification device localization algorithm three-dimensional space sphere trilateration method of least squares

Introduction

Localization plays an important role in the moving object tracking and target existence sensing.^1–3 At present, the solutions include global positioning system (GPS),⁴ Wi-Fi,⁵ Bluetooth,⁶ and ZigBee.⁷ As a possible solution, radio-frequency identification device (RFID),⁸ RFID Ultra Wideband (UWB),⁹ and other technologies have attracted a lot of research attention. Each technology has its advantages and disadvantages. We can determine the optimal solution based on different requirements at the aspects of cost, accuracy, range, and energy consumption. Table 1 shows a comparison between the different technologies on those aspects.^10,11 Values are in the range of 0–4, while 4 is the best classification and 0 is the worst. It should be noted that these data are from three different studies.^8,11,12

Table 1.

Comparison between different localization technologies.

Technology	Features
Technology	Cost	Accuracy	Range	Energy consumption
Thermal signature	0	4	1	1
Multicamera	0	4	2	2
GPS	4	3	4	0
IMU	3	2	0	1
Bluetooth	3	1	1	0
Wi-Fi	3	2	2	1
ZigBee	3	3	1	4
RFID	3	2	0	3
RFID UWB	1	4	3	4

GPS: global positioning system; IMU: inertial measurement unit;RFID: radio-frequency identification device; UWB: ultra wideband.

From Table 1, the RFID UWB technology shows the superiority, despite its higher cost compared to other technologies (e.g. GPS and ZigBee). RFID UWB develops from early RFID technology, with a higher bandwidth and higher localization accuracy.

The general RFID system consists of RFID tags, antennas, and readers. The basic working principle is that RFID system generates a magnetic field through the readers’ antennas; when the RFID tags enter the magnetic field, they will receive the radio-frequency (RF) signal emitted by the readers; then, the RFID tag will send the product information stored in the chip to the readers (passive RFID tag), or actively send an RF signal to the readers (active RFID tag). Active RFID tag, with built-in battery, transmits the identification signal periodically and has a longer reading range compared to passive RFID tag; however, it is bigger in size, more expensive in price, and shorter in service life. Localization with active tag is generally used to locate an RFID reader, such as the tracking of person or object with RFID reader, while localization with passive tag is generally used to locate a target object of a passive tag.

At present, there are extensive studies on RFID-based two-dimensional (2D) localization.^13–17 A localization system called LANDMARC (LocAtioN iDentification based on dynaMic Active RFID Calibration) is published by LM Ni et al.¹⁸ LANDMARC algorithm arranges a number of reference RFID tags and readers, which have several power levels, and then estimates the coordinates of the target tag using k-NN method according to the power levels. However, the LANDMARC algorithm can only provide coarse-grained localization information; therefore, it cannot meet the need for three-dimensional (3D) indoor localization. Some methods use the phase information to locate the objects;^19–21 Zhou et al.¹⁹ proposed a localization method using the changing of phase affected by new tags; Yang et al.²⁰ developed a phase hologram to track mobile objects, the method can get a high precision in known track. Some other machine-learning-based methods^22–25 have been proposed; Kuo and Chang²² uses an immune-based feed-forward neural network to build the RFID positioning system; Lin et al.²³ combined the neural network and genetic algorithm to locate the objects; Soltani et al.²⁴ used artificial neural networks and virtual reference tags to do RFID-based localization.

RFID-based 3D localization has also been developed in some ways, but the methods still have some drawbacks. Hightower et al.²⁶ proposed a 3D localization system called SpotON, the system directly uses rich site summary (RSS) information to estimate the distances and aggregation algorithm to get the location; the algorithm is simple, but it is not useful to estimate the distance simply through RSS values in real environment. Bouet and Pujolle²⁷ uses multiple readers and the binary information to get the object region, but the method needs a lot of readers to reach a high accuracy, and it is hard to get the reachable region of the readers in real environment. Another solution is presented in Han et al.,²⁸ where the location is predicted by the tag array, but this method works only in simulation environment. Maneesilp et al.²⁹ proposed an RFID 3D localization method. Its basic principle is to use Nelder–Mead method to minimize the error function. This algorithm needs a large number of readers; hence, the cost is expensive. Almaaitah et al.³⁰ proposed a method called APM (Adaptive Power Multilateration) to dynamically adjust the transmit power of RFID readers and use multi-point method to estimate the location of the target tag; but the APM method requires the reader to dynamically adjust the power level to achieve fine-grained resolution, which is harsh.

