Radio frequency identification tag location estimation and tracking system and method

0001(Cross-reference of related applications) This application claims the interests of US Provisional Patent Application No. 61 / 124,294 filed on April 14, 2008, and US Provisional Patent Application No. 61 / 044,904 filed on April 14, 2008. The disclosures of these patents are hereby incorporated by reference, as they are all described herein.

0002The present application relates to estimating and tracking the position of passive or active sensors, in particular the position of sensors and / or RFID tags using phased array antenna systems and radio frequency identification (RFID) systems. Regarding identification.

0003RFID systems typically include a set of fixed or movable RFID tags that are processed by a reader / interrogator system. Each sensor can be passive or active, i.e. with or without batteries. In conventional systems, the reader and RFID tag usually need to be in close proximity, so the tag can operate in close proximity to the reader antenna.

<p num="0004"> The limited transmission distances available in conventional RFID systems limit their use in automated factory and / or indoor wireless environments. Such systems are often unreliable due to interference and collisions, even within the intended operating range.</p><p num="0005"> Typical RFID systems are also not designed to cover extremely large areas, as a large number of base stations are required to provide sufficient coverage of the area. This is extremely expensive and can be exorbitant. Also, compensating for a large area at a cost is done by selecting a high usage area, for example, a dock door and limiting the receivable area. Moreover, such systems are often unable to pinpoint RFID location due to the size of the environment and the complexity of the space. Therefore, there is a need for RFID systems that solve the above obstacles and shortcomings in the art.</p>

<p num="0006"> In one aspect, RFID tag / sensor location is determined using both single and multiple reading points.</p><p num="0007"> In one embodiment, the method of locating one or more radio frequency identification (RFID) tags involves a step of irradiating at least one RFID tag with an exciter and information from at least one illuminated RFID tag. Based on the step of receiving the signal by a plurality of receiving antennas, the step of obtaining the phase differential with respect to the received information signal from at least one irradiated RFID tag received by each of the plurality of receiving antennas, and the phase differential obtained by the received information signal. Includes a step to locate at least one RFID tag. Also, in one embodiment, the method further comprises locating at least one RFID based on the ratio of phase differential to frequency differential.</p><p num="0008"> In another embodiment, the radio frequency identification (RFID) system for locating one or more RFIDs comprises at least one exciter and reader. At least one exciter has multiple antennas, selectively transmits a ringing signal through at least two of the antennas, and from at least one RFID tag, at least two of the antennas. It is configured to selectively receive reception information signals through one of a plurality of antennas, which is different from one antenna. The reader is configured to communicate with at least one exciter and activate at least one exciter. The reader locates at least one RFID tag based on the phase differential of the received information signal.</p><p num="0009"> According to yet another embodiment, a radio frequency identification (RFID) system that locates one or more RFIDs includes at least one RFID tag, antenna array, transmitter and reader. The antenna array is configured to illuminate at least one RFID tag. The transmitter is connected to the antenna array and is configured to operate the antenna array to repeatedly irradiate at least one RFID tag in a specific time frame and a specific space. The reader is configured to create a probabilistic model based on the information signal received by communicating with the transmitter and repeatedly irradiating at least one RFID tag, applying the created probabilistic model to the particle filter. Determine the position of at least one RFID tag based on the results of the applied particle filter.</p><p num="0010"> In a further embodiment, the method of locating one or more radio frequency identification (RFID) tags is to place at least one receiver and at least one transmitter so that they share geometric features with each other. And, using the step of obtaining the position identification measurement value based on the received information signal from at least one RFID tag emitted by at least one transmitter, and the probability model and the obtained position identification measurement value, Includes a step of estimating the position of at least one RFID tag.</p><p num="0011"> For a more complete understanding of the methods and systems described, reference is made herein to the following description made in connection with the accompanying drawings, where the same reference numbers refer to the same parts.</p>

0012<figref num="1">FIG. 1 is a conceptual diagram of a distributed exciter architecture showing TX and receivable areas for two readers similar to the exciter calling space, according to various aspects of the invention.</figref><figref num="2">FIG. 2 is a conceptual diagram of a distributed exciter architecture showing TX and receivable areas for two readers similar to the exciter calling space, according to various aspects of the invention.</figref><figref num="3">FIG. 3 is a diagram of a wired (single) distributed exciter architecture system showing TX and receivable areas for two readers similar to the exciter call space, according to various aspects of the invention. Is.</figref><figref num="4A">It is a figure explaining an exciter layout and an exemplary frequency plan which concerns on various aspects of this invention.</figref><figref num="4B">It is a figure explaining an exciter layout and an exemplary frequency plan which concerns on various aspects of this invention.</figref><figref num="5">FIG. 5 is a diagram showing an algorithm for estimating an arrival angle from a received signal according to various aspects of the present invention.</figref><figref num="6">FIG. 6 is a block diagram of a reader system showing RFID tags, interferers and readers (eg, transmit and receive chains) according to various aspects of the invention.</figref><figref num="7">FIG. 7 is a block diagram of a reader showing an antenna array, RF / IF, signal processing and synthesizer subsystem according to various aspects of the invention.</figref><figref num="8">FIG. 8 is a graphical representation illustrating a leader array and a single excitation point topology according to various embodiments of the present invention.</figref><figref num="9A">It is a graphic representation that provides an elliptical representation for finding the position of an RFID tag according to various embodiments of the present invention.</figref><figref num="9B">It is a graphic representation that provides an elliptical representation for finding the position of an RFID tag according to various embodiments of the present invention.</figref><figref num="9C">It is a graphic representation that provides an elliptical representation for finding the position of an RFID tag according to various embodiments of the present invention.</figref><figref num="10">FIG. 10 is a block diagram of a 4-port excitation node according to various embodiments of the present invention.</figref><figref num="11">FIG. 11 is a detailed block diagram of a 4-port excitation node according to various embodiments of the present invention.</figref><figref num="12">FIG. 12 is a graphical representation illustrating a distributed array of readers and a single excitation point topology according to various embodiments of the present invention.</figref><figref num="13A">It is a graphic representation explaining the 4-port excitation node in the chandelier configuration which concerns on various embodiments of this invention.</figref><figref num="13B">It is a graphic representation explaining the 4-port excitation node in the chandelier configuration which concerns on various embodiments of this invention.</figref><figref num="13C">It is a graphic representation explaining the 4-port excitation node in the chandelier configuration which concerns on various embodiments of this invention.</figref><figref num="13D">It is a graphic representation explaining the 4-port excitation node in the chandelier configuration which concerns on various embodiments of this invention.</figref><figref num="14A">It is a graphic representation explaining the 4-port excitation node in the offset linear array configuration which concerns on various embodiments of this invention.</figref><figref num="14B">It is a graphic representation explaining the 4-port excitation node in the offset linear array configuration which concerns on various embodiments of this invention.</figref><figref num="14C">It is a graphic representation explaining the 4-port excitation node in the offset linear array configuration which concerns on various embodiments of this invention.</figref><figref num="14D">It is a graphic representation explaining the 4-port excitation node in the offset linear array configuration which concerns on various embodiments of this invention.</figref><figref num="15A">It is a graphic representation illustrating a 4-port excitation node in an offset linear array configuration for the positions of six different RFID tags according to various embodiments of the present invention.</figref><figref num="15B">It is a graphic representation illustrating a 4-port excitation node in an offset linear array configuration for the positions of six different RFID tags according to various embodiments of the present invention.</figref><figref num="15C">It is a graphic representation illustrating a 4-port excitation node in an offset linear array configuration for the positions of six different RFID tags according to various embodiments of the present invention.</figref><figref num="15D">It is a graphic representation illustrating a 4-port excitation node in an offset linear array configuration for the positions of six different RFID tags according to various embodiments of the present invention.</figref><figref num="15E">According to various embodiments of the present invention, six different Do a graphical representation for explaining the 4-port excitation node in the offset linear array configuration with respect to the position of that RFID tag.</figref><figref num="15F">It is a graphic representation illustrating a 4-port excitation node in an offset linear array configuration for the positions of six different RFID tags according to various embodiments of the present invention.</figref><figref num="16">FIG. 16 is a graphical representation illustrating a generalized multi-port excitation node, or array of nodes, in any or irregular configuration according to various embodiments of the present invention.</figref><figref num="17">FIG. 17 is a flowchart of particle filter position processing for specifying the position of an RFID tag according to various embodiments of the present invention.</figref><figref num="18">FIG. 18 is a flow chart showing an outline of the position estimation process performed by the reader according to various aspects of the present invention.</figref><figref num="19">FIG. 19 is a flow chart showing an outline of a position estimation process performed by a reader according to various aspects of the present invention.</figref><figref num="20">FIG. 20 illustrates a simplified four-element array in a two-dimensional implementation according to various aspects of the invention.</figref><figref num="21">FIG. 21 illustrates an analysis setup for DOA analysis showing RFID tags and antennas according to various aspects of the invention.</figref><figref num="22">FIG. 22 is a diagram of the variance of the calling space and the two-dimensional (x, y) Euclidean space considered to be two-dimensional Gaussian densities with known means, according to various aspects of the invention.</figref><figref num="23">FIG. 23 is a diagram illustrating Markov chain assumptions relating to various aspects of the present invention.</figref><figref num="24">FIG. 24 is a flow diagram showing steps in the general form of the sequential Monte Carlo method according to various aspects of the invention.</figref><figref num="25">FIG. 25 is a flow diagram showing steps of a general solution for searching the position of an RFID tag / sensor according to various aspects of the present invention.</figref><figref num="26">FIG. 26 is a flowchart of differentiation and gene expansion processing according to various aspects of the present invention.</figref><figref num="27">FIG. 27 is a conceptual diagram showing the automatic permanent use of inventories stored on shelves with racks arranged vertically.</figref>

