1 Introduction

In backscatter radio frequency identification (RFID) systems, the objects that have to be identified are equipped with RFID transponders (tags). The tags modulate the backscattered signal that was sent by an interrogator (reader) [1]. Especially, in applications with a plurality of objects to be identified, the use of batteryless, or rather passive tags is very economical, as they are low cost and low maintenance devices [1]. The possibility of using the backscatter RFID technology not only for the identification of objects but also for the localization of these objects, has led to the proposal of several new applications in recent years. A small excerpt of the possible applications is the localization of pallets in warehouses [2], the localization of misplaced books in libraries [3], and the localization of every day products in smart homes [4]. However, the performance of a backscatter RFID system and its localization capabilities strongly depend on the environment where it is used in. Operating RFID systems in metallic environments suffers from multipath propagation of the radio signals. Consequences of this multipath propagation are the distorted power transfer to the passive tags at some regions as well as the reception of distorted tag signals at the reader leading to a distorted localization. Several localization approaches have been proposed in recent years to combat these distortions that differ in terms of localization accuracy, implementation effort, and material costs. These approaches reach from using systems (e.g., ultra wide band system [5], frequency modulated continuous wave system [6]) parallel to an RFID system over the use of reference tags [7] and multiple input multiple output systems [8] up to RFID signals with superimposed spread spectrum signals [9].

Our approach is to implement a narrow band and low cost UHF RFID localization system that does not rely on the use of reference tags. A channel measurement based system hardware dimensioning assures a sufficient power transfer to the tags and thus a reliable communication in a selected environment. In order to combat distorted localization in multipath environments, the localization system exploits the use of three operating frequencies. In particular, this paper presents an RFID reader for tag localization in an AVL engine test bed. The starting point for the reader implementation is a measurement based characterization of the radio channel in this engine test bed. Generally, this approach allows a decoupled examination of the radio channel and the RFID system components and thus the exploration of the individual influence of various RFID system components (different tag and reader parameters) on the system performance in selected environments. For the reader implementation presented here, the channel measurements are used to estimate the minimum reader hardware requirements—with respect to the radio signal power (minimum transmit power, etc.)—that are necessary to perform a reliable tag localization in the engine test bed. The final reader implementation exploits phase difference of arrival (PDoA) techniques for the localization of RFID tags as they are more robust in multipath environments compared to state-of-the-art techniques that are based on the received signal strength [10]. For the exploitation of the PDoA techniques, the reader antennas are setup in a single input multiple output (SIMO) configuration. The core of the reader is a software defined radio (SDR) that makes the reader flexible in order to achieve a rapid prototyping and a fast verification of the implementation.

The paper is structured as follows. Section 2 describes the communication links of the used SIMO RFID system. Section 3 shows the RFID system feasibility in the AVL engine test bed by assuming state-of-the-art system components. The reader implementation is presented in Sects. 4 and 5 and is based on the evaluation of the channel measurements in the engine test bed. Sections 6 and 7 present the functional verification of the implemented reader and the conclusions, respectively.

2 SIMO RFID system

A backscatter RFID system can be classified into three communication links, the forward link, the backward link, and the backscatter link. In the forward link, the reader consecutively sends commands and a continuous wave (CW) signal to the tag that consists of an antenna and a chip. The CW signal provides the power for the tag that is necessary to supply its chip. Once the tag has received enough power, it starts to modulate the backscattered signal with its information depending on the reader command, e.g., its identification number ID. This link from the tag to the reader is denoted as backward link. If the backscattered tag signal can be correctly interpreted by the reader, the RFID system is reliable. The backscatter link describes the combination of the forward link and the backward link (reader-tag-reader) [1]. The reliability of a backscatter RFID system strongly depends on the signal power within the mentioned communication links. The minimum power of the CW signal received at the tag is limited by the sensitivity of the tag chip (chip sensitivity). The minimum power of the tag signal received at the reader is limited by the sensitivity of the reader receiver (reader sensitivity) [11].