In this article, we propose a 3D localization method based on passive RFID tags. In the method, the signal strength of the readers is divided into several levels. First, we estimate the distances between the readers and the RFID tags, according to the reference tag collection in 2D space. Then, we estimate the location of the RFID tag in 3D space through sphere trilateration.³¹ Finally, we use the method of least squares³² to decrease the error. In this article, we use MATLAB³³ to simulate the RFID-based localization system and compare the proposed method with other methods. Moreover, we have carried out testbed experiments in realistic environment to further demonstrate the appropriateness of the proposed method. Both the simulation and the testbed experiment results have shown that, compared to other existing methods, the proposed method can reach a higher positional accuracy using less RFID readers.

The rest of this article is organized as follows: section “Algorithms” introduces the principles of the algorithms; section “Simulation” explains the algorithm simulation and the simulation results; section “Experimental evaluation” shows the experimental evaluation results in real environment; and section “Conclusion and future work” gives a conclusion of this article and explains our future work.

Algorithms

Power level

In this method, the signal strength of the readers is divided into several levels. By adjusting the power level of the reader, we can get different tags sets. The basic model of this method is shown in Figure 1, and the system consists of reference readers, reference tags, and target tag.

Figure 1.

Compositions of the localization system.

By adjusting the power levels of reader i, two key power levels of the target tag can be determined: $K_{\min}$ (when the target tag just cannot be reached) and $K_{\max}$ (when the target tag just can be reached). We denote the reference tags set in power $K_{\min}$ as $Q_{\min}$ , in power $K_{\max}$ as $Q_{\max}$ ; according to $Q_{\min}$ and $Q_{\max}$ , we can get a valid reference tags set $Q_{i}$ in 2D space, as shown in Figure 2

$Q_{i} = Q_{\max} - Q_{\min}$ (1)

Figure 2.

Reference tags set $Q_{i}$ between $Q_{\min}$ and $Q_{\max}$ .

The distance set between the readers i and the target tag t can be evaluated by the following steps:

Calculate the distance $D_{ij}$ between reference tag $j (x_{j}, y_{j}, z_{j})$ in set $Q_{i}$ and reader $i (x_{i}, y_{i}, z_{i})$

$D_{ij} = \sqrt{((x_{j} - x_{i})^{2} + {(y_{j} - yi)}^{2} + {(z_{j} - z_{i})}^{2})}$ (2)

Get the minimum distance $D_{\min}$ and maximum distance $D_{\max}$ in set $Q_{i}$

$\begin{matrix} {\begin{matrix} D_{\min} = \min {D_{ij}} \\ D_{\max} = \max {D_{ij}} \end{matrix} & D_{ij} \in D \end{matrix}$ (3)

Repeatedly get the signal strength of the reference tag in set $Q_{i}$ and the target tag t received by the reader i, totally n times, denote the $k th$ signal strength as $RSS I_{ik}$ ;

Calculate the variance $σ^{2}$ of the received signal strength indication (RSSI)

$\begin{matrix} σ^{2} = \frac{1}{n - 1} \sum_{k = 1}^{n} {(RSS I_{ik} - A)}^{2} \\ where A = \frac{1}{n} \sum_{k = 1}^{n} RSS I_{ik} \end{matrix}$ (4)

For the $k th$ signal strength $RSS I_{ik}$ , remove it if $RSS I_{ik} - A > 3 σ$ and get RSSI set with size of m, and get the average signal strength $\bar{RSS I_{i}}$

$\bar{RSS I_{i}} = \frac{1}{m} \sum_{k = 1}^{m} RSS I_{ik}$ (5)

Extend the set $Q_{i}$ to $\hat{Q_{i}}$ according to the missing reference tags $ta g_{ik}$ between $D_{\min}$ and $D_{\max}$

$\hat{Q_{i}} = Q_{i} + ta g_{ik}, where D_{\min} < D_{ik} < D_{\max}, ta g_{ik} \notin Q_{i}$ (6)

Complement the signal strength $\bar{RSS I_{ik}}$ according to the $\hat{k} - nearest neighbors$ in set $Q_{i}$