0013The systems and methods for locating one or more radio frequency identification (RFID) tags are described below with reference to the drawings. The system utilizes various geometric arrangements of transmitters and receivers to obtain observations about the location of RFID tags. The observation result can be supplied to any various estimators capable of generating an estimate of the position of the observed RFID tag.

0014The geometry of the transmitter and receiver of an RFID system affects the accuracy with which the system can estimate the location of RFID tags. Various architectures according to embodiments of the present invention will be described. In some embodiments, one or more exciters transmit a call signal to illuminate the RFID tag, and the reflected signal is received by the plurality of receivers. Each receiver can be an independent receiver and / or a single receiver system, or an independent receiver antenna connected to a multi-port exciter system. In some embodiments, the multi-port exciter system functions as both an exciter and a receiver. One of the antennas is selectively configured to transmit a ringing signal and the remaining antennas are configured to receive backscattered signals with RFID tags. In some embodiments, the multiport exciter does not have the ability to read data from the RFID tag, but simply has the ability to make observations that help locate the RFID tag.

0015Due to the instability of the RFID backscattering process, the observables selected when estimating the position can affect the accuracy of the estimates obtained. In various embodiments, the system observes the phase difference of the backscattered signal from the illuminated RFID tag. In some embodiments, phase differences are observed at various transmission frequencies to provide range information. The ratio of phase difference to frequency difference is also called group delay. In many embodiments, the system observes the read rate of RFID tags in response to different exciters irradiating different calling spaces. The read rate is the number of times a tag has been read, as a ratio to the number of opportunities the tag has read. In systems that utilize highly sensitive receivers, the read rate can be thought of as indicating the distance from the RFID to the exciter. In such a system, the majority of tags activated by the exciter are read. Therefore, the reading rate approximately indicates the operating rate of the RFID tag by the transmitter.

0016As further described below, various techniques can be used to estimate position based on one or more observable quantities according to aspects of the invention. Given the complexity of the system and the large number of RFID tags that can exist in a given space, statistically modeling the location of the RFID tags estimates the exact location for each RFID tag in the space. can do. Therefore, in various embodiments, a large number of observations of RFID tags are used to create a probability distribution model. Use one or more algorithms and / or filters to further refine the model and determine the location of RFID tags. In some embodiments, particle filters are used to create a probability distribution model to further refine it. In other embodiments, a variety of other techniques can be used to further refine the position estimates obtained using observables.

0017(System architecture) The ability to locate an RFID tag in a given space is that of the antenna used to send the ringing signal to the RFID tag and the antenna used to receive the backscattered signal on the RFID tag. It depends almost entirely on the position. Various geometric arrangements can be used according to embodiments of the present invention, including geometric arrangements in which the transmitting and receiving functions are separated and independent exciters and receivers can perform those functions.

0018In US Patent Application No. 12 / 054,331, which was filed on March 23, 2007 and is pending at the same time, entitled "RFID Systems Using Distributed Exciter Network", Separation of the receiving and transmitting systems that process passive RFID tags enhances the performance and capacity of the RFID system, the disclosure of which patent is incorporated by reference, as fully described herein. .. This functionality presents a population of RFID tags / sensors to a set of calling spaces (1-16, 1-32, 1-38, 1-44, 1-40, 1), as shown in Figures 1-2. Realized by decomposing into -28, 1-24, 1-48, 1-54, 1-56), the exciter is each target calling space (1-18, 1-34, 1-36, 1-46). , 1-42, 1-30, 1-26, 1-58, 1-52, 1-50).

0019The size of each calling space can be adjusted by controlling the total radiation output from the exciter. On the other hand, of course, the radiated power of RFID systems is usually limited by law, limiting the calling range of exciters to, for example, 20 to 30 feet. Radiation output control is performed through the RFID reader (1-2) exciter output management and gain controller subsystem (3-18, 3-30). In addition to adjusting the size of each call space, by selecting the type of each exciter transmit antenna to provide the desired level of directivity, thereby controlling the beamwidth towards the target call space. , The overall performance of the system can be further improved.

0020In some embodiments, the reader comprises a phased array antenna capable of forming a beam. The reader-received phased array antenna beam (1-4) can be formed to be concentrated in a particular calling space (1-17, 1-21) or as a broad beam (1-20). The network of transmitting antennas, also called distributed exciters, can be controlled by wire (2-24, 2-30, 2-36, 2-12, 2-56) or wireless connection. The "return signal" transmitted from the controller to the exciter contains all the signal characteristics and parameters necessary to generate the desired waveform output from the exciter module to the tag. Figures 1 and 2 also show RFID application management and compute servers (1-7, 2-51) connected to the reader via a local area network (1-11, 2-50).

0021Figure 3 shows the receiving system (3-2), receiving antenna array (3-4), and, in one aspect, coaxial cables (3-10, 3-9, 3-16, 3-22, 3-26). Shows the layout of a distributed exciter / transmitter RFID system showing distributed exciters (3-6, 3-14, 3-18, 3-24, 3-30) connected to the system via. The call space and transmission output of each exciter are controlled by the central unit (3-2). Figure 3 shows exciter calling spaces of various sizes (3-8, 3-15, 3-20, 3-23, 3-28). The full receivable area (3-11) of the receive array is also shown. Each call space (3-8, 3-14, 3-20, 3-23, 3-28) can be used sequentially or simultaneously, depending on the number of beams the receiving array can handle.

0022The controller (3-30) in the system (3-2) plans the time, frequency, and size of space in which each exciter operates. The scheduler for S / T / FDM (spatial, time, and frequency division multiplexing) utilizes optimization algorithms to maximize the probability of reading all tags in the target call space. The controller can use frequency hopping (while meeting regulatory requirements) to schedule the use of frequency channels per exciter. Figures 4A and 4B show an example of an exciter layout (4-6) and a time series (4-4) showing the assigned frequency channels (4-8). The time series (4-4) shows frequency hopping channelization in the 900MHz ISM band (4-10). In each time series, a different hopping sequence assigns a different set of random frequencies to each active exciter (4-8). The algorithm shown in Figure 14 manages and optimizes this behavior. Various exciters are activated according to the schedule, and the RFID system can collect observation results regarding RFID tags. In many embodiments, the schedule plays an important role in preventing interference from multiple exciters during the collection of observations. In addition, the availability of different excitation frequencies in the schedule, as described below, allows RFID systems to collect observations about group delays of different RFID tags.

