Label anti-serial-reading identification method and device based on multiple ports, medium and product

Abstract

The invention discloses a multi-port-based tag anti-serial-reading identification method and device, a medium and a product, and relates to the field of data processing. The method comprises the following steps: by taking minimization of a signal radiation overlapping region between antennas as a target, adjusting real-time transmitting power of a plurality of antenna ports; after the antenna ports transmit detection signals at the real-time transmitting power, backscattering signals returned by the read tag are obtained through the antenna ports; according to the feature data of the backscattering signal, determining a key feature for judging the affiliation of the read tag; based on the key features, calculating the posterior probability of each antenna port to which the tag belongs through a Bayesian probability framework; and determining the antenna port with the maximum median value of the posterior probabilities as the attribution port corresponding to the tag. The problem of label serial reading can be relieved under the condition that the label reading rate is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a multi-port-based tag anti-cross-read identification method, device, medium, and product. Background Technology

[0002] With the rapid development of IoT technology, UHF RFID (Ultra-High Frequency Radio Frequency Identification) technology has been widely used in many fields such as smart shelves, high-density warehousing, and logistics tracking due to its advantages such as non-contact, long-distance, and simultaneous identification of multiple tags.

[0003] In these practical applications, because the antennas of multi-port readers are densely installed on shelf panels or inbound / outbound aisles, the electromagnetic radiation fields of each antenna inevitably overlap, such as... Figure 1 As shown. This makes the reader prone to cross-reading during operation, meaning that a single antenna may incorrectly read tags that should belong to the coverage area of ​​other antennas.

[0004] To address the aforementioned cross-reading problem, related technologies typically reduce the transmission power of each antenna in the reader to narrow the electromagnetic coverage of each antenna, segment the radiation fields of adjacent antennas, and avoid overlapping of the radiation fields of adjacent antennas, thereby achieving the purpose of preventing cross-reading.

[0005] However, as Figure 2 As shown, reducing the signal transmission power may cause tags located at the edge of antenna coverage or in areas with weak signals to be unreadable, which in turn makes it impossible to guarantee the tag read rate. Summary of the Invention

[0006] To address the aforementioned technical problems and deficiencies, the purpose of this invention is to provide a multi-port-based tag anti-cross-read identification method, device, medium, and product that can alleviate the problem of tag cross-read while ensuring tag reading rate.

[0007] To achieve the above objectives, in a first aspect, the present invention provides a multi-port-based tag anti-cross-read identification method, comprising: adjusting the real-time transmission power of multiple antenna ports with the goal of minimizing the signal radiation overlap area between antennas; after the antenna ports transmit detection signals at the real-time transmission power, acquiring the backscattered signals returned by the read tag through each of the antenna ports; determining key features for determining the affiliation of the read tag based on the feature data of the backscattered signals; calculating the posterior probability of the tag affixing to each of the antenna ports using a Bayesian probabilistic framework based on the key features; and determining the antenna port with the largest median posterior probability as the affiliation port corresponding to the tag.

[0008] This invention does not simply reduce power, but rather aims to minimize the signal radiation overlap area between antennas by intelligently adjusting the real-time transmit power of each antenna port. This optimizes coverage to some extent and reduces unnecessary and excessive overlap, while avoiding excessive power reduction that sacrifices readability. More importantly, upon receiving the tag's backscattered signal, this invention does not directly perform a hard assignment determination. Instead, it deeply analyzes signal strength, phase, and other characteristic data, and based on these key features, uses a Bayesian probabilistic framework to calculate the posterior probability of the tag belonging to each antenna port. Through this probabilistic reasoning approach, it comprehensively considers signal information from different antennas, assesses which antenna coverage area the tag is most likely to be located in, and ultimately determines the antenna port with the highest posterior probability value as the tag's assigned port. This determination method based on signal characteristics and probabilistic models can more accurately "decouple" the tag from a specific antenna, even in the presence of some signal overlap or interference, improving anti-crosstalk capabilities. By optimizing power and intelligent determination, it solves the crosstalk problem while ensuring the tag's readability.

[0009] Optionally, in some embodiments, with the goal of minimizing the signal radiation overlap area between antennas, the real-time transmit power of multiple antenna ports is dynamically adjusted according to the environmental changes. This includes: obtaining an objective function with the transmit power of each antenna port as a variable, the objective function being used to minimize the field strength difference integral within the signal radiation overlap area formed by any two antennas in space; using an iterative optimization algorithm to calculate the gradient of the objective function with respect to the transmit power of each antenna port; and iteratively updating the transmit power of each antenna port according to the direction of the gradient until the objective function converges to a minimum value to obtain the real-time transmit power.