$\bar{RSS I_{ik}} = \frac{\sum_{j = 1}^{\hat{k}} RSS I_{jk}}{\hat{k}}$ (7)

Calculate the difference in signal strength between every reference tag j in set $\hat{Q_{i}}$ and target tag t, denote as $E_{ij}$

$E_{ij} = \sqrt{{(\bar{RSS I_{ij}} - \bar{RSS I_{it}})}^{2}}$ (8)

Choose $\hat{k} - nearest neighbors$ based on $E_{ij}$ , and calculate the weight, denote as $w_{ij}$

$w_{ij} = \frac{1 / E_{ij}^{2}}{\sum_{j = 1}^{\hat{k}} 1 / E_{ij}^{2}}$ (9)

Calculate the distance between target tag t and reader i, denote as $L_{it}$

$L_{it} = \sum_{j = 1}^{k} w_{ij} D_{ij}$ (10)

Denote the distance set as $\bar{D}$

$\bar{D} = {L_{1 t}, L_{2 t}, \dots, L_{nt}}$ (11)

where n is the number of readers.

Sphere trilateration

After we get the distance set $\bar{D}$ between the readers and the target tag t, the coordinates of the target tag can be solved by the following three sphere equations, as shown in Figure 3

${\begin{matrix} {(x_{1} - x_{0})}^{2} + {(y_{1} - y_{0})}^{2} + {(z_{1} - z_{0})}^{2} = L_{1}^{2} \\ {(x_{2} - x_{0})}^{2} + {(y_{2} - y_{0})}^{2} + {(z_{2} - z_{0})}^{2} = L_{2}^{2} \\ {(x_{3} - x_{0})}^{2} + {(y_{3} - y_{0})}^{2} + {(z_{3} - z_{0})}^{2} = L_{3}^{2} \end{matrix}$ (12)

where $(x_{i}, y_{i}, z_{i}) (i = 1, 2, 3)$ is the coordinates of readers, and $(x_{0}, y_{0}, z_{0})$ is the coordinate of target tag to be calculated.

Figure 3.

Sphere trilateration.

We can get $(x_{0}, y_{0}, z_{0})$ by solving the equations. There can be two results: the complex solutions or the real solutions; however, we only record the real solutions and ignore the complex ones in this method. In the actual calculations, different numbers of readers can lead to different numbers of sphere equations. For example, if there are n readers and every three sphere equations can make a sphere equation set, then there will be $C_{n}^{3}$ combinations. To fasten the calculating speed, when the number of combinations is more than k, we choose k of them to calculate. Feasible solution set of sphere equations is denoted by R.

The method of least squares

Through sphere trilateration, we can get the feasible solution set R. After several experiments, we found that the error in x axis and y axis in feasible solutions is relatively small, so we choose planar fit method using the method of least squares to minimize the error in z axis, a case of fitting results is shown in Figure 4.

Figure 4.

Planar fit method using the method of least squares.

The method of least squares is a mathematical optimization technique; it looks for the best matching function of the data by minimizing the squares of errors. Using the method of least squares, we can easily get the solution and minimize the squares of errors between the solution and the actual data. The general linear model with the method of least squares can be expressed as following:

Irrelevant model variables $t_{1}, \dots, t_{q}$ can be composed of the form of a linear function

$y (t_{1}, \dots, t_{q}; b_{0}, b_{1}, \dots, b_{q}) = b_{0} + b_{1} t_{1} + \dots + b_{q} t_{q}$ (13)

Equation (13) is the same as the linear equations (14)

$\begin{matrix} b_{0} + b_{1} t_{11} + \dots + b_{j} t_{qj} + \dots + b_{q} t_{1 q} = y_{1} \\ b_{0} + b_{1} t_{21} + \dots + b_{j} t_{2 j} + \dots + b_{q} t_{2 q} = y_{2} \\ ⋮ \\ b_{0} + b_{1} t_{i 1} + \dots + b_{j} t_{ij} + \dots + b_{q} t_{iq} = y_{i} \\ ⋮ \\ b_{0} + b_{1} t_{n 1} + \dots + b_{j} t_{nj} + \dots + b_{q} t_{nq} = y_{n} \end{matrix}$ (14)

Generally, $t_{ij}$ is denoted by data matrix A, parameters $b_{j}$ by parameter vector b, observed values $y_{i}$ by Y, then the linear equations can be written as