0023Referencing Figure 1 again shows the wireless exciter layout and deployment system. In the illustrated embodiment, two RFID systems (1-2, 1-14) are provided. Each system has an independent receiver (1-4, 1-12) and transmitting antenna (1-6, 1-13). The transmitting antenna radiates the transmitting link to the exciter, while the receiving antenna receives a signal from a tag in the calling space of each exciter. In one aspect, the transmit link carries additional information such as an exciter identification (ID) number, command, control and management information. The figure shows the receivable areas (1-22, 1-20) of the two systems. As shown in the figure, the system receives incoming call spaces (1-24, 1-48, 1-28, 1-40, 1-54, 1-56, 1-44, 1-16, 1-32, It corresponds to 1-38). The overlapping region between the receivable regions (1-22, 1-20) is managed by the interaction of the beam formation of the receive array with the frequency or time of the exciter operation. The two systems are connected to the LAN (1-3, 1-8, 1-9) in a manner similar to the wired exciter system in Figure 2, and the wireless exciter management server (1-7) is two by LAN. It connects to the system (1-2, 1-14) and controls the operation of the exciter, including control, instruction, coordination and calibration of the exciter, as well as optimization of the calling space.

0024(Call RFID tag) The sensor or FRID tag can be called any number of times during a fixed time interval. For each of these calls, the RFID tag is combined with signals that act on the array to form a single (beam-formed) signal for detecting the sequence of codes transmitted by the RFID tag. It is possible to detect sensor data or information incorporated in. Each ringing period usually consists of multiple packets (eg, two, called RN16 and EPC packets) from a strip of RFID tag. The payload in these packets usually contains a temporary address (eg, a 16-bit random number in a packet of type RN16), and after an acknowledgment, the tag then has its information content (eg, electronic product code (EPC)). Packets containing can be forwarded.

0025During each ringing period, a large number of RFID tag ringing signals can be transmitted at different frequencies. By calling the tag using different frequencies, further observations of the tag can accurately model the phase and amplitude trajectories of the received signal over time and characterize the signal scatter with multiple reflections of the transmitted signal. It will be possible.

0026(Summary of the fact that the position can be estimated by using the call) The processing work during the ringing period is not limited to them, but the relative phase difference between the signal from each antenna element and the reference signal is evaluated, and the adjusted phase for each such antenna element is evaluated. It can include deriving the relative range from each antenna element to the RFID tag based on the delay difference. Estimating the location of RFID tags can then be addressed by treating the aggregate of each call period as a single database forming a "sample space". It should also be noted that reading the same RFID tag at multiple frequencies makes it possible to estimate the range of the signal source (distance from the tag to the reading point) by "sequential ranging". For applications that use only a single reader (reading point), the reader system can estimate the position without the need to "triangulate".

0027To process the signal, the adjusted phase difference between the signal from each of the other antenna elements and the reference signal is derived, and each antenna element derives the relative arrival direction of the received signal from the RFID tag. Can include that. The direction of arrival of the signal from a single RFID tag at multiple reading points can be used to further improve the estimation of RFID tag position. This can be done by combining the relative directions of the arrival information obtained from the signals received by each array element at each reading point. By using multiple call cycles and arrays, the number of reading points is each multiplexed in time, further improving the overall estimation of tag position.

0028In repetitive work using multiple reading points, RFID tag information obtained from multiple repetitions is combined in order to form a probability distribution of tag positions, and the arrival direction of the signal from the signal source to each antenna element is determined. To assess the impact of being multipathic and mitigate this, it can be included to apply an algorithm.

0029The algorithm for estimating the arrival angle from the received signal in the system, shown in Figure 5, is addressed in co-pending U.S. Patent Application No. 11 / 770,712, filed June 28, 2007. The disclosure of this patent is incorporated above by reference. In that application, a method of using time-separated signals sampled from each received packet from a tag to estimate the AOA (arrival angle) using the relative phase and amplitude of the received signal from each antenna element. Is described. The algorithm shown in Figure 5 performs an FFT (5-14) on the array input (5-16), followed by a mutual spectrum matrix calculation (5-12) and decomposition processing (5-10). ), Calculate the beam scan response (5-8). Scan vector data (5-6) is used for scan response calculation (5-8), and this scan vector data is generated and adjusted using the array response and calibration database (5-2, 5-). 7). The beam scan response calculation (5-8) is repeated in various directions by changing the mutual spectral matrix (5-12).

0030Among some challenges, each antenna element is periodically calibrated to address the practical challenges caused by the electronic components used in the wireless circuitry of the antenna array, with the relative phase of each antenna element and the relative phase of each antenna element. Eliminate the amplitude imbalance and its respective in-phase and orthogonal phase components. Calibration is performed on one or more test signals and the processing of the signals received by each antenna element can be corrected to compensate for such imbalances.

0031In one embodiment, if the position of the exciter is unknown, or to ensure that the position of the exciter has not moved, a series of calibration operations are performed before or during the position estimation operation. Go to find the location of the exciter or dummy RFID tag. In one embodiment, the position of the exciter or dummy RFID tag is determined by an RFID receiving system similar to the "real" RFID tag position estimation.

0032The reader of the wireless automatic identification system in one embodiment is provided using an antenna array. The transmit channel (the transmit path between the reader and the tag) allows the transmit antenna array to be distributed across several physical arrays. In the case of a distributed transmitting antenna, the receiving antenna array can capture the collision energy from the tag signal excited by the antenna elements of the distributed array. This technique can use spatial multiplexing to significantly enhance bandwidth utilization compared to a single antenna system. The antenna array can support a plurality of frequency bands. A typical array element configuration includes an aperture-coupled fed tiled patch antenna. The tiled structure contains the same elements arranged in a matrix in a two-dimensional plane. A low noise amplifier (LNA) can be embedded in the antenna element itself to enhance the overall performance of the system.

0033When using a transmit array antenna, beam formation can be used to scan the transmit beam to a desired position in space. This beam scanning reduces collisions and interference between signals received from the responding tags. Various transmission methods can be adopted, for example, to maximize the received isotropic power to the RFID tag, while radiating a "spatial hopping" pattern that meets the conventions for maximum power and residence time. , The transmission beamformer coefficient can be updated for each time zone.

0034Through regular calibration, the beamformer is placed (between the antenna and the analog-to-digital converter (ADC) for the receiving path, and between the digital-to-analog converter (DAC) and the antenna for the transmitting path. It can compensate for the inconsistency and imperfections of the RF microwave device at the front end, as well as the phase and amplitude inconsistencies from radio frequency to baseband that occur in the independent parallel array element paths.

0035Here, with reference to FIG. 6, an RFID reader that calls a group of RFID tags placed on a large number of inventories placed on a pallet according to an embodiment of the present invention is shown. RFID systems operate in the presence of interference from exemplary interferers 6-10. A pallet of goods 6-1 includes a number of cases or articles with RFID passive tags. The transmission call signal 6-4 from the antenna 6-6 collides with the pallet 6-1. In response to the signal energy detected by each tag, each tag can use the transmit call signal or the power received from beam 6-4 to backscatter a set of information. In this environment, there can be artificial or natural interference illustrated as interferers 6-10. Since the receiving antenna array 6-12 forms a beam on the backscattered signal from the tag, the power received from the tag is maximized and the power received by the antennas 6-6 from the interfering bodies 6-10 is minimized.

0036In Figure 7, a functional block of a reader system (eg, Figure 6) that includes an antenna array subsystem (17-1), an RF / IF subsystem (17-2), and a signal processing subsystem (17-3). The figure is shown. The synthesizer subsystem (17-4) also supplies the clock frequency and local frequency to the RF / IF subsystem of the reader system. The reader system calls the RFID tag on pallet 1-1 in the presence of interference (eg, Figure 1).

0037(Geometric arrangement of antenna) RFID systems can include multiple transmit antennas and multiple receive antennas. In a distributed exciter architecture, the multiple transmitting antennas are the exciter's antennas. As described below, the plurality of receiving antennas can be the antenna array of the RFID receiver and / or the antenna of a multi-port exciter that is switched for observation for the purpose of estimating the position. When collecting observations to estimate the location of RFID tags, the quantity and location of receiving antennas with respect to the exciter and with respect to each other can greatly affect the accuracy with which individual observations can be made. Therefore, the geometry of the transmitter and receiver can affect the number of antennas required for applications that require accuracy in estimating a particular position.