As mentioned above, the UHF RFID reader is based on a SIMO antenna configuration, consisting of one transmitting antenna TX and two receiving antennas RX1, RX2 of the type Motorola AN480 [12] (see Fig. 1). As a result of the SIMO configuration, the communication links expand to one forward link FL, two backward links \(\mathrm{BL}_{1}\), \(\mathrm{BL}_{2}\) and two backscatter links \(\mathrm{BSL}_{1}, \mathrm{BSL}_{2}\).

Fig. 1.
Fig. 1.
Full size image

SIMO antenna configuration: The spacing of TX and RX1 is 60 cm, while the spacing of TX and RX2 is 90 cm

3 System reliability

The reliability of implementing a SIMO RFID system performing tag localization in the engine test bed at AVL is in the following investigated with respect to the signal power and state-of-the-art RFID system parameters. The investigation is based on channel measurements that characterize the communication links within this challenging multipath environment. In particular, the system reliability is investigated with respect to the measured power availability in the communication links at 890 MHz [11, 13]. Figure 2 shows the floor plan of the AVL engine test bed (test bed 1) and a neighboring engine test bed (test bed 2). The reader antennas (TX, RX1, RX2) have been mounted on a window to the control room of test bed 1 at a height of 1.5 m. For the measurement of the communication links, a tag antenna has been positioned at several positions within test bed 1 (positions 1 to 30) at the same height as the reader antennas. To investigate the interference of tags that are located in neighboring engine test beds, additional measurements have been performed where the tag antenna has been positioned in test bed 2 (positions 31 to 34). Two different scenarios have been considered for the measurement of the power in the communication links. For the tag antenna, a custom built patch antenna—with typical tag antenna gain—has been used in the tag scenario, while a Motorola AN480 antenna—the same as for the reader antennas—has been used as the tag antenna in the reference scenario. The measurements of the reference scenario are used to obtain a best case scenario in test bed 1 and a worst case scenario in test bed 2.

Fig. 2.
Fig. 2.
Full size image

AVL engine test bed floor plan: Test bed 1 consists of metallic walls and several metallic objects like the engine test bench, boxes and objects. The reader antennas are mounted on the window to the control room of test bed 1. For the measurements, the tag antenna has been positioned at several points in test bed 1 (positions 1 to 30). Also, interference measurements have been performed by positioning the tag antenna in the neighboring test bed 2 (positions 31 to 34)

The channel measurements have been performed by measuring the scattering (S) parameters with a vector network analyzer (VNA) [14]. For every tag position, the S parameters have been measured 20 times in a frequency range from 700 MHz to 1200 MHz. The reference plane for the measurements has been set to the input of the antennas. Consequently, the antennas are part of the measured channel. The communication links of the system are described by the transmission coefficients of the S parameters. The amount of power that is transmitted at a certain frequency from one to another antenna, or rather the channel gain can be calculated as the squared magnitude of the transmission coefficients. The backscatter link channel gain \(g_{\mathrm{BSL}}\) is the sum of the forward link channel gain \(g_{\mathrm{FL}}\) and the backward link channel gain \(g_{\mathrm{BL}}\) (\(g_{\mathrm{BSL}}|_{\text{dB}}=g_{\mathrm{FL}}|_{\text{dB}} +g_{\mathrm{BL}}|_{\text{dB}}\)) [11].