$\begin{matrix} (\begin{matrix} 1 & t_{11} & \dots & t_{1 j} & \dots & t_{1 q} \\ 1 & t_{21} & \dots & t_{2 j} & \dots & t_{2 q} \\ ⋮ & ⋮ & ⋮ & ⋮ \\ 1 & t_{i 1} & \dots & t_{ij} & \dots & t_{iq} \\ ⋮ & ⋮ & ⋮ & ⋮ \\ 1 & t_{n 1} & \dots & t_{nj} & \dots & t_{nq} \end{matrix}) \cdot (\begin{matrix} b_{0} \\ b_{1} \\ ⋮ \\ b_{j} \\ ⋮ \\ b_{q} \end{matrix}) = (\begin{matrix} y_{1} \\ y_{2} \\ ⋮ \\ y_{i} \\ ⋮ \\ y_{n} \end{matrix}) \\ Namely, Ab = Y \end{matrix}$ (15)

Derive the equations in the form of linear squared difference using the method of least squares

$min_{b} ‖ Ab - Y ‖_{2}$ (16)

Through sphere trilateration, we can get the feasible solution set R, and the discrete points in R are denoted by $(x_{i}, y_{i}, z_{i})$ . We use the method of least squares to fit the plane equation

$z (x_{1}, y_{1}; b_{0}, b_{1}, b_{2}) = b_{0} + b_{1} x_{1} + b_{2} y_{1}$ (17)

For the discrete points in feasible solution set R, we can get the equations

${\begin{matrix} b_{0} + b_{1} x_{1} + b_{2} y_{1} = z_{1} \\ b_{0} + b_{1} x_{2} + b_{2} y_{2} = z_{2} \\ ⋮ \\ b_{0} + b_{1} x_{n} + b_{2} y_{n} = z_{n} \end{matrix}$ (18)

The target equation of the method of least squares is $min_{b} (\sum_{i = 1}^{n} (b_{0} + b_{1} x_{i} + b_{2} y_{i} - z_{i})^{2})$ , we need to solve the coefficients of the fitting equation $(b_{0}, b_{1}, b_{2})$ and get the fitting equation

$z (x_{1}, y_{1}; b_{0}, b_{1}, b_{2}) = b_{0} + b_{1} x_{1} + b_{2} y_{1}$ (19)

Then, we sort the x values and y values in feasible solution set R, use mean values of the middle third of sorted $x, y$ as the input for the fitting equation, and finally get the estimated target coordinates, denoted as $(x_{e}, y_{e}, z_{e})$ .

Flowchart

The flowchart of the algorithms is shown in Figure 5, it contains the following steps:

Get the locations of reference tags;

Adjust the power level to get valid reference tags set;

Repeatedly get RSSI information and use Gaussian filter to preprocess the data;

Complete the missing reference tags and get the corresponding RSSIs by $\hat{k} - nearest neighbors$ ;

Evaluate the distance by the relation between RSSI and power level;

Get the feasible coordinates of the target tag through sphere trilateration;

Evaluate the coordinates of target tag through method of least squares and get the location of target tag;

If there is any target tag left, go to step 2; otherwise, finish.

Figure 5.

Flowchart of the proposed algorithm.

Simulation

In the simulation, we simulate a hexahedron with the size of 10 m × 10 m × 10 m.

We use the log-distance path loss model in our simulation experiment. It provides the attenuation of signal from distance in the indoor environment

$PL (d) = PL (d_{0}) + 10 nlo g_{10} (\frac{d}{d_{0}}) + X_{δ}$ (20)

where $d_{0}$ is the reference distance, n denotes the path loss exponent, d is the distance between the sender and receiver, and $X_{δ}$ is zero mean Gaussian random variable with standard deviation $δ$ .

The RSSI can be expressed by the following equation

$RSSI (d) = P_{t} + G_{t} - PL (d)$ (21)

where $P_{t}$ is the transmit power, and $G_{t}$ denotes the antenna gain of transmit node.

Since the $G_{t}$ is fixed, combine equation (20) and equation (21), we can get the equation (22)

$RSSI (d) = RSSI (d_{0}) - 10 nlo g_{10} (\frac{d}{d_{0}}) + X_{δ}$ (22)

where $d_{0}$ is the reference distance, and n denotes the path loss exponent. In this simulation experiment, we choose $RSSI (d_{0}) = - 45 dBm$ , $n = 2$ , and $δ = 0.5$ .