[0010] By employing the technical solution described in the above embodiments, and by minimizing the integral of the field strength difference within the overlapping signal radiation region between antennas, and using an iterative optimization algorithm to adjust the transmit power of each antenna port, the antenna coverage range can be effectively controlled and optimized, reducing spatial interference and overlap between adjacent antenna signals. This helps to more clearly distinguish tags within different antenna coverage areas, reduces the aliasing effect caused by multipath propagation, and provides more accurate raw data for subsequent attribution determination based on signal characteristics, thereby improving the overall accuracy and robustness of identification.

[0011] Optionally, in some embodiments, the objective function includes: Among them, P i Let G be the dynamic power of the i-th antenna, (θ,φ) be spherical coordinates, and G be the dynamic power of the ith antenna. i (θ,φ) is the antenna pattern function, P j Let G be the dynamic power of the j-th antenna.j (θ,φ) is the antenna pattern function, V is the volume of the overlapping area covered by the i-th antenna and the j-th antenna, and N represents the total number of antennas.

[0012] The technical solution described in the above embodiments specifically provides an objective function expression for minimizing the signal radiation overlap region between antennas. This objective function quantifies the field strength difference between any two antennas within the overlap region. By minimizing this integral value, antenna power allocation can be directly optimized, thereby minimizing the field strength difference within the overlap region. This provides a clear mathematical model to guide the power adjustment process, ensuring that power optimization is operable and measurable.

[0013] Optionally, in some embodiments, the feature data includes signal strength, phase, and wavelength; determining key features for determining the affiliation of the read tag based on the feature data of the backscattered signal includes: calculating the phase difference of the backscattered signal between each of the antenna ports based on the phase, and calculating the phase distance difference based on the wavelength and the phase difference; determining the propagation delay difference of the backscattered signal reaching each of the antenna ports, and calculating the delay distance difference based on the propagation delay difference; generating a comprehensive distance difference based on the phase distance difference and the delay distance difference; and constructing the key features based on the comprehensive distance difference, the signal strength, and the phase.

[0014] By employing the technical solution described in the above embodiments, and combining signal strength, phase, and wavelength information, and further calculating phase difference, time delay difference, and the resulting comprehensive distance difference, key features for determining tag affiliation are constructed. These features comprehensively utilize the spatial location information (distance, angle, etc.) carried by the radio frequency signal during propagation, and compared to a single feature (such as signal strength), they can more comprehensively and accurately reflect the tag's positional relationship relative to different antennas. This provides richer and more discriminative inputs for the subsequent Bayesian probabilistic framework, significantly improving the accuracy and reliability of affiliation determination.

[0015] Optionally, in some embodiments, the relationship between the signal strength and the tag antenna distance is characterized by a modified free-space path loss model, which includes: RSSI = P TX +G TX +G RX -20log 10 (4πd / λ)–L env ; Where RSSI is the signal strength, P TX For transmission power, G TX For the transmit antenna gain, G RXLet L be the receiving antenna gain, λ be the signal wavelength, d be the tag antenna distance, and L be the distance to the receiving antenna. env Environmental depletion factor; The relationship between the phase and the tag antenna distance is characterized by a two-way path model, which includes: Where φ represents the phase angle of the backscattered signal, λ represents the signal wavelength, and φ offset This indicates a fixed phase offset of the system, and mod2π indicates a phase periodic constraint.

[0016] The technical solutions described in the above embodiments, by introducing modified free-space path loss models and two-way path models, specifically characterize the relationship between signal strength and phase and the distance between the tag and antennas. These models provide mathematical tools for transforming easily measurable signal characteristics into those related to physical distance. The modified models consider environmental loss factors, making the models closer to the complex electromagnetic environment in reality. The application of these models enables more accurate estimation of the distance or relative position information between the tag and each antenna based on the received signal characteristics, providing a theoretical basis and computational methods for constructing key features.