3.1 Evaluation method

The system reliability is determined by combining the knowledge of the measured communication links and specified RFID system parameters [13]. More precisely, the measured channel gain of the communication links at 890 MHz is compared to channel gain thresholds that are defined by the RFID system components (reader and tag parameters). The evaluation is based on 600 (30 positions in test bed 1 times 20 repetitions) measurement points in the reference and tag scenarios and 80 (4 positions in test bed 2 times 20 repetitions) measurement points in the interference reference scenario. Here, the feasibility of the system is evaluated with respect to state-of-the-art RFID system parameters. For the RFID reader, a transmission power of \(P_{\mathrm{TX}}=30~\text{dBm}\) and a sensitivity of \(T_{\mathrm {RX}}=-105~\text{dBm}\) for both receivers are assumed (Zebra FX9500 RFID Reader [15]). The tag provides a chip sensitivity of \(T_{\mathrm{Chip}}=-17.5~\text{dBm}\) (NXP UCODE G2iM [16]), a modulation efficiency of \(\eta=-7~\text{dB}\) (amplitude modulation [17]) and a power transmission coefficient of \(\tau=0~\text{dB}\). \(\tau=0~\text{dB}\) means a perfect matching between tag antenna and tag chip. The measured matching of the antennas (reflection coefficients) allows the latter assumption. The antenna matching is \(-13~\text{dB}\) in the tag scenario and \(-20~\text{dB}\) in the reference scenario. The above mentioned state-of-the-art parameters lead to a forward link threshold \(F_{\mathrm{Th}}\) and to a backscatter link threshold \(B_{\mathrm{Th}}\) for both backscatter links of [11]:

$$ F_{\mathrm{Th}}=T_{\mathrm{Chip}}-\tau- P_{\mathrm{TX}}=-47.5~\text{dB}\quad \text{and} $$
(1)
$$ B_{\mathrm{Th}}=T_{\mathrm{RX}}-\eta- P_{\mathrm{TX}}=-128~\text{dB}. $$
(2)

The RFID system is forward link limited, if the forward link channel gain is lower than \(F_{\mathrm{Th}}\). The system is denoted as backward link limited, if the forward link channel gain is higher than \(F_{\mathrm{Th}}\), but the backscatter link channel gain is lower than \(B_{\mathrm{Th}}\) [10].

3.2 Evaluation

The system reliability of the two mentioned scenarios is evaluated by means of the reliability of the individual communication links (FL, BSL1, BSL2). The reliability of the communication links is indicated in percent and is illustrated by means of bar charts (see Fig. 3). The forward link is 100 % reliable, if \(g_{\mathrm{FL}}\) is higher than \(F_{\mathrm{Th}}\), while the backscatter link is 100 % reliable, if \(g_{\mathrm{BSL}}\) is higher than \(B_{\mathrm{Th}}\), for every measured tag position. However, the system is only 100 % reliable, if FL, BSL1, and BSL2 are 100 % reliable and thus, the system is neither forward link limited nor backward link limited.

Fig. 3.
Fig. 3.
Full size image

System feasibility: The system feasibility of the reference and the tag scenario is depicted by the reliability of their communication links. The communication links of the reference scenario (Ref) are 100 % reliable, while the system is forward link limited in the tag scenario (Tag). The possibility for detecting an interfering tag is 25 % in the worst case (Ref Interference). The forward link limitation at some positions in the room in the tag scenario can be overcome for example by using semi passive tags (Tag Semi)

Figure 3 shows the reliability of the communication links for the reference (Ref, Ref Interference) and the tag (Tag) scenario. The communication links in the reference scenario in test bed 1 (Ref) are 100 % reliable and thus, the system is 100 % reliable in this best case scenario. The interference measurements (Ref Interference) exhibit 25 % reliability of the forward link and 100 % reliability of the backscatter links. This results in the detection of an interfering tag with a probability of 25 % in this worst case. However, by reducing the reader transmission power by 2 dB (\(P_{\mathrm{TX}}=28~\text{dBm}\)), no interferences occur anymore, while the system still operates reliable in test bed 1. In the tag scenario (Tag), the forward link is 77 % reliable, while both backscatter links are 100 % reliable. Thus, the tag does not receive enough power at some positions to respond. This problem can be overcome by using battery-assisted, or rather-semi passive RFID tags with higher chip sensitivities. Figure 3 shows that using semi-passive tags (Tag Semi) with a chip sensitivity of \(T_{\mathrm{Chip}}=-40~\text{dBm}\) (Intelleflex XC3 IC [18]) leads to a 100 % reliable system. However, using battery-assisted tags would increase the acquisition and maintenance cost of the localization system. The future goal is to exploit the advantages of passive tags. Thus, the placement of the reader antennas has to be chosen carefully to increase their coverage in the test bed.