0038(Linear array) With reference to FIGS. 8-9C, the reader array 18-1 associated with the single exciter 18-3 and RFID tag 18-2 is shown here. A reader array is a linear array that includes four antennas that provide four receiving points. Each antenna is offset from each other. The exciter is located at a specific distance d1 from the reader array and is further located at a specific distance d2 with respect to the RFID tag. RFID tags are also located at a specific distance d3 with respect to the reader array. Observations using a receiver array

00399A-9C show the distance from the RFID receiver system to the exciter (d1), the distance from the exciter to the RFID tag (d2), and the distance from the RFID tag to the RFID receiver system according to the embodiment of the present invention (d1). As an alternative to d3), observation of RFID tag position using a calibrated gradient of group delay has been shown. Group delay observations are described in great detail below, but here they can be considered to provide an estimate of the path length between the transmitting antenna, RFID tag and receiving antenna. An ellipse is a locus in the plane of a point such that the sum of the distances to two fixed points is constant. To be precise, the above is the case where the distance between the reader and the exciter is known (d1) and the total distance d2 + d3 is known. Calibration allows us to determine the time (distance) between the reader (marked with a star) and the exciter (marked with a diamond). We can also find the total time d1 + d2 + d3. Taken together, these two distance measurements (d1 and d1 + d2 + d3) show that the tag must be in one of the positions on the ellipse 18-4. In addition, it may be degenerated as 2 * d1 = d1 + d2 + d3. This indicates that the tag is somewhere on the line segment between the reader and the exciter, and to be precise, in this case somewhere cannot be determined (as explained by another exciter). Resolve this case). An equation can be derived to find the position of the intersection of the line from one focal point of the ellipse (in this case, the reader) and the position on the ellipse.

0040FIG. 9B shows the case where the tag is on the line segment between the reader and exciter 1. This is the case of degeneracy where 2d1 = d1 + d2 + d3. In this case, the region of possible tag positions by the exciter 1 is the line segment between the reader and the exciter 1. If another exciter, Exciter 2 (18-4), can illuminate the tag, this case is from the ellipse (18-5) predicted from the focal leader and Exciter 2 positions, and from Exciter 2 and the tag. It can be solved by finding the intersection between the distance d22, which is the distance, and the distance d32, which is the distance from the leader to the exciter 2. Of course, in this case, the arrival angle (AOA) information is not needed to determine the tag position. However, this angle can be used as a further measurement by the above process.

0041Figure 9C shows two non-degenerate position predictions by Exciter 1 and Exciter 2. Both the distance d11 and the distance d12, which are the distances from the reader to each receiving exciter, can be obtained by calibration work. The distances d21 + d31 and the distances d22 + d32 define the left and right ellipses, respectively. Angle φ₁, Φ₂From, the line segment leaving the leader is predicted, and the intersection of this line segment and each ellipse indicates the tag position. Shown angle φ₁, Φ₂Looks the same. However, each angle is an independent observation result obtained as a result of reading the target tag by excitation from exciters 1 and 2, respectively.

0042(Geometric arrangement of multi-port exciter) The exciter-distributed architecture separates the transmit and receive functions within the RFID system, and the exciter is responsible for performing the transmit function. The advantage of separating the system in this way is that you can use a low-cost exciter and place it in more positions than a traditional RFID receiver system, and if you have RFID receivers in multiple positions, it is usually The cost is too high. In some embodiments, the exciter has multiple ports so that a single exciter can use multiple antennas (ie, ports) to activate RFID tags. In many embodiments, the multi-port exciter can switch some antennas to receive backscattered signals with RFID tags. By switching the function of the antenna in this way, the multi-port exciter can collect the observation result of the backscattered signal by the RFID tag. Exciters can make these observations without the need for complex decoding circuits used in RFID receivers. Allowing multi-port exciters to collect observations about RFID tags greatly increases the number of receiving antennas in an RFID system that can be used to collect information for use in position estimation. In addition, the antennas of the multi-port exciter are usually distributed farther away from each other than the antennas in the linear array of RFID receivers. The multi-port exciter that can collect observations about RFID tags and the various geometric arrangements of antenna positions on the multi-port exciter are further described below.

0043With reference to FIG. 10, FIG. 10 shows a block diagram of a multi-port exciter or excitation node (eNode) having four ports (4-port exciter) according to various embodiments of the present invention. .. Much of the description below relates to a multi-port exciter that includes four ports, but according to embodiments of the present invention, a multi-port exciter that includes any number of ports can be utilized. The 4-port exciter includes a switching circuit 19-1 for selecting one or more antennas 19-2. Antenna selection is controlled by processor 19-3, utilizing the associated transmit circuit 19-4 or receive circuit 19-5. The transmitter circuit handles all output communications, such as call and calibration signals, to the vicinity of RFID. The receiving circuit handles all input communications such as response data from the called RFID. The receiving circuit selects one of the receiving antennas of the input signal through the switching circuit. Similarly, the transmitting circuit selects one of the transmitting antennas of the output signal through the switching circuit. In most cases, multiple antennas are selected for transmission to reach the closest of the most RFIDs, and to locate RFIDs quickly and accurately, as described throughout this application. Guarantee maximum call range.

0044FIG. 11 shows a detailed block diagram of a multi-port exciter or excitation node (eNode) according to various embodiments of the present invention. The excitation node includes an antenna array that sends and receives data from RFID. The antenna array includes a plurality of antenna elements 7-1 to 7-4. Each antenna element is configured by a switch (7-5) so that one or more of the antenna elements transmit and the remaining antenna elements receive. Therefore, the signal received via the antenna element 7-1 is supplied to the amplifier 7-10. The received path signal is thus amplified (7-10), processed with a passband filter (7-14), and mixed directly with the baseband signal supplied by the local oscillator (7-60) (7-20). .. The carrier frequency used to mix the received signal with the baseband signal is the same as the frequency (7-62) used to mix with the transmitting RF. The received baseband signal is then processed by a lowpass filter (7-24) and amplified (7-28). The signal is sampled by the receiver signal processing circuit, associated with the extraction phase, demodulated, and decoded (7-70). Command and control messages (not RFID tag related data) are decoded by the control circuit (7-74). Based on the decrypted commands and control messages, the processor (7-80) issues commands to control the transmit output level calibration (7-86) and other maintenance characteristics. The received RFID data is sent to a data encoder and modulator (7-31) to meet the aspects of different RFID protocols. Packets for transmission are up-converted (7-32), processed with a passband filter (7-36), variably attenuated (7-42), and amplified (7-44). The final band-pass filter ensures that out-of-band radiation requirements are applied (7-52) before radiating through one of multiple (in this case four) available antennas. In one embodiment, the coaxial cable provides frequency reference, DC power and command control. This embodiment

0045See here in Figure 12, where the multi-port exciter antennas are as a distributed array of readers phase-locked at reading points 20-1, 20-2, 20-3, and a single exciter 20-5. It is configured and shown. The placement of RFID tags 20-4 within the distributed array is also shown. The multi-port exciter receives signals backscattered by RFID tags on multiple ports with antennas located at reading points 20-1, 20-2, and 20-3, and uses a single port to excite. It is configured to transmit a ringing signal from antennas 20-5. Each reading point is at a known distance from the excitation node. For example, reading points 1, 2, and 3 are separated from the excitation node by specific distances d11, d21, and d31, respectively. In addition, the distances d13, d23, and d33 for each reading point 1, 2, and 3 of the RFID tag can be obtained in the same manner as the distance d22 for the excitation point of the RFID tag (see the explanation of the group delay below). ). Therefore, the multi-port exciter uses the total distances d12 + d13, d22 + d23, d32 + d33 to determine the ellipse, along with the known reading and excitation point positions, and collects observations about the RFID position. be able to. The angle in the arrival information is not used, or required, to locate the RFID tag.

0046Here, with reference to FIGS. 13 to 16, an exemplary excitation node or eNode configuration is shown. For example, FIGS. 13A-13D show RFID tag observations obtained by a "4 port" eNode configured as a "chandelier" and various ports that function as exciters. In the chandelier configuration, the antenna elements are set equidistant from each other and set in a square shape. The 4-port eNode maintains phase synchronization between the transmit and receive points using a local oscillator or by external reference. The figure is viewed from above, with three of the four ports receiving (21-1, 21-2, and 21-3 in Figure 13A) and one (21-5 in Figure 13A) receiving. Indicates an eNode that is set to send. The target tags are indicated by squares (21-4). In each figure, three ellipses are displayed. Each ellipse shares a transmitting antenna as one focal point and has a different receiving antenna as another focal point. The intersection of the three ellipses can be regarded as the observation result of the tag position. If there is more than one intersection (not shown here), the information that the antenna excited the tag can be used to identify the most likely three-way intersection. In one embodiment, the phase of the received tag signal is determined from its correlation with the preamble sequence. Such a technique is described in U.S. Patent Application No. 11 / 770,712, filed June 28, 2007, entitled "RFID Beam Forming System" of this patent. Disclosure is incorporated herein by reference in the same manner as described herein in its entirety.