The error is defined as the Euclidean distance between the estimated target coordinate $(x_{e}, y_{e}, z_{e})$ and the actual target coordinate $(x_{t}, y_{t}, z_{t})$ , denoted the $i th$ error result as $ρ_{i}$

$ρ_{i} = \sqrt{{(x_{e} - x_{t})}^{2} + {(y_{e} - y_{t})}^{2} + {(z_{e} - z_{t})}^{2}}$ (23)

The mean error in location l is the average of the errors by 216 sample points uniformly scattered in the hexahedron, denoted as

$\bar{ρ_{l}} = \frac{\sum_{i = 1}^{n} ρ_{i}}{n}$ (24)

In the default case, the reference tags are placed every 0.5 m on the ceiling of the hexahedron, eight readers are placed on the upper side, and every reader has 16 power levels.

Influence factors

Factors that affect the localization results include the number of power levels $(K)$ , the number of readers $(N)$ , the density of reference tags $(Δ)$ , and the location of target tag. In the following, we will analyze the factors, respectively, and compare the mean error of the proposed method with the method in research²⁹ (denote as partial least squares (PLS) method) and the method in research³⁰ (denote as APM method) in the same environment.

The number of power levels

In this experiment, we, respectively, divide the power strength into 4, 8, 12, and 16 levels.

As shown in Figure 6, the mean error of the proposed method always decreases when the number of power levels increases. As the number of power levels increases, the signal strength between reference tags and target tag in $K_{\min}$ and $K_{\max}$ is closer, then the estimated distance between reader and target tag becomes more accurate.

Figure 6.

Error affected by the number of power levels $(K)$ .

Density of reference tags

The density of reference tags $(Δ)$ is divided into 0.5, 0.8, 1.4, and 2.0 m per reference tag.

As shown in Figure 7, the average localization error increases when the density of reference tags decreases. As density of reference tags decrease, the reference tags in $K_{\min}$ and $K_{\max}$ are less, then the estimated distance between reader and target tag becomes less accurate.

Figure 7.

Error affected by density of reference tags $(Δ)$ .

The number of readers

In this simulation, we have studied the influence of the number of readers. The readers are placed in the corners and the edges (when the number of readers is 9, one is placed in the center of the ceiling).

As shown in Figure 8, the average error decreases when the number of readers $(N)$ increases, when $N = 8$ , the average error is smaller than the error when $N = 6$ , but the average error almost stays the same when $N = 9$ , hence $N = 8$ is the best choice.

Figure 8.

Error affected by the number of readers $(N)$ .

Location of target tag

In this part, we have studied the influence of the location of target, including the x, y, and z axes. First, we fix the z value to 5 and change the x and y values in sequence. Then, we made the x, y fixed to 5 and change the z values. The experiment results are the average errors of repeated tests.

As shown in Figures 9 and 10, the average error increases significantly when target tag is near the upper side, lower side, or the boundaries of the hexahedron. When the target tag is near the lower side or the boundaries, the decrease in the reference tags leads to the increase in the error. When the target tag is near the upper side, since the existence of error, the number of complex solutions (invalid solutions) increases when using sphere trilateration to get feasible solutions, then the average error increases.

Figure 9.

Error distribution when $z = 5$ .

Figure 10.

Error distribution when $x = 5, y = 5$ .

Comparison with other methods

Compare influence factors

In this part, we have compared the proposed method with PLS method in research²⁹ and APM method in research.³⁰ We have studied the influence of the number of power levels, the density of reference tags, and the number of readers by the three methods, respectively.

As shown in Figure 11, the mean errors of the three methods keep decreasing when power levels $(K)$ increase; as shown in Figure 12, the average localization errors of the proposed method and PLS method increase when the density of reference tags $(Δ)$ decreases and APM method stays the same; as shown in Figure 13, the average errors of the three methods decrease as the number of readers $(N)$ increases. Comparing with PLS method and APM method, the mean error of the proposed method is smaller, for example, when $K = 16$ , $Δ = 0.5$ , and $N = 8$ , the mean error of the proposed method is 0.19 m while the mean error of PLS method is 0.83 m and APM method is 0.55 m.

Figure 11.