Additionally, the system reliability is investigated with respect to the influence of varying the tag antenna height and the influence of a running test engine. The respective measurements have been repeated at 5 tag positions (1, 17, 18, 19, 27) in the tag scenario. First, the height of the tag antenna has been reduced to 1 m (Tag 1 m) and second, a test engine has been run in test bed 1 (Tag Engine). Figure 4 shows a comparison of the link reliability in the tag scenario with the three different measurement conditions, i.e., Tag, Tag 1 m and Tag Engine at the 5 tag positions. For Tag and Tag 1 m, the tag does not receive enough power at one position to respond (80 % reliability of the forward links), while no limitations can be observed for Tag Engine. The differences can be explained by slightly differing tag antenna positions caused by the manual positioning. However, no degradation of the system can be observed by changing the measurement conditions in the tag scenario.

Fig. 4.
Fig. 4.
Full size image

Link reliability comparison of measurements with different conditions in the tag scenario at 5 tag positions (1, 17, 18, 19, 27): Varying the tag antenna height and the influence of a running test engine do not degrade the link reliability. The differences in the forward links can be explained by slightly differing tag antenna positions due to manual positioning of the tag antenna

4 Reader hardware implementation

The previous section investigates the system reliability in the AVL engine test bed with respect to state-of-the-art RFID system parameters, while this section presents the SIMO UHF RFID reader hardware implementation based on the performed channel measurements. For the implementation, some hardware requirements have to be considered in order to perform a reliable localization in the AVL engine test bed. On the one hand, the reader has to fulfill requirements with respect to the signal power in the communication links, i.e., the reader has to provide a sufficient transmission power and sensitivity. On the other hand, the reader has to fulfill requirements in order to perform localization based on phase difference of arrival (PDoA) techniques [10]. The core of the reader is chosen to be an SDR, to guarantee a fast implementation of the hardware and the PDoA algorithms. In the following, a estimation of the requirements for the SDR and additional hardware components is presented. Especially, the tag scenario using semi-passive tags with a chip sensitivity of \(T_{\mathrm{Chip}}=-40~\text{dBm}\) is assumed for these considerations.

4.1 Minimum transmission power

A minimum transmission power \(P_{\mathrm{TX},\mathrm{min}}\) of the reader is required to provide the tag with sufficient power at all positions in test bed 1. \(P_{\mathrm{TX},\mathrm{min}}\) is determined by the minimum channel gain of the forward link \(g_{\mathrm{FL},\mathrm{min}}\) at 890 MHz and the tag parameters, i.e., the chip sensitivity and the power transmission coefficient. The minimum transmission power is [11],

$$ P_{\mathrm{TX},\mathrm{min}}=T_{\mathrm{Chip}}-\tau- g_{\mathrm{FL},\mathrm{min}}=24.5~\text{dBm}, $$
(3)

with a minimum measured channel gain of \(g_{\mathrm{FL},\mathrm{min}}=-64.5~\text{dB}\) and a power transmission coefficient of \(\tau=0~\text{dB}\).

4.2 Minimum reader sensitivity

To correctly detect the backscattered tag signals from all tag positions in test bed 1, the reader has to provide a minimum reader sensitivity \(T_{\mathrm{RX},\mathrm{min}}\). \(T_{\mathrm{RX},\mathrm{min}}\) is determined by the minimum measured channel gain of the backscatter links \(g_{\mathrm{BSL},\mathrm{min}}\), the reader transmission power, and the modulation efficiency of the tag. For the following, only the minimum transmission power of \(P_{\mathrm{TX},\mathrm{min}}=24.5~\text{dBm}\) is used for the calculations to obtain a lower bound of the reader requirements. With a minimum measured channel gain of \(g_{\mathrm{BSL},\mathrm{min}}=-114~\text{dB}\) and a modulation efficiency of \(\eta=-7~\text{dB}\), the minimum reader sensitivity is calculated as [11],

$$ T_{\mathrm{RX},\mathrm{min}}=P_{\mathrm{TX},\mathrm{min}} + g_{\mathrm{BSL},\mathrm{min}} + \eta =-96.5~\text{dBm}. $$
(4)