[0017] Optionally, in some embodiments, the calculation formula in the Bayesian probability framework includes: Among them, P(A) i |x) is the posterior probability, representing the probability that the label belongs to antenna A given feature vector x. i The probability of P(x|A) i ) and P(x|A j ) represent the likelihood probabilities, P(x|A) i ) indicates that when antenna A i When covering labels, the probability of observing feature vector x is P(x|A). j ) indicates that when antenna A j The probability of observing feature vector x when the label is overlaid; P(A i ) and P(A j P(A) represents the prior probability, and P(A) represents the prior probability. i This indicates that the tag belongs to antenna A. i The initial probability, P(A) j This indicates that the tag belongs to antenna A. j The initial probability; N is the total number of antennas.

[0018] The technical solution described in the above embodiments provides a specific formula for calculating the posterior probability of tag assignment based on a Bayesian probabilistic framework. This formula utilizes prior probability and likelihood probability, linking observed signal features with the probability of a tag belonging to a specific antenna through Bayes' theorem. By calculating the posterior probability of a tag belonging to each antenna and selecting the antenna with the highest probability as the assigned antenna, this method provides a decision-making mechanism based on probabilistic reasoning. It can comprehensively consider multiple factors (including prior knowledge and observational data) to make optimal judgments, improving the accuracy and robustness of tag assignment determination in uncertain environments.

[0019] Optionally, in some embodiments, after determining the antenna port with the largest posterior probability as the home port corresponding to the tag, the method further includes: receiving truth home port information associated with the actual location of the tag being read; comparing the truth home port information with the home port to generate a loss function value; and backpropagating and updating the model parameters in the Bayesian probabilistic framework based on the loss function value.

[0020] The technical solution adopted in the above embodiments introduces a model optimization mechanism based on truth-based port information. By receiving the true port information of the labels, comparing it with the port determined by the model, a loss function value is generated, and the model parameters in the Bayesian probabilistic framework are updated based on this through backpropagation. This constitutes a closed-loop training and optimization process. By continuously using truth data to "correct" and adjust the model, the Bayesian model can better learn and adapt to the signal propagation characteristics in the real environment, improve the model's generalization ability and judgment accuracy in practical applications, and enable the system to continuously improve its performance.

[0021] In a second aspect, embodiments of the present invention provide an electronic device, including: one or more processors and a memory; the memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors calling the computer instructions to cause the electronic device to perform the method described in the first aspect and any possible implementation thereof.

[0022] Thirdly, the present invention provides a computer-readable storage medium including instructions that, when executed on the electronic device, cause the electronic device to perform the method described in the first aspect and any possible implementation thereof.

[0023] Fourthly, the present invention provides a computer program product containing instructions that, when the computer program product is run on the electronic device, cause the electronic device to perform the method described in the first aspect and any possible implementation thereof.

[0024] It is understood that the electronic device provided in the second aspect, the storage medium provided in the third aspect, and the computer program product provided in the fourth aspect are all used to execute the method provided by this invention. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0025] The present invention provides one or more technical solutions, which have at least the following technical effects or advantages: 1. Improved accuracy and anti-cross-reading capability of tag identification. Through a tag attribution determination method based on signal features and a Bayesian probabilistic framework, and through detailed construction and modeling of key features, the present invention can comprehensively utilize multi-dimensional signal information (such as signal strength, phase, distance difference, etc.) and use probabilistic reasoning to determine the true attribution port of the tag. This method overcomes the limitations of traditional methods that rely solely on signal strength or simple power segmentation. Even in complex environments where antenna signals overlap and interfere, it can more accurately distinguish tags in different areas, reduce cross-reading rate, and improve the accuracy of multi-tag identification in densely deployed scenarios.

[0026] 2. Optimizing Antenna Coverage While Maintaining Read Rate: The proposed power adjustment method, aimed at minimizing signal radiation overlap, along with its specific objective function, enables the system to intelligently optimize the transmit power of each antenna port. This optimization is not simply about reducing power, but rather about minimizing interference and overlap between adjacent antennas while ensuring a certain coverage area. This avoids the tag miss problem caused by excessive power reduction in traditional anti-crosstalk methods, effectively balancing the trade-off between anti-crosstalk performance and tag read rate, and improving the overall performance and reliability of the system in practical applications.