Compare the number of power levels $(K)$ .

Figure 12.

Compare the density of reference tags $(Δ)$ .

Figure 13.

Compare the number of readers $(N)$ .

Cumulative distribution function curves

We used the cumulative distribution function (CDF) curves of localization error to compare our proposed method with the PLS method and APM method. All of the results are gotten from the same case, where the number of readers is 8, the number of power level is 16, and tag density is 0.5 m.

From Figure 14, we have learnt that the localization error of the proposed method is less than 0.35 m with the probability of 90% and all of the errors are under 0.5 m. While the APM method has an error of localization less than 0.65 m with the probability of 90% and the same probability of PLS method is 1.3 m. The results show that the proposed method is better than both the APM method and the PLS method in the same simulation environment.

Figure 14.

CDF curves of proposed method and other methods.

Experimental evaluation

Experiment setting

To further demonstrate the appropriateness of the proposed method compared to the other methods, we have conducted several realistic experiments in Guangzhou Research Institute of O-M-E Technology.

The reader model we used is Alien ALR-9900+,³⁴ the main working frequency is 920 MHz, the maximum power strength is 30.7 dBm and the minimum power strength is 15.7 dBm. We use two kinds of antennas, their model are Alien ALR-8696-C³⁵ (8.5 dBic gain) and ALR-9611-CR³⁶ (6 dBic gain). Our experiment setting is shown in Figure 15, we finish the experiments under a realistic environment with 4.00-m long, 3.63-m wide, and 2.60-m high. The computer communicates with the ALR-9900+ reader via transmission control protocol/Internet protocol (TCP/IP). We have written a C# program to adjust the power level and get the tag ID and corresponding RSSI values. The power level can be changed in millisecond level. The readers and the reference tags are deployed in the ceiling of the room. The numbers of readers we use are 4, 6 and 8. The passive reference RFID tags are placed every 0.33 m and the total number of reference tags is 72.

Figure 15.

Experimental setup in realistic environment.

Experiment results

In the experiment, we have divided the power strength into 4, 6, 8, and 16 levels and the number of readers into 4, 6, and 8. All of the results are gotten from the 16 sample points distributed uniformly in the space. The performance comparison results between the proposed method, the APM method, and the PLS method are presented in Figure 16. As shown in Figure 16, the mean errors of the three methods decrease with the increase in the number of power levels and the number of readers. When the number of power levels is 16 and the number of readers is 8, the average localization accuracy of all the sample points with PLS method, APM method, and the proposed method is, respectively, 0.75, 0.63 and 0.28 m. The proposed method enhances the precision of localization accuracy by 62% over PLS method and 55% over APM method.

Figure 16.

Experiment results with no sheltering: (a) the number of power levels (K) and (b) the number of readers (N).

In summary, the proposed method can provide lower localization errors than PLS method and APM method in the same circumstance.

Impacts of sheltering

The environment of the working space is not always the same, and as the time passes, the tags may be sheltered by water or metal. In order to study the impacts of sheltering, we have conducted several experiments in realistic environment. We have put several water bottles or metal between the readers and the tags, respectively, and repeat the experiments to get the results. The impact results of water sheltering are shown in Figure 17; when the number of power levels is 16 and the number of readers is 8, the mean errors of PLS method, APM method, and the proposed method are, respectively, 0.90, 0.79 and 0.30 m. Compared to the results of no sheltering (Figure 16), the average localization accuracy of PLS method, APM method, and the proposed method decreases 20%, 25%, and 7%, respectively.

Figure 17.

Experiment results with water sheltering: (a) the number of power levels (K) and (b) the number of readers (N).

When the sheltering factor changes to metal sheltering with other factors unchanged, the results are shown in Figure 18. The mean errors of PLS method, APM method, and the proposed method are, respectively, 0.95, 0.88, and 0.33 m. Compared to the results of no sheltering (Figure 16), the average localization accuracy of PLS method, APM method, and the proposed method decreases 27%, 39%, and 18%, respectively; compare with water sheltering (Figure 17), the average accuracy decreases 6%, 11%, and 10%, respectively.

Figure 18.

Experiment results with metal sheltering: (a) the number of power levels (K) and (b) the number of readers (N).

We can conclude from the above that the proposed method is more adaptable to the environment than PLS method and APM method.