Providing this sensitivity means that signals with a minimum power of \(-96.5~\text{dBm}\) can be resolved at the reader. From this it follows that the weakest signals at the reader receiver have an amplitude of about 150 μV, assuming a \(50~\Omega\) system and a gain of 31.5 dB of a low noise amplifier in the receiving chain of the reader [1]. Thus, at least 12 bit analog to digital converters (ADCs) are necessary to detect a change of the amplitude modulated tag signal at the receiver, when assuming a full scale range of 1 V [19]. This is only a rough estimation not taking any losses in the receiving chains into account.

4.3 PDoA-related requirements

The SDR has to fulfill some requirements for the exploitation of PDoA techniques. As the SIMO configuration of the reader antennas is necessary to perform spatial domain (SD)-PDoA techniques (see Sect. 5.2), the SDR has to provide at least one transmitting chain and two receiving chains. For detecting phase differences of the signals that are received at the two receiving antennas, it is important that the receiving chains provide phase coherence and phase alignment [10].

4.4 Final reader setup

Based on the mentioned hardware requirements, a SDR from National Instruments (NI USRP-2942R [20]) has been chosen that fulfills almost all requirements. It is a \(2 \times 2\) MIMO system and hence provides two separate transmitting chains and two separate receiving chains. It can operate in a full duplex mode (transmitting and receiving at the same time) and provides 14 bit ADCs. The maximum output power of the SDR is 17 dBm (measured). The two receiving chains are phase coherent, but not phase aligned. However, this phase misalignment can be corrected by an initial calibration (see Sect. 5.1). The final RFID reader is composed of the NI USRP-2942R SDR and the necessary additional hardware components to meet the requirements (see Fig. 5). A switch matrix is used to perform an initial automatic calibration of the two receiver chains of the SDR, i.e., to measure the phase misalignment of the receiving chains. A power amplifier increases the transmission signal to a sufficient power level. A laptop computer controls the hardware components of the reader and processes the raw inphase (I) and quadrature (Q) data received from the SDR. One transmitting antenna and two receiving antennas complete the setup of the SIMO UHF RFID reader.

Fig. 5.
Fig. 5.
Full size image

Final SIMO UHF RFID reader setup: The reader is composed of the SDR (NI USRP-2942R), the switch matrix, the power amplifier, one transmitting antenna and two receiving antennas and is controlled by the laptop computer (reader control)

5 Reader software implementation

The previous section presented the reader implementation with respect to hardware aspects. This section deals with the procedure of the initial calibration, the implementation of the PDoA algorithms, and the sensor data fusion. For the two dimensional localization of tags, two PDoA algorithms are implemented that estimate the direction (SD-PDoA) and the range (frequency domain (FD)-PDoA) of the tag, respectively. The definitions of the estimators are based on the assumption that the tag is far away from the reader, compared to the spacing of the receiving antennas, the tag is at rest within the time period of an estimation, and the tag is frequency independent within the range of the operating frequencies [10].

5.1 Initial calibration

As stated in Sect. 4.4, the receiving chains are not phase aligned. However, this phase misalignment can be measured in advance and afterwards be incorporated into the localization estimates. The phase misalignment measurement is done in an automatic initial calibration phase. In this phase, the switch matrix switches the transmitting port via a splitter to both receiving ports. Thus, the same CW signal is provided to the two receiving chains of the SDR and allows to determine the phase misalignment. Figure 6 exemplarily shows the baseband IQ vectors \(\mathrm{RX}_{1,\mathrm{BB}}\) and \(\mathrm{RX}_{2,\mathrm{BB}}\) of the received signals at both receiving ports. The angle \(\beta\) between \(\mathrm{RX}_{1,\mathrm{BB}}\) and \(\mathrm{RX}_{2,\mathrm{BB}}\) is equivalent to the phase misalignment of the receiving chains. The system is calibrated by rotating \(\mathrm {RX}_{2,\mathrm{BB}}\) by \(\beta\) towards \(\mathrm{RX}_{1,\mathrm{BB}}\) and phase differences can be correctly detected.