0047The reliability of being able to make position estimation observations depends on the noise in the system. When the observation results are the intersections of the ellipses in the manner outlined above, the reliability of the observation results can be evaluated by the degree to which they are approximately parallel at the points where the ellipses intersect (see, for example, Figure 3D). thing). When the ellipses are nearly parallel, slight variations in phase noise can cause large offsets in the observed RFID tag positions. As described below, the number and position of multi-port exciter antennas can significantly increase the reliability of position observations performed using the exciter.

0048In Figures 14A-14D, the 4-port eNode is configured as an "offset linear array". In an offset linear array configuration, the antenna element pairs are linearly aligned and equally separated from each other. Also, the first pair is offset by a set distance from the second pair. The offset configuration is used to increase the total percentage of the region where backscattered signals from RFID tags allow for position observation where the RFID has only one possible position. As shown, the three ellipses are placed for each of the four ports of the eNode, and the intersection of the ellipses identifies the tag position. As further explained below, each of the four sets of observations consisting of three ellipses can be sent to the particle filter process, and any one result is sufficient to locate the tag.

0049Figures 15A-15F present another example of a 4-port eNode configured as an offset linear array, showing a possible range for locating one or more RFID tags. To simplify the reader of this application, the array is shown in a single transmitter configuration, three ellipses are drawn, and their intersections locate six different RFID tags.

0050Figure 16 shows an example of using an "irregular" array to locate RFID tags. In such a configuration, the antenna elements are set in a pseudo-random pattern. In the illustrated example, using six total receive patches (21-1, 21-2, 21-3, 21-6, 21-7, 21-8) and one transmit patch (21-5), Identify the location of the RFID tag (21-4). Any array configuration can also be used to elucidate the position based on group delay.

0051(Observable amount used for position estimation) Backscattered signals from RFID tags provide various observable quantities that can be used in position estimation. The observable amount used as a substitute for the distances described above for the geometry of the transmitting and receiving antennas is the calibrated gradient of group delay. Group delay refers to the phase difference observed at various frequencies. An embodiment in which group delay can be used for position estimation according to an embodiment of the present invention will be described below. In some embodiments, readability observations are used in position estimation. The RFID tag read rate can usually be described as the number of times the RFID tag has been read, as a ratio to the number of opportunities the RFID tag has been read. Other observables available for position estimation include, but are not limited to, phase, magnitude of phase constants, read rate, carrier frequency, excitation node index (index), and receive antenna index.

0052(Group delay as observable amount) FIG. 18 shows a flowchart illustrating one aspect of the position estimation process. Processing begins when the reader sends a start signal to a given exciter (8-1). The exciter irradiates (sends a call signal) the RFID tag in the exciter's transmission field (8-2). RFID tags respond by sending an information signal (8-3), and the reader asks for the group delay of that signal (8-4). RFID receiver systems cause RFID tags to respond with different frequencies of information signals only, and the system seeks phase differences between various information signals. Using the ratio of phase difference to frequency difference (ie, group delay), the reader finds the distance to the RFID tag. In one embodiment, the reader also finds the total round trip time (8-5). Group delay can be used to determine the distance to an RFID tag, as described below. It should be noted that throughout this process, the signal sent to each system is frequency and phase locked.

0053If the position of the exciter from the reader is known, the phase of the tag signal received by the reader is measured. When using different tone frequencies, measure different relative phases. The measured relative phase difference of two tones at two different frequencies due to the reciprocating delay is associated with the frequency difference by the following equation (assuming the exciter is placed with the reader):<maths num="1"><img id="000003" he="16" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> In the above equation, Δφ is the relative phase difference, Δf is the frequency difference, d is the distance, and c is the speed of light. Tone frequency f₁Phase θ at₁Can be measured with an ambiguity of 2 mπ. Similarly, the tone frequency f₂The phase at can be measured with the ambiguity of 2nπ. As long as the phase difference is less than 2π, the range d can be determined given Δf using the phase difference of the measured values relative to 2π. This is true as long as Δφ is less than 2π. The conditions can be satisfied by selecting an appropriate frequency separation in consideration of the predicted operating range. In the case of the two-dimensional example, the tag position can be obtained from the range d and the azimuth angle θ. Those skilled in the art will appreciate that the extension to three dimensions is feasible and intended. The exciter is not placed with the reader and the distance to the tag is d₁In the case of, the following formula is obtained.<maths num="2"><img id="000004" he="12" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>

0054(Arrival Angle as Observable (AOA)) In systems that include special cases such as linear antenna arrays, RFID tags can be triangulated using the results of arrival angle (AOA) observations from multiple linear arrays. In addition, RFID tags can be triangulated using multiple observations at different frequencies from a single linear array.

0055An example of a technique for observing a position using AOA is based on a group of techniques known as a Multiple Signal Classification (MUSIC) algorithm with spatial smoothing. In particular, to simplify the notation, we consider techniques applied to 4-element linear arrays using the MUSIC algorithm with forward and reverse filters. Those skilled in the art will appreciate that the extension of the algorithm to any array is feasible and intended.

0056The signal r received by the i-th element of the M-element linear array, each separated by a certain distance, i.e. λ / 2._i(t) is given by the following equation.<maths num="3"><img id="000005" he="17" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> In the above formula, a_kIs the amplitude of the kth multipath signal, s₁(t) is the desired signal, s_k(t) (k = 2,3, ..., N) is a multi-path reception signal, θ_kIs the angle of the AOA with respect to the antenna aiming for the kth signal, and n (t) is the additional noise or interference. In-phase and quadrature phase components, ie I_nAnd Q_nIs the received signal r_iRepresents the real and imaginary parts of (t). In vector notation:<maths num="4"><img id="000006" he="37" wi="170" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> In the above equation, θ is the AOA for antenna aiming. The signal s (t) includes the desired signal and the (N-1) multipath signal.

0057With reference to FIG. 21, the location of RFID tag 10-1 is shown through multiple AOA measurements. In this two-dimensional diagram, when the positions of the two array antennas 10-2 and 10-3 are known, the tag position can be obtained from the two AOA measurements. In particular, the tag position (x, y) can be obtained from the following equation.<maths num="5"><img id="000007" he="16" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>

0058Here, with reference to FIG. 20, one aspect of an array antenna RFID system operating in short range mode is shown in terms of its ability to locate RFID tags. A simplified four-element array is shown in a two-dimensional diagram as an example of RFID tag positioning performed by the system. Those skilled in the art will appreciate that extensions from 2D to any 3D array are feasible and intended. RFID tag positioning technology is based on measuring the phase difference 9-1 of an arrival signal or preamble signal between a specific element 9-3 and a reference element 9-2. The phase difference 9-1 is the range difference (r) of the paths 9-4, 9-5 between the RFID tag and the two array elements 9-2, 9-3.₂° -r₁°) is proportional. In particular, the range difference is given by the following equation.<maths num="6"><img id="000008" he="16" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> In the above equation, f is the carrier frequency of the RFID tag. x₁, X₂, X₃, X₄And range difference (r₂-r₁), (R₃-r₁), (R₄-r₁The position of the RFID tag that can be uniquely obtained from) is the known position (x) of the array element._i, y_i) And the measured range difference can be calculated by a very efficient algorithm.<maths num="7"><img id="000009" he="47" wi="159" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> y_iEven if = 0, generality is not lost. This solution is based on a linear least squares (LS) solution weighted to find the intersection of hyperbolas that define the range difference. The accuracy of the solution approaches that predicted by the Cramer-Rao Bound (CRB).