[0027] 3. Enhanced System Adaptability and Robustness. The introduction of a model backpropagation and parameter update mechanism based on truth information imbues the entire Bayesian probabilistic framework with learning and adaptive capabilities. By receiving actual label assignment truth values, the internal model parameters can be continuously optimized and adjusted to better adapt to complex electromagnetic environments and deployment scenarios. This continuous optimization process enhances the system's robustness in different environments, enabling it to maintain high recognition performance over time and with environmental changes, reducing reliance on manual debugging and calibration, and improving practicality and ease of use. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of a typical UHF RFID technology application scenario; Figure 2 This is a schematic diagram illustrating application scenarios for solving the problem of cross-reading in related technologies; Figure 3 This is a schematic diagram illustrating an application scenario for solving the problem of serial reading in an embodiment of the present invention; Figure 4 This is a flowchart of a multi-port-based tag anti-cross-reading identification method according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the architecture of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0029] The terminology used in the following embodiments of the present invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification of the invention, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in the invention refers to any or all possible combinations comprising one or more of the listed items.

[0030] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of the present invention, unless otherwise stated, "a plurality of" means two or more.

[0031] It should also be noted that, unless otherwise explicitly specified and limited, the terms "setting" and "connection" in the embodiments of the present invention should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components; it can be a wired communication connection or a wireless communication connection. Those skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific circumstances. The embodiments of the present invention will be described in detail below.

[0032] To address the issue of tag cross-reading, relevant technologies generally employ a fixed power threshold method. This method reduces the transmission power of each antenna on the reader, forcibly narrowing the electromagnetic coverage area of ​​each antenna. This separates the radiation fields of adjacent antennas, preventing overlap and thus reducing the probability of tags from different areas being read simultaneously, thereby minimizing cross-reading. However, to narrow the antenna coverage area, the transmission power must be reduced, which directly results in tags located at the edge of antenna coverage or in areas with weak signals being unreadable, leading to a low tag read rate.

[0033] Related technologies attempt to prevent cross-reading by reducing power or using timing isolation, but this often comes at the cost of performance. This invention takes the opposite approach, providing a multi-port-based tag anti-cross-reading identification solution, such as... Figure 3 As shown, cross-reading is allowed to a certain extent, but subsequent algorithm analysis is needed to accurately identify and clarify the antenna port to which the tag belongs.

[0034] This invention does not rely on reducing signal power to prevent crosstalk, allowing the antenna to operate at power levels that ensure a high read rate. It addresses the crosstalk problem at the software and algorithm level, decoupling the tasks of "reading" and "distinguishing," thereby achieving accurate attribution determination while ensuring that the vast majority of tags can be successfully read, thus guaranteeing system efficiency.

[0035] This invention first aims to minimize the overlapping area of ​​signal radiation by actively and dynamically optimizing the transmission power of each antenna to create the optimal physical environment for subsequent identification, rather than simply reducing power. More importantly, this embodiment does not completely avoid cross-reading; instead, it allows multiple antennas to read simultaneously by deeply analyzing the backscattered signals returned by the tags and extracting key features that uniquely identify their spatial attributes. Finally, using a Bayesian probabilistic framework, these key features are calculated to intelligently and with high probability determine the final port of each read tag. This shift from "physical isolation" to "intelligent decision-making at the information level" effectively resolves the technical contradictions of related technologies. It improves the accuracy of tag attribution identification in complex and dynamically changing real-world application scenarios without sacrificing tag reading rates and inventory efficiency, thus providing a highly reliable, efficient, and cost-free solution for applications such as smart shelves and high-density warehousing.

[0036] The following is combined with Figure 4 This embodiment provides a multi-port-based tag anti-cross-read identification method, which includes the following steps: Step 201: Adjust the real-time transmit power of multiple antenna ports with the goal of minimizing the signal radiation overlap area between antennas.

[0037] In this step, the control unit inside the RFID reader or the connected host computer system runs a power adjustment algorithm. This algorithm can decide how to adjust the power based on a variety of input information.

[0038] One implementation involves the reader periodically or upon detecting environmental changes (e.g., by monitoring channel status, tag read activity, or the strength of received interference signals) performing power scans or evaluations. For example, the reader can attempt to activate each antenna port in turn with different power combinations, monitoring the background noise or the signal strength of the test tag received at each port. By analyzing this data, the reader can estimate the coverage area of ​​each antenna and the degree of overlap between them under the current power settings. Based on these evaluation results, the algorithm iteratively adjusts the transmit power settings of each antenna. For instance, if excessive signal overlap is found between adjacent antennas, the power of one or two antennas can be reduced; if the tag read rate in a certain area is low, the power of the relevant antennas can be moderately increased. This adjustment process can be based on a preset optimization model (such as minimizing mutual information, maximizing area exclusivity, etc.) or on machine learning algorithms, continuously trying and learning to find the optimal power allocation strategy.