Fig. 6.
Fig. 6.
Full size image

Determination of the phase misalignment: When receiving the same CW signal at both receivers, the angle \(\beta\) between the baseband vectors \(\mathit{RX}_{\mathit{1},\mathit{BB}}\) and \(\mathit{RX}_{\mathit{2},\mathit{BB}}\) is equivalent to the phase misalignment

5.2 Localization estimators

The direction estimator \(\varTheta\) is based on different path lengths of the backscattered tag signal received at the two receiving antennas at a certain frequency. The different path lengths cause a phase difference \(\Delta\varphi_{\mathrm{D}}\) of the two signals that is related to the direction of the tag, where the reference point is in the center of the receiving antennas. \(\varTheta\) can be calculated by using a trigonometric relationship as [10],

$$ \varTheta=\cos^{-1}\frac{c}{2\pi f}\frac{\Delta\varphi_{\mathrm{D}}}{d}, $$
(5)

where \(c\) is the light velocity, \(f\) the operating frequency and \(d\) the distance between the receiving antennas.

The range estimator \(R\) is based on the phase difference \(\Delta\varphi_{\mathrm{R}}\) of two consecutively received tag signals at different operating frequencies at one receiving antenna. Assuming a monostatic antenna configuration, i.e., one antenna for transmission and reception, \(R\) can be calculated as [10],

$$ R=\frac{c}{4\pi}\frac{\Delta\varphi_{\mathrm{R}}}{\Delta f}, $$
(6)

where \(\Delta f\) is the frequency difference of the operating frequencies. For a monostatic antenna configuration, the estimator \(R\) is based on the fact that half of \(\Delta\varphi_{\mathrm{R}}\) transpires in the backward link (see Fig. 7a). This fact does not hold for the SIMO RFID reader presented here and Eq. (6) would lead to imprecise estimates. As transmitting and receiving antennas are in a bistatic configuration, the estimation of the range expands to the calculation of the side \(R_{\mathrm{b}}\) of a triangle (see Fig. 7). Figure 7b depicts the triangle formed by TX, the tag, and RX and the sides \(a\), \(R_{\mathrm{b}}\), and \(c\), where RX is the reference point in the center of \(\mathrm{RX}_{1}\) and \(\mathrm{RX}_{2}\). We assume that the sum of the triangle sides \(a\) and \(R_{\mathrm{b}}\) is approximately given by \(2\cdot R\), i.e., \(a+R_{\mathrm{b}}\approx2\cdot R\) (as the reference point is not exactly at a receiving antenna). By incorporating the knowledge of the tag direction estimate \(\varTheta\) that is an angle of the triangle, the range in the bistatic configuration \(R_{\mathrm{b}}\) can be calculated by rearranging the law of cosines as,

$$ R_{\mathrm{b}}=\frac{c^{2}-4R^{2}}{2c\cos\varTheta-4R}. $$
(7)
Fig. 7.
Fig. 7.
Full size image

Range estimation: The range estimation in the monostatic configuration is based on the bisection of the distance TX/RX-Tag-TX/RX. In the bistatic configuration, the range estimation is based on the calculation of side \(R_{\mathrm{b}}\) of the triangle, where the direction estimate \(\varTheta\) is an angle of the triangle

5.3 Sensor data fusion

The reader is implemented to operate at three different operating frequencies \(f_{0}\), \(f_{1}\), and \(f_{2}\) in order to make the tag localization more robust in multipath environments [21]. The reader consecutively sends commands at the different operating frequencies and stores the received data for the post processing. This procedure leads to three direction estimates \(\varTheta_{0}\), \(\varTheta_{1}\), \(\varTheta_{2}\) and to two range estimates for each receiving antenna \(R_{\Delta f_{1}, \mathrm{RX}_{j}}\), \(R_{\Delta f_{2}, \mathrm{RX}_{j}}\), where \(\Delta f_{1}=f_{1}-f_{0}\) and \(\Delta f_{2}=f_{2}-f_{0}\). A fusion of the estimated direction and range data of the RFID tag leads to the final two dimensional localization capability of the implemented RFID reader. However, the reader observes the estimates for a period of time and then statistically evaluates the data. The most trustful estimates of the direction (\(\varTheta_{0}\), \(\varTheta_{1}\) or \(\varTheta_{2}\)) and the range (\(R_{\Delta f_{1}, \mathrm{RX}_{j}}\) or \(R_{\Delta f_{2}, \mathrm{RX}_{j}}\)), i.e., the estimates with the smallest standard deviation are fused to find the most probable tag position.