0059(Reading rate as observable amount) The read rate is the ratio of the number of times the RFID tag was read to the number of times the RFID tag could be read during the exciter's excitation. Systems utilizing a distributed exciter architecture can have very high receive sensitivity, so the main factor affecting tag readability is path loss between the transmitter and the tag. Therefore, the read rate is expected to correlate with the position of the tag relative to the exciter. For example, hypothesis region X_aIs an exciter e₁, E₂Given that they are equidistant from, in that case, their respective read rates RR for RFID tags placed within the hypothetical domain._e1, RR_e2Are expected to be equal. Collisions can affect the data when determining readability. There is usually a balance between avoiding collisions and ensuring that the number of slots provided to avoid collisions is not large enough to significantly affect the performance of the system.

0060Using the excitation line margin, the tag has inventory area x_aCreates a probability mass function (pmf) that represents the likelihood that a tag will be read a given number of times (reading rate) when placed inside. Read rate is a specific time interval, dividing this amount by the total number of possible reads that can be achieved with the same duration. Read rate (RR) is the notation RR_eExciter using (e^j) Is expressed as a subscript. Based on the above definition, it is possible to specify the probability as a point on Gauss's probability mass function.<maths num="8"><img id="000010" he="16" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> In the above equation, μ and σ are obtained according to the excitation power, the angle from the exciter to the hypothetical region, the distance from the exciter to the hypothetical region, the exciter radiation pan-turn, and the tag radiation pattern. Note that all probabilities associated with a given exciter e are normalized as follows before finding the probabilities of reading an RFID tag at a given position.<maths num="9"><img id="000011" he="18" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>

0061As further described below, various estimators can be used to determine the position of the hypothetical region and obtain observable RFID tag position estimates within the various hypothetical regions.

0062(Position estimation using observable amount) With reference to FIG. 19, the work shown in FIG. 18 is repeated (8-7). By repeating the work, a plurality of reading points are generated. Using these multiple reads, and thus the RFID tag information or distance estimates, a probability distribution model is formed (8-8). An algorithm can be selected and applied to obtain RFID tag position estimates (8-9). In addition, the confidence level and accuracy factor can be determined. As a result of applying the algorithm, for example, the influence of multiple paths such as the arrival direction of the signal from the signal source to each antenna element is clarified and mitigated. Various estimators are described below.

0063(Estimator used to locate from observables) The effect of noise when observing RFID tag positions can be limited using an estimator. Several different estimators that can be used to estimate RFID tag positions using any observable amount outlined above are described below.

0064(Estimator based on particle filter) In one embodiment shown in FIG. 18, phase, magnitude of phase constant, read rate, carrier frequency, excitation node index, receive patch antenna index phase, and / or (the observables have been described above). Other observables associated with a given tag reading are passed through a Monte Carlo hypothesis testing algorithm known as a particle filter (25-4). Since the generalized three-dimensional probability distribution function is continuous and therefore the complexity is technically infinite, a finitely compressed description of this distribution has been found. The Kalman filter provides only the second moment description of this general distribution. Unscented Kalman filters are good for third-order moment complexity. It is also possible to consider the grid point hypothesis, which can prevent a set of hypotheses from covering the entire state region and extend to higher point densities. However, these can be extremely wasteful in the number of hypotheses needed to express local likelihood.

0065Particle filters are an adaptive hypothetical approach to estimation using a non-uniform time adaptive lattice. Particles representing the state-space test hypothesis are generated based on the previous state distribution. For each observation, the observation is evaluated by the likelihood (state potential) produced by a given particle. Particles with high likelihood are replicated and particles with low likelihood are deleted. Finally, the replicated particles are randomly moved within the state space by a small amount, similar to gene mutation or annealing.

0066For this position estimation problem, the particles correspond to (x, y, z) positions and are optionally<maths num="10"><img id="000012" he="10" wi="20" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Corresponds to speed. Particle filtering can be divided into an initialization process and an iterative process set, each used in a new measurement. The first measured distribution or other previous state is used to generate a seed particle cloud. If the states have a uniform distribution over a finite range, these early particles can be selected from the lattice. However, the previous distribution is usually more complex, with random state values selected to sow the particle set.

0067The time update process (25-5) takes place each time a new observation (25-3) is entered into the system. This corresponds to the propagation of the particle state and the dynamic uncertainty due to the fact that some time has passed since the last update. This step is carried out by a physical process and is deterministic and stochastic. Considering that some time has passed since the last update, there is some uncertainty about the current position and velocity of each particle. Randomly move each particle to a new position and new velocity. The distribution used for this process depends on the environment and the current state of the particles. As an example, if one wants to estimate the position of a tag moving on a forklift, the new speed is limited to the speed that can be obtained with an acceleration of 1G or less in each direction. There is also the maximum absolute value of speed that a forklift can have. The time update process is, in principle, independent of the regularization step, but both time update and regularization follow the implementation as they add noise to the state particles.

0068The measurement update (25-6) process then calculates the likelihood corresponding to each particle based on the new measurement. The likelihood obtained is the likelihood that each observation (eg, phase vector or readability measurement) corresponds, given the predicted phase between the eNode to the tag and the distance from the tag to the antenna element. Generated from degrees. These probabilities can, in one embodiment, be evaluated by a Gaussian distribution, which depends on the received power of the antenna at the time of the observation and, in addition, the reliability associated with the estimated calibration factor. Use a standard deviation that depends on. The calibration factor is used in each tag reading measurement to eliminate any effects that do not correspond to wave propagation. In one embodiment, if the distance between the excitation point and the receiving patch is known, the extra phase rotation of each frequency is referred to as a "back channel" waveform or a reference tag placed with the excitation point. It can be removed by comparison with the observed phase used (25-10). The extra rotation removed at each frequency is recorded and "cancelled" from the next receive tag measurement phase data to compensate for phase rotation effects that are not due to radio propagation (such as electrical delay).

0069The resampling process (25-7) is responsible for particle removal / replication based on particle likelihood. This is done by taking the cumulative distribution produced by the likelihood of the particles and using it to generate new particles. The higher the likelihood of a particle, the more it replicates. The clones of the particles have the same position and velocity (in other words, they are the exact clones at this time, and the next step (regularization) will be subject to carefully selected changes).

0070The final step of the particle filter process, regularization (25-8), is involved in maintaining some memory of measurement likelihood. Previous probabilities of particles are preserved by duplication and modification. In this way, high level particles are replicated. The regularization process is similar to gene mutation or simulated annealing. Its purpose is to fluctuate the clones that fill the gaps in the particle set. One of the known problems with particle filters is that each point can be degenerated into a small number of hypotheses. If the particle cloud is degenerate, there are too few hypotheses to test in future measurements. The regularization process is responsible for avoiding this problem by introducing its random variation.

0071The result is finally output to a higher layer that is statically assembled (25-9). In this layer, it is possible to calculate the probability density of the position-specific solution over time. Usually, the variance measurement of this final layer statistic can report the quality of the final solution.

0072(Bayesian inference device) In one embodiment, the signal for the selected RFID tag from which information is retrieved is by multiple RFID tags based on the spatial position of the selected RFID tag relative to the other spatial position of the plurality of RFID tags. You can choose from the signals. That is, for a given calling space, only a specific population of tags is irradiated, as shown in FIG. 3 and, as an example, calling space 3-8. When estimating the position of the tag, all available information such as AOA, the position of the excitation node and the known position of the sensor or tag population, the multipath propagation environment, the measurement frequency, and the array response (beam pattern). Based on this, the Bayesian method is used to model the probability density function of RFID tag positions, and measurements and any other auxiliary information are further introduced into the apriori Bayesian model. In one aspect, the system recursively estimates its position in 3D Euclidean space (positions in x, y, z, ie height, roll, yaw).

0073Observation vector Y (for the jth tag) measured at each antenna element^j_tIs a discrete complex-valued received signal sample r for each antenna element, each with a real and complex part._t, Or homeomorphic and quadrature phase components I_n, Q_nAnd known exciter position (x, y, z), beamformer coefficient a, signal-to-noise ratio (SNR) estimation, gain setting α, soft metric, external information β (I) for each call space._n, Q_n), And a packet (eg, RN16 + EPC code). The model measurements used are time t, Y^j_tIs a single vector of. Observed L-dimensional vector Y^j_tMaps the 3D Euclidean space of the tag position to an L-dimensional observable vector (R)³ R^L) It is assumed that it can be obtained. Probability distribution P (x^j_t| Y^j_t) Is recursively implemented in various methods to estimate x^j_tIs the position coordinate of the jth tag in 3D. The conditional expectation of this density (ie, the mean E (x | Y)) corresponds to the tag position or equivalent being isomorphic to the estimation of this sequence.