[0039] Furthermore, the real-time transmission power in this embodiment is dynamically adjusted, meaning that the power is not constant and can change in real time according to factors such as the needs of the actual application scenario, tag density, and environmental interference levels. For example, in densely tagged areas, the system may employ more refined power control to reduce crosstalk; in sparsely tagged areas, the power may be appropriately increased to expand the coverage area.

[0040] The entire process aims to optimize the signal distribution between antennas at the physical level through intelligent power management, creating ideal working conditions with less signal overlap and less interference for subsequent accurate identification.

[0041] Step 202: After transmitting the detection signal at the real-time transmission power at the antenna port, the backscatter signal returned by the read tag is obtained through each of the antenna ports.

[0042] Specifically, the RFID reader transmits probe signals sequentially or in parallel (depending on the reader's architecture and operating mode) through its connected antenna ports, according to an adjusted real-time transmission power. These probe signals are typically radio frequency pulses or continuous wave signals conforming to the UHF RFID protocol standard. When these probe signals illuminate an RFID tag within its coverage area, the tag is activated and uses the energy extracted from the signal transmitted by the reader to modulate and reflect a backscattered signal by changing its antenna impedance.

[0043] The core task of this step is to accurately capture the backscattered signals returned by the tags being read by transmitting detection signals through the same or different antenna ports (depending on whether the reader supports full-duplex or half-duplex mode). The reader's receiving module receives these weak backscattered signals and processes them by amplification, filtering, demodulation, etc.

[0044] During this process, the reader extracts key feature data from these backscattered signals. This feature data includes not only the tag's ID information, but more importantly, physical layer information reflecting the signal propagation path and tag status, such as the received signal strength (RSSI), signal phase information, signal time of arrival (ToA / TDoA), and even the Doppler shift of the signal (if the tag or reader is moving).

[0045] The reader needs to have a high-sensitivity and high-resolution receiving circuit to ensure that it can accurately capture and digitize these signal characteristic data. For multi-port readers, especially UHF (ultra-high frequency) multi-port readers, this means that the reader needs to be able to distinguish and record which antenna port the backscattered signal of the same tag was received from, as well as the specific characteristic values ​​of the signal received at each port.

[0046] The acquired signal feature data, associated with each read tag and its receiving antenna port, will serve as crucial input for tag attribution determination in subsequent steps. The reader typically stores this raw or preliminarily processed signal feature data in its internal cache or transmits it to a connected host computer system for further analysis and processing.

[0047] Step 203: Based on the characteristic data of the backscattered signal, determine the key features used to determine the affiliation of the read tag.

[0048] After receiving the backscattered signals captured by each antenna port, the RFID reader will immediately digitize and perform preliminary processing on these signals.

[0049] Based on this raw signal data, the key features most helpful in determining the actual physical affiliation of the tag (i.e., which antenna it is closest to or has the strongest association with). The signal processing module inside the reader calculates and extracts a series of feature values ​​for each detected tag and each antenna port that receives the tag's signal.

[0050] These characteristic data may include, but are not limited to, Received Signal Strength Indication (RSSI), which intuitively reflects the degree of signal attenuation from the tag to the antenna; generally, the closer the distance, the stronger the signal. Signal phase information, the phase is closely related to the path length of signal propagation, and the phase difference of the same tag signal received by different antennas can be used to locate or distinguish the tag. Signal arrival time (ToA) or time difference of arrival (TDoA), this time information can also directly or indirectly reflect the distance relationship between the tag and each antenna. It may even include the Doppler frequency shift of the signal if there is movement of the tag or the environment.

[0051] The reader analyzes this multi-dimensional, multi-antenna-view feature data and selects the feature combinations with the highest discriminative power for distinguishing label affiliation—these are the key features. For example, if RSSI is significantly higher on one antenna port than on others, then RSSI is a key feature. If phase information exhibits a specific spatial distribution pattern across multiple ports, phase information may also be identified as a key feature. The reader may use pre-defined rules, feature selection algorithms, or a trained model to automatically identify these key features.

[0052] Ultimately, the reader generates a dataset containing these key features for each tag being read. This dataset records in detail the key signal performance of the tag at each receiving antenna port, providing accurate input for subsequent probability calculations.