Conclusion and future work

In this article, we have proposed a novel RFID-based 3D localization method. In the simulation section, we have studied the influence factors of the localization results, including the number of power levels, the number of readers, the density of reference tags, and the location of target tag. Besides, we have given simulation results to compare our proposed method with the other methods. Moreover, we have carried out testbed experiment to evaluate our proposed method. Both the simulation and testbed experiment results have shown that, compared to other methods, our proposed method can reach a higher positional accuracy using less RFID readers and it is more robust to the environment.

However, the proposed method does not solve the problem that the error increases significantly with the decrease in the distance between the target tags and the reader. Our further work will focus on solving these problems and we will also carry out experiments to test the robustness of the proposed algorithm when varying the hardware.

Footnotes

The authors thank the anonymous reviewers and editors for their valuable comments on improving this paper.

Handling Editor: Luca Catarinucci

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research,authorship,and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research,authorship,and/or publication of this article: This paper was partially supported by the Engineering and Technology Research Center of Guangdong Province for Logistics Supply Chain and Internet of Things (Project No. GDDST[2016]176);the Provincial Science and Technology Project in Guangdong Province (Project No. 2013B09020 0055);the Key Laboratory of Cloud Computing for Super-integration Cloud Computing in Guangdong Province (Project No. 610245048129);and the Engineering and Technology Research Center of Guangdong Province for Big Data Intelligent Processing (Project No. GDDST[2013]1513-1-11).

References

Hahnel

Burgard

Fox

et al . Mapping and localization with RFID technology. In: 2004 IEEE international conference on robotics and automation, New Orleans, LA, 26 April–1 May 2004, vol. 1, pp.1015–1020. New York: IEEE.

Sanpechuda

Kovavisaruch

. A review of RFID localization: applications and techniques. In: 5th international conference on electrical engineering/electronics, computer, telecommunications and information technology, Krabi, Thailand, 14–17 May 2008, vol. 2, pp.769–772. New York: IEEE.

Wang

Tzeng

NF.

RFID-based 3-D positioning schemes. In: IEEE INFOCOM 2007-26Th IEEE International Conference on Computer Communications, Barcelona, 6–12 May 2007, pp.1235–1243. New York: IEEE.

Hofmann-Wellenhof

Lichtenegger

Collins

Global positioning system: theory and practice. Vienna: Springer-Verlag, 2012.

LaMarca

Chawathe

Consolvo

et al . Place lab: device positioning using radio beacons in the wild. In: Third International Conference, PERVASIVE 2005, Munich, 8–13 May 2005, pp.116–133. New York: Springer.

Bargh

de Groote

Indoor localization based on response rate of bluetooth inquiries. In: Proceedings of the ACM international workshop on mobile entity localization and tracking in GPS-less environments, MELT, San Francisco, CA, USA, 19 September 2008, pp.49–54. ACM.

Cho

Jang

Baek

Practical localization system for consumer devices using ZigBee networks. IEEE T Consum Electr 2010; 56(3): 1562–1569.

Abreu

Garganta

Reis

et al . Real-time wireless location and tracking System with motion pattern detection. In: Yussof

(ed.) Robot localization and map building. Rijeka: InTech, 2010.

Dardari

Errico

Roblin

et al . Ultrawide bandwidth RFID: the next generation?P IEEE 2010; 98(9): 1570–1582.

10.

Henriques Abreu

Xavier

Castro Silva

et al . Using Kalman filters to reduce noise from RFID location system. Sci World J 2014; 2014: 796279.

11.

Santiago

Sousa

Estriga

et al . Survey on team tracking techniques applied to sports. In: 2010 international conference on autonomous and intelligent systems (AIS), Povoa de Varzim, 21–23 June 2010, pp.1–6. New York: IEEE.

12.

Inc

Comparison of wireless indoor positioning technologies—an Ekahau whitepaper, 2005, https://www.rfidsolutionsonline.com/doc/comparison-of-wireless-indoor-positioning-tec-0001

13.

Baglietto

Cannata

Capezio

et al . A multi-robot coordination system based on RFID technology. In: 2009 international conference on advanced robotics, Munich, 22–26 June 2009, pp.1–6. New York: IEEE.

14.

Munishwar

Singh

Wang

et al . On the accuracy of RFID-based localization in a mobile wireless network testbed. In: 2009 IEEE international conference on pervasive computing and communications, Galveston, TX, 9–13 March 2009, pp.1–6. New York: IEEE.