0074Referring here to FIG. 22, for the entire space or volume 12-1, the exciter activated by the reader creates the RFID position estimation illustrated in 12-4. Position estimation of multiple RFIDs when the reading is repeated multiple times is illustrated in 12-5. The peak of the cone or the center of gravity identifies the position of the RFID. The perimeter of the cone specifies the accuracy or accuracy of the identified position. For example, a steeper peak 12-5a presents a very accurate position of an RFID tag, while a flatter peak 12-5b presents a less accurate estimated position of an RFID tag.

0075If the location of the transmitter / exciter (11-2) is known, the tag position estimation problem can be changed to finding the position of the tag in the cube as shown in 12-1. Hypothesis testing can be used to quantize a cube into a smaller cube, as shown in 12-2, for each tag position, and each position is treated as a sphere 12-3. The probability distribution for the position of the tag population within the calling space can be considered as a two-dimensional Gaussian density with known means and variances in two-dimensional (x, y) Euclidean space. The spheres projected onto the circle can also be considered as two-dimensional Gaussian densities with known means and variances in two-dimensional (x, y) Euclidean space, respectively, as shown in Graph 12-5. In the case of the 3D sphere 12-3 in 3D Euclidean space, each point in the (x, y, z) dimensions corresponds to the 3D Gaussian density. Thus, for a particular type of algorithm described below, that algorithm can be initialized with a known prior probability density model as shown in Graph 12-5.

0076Here, with reference to FIGS. 23-25, exemplary processing and formulation is presented that further improves and / or makes the RFID tag position estimation more accurate. In the following, the observed sample space Ω = {Y^j_t, t, j} is shown. P (x)^j_t| Y^j_t) Represents the probability distribution function position at time t, based on all past AOA measurements. Bayesian model is a recursive estimation p (x)^j_tProvides a stochastic framework for | θ) and vector θ = (θ)_j... θ_t) Specifies the angle of the arrival vector.

0077In an indoor propagation environment, the direction of the main signal may, in some cases, be due to the reflected signal rather than the direct path. Address this situation to avoid false estimates of the actual location of the signal source. Tag position {x_t; t N}, x_t X (t may represent an iterative exponent) has an initial distribution of p (x)₀) In Markov's relationship P (x_t| x_t-1) Is modeled as a primary Markov process. Observed series of tag signals Y₁ Ω can contain both complex and real measurements, and the estimates made by the reader for each array element are shown in Figure 23.

0078The observation vector from the jth tag is Y^j_t= (y_t^j, y_t^j_-1, y_t^j_-2, ..., Y^j₀), Each y_t^jIs a vector. P (x in one aspect^j_t| Y^j_tThe probability density function is recursively obtained in two stages: the prediction stage and the update stage. x_tThe prior probability density function at time step t (excluding the part that depends on j for the sake of clarity) used to predict is:<maths num="11"><img id="000013" he="10" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> The update by Bayes' law is:<maths num="12"><img id="000014" he="11" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> In the above formula, P (y_t| y_t-1) = P (y_t| x_t) P (x_t, | y_t-1) dx_tAnd the initial state is P (x)₀| Y₀). Equation (8) is P (x)_t| Y_t) = W_tP (x)_t| Y_t-1,), And the weight is defined by the following equation.<maths num="13"><img id="000015" he="11" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>

0079Recursive estimation P (x)^j_t| Y^j_t) Are provided in various aspects. FIG. 24 shows each step of the general form of the sequential Monte Carlo method. Conditional density function P (x)^j_t| Y^j_t) Is updated every iteration (14-1), the weight is 14-2, the function of the previous weight value (defined later), and in some cases the random parameter p (defined immediately). And will be updated based on. Prediction step 14-3 is P (x)^j_{t + 1}| Y^j_t), And then a new sample Y from the sample space^j_{t + 1}Including getting. This last step is called resampling.

0080One resampling technique is to evaluate the density using a point approximation. Using the classical Monte Carlo method,<maths num="14"><img id="000016" he="6" wi="39" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>By applying the histogram averaging by x_tGiven the empirical distribution of, in this case {x (i)} is extracted from the random source of the probability distribution P (x). Each time a set of measurements is made, the likelihood of each of the previous measurements can be estimated.

0081According to various aspects, the system utilizes multiple problem-solving means, as previously described. FIG. 25 shows a general method of locating RFID tags / sensors in various ways. First, there are multiple problem-solving means candidates, and each problem-solving means set itself is 15-1, and by sampling from Ω, the likelihood is estimated as shown in 15-2, and the probability density P (x).^j_{t + 1}| Y^j_t) Applies to one of many choices to calculate. Various techniques shown in 15-3 can be used, such as rejection sampling, importance sampling and sampling importance resampling (SIR), annealing, particle filtering, and unscented conversion techniques. Each of these techniques uses a slightly different technique to calculate the weight sequence when resampling over time, and resampling is done by restoring from Ω N times in all cases.

0082(Sampling Importance Resampling Estimator) The recursive SIR method is performed as follows. 1. Set t = 0 and M samples xⁱ_{o o}We get Ω (i = 1, ..., M). 2. Weight update: Likelihood weight w_i= P (y_t| x^j_t) (I = 1, ..., M) is calculated. 3.<maths num="15"><img id="000017" he="14" wi="51" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Normalize the weights by. 4. Resampling: Discrete set<maths num="16"><img id="000018" he="11" wi="39" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths><maths num="17"><img id="000019" he="11" wi="39" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Restore N times from and a new set<maths num="18"><img id="000020" he="9" wi="20" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>, I = 1, ···, N is generated. 5. Prediction: Predict each resampling state k times independently, where<maths num="19"><img id="000021" he="9" wi="51" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>, I = 1, ···, N, m = 1, ···, k. 6. Go to step 2 and repeat with t = t + 1.

0083After several iterations, modify the resampling step with a known distribution surrounding the predicted position of the RFID tag near the exciter to avoid a degenerate solution with only one candidate state vector value. can do. As an important function of each coordinate, the mean m and variance σ of the density²Is equal to the position of the exciter plus the correction term (the intermediate range between the exciter and the farthest tag illuminated by the exciter), and the variance σ²An independently and equally distributed Gaussian distribution density N (m, σ) is selected, where is equal to one of the diameters of the ellipses in 3D Euclidean space.

0084In this case, the importance sampling is the proposed distribution q (x)._t| Y_t) = q (x)_t| x^j_t-1,Y_t) q (X_t-1| Y_t-1) To generate a sample. In this version of the particle cloud degeneracy, the relative efficiency of the importance sampling method is related to the ratio between the variance of the critical sampling estimates and the variance of the estimates if a complete Monte Carlo simulation was possible. is there. The amount is<maths num="20"><img id="000022" he="15" wi="20" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Can be estimated by N_threshIs a preselected threshold at which the resampling technique is applied to the particle set.

0085(Expanded particle filter estimator) In one embodiment, the extended particle filter technique begins with the generation or selection of N inputs or samples (t = 0 and).<maths num="21"><img id="000023" he="15" wi="20" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Set and q (_xo|_{yo yo})) N samples xⁱ₀ Ω (i = 1, ..., N) is obtained). Then, the weights for each sample (i = 1, ..., N) are calculated according to the following function.<maths num="22"><img id="000024" he="12" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Normalize according to the following formula.<maths num="23"><img id="000025" he="14" wi="20" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths><maths num="24"><img id="000026" he="13" wi="20" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> Relative efficiency is a preselected threshold<maths num="25"><img id="000027" he="7" wi="26" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>If it is larger than, skip resampling. Otherwise, a discrete set<maths num="26"><img id="000028" he="8" wi="26" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths><maths num="27"><img id="000029" he="6" wi="26" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Restore N times from and a new set<maths num="28"><img id="000030" he="7" wi="13" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Resampling by generating, i = 1, ..., N and weighting<maths num="29"><img id="000031" he="11" wi="13" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Reset to. The prediction is then made k times independently for each of the states or resampled states, where x^j_{t + 1}q (x)_{t + ι}| x^j_t, Y_{t + l}) (I = 1, ..., N, m = 1, ..., k). Then, the process is repeated for the next set (t = t + 1), and the weight for the new sample is calculated.