6 Reader verification

The functionality of the implemented reader has been verified in an anechoic chamber. The anechoic environment especially allows the evaluation of the reader software implementation, i.e., the initial calibration phase, the PDoA algorithms, and the sensor data fusion. For the measurements, a passive tag (Sokymat InLine UHF Tag [22]) was placed on a styrofoam carrier in the middle of the anechoic chamber at the same height as the reader antennas (see Fig. 8).

Fig. 8.
Fig. 8.
Full size image

Test setup in the anechoic chamber: The tag is positioned on a styrofoam carrier in the middle of the chamber. Tag and reader antennas are at the same height

Figure 9 shows the coordinates of the tag positions (true pos.) in the anechoic chamber, where the center of the reader receiving antennas is the origin of the coordinate system. The position of the tag has been estimated in 20 cm steps for a fixed \(x\) coordinate (vert. pos.) on the one hand and for a fixed \(y\) coordinate (hor. pos.) on the other hand. The adjustable operating frequencies have been set to \(f_{0}=869~\text{MHz}\), \(f_{1}=866~\text{MHz}\), and \(f_{2}=872~\text{MHz}\). The shown position estimates are composed of the direction estimate \(\varTheta_{0}\) and the mean of the range estimates \(R_{\Delta f_{2}, \mathrm{RX}_{1}}\) and \(R_{\Delta f_{2}, \mathrm{RX}_{2}}\) (\(\Delta f_{2}=3~\text{MHz}\)). The maximum deviation of the position estimate to the true position is about 15 cm. The estimation errors can be explained by a combination of the following reasons. The manual placement of the tag introduces slight deviations from the true positions. Small errors may be introduced by deviations of the generated operating frequencies. The assumptions for the definition of the estimators that are mentioned in Sect. 5 do not completely hold for this environment, e.g., the tag is not far away from the reader, compared to the spacing of the receiving antennas. Finally, the range estimation is based on not faultless direction estimates [23].

Fig. 9.
Fig. 9.
Full size image

Functional verification in the anechoic chamber: The true and the estimated positions show a maximum deviation of 15 cm. The position estimates are composed of \(\varTheta_{0}\) and the mean of \(R_{\Delta f_{2}, \mathrm{RX}_{1}}\) and \(R_{\Delta f_{2}, \mathrm{RX}_{2}}\)

7 Conclusions

This paper presents a method for implementing a custom-built UHF RFID reader for tag localization for the use in a specific environment that is in our case an AVL engine test bed. The method is based on radio channel measurements in the engine test bed that allows the determination of reader hardware requirements for the implementation of a reliable tag localization system. The final reader hardware implementation is able to perform reliable localization in the engine test bed with respect to power availability in the radio channel. The localization capability of the reader is provided by the exploitation of PDoA techniques. This first prototype of the reader achieves a localization accuracy of 15 cm in an anechoic chamber. Recently, the localization capability of the realized reader was demonstrated within an AVL engine test bed. The demonstration showed that the position of a passive tag can be correctly assigned to one of four equally sized regions in the test bed, i.e., detecting in which of the four regions the tag is located. The reader operates at three different operating frequencies to make the tag localization more robust in difficult multipath environments. In the future, the localization robustness will be improved by exploiting more than three operating frequencies [21]. In addition, the implemented SDR based RFID reader can be favorably exploited for the prototyping of future RFID systems, e.g. RFID systems that can detect sensor information of passive RFID tags in addition to their ID and location [24].