[0053] In some embodiments, the feature data includes signal strength, phase, and wavelength. The relationship between the signal strength and the tag antenna distance is characterized by a modified free-space path loss model, which includes: RSSI = P TX +G TX +G RX -20log 10 (4πd / λ)–L env ; Wherein, RSSI is the signal strength, representing the power of the signal returned from the tag that is actually measured at the reader receiver; P TX Transmit power, representing the original signal power emitted by the reader antenna; G TX The transmit antenna gain represents the ability of the transmit antenna to concentrate its input power in a specific direction. It does not amplify the energy, but rather redistributes the energy, resulting in stronger energy in the main radiation direction. G RX The receiving antenna gain represents the ability of the receiving antenna to capture electromagnetic wave energy from a specific direction. λ is the signal wavelength, which is a value within a fixed range for UHF RFID; d represents the tag-antenna distance, indicating the distance between the tag and the antenna; L env The environmental loss factor is a key correction term that enables theoretical formulas to be applied in practice. It is an empirical or semi-empirical value used to quantify all the additional signal losses introduced by the real environment that do not exist in ideal free space. The environmental loss factor can take into account factors such as multipath effects (interference caused by signal reflection from walls and shelves), diffraction, absorption (attenuation of signal as it passes through goods and people), and polarization mismatch.

[0054] 20log 10 (4πd / λ) describes the natural attenuation of electromagnetic wave energy due to diffusion when propagating in ideal, unobstructed free space. It represents the physical law that signal energy decreases with increasing distance. Under ideal free-space conditions, wireless signals attenuate with increasing propagation distance. This attenuation is caused by the signal energy diffusing onto an increasingly larger sphere, and its attenuation is proportional to the square of the distance. On a logarithmic scale (in dB), this square relationship is proportional to the logarithm of the distance, i.e., 20log₂(λ / d). 10 (d). The 4π / λ in the formula is a frequency-related constant, ensuring unit consistency. Therefore, placing the distance d into the logarithmic term and multiplying it by 20 accurately reflects the fundamental physical law of signal attenuation with distance in free space.

[0055] The environmental loss factor (Lenv) is a significant modification to the standard free-space model and is key to its rationality in practical RFID applications. The free-space model assumes that signals propagate in an environment without any obstructions or reflections, which is virtually nonexistent in the real world. Actual RFID operating environments are often complex and diverse, containing walls, metal objects, people, and other electronic devices, all of which can cause absorption, reflection, scattering, and diffraction of radio frequency signals, leading to additional signal loss.

[0056] Lenv can be an empirical value, or it can be a value obtained through actual measurement or more complex environmental modeling. By introducing Lenv, the model can better approximate the complexity of signal propagation in real-world environments, making it more suitable for predicting and analyzing the signal strength-distance relationship in practical RFID systems. For example, in a warehouse environment with many metal objects, Lenv will be larger than in an open outdoor environment, reflecting the strong absorption and reflection effects of metal on radio frequency signals. Therefore, the presence of the Lenv term transforms this model from an ideal free-space model into a more practically applicable "corrected" model, improving its ability to describe and its reasonableness in the context of real-world RFID environments.

[0057] In summary, this model combines solid physical principles (free space loss) with flexible reality corrections (environmental loss factor) to construct a mathematical model that is both theoretically profound and practically valuable, providing strong and reasonable theoretical support for anti-reading identification methods.

[0058] The relationship between the phase and the tag antenna distance is characterized by a two-way path model, which includes: Wherein, φ represents the phase angle of the backscattered signal, specifically the phase angle of the tag's backscattered signal received by the reader. This is the core value measured, and its unit is radians, which directly reflects the time delay of the signal's round-trip propagation. λ represents the signal wavelength; d represents the tag-antenna distance, indicating the distance between the tag and the antenna; φ offset This indicates a fixed phase shift in the system, including inherent deviations introduced by hardware circuit delays, antenna group delays, and environmental reflections, which usually need to be compensated for through calibration measurements. mod2π represents the phase periodicity constraint, which takes the total phase value modulo 2π to ensure that the output phase is within the range of [0, 2π). This is the key operation to solve the phase ambiguity problem.

[0059] Step 204: Based on the key features, calculate the posterior probability of the tag belonging to each of the antenna ports using a Bayesian probability framework.