0086(Metropolis-Hasting Algorithm Estimator) The Metropolis-Hasting algorithm uses a Markov chain model for the observed sequences and estimates when generating samples using the proposed distribution, and the candidate sample z is the proposed q (z | x). It is extracted from and accepted with the probability given by the following equation (p, q, π have different distributions).<maths num="30"><img id="000032" he="13" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> When the Markov chain moves to a new dataset, the candidate is accepted or rejected, and the rejected object leaves the Markov chain at the current data point in the state space. When π (x) = p (x | y) is selected, the acceptance probability becomes simple as in the following equation.<maths num="31"><img id="000033" he="15" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> The Metropolis-Hasting algorithm is summarized below. 1. Set t = 0 and x₀Are randomly or deterministically selected. 2. z ~ q (z | x_t) And u ~ U (0,1). 3. Calculate the acceptance probability: α (x, z). 4. Prediction: If u a (x, z), then a new sample x_{t + 1}Accept = z, otherwise x_{t + 1}= x_tAnd. 5. Go to step 2 and repeat with t = t + 1. In step 4, by adopting a statistical mechanics method that introduces an energy or goodness-of-fit function for the state of the system, in this case x_tThe probability density in the topological space of is e^{-βE (Xt)}Is proportional to, where β = 1 / kT, where T is the absolute Kelvin temperature, and k is the Boltzmann constant 1.38 × 10.^-23J / Kelvin. Improving energy or goodness of fit by transitioning from one state to another is the difference between two energy states such that the energy is reduced at each iteration, ie ΔE = E (x)._{t + 1})-E (x)_t), That is, the transition probability of the state is<maths num="32"><img id="000034" he="18" wi="77" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> And here,<maths num="33"><img id="000035" he="7" wi="13" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> Is. Initial state T₀T_n&lt;T_n-1&lt;T_n-2And in each iteration T for the dataset_n&lt;T_n-1&lt;T_n-2By using the topological locus of a solution that is aperiodic and irreducible and follows the state of the irreducible Markov chain, when further restrictions are applied to monotonically reduce the temperature T, the solution is approximately It is expected to converge to the optimum estimate.

0087(Unscented conversion estimator) Unscented conversion is another method of estimating the position of FRID tags. Covariance matrix<maths num="34"><img id="000036" he="11" wi="51" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>(here,<maths num="35"><img id="000037" he="10" wi="7" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Represents the mean value of a random variable x) to address the problem of approximating the distribution of N-dimensional random variables using means and covariances. Affine transformation<maths num="36"><img id="000038" he="8" wi="32" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Is defined, in this case,<maths num="37"><img id="000039" he="9" wi="11" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Is a characteristic<maths num="38"><img id="000040" he="7" wi="26" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Is the square root of the matrix of X with. Therefore, the unscented conversion method is summarized as follows. 1. Initialization<maths num="39"><img id="000041" he="30" wi="127" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> 2. Definition<maths num="40"><img id="000042" he="9" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> 3. Time update<maths num="41"><img id="000043" he="39" wi="119" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> 4. Weight update<maths num="42"><img id="000044" he="8" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> λ is a composite scaling factor, n_a= n_x+ n_v+ n_n, Q is the process noise covariance. R is the measurement noise covariance matrix.

0088(Estimator based on differential expansion) Another layer that optimizes RFID tag location discovery starts with a population of possible solutions (rather than a single solution). The first population is selected with consideration for spanning the space of the exciter region as large as possible. In one aspect, a uniform probability distribution is first utilized for all random locations. If a preliminary solution is available, the first population will reference a normally distributed random deviation x_nominalOften produced by adding to. Derivative expansion (DE) provides a method for generating experimental parameter vectors. DE creates a new parameter vector by adding a weighted difference vector between the two population elements to the third element. If the resulting vector yields an objective function value lower than a given population element, the newly generated vector will be compared and replaced by the next generation. The vector comparison can be part of the above generation process, but it does not have to be. In addition, the best parameter vector x_{Best, G}Is evaluated on a per-generation G basis to maintain the flow of progress formed during the minimization process. A convergent solution is obtained by extracting distance and direction information from the population to generate random deviations. Test vector is generated for each<maths num="43"><img id="000045" he="9" wi="64" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Introduced every r₁, R₂, R₃Is randomly selected between 1 and L, L is the number of subsequent generations, is a fixed parameter throughout the expansion, and μ is the step size of the delta change from one generation to the other. To control. This process is summarized in Figure 26, starting with a possible solution population at 16-1, then performing some goodness-of-fit calculation at 16-2, and sorting and population selection at 16-3. Change and check the stop rule at 16-3.

0089(Estimator based on ant colony optimization) In various other embodiments, other nonlinear stochastic optimization algorithms are utilized by considering the population of solutions and updating the likelihood of each solution with some selected metrics. Frequently used, called ant's nest optimization, a metric called a pheromone is defined by:<maths num="44"><img id="000046" he="24" wi="153" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths> For iterative t, consider the N possible solutions (rather than the next one solution) and the next k solutions, and the initial condition τ_ij (1-p) τ_ij,Solution metric τ with (i, j) A_ij τ_ij+ Δτ^kIs set to calculate the probability j of the solution. N^k_iIndicates the number of solutions in the vicinity of the k-th solution of the i-th iteration. In this way, the possible degeneracy problem is avoided at the algorithm level by considering a plurality of solutions at the same time in the solution space. This can be applied to any of the above algorithms by treating each solution as a single point in the planar graph and finding the best path in the graph by the solution metric described here. This technique is similar to so-called "gene programming" or "ant's nest optimization".

0090(Stop rule) According to various aspects, the presented approach has no specific restrictions on the form or type of suspension rule. For brevity here, several different stopping rules are presented as used according to various aspects of the invention. Where the distance is calculated d (x_t, x_{t + 1}), For example | x_t-x_{t + 1}| And d (x_t, x_{t + 1}) &lt;&lt; If ε, the algorithm is stopped. The suction area is A = {x_t, d (x_t, x^*) Defined by ε}, x^*Indicates the optimal solution, and ε is a small positive number. For algorithms that utilize discrete Markov chain techniques such as the Metropolis-Hasting method, a portion of the unobstructed state space is minimized to reach the suction region. This technique uses a count to visit each state of Ω, so the count is incremented each time a state is visited. The stop rule is<maths num="45"><img id="000047" he="9" wi="20" file="JP2017122735A_D0001.tif" img-format="tif" img-content="drawing" /></maths>Further, it is assumed that the distance standard is satisfied.

0091(Position estimation for shelves with racks arranged vertically) The final use of the present invention is shown in FIG. The figure details a series of adjacent, vertically arranged shelves used to store inventory. Such shelves are usually in warehouses that store inventory. In current technology, inventories often use bar code readers to record a series of distinguished names of goods along with the x, y, z positions of the shelves, moving from one part of the shelf to the other. Therefore, it must be counted manually. In contrast, throughout the specification, the RFID locating systems described in various embodiments are continuous states relating to articles stored on shelves without the machine or human moving from one area to the next. Can inform you. This is achieved in one embodiment by arranging the radio excitation points at high positions in the area between the passages so that they do not move regularly. Along with the particle position identification method, the excitation output control and / or the RFID tag reading rate per exciter can be used to specify the position of the article in cases such as storage in which racks are arranged vertically. In particular, the read rate of the RFID tag per exciter and per excitation output level provides dimensional information in the z direction for identifying the position of the RFID tag. On the other hand, of course, as a special example of vertical rack storage, where the vertical position (z) of the goods is known in advance as a known constant, the floor (also known as the flooring of the goods). There is also storage at height. Of course, simply stacking articles on top of each other is also a special case of vertical rack storage. In this case, the physical shelves do not separate the articles in the "Z direction" dimension, but to be precise, the heights of the articles are delineated in the order in which they are stacked.

0092Although the above description includes many specific embodiments of the present invention, it should not be construed as limiting the scope of the present invention, but rather as an example of one embodiment of the present invention. Therefore, the scope of the present invention should be determined not by the illustrated embodiment but by the appended claims and their equivalents.