15.

Wada

Uchitomi

Ota

et al . A novel localization scheme for passive RFID tags; communication range recognition (CRR). In: 2009 IEEE international conference on RFID, Orlando, FL, 27–28 April 2009, pp.163–169. New York: IEEE.

16.

Wilson

Prashanth

Aghajan

. Utilizing RFID signaling scheme for localization of stationary objects and speed estimation of mobile objects. In: 2007 IEEE international conference on RFID, Grapevine, TX, 26–28 March 2007, pp.94–99. New York: IEEE.

17.

Alsalih

Almaaitah

Alkhater

RFID localization using angle of arrival cluster forming. Int J Distrib Sens N 2014; 2014: 269596.

18.

Liu

Lau

et al . LANDMARC: indoor location sensing using active RFID. Wirel Netw 2004; 10(6): 701–710.

19.

Zhou

Zhang

Zhou

HL.

Localization of pallets in warehouses using passive RFID system. J Cent South Univ 2015; 22(8): 3017–3025.

20.

Yang

Chen

et al . Tagoram: real-time tracking of mobile RFID tags to high precision using COTS devices. In: 20th annual international conference on mobile computing and networking, Maui, HI, 7–11 September 2014, pp.237–248. New York: ACM.

21.

Zhao

Liu

et al . An improved k-NN algorithm for localization in multipath environments. EURASIP J Wirel Comm 2014; 2014(1): 208.

22.

Kuo

Chang

Intelligent RFID positioning system through immune-based feed-forward neural network. J Intell Manuf 2015; 26(4): 755–767.

23.

Lin

Chen

Lin

. An indoor location identification system based on neural network and genetic algorithm. In: 2011 3rd international conference on awareness science and technology (iCAST), Dalian, China, 27–30 September 2011, pp.193–198. New York: ACM.

24.

Soltani

Motamedi

Hammad

Enhancing cluster-based RFID tag localization using artificial neural networks and virtual reference tags. In: 2013 international conference on indoor positioning and indoor navigation (IPIN), Montbeliard, 28–31 October 2013, pp.1–10. New York: IEEE.

25.

Wang

Shi

et al . Indoor positioning technique by combining RFID and particle swarm optimization-based back propagation neural network. Optik 2016; 127(17): 6839–6849

26.

Hightower

Want

Borriello

SpotON: an indoor 3D location sensing technology based on RF signal strength. Technical Report, University of Washington, Seattle, WA, 18February2000.

27.

Bouet

Pujolle

. A range-free 3-D localization method for RFID tags based on virtual landmarks. In: 2008 IEEE 19th international symposium on personal, indoor and mobile radio communications, Cannes, 15–18 September 2008, pp.1–5. New York: IEEE.

28.

Han

Zhao

Cheng

et al . Improving accuracy for 3D RFID localization. Int J Distrib Sens N 2012; 8: 184–195.

29.

Maneesilp

Wang

et al . RFID support for accurate 3D localization. IEEE T Comput 2013; 62(7): 1447–1459

30.

Almaaitah

Ali

Hassanein

et al . 3D passive tag localization schemes for indoor RFID applications. In: 2010 IEEE international conference on communications (ICC), Cape Town, South Africa, 23–27 May 2010, pp.1–5. New York: IEEE.

31.

Manolakis

DE.

Efficient solution and performance analysis of 3-D position estimation by trilateration. IEEE T Aero Elec Sys 1996; 32(4): 1239–1248.

32.

Nocedal

Wright

Numerical optimization. New York: Springer, 2006.

33.

The MathWorks I. Matrix laboratory (MATLAB), 2015, http://www.mathworks.com/

34.

Corp

AT.

Alien technology ALR-9900 enterprise RFID reader, 2013, http://www.alientechnology.com/wp-content/uploads/Alien-Technology-ALR-9900-Enterprise-RFID-Reader.pdf

35.

Corp

AT.

Alien technology ALR-8696-c antenna, 2014, http://www.alientechnology.com/wp-content/uploads/Alien-Technology-ALR-8696-C-Antenna.pdf

36.

Corp

AT.

Alien technology ALR-9611-cr antenna, 2014, http://www.legacyglobal.com/Alien-Technology-Part-Number-ALR-9611-CR