[0060] The Bayesian probabilistic framework allows the system to update its confidence in tag attribution by combining prior knowledge with evidence observed through key features. In practice, the reader's internal or associated processing system maintains a probabilistic model. This model includes: 1) a prior probability P(attribution|antenna i) that a tag might belong to a particular antenna, which can be a uniform distribution (initially assuming the tag belongs to any antenna with equal probability) or a distribution based on historical data; 2) and most importantly, a likelihood model P(key features|attribution|antenna i), which describes the probability of observing the set of key features (e.g., a specific combination of RSSI, phase values) extracted in step 203, assuming the tag does indeed belong to a specific antenna i. This likelihood model typically needs to be built through offline training or field calibration, reflecting the typical distribution patterns of signal features within different antenna coverage areas.

[0061] Based on Bayes' theorem, the reader calculates the posterior probability: P(attribute|antenna i|key feature) = [P(key feature|attribute|antenna i) * P(attribute|antenna i)] / P(key feature). P(key feature) is the total probability of observing these features, serving as a normalization factor. The reader repeats this calculation for each possible antenna port (i from 1 to N, where N is the number of antenna ports), obtaining a set of posterior probability values. This set of probability values ​​intuitively represents the likelihood that the tag belongs to each antenna port based on the currently observed signal features.

[0062] The specific calculation process includes: 1) For each possible antenna port, calculate the product of its likelihood probability and prior probability to generate an assignment weight value. 2) Summate the assignment weight values ​​of all antenna ports to obtain an evidence value as a normalized denominator. 3) Divide the assignment weight value of each antenna port by the evidence value to determine the final posterior probability that the tag belongs to each antenna port.

[0063] Specifically, the first step of the calculation process involves a preliminary quantitative assessment of each possible antenna affiliation. Specifically, the system iterates through every antenna port that might read the tag, for example, antennas 1 through N. For any antenna i, the computer performs a multiplication operation. The two factors in this multiplication are the likelihood probability and the prior probability. The likelihood probability is calculated using a pre-trained Gaussian Mixture Model (GMM), answering the question, "If the tag truly belongs to antenna i, how likely is it that this feature vector can be observed?" The prior probability is an initial guess based on simple metrics such as signal-to-noise ratio (SNR). Multiplying these two yields an affiliation weight value, which combines the model's complex judgment with the initial simple judgment; this can be seen as a raw score indicating that the tag belongs to antenna i.

[0064] The calculation process in step 2 aims to provide a common benchmark for subsequent probability normalization. In step 1, an independent "attribution weight value" has been calculated for each antenna port (from 1 to N). This step is straightforward: all these weight values ​​are summed together to obtain a total. This sum is called the "evidence value" in Bayesian theory; it represents the overall probability of observing this specific feature vector in the current system environment, regardless of which specific antenna it originates from. This evidence value itself is not used for direct determination; its sole and crucial role is to serve as the common denominator in the next step of the calculation, ensuring that the final sum of the attribution probabilities of all antennas is exactly equal to 1.

[0065] Step 3 is crucial for arriving at the final decision. It requires iterating through all antenna ports (from 1 to N) again. For each antenna i, the antenna's unique attribution weight calculated in step 1 is retrieved, and this value is divided by the unique, shared evidence value calculated in step 2. The result of this division is the final posterior probability that the tag belongs to antenna i. This posterior probability is a normalized, precise value between 0 and 1, representing the final confidence level that the tag belongs to antenna i after considering all known information and evidence. After this step, a complete set of posterior probability values ​​that can be directly used for comparison is obtained.

[0066] Step 205: Determine the antenna port with the largest median posterior probability as the home port corresponding to the tag.

[0067] This step can be based on the "Maximum A Posteriori" (MAP) principle. For each tag that has been successfully read and processed through the first two steps, the reader now has a list of posterior probabilities for that tag belonging to each antenna port.

[0068] For example, for tag X, the reader might calculate that it has a probability of 0.8 belonging to antenna 1, a probability of 0.15 belonging to antenna 2, and a probability of 0.05 belonging to antenna 3. In step 205, the reader simply compares these posterior probability values ​​and finds the maximum one.

[0069] In the example above, the maximum value is 0.8, corresponding to antenna 1. Therefore, the reader determines that the home port of tag X is antenna 1. Even if the backscattered signal of tag X might also be received by antenna 2 or antenna 3 (leading to the possibility of cross-reading), through this intelligent discrimination based on signal characteristics and probability models, the reader can determine the most likely physical location or associated area of ​​the tag. This determined home port information will then be used in inventory counting, location applications, or other business logic.

[0070] In this way, the reader effectively solves the problem of tag cross-reading in multi-antenna environments, uniquely and accurately associating tags that might otherwise be repeatedly or vaguely identified by multiple antennas with the most likely physical location or reading area, thereby improving inventory efficiency and data accuracy.

[0071] This embodiment employs the aforementioned method and steps. First, aiming to minimize the signal radiation overlap area between antennas, the real-time transmit power of each antenna port is dynamically adjusted. This means that a relatively optimized power level can be flexibly found based on environmental changes or preset strategies, mitigating physical overlap to some extent while maximizing signal coverage. More importantly, even with signal overlap, the backscattered signal returned by the read tag is collected, and key features are extracted from it. These key features constitute a unique spatial relationship "fingerprint" between the tag and each antenna port. Using a Bayesian probabilistic framework, the posterior probability of the tag belonging to each antenna port is calculated based on these key features. This probabilistic model can comprehensively consider signal information from different antennas, and even if a tag is read by multiple antennas simultaneously, it can determine through probability calculation which antenna's coverage area the tag is most likely to belong to. Finally, the antenna port with the highest posterior probability is selected as the tag's assigned port.

[0072] The core advantage of this method lies in its elevation of the anti-cross-read problem from simple physical isolation to the level of signal processing and intelligent discrimination. It optimizes the physical environment by dynamically adjusting power and then addresses the cross-read problem at the software level by analyzing multi-antenna signal characteristics and probabilistic inference. This allows the system to effectively handle tag cross-read in complex environments while maintaining a high tag read rate, improving recognition accuracy and overall system performance, and overcoming the inherent limitations of traditional methods.

[0073] This invention also provides a multi-port-based tag anti-cross-read identification method, which specifically includes the following steps: S301, obtaining an objective function with the transmit power of each antenna port as a variable.

[0074] The objective function is used to minimize the integral of the field strength difference within the overlapping region of signal radiation formed by any two antennas in space.

[0075] The first step in developing an RFID reader is to establish a mathematical model to guide its power adjustment behavior. The core of this model is an objective function that takes the transmit power of each antenna port in the current system as input and outputs a value representing the "degree of poorness" of signal overlap between antennas under the current power configuration.

[0076] Initially, this objective function might be based on an ideal physical model, such as quantifying the integral of the signal strength difference between any two antennas in a spatially overlapping region. Theoretically, the smaller the difference in field strength, the more similar the two signals are in that region, resulting in greater overlap and increased interference or uncertainty for tag reading. However, relying solely on an ideal physical model is insufficient because real-world radio frequency environments are complex and dynamically changing.

[0077] Therefore, RFID readers adjust or parameterize this objective function based on real-time monitored environmental changes. Environmental changes may include: sensing multipath, fading, and interference through received signal strength index (RSSI) or channel state information (CSI); and analyzing tag reading performance, such as overall read rate, individual tag reading stability, tag collision rate, and cross-reading (the same tag being read by multiple antennas).

[0078] For example, if a reader detects an abnormally high tag collision rate in a certain area, it may indicate severe antenna signal overlap in that area. In this case, the reader will dynamically adjust its objective function, increasing the penalty weight for areas with high collision rates, or incorporating the collision rate itself as part of the objective function.

[0079] The objective function can evolve from a simple field strength difference integral into a composite function that comprehensively considers field strength difference, tag collision rate, cross-read rate, and even tag reading stability. This dynamic adjustment allows the objective function to reflect the optimization needs of the current specific RF environment in real time, ensuring that power adjustments are made to address the most prominent issues, thereby improving the effectiveness and environmental adaptability of the optimization process. In this way, the reader "learns" and adapts to the environment, constructing an "environment-aware" objective function that guides its subsequent power optimization direction.

[0080] In some embodiments, the objective function may specifically include: Among them, P i Let G be the dynamic power of the i-th antenna, (θ,φ) be spherical coordinates, and G be the dynamic power of the ith antenna. i (θ,φ) is the antenna pattern function, P j Let G be the dynamic power of the j-th antenna. j (θ,φ) is the antenna pattern function, V is the volume of the overlapping area covered by the i-th antenna and the j-th antenna, and N represents the total number of antennas.

[0081] The use of double summation in the objective function ensures that it considers the interaction between any two different antennas (antenna i and antenna j) in the system. This is necessary because overlap occurs between antenna pairs.

[0082] The integral symbol represents accumulation within a specific spatial region V. Here, V is defined as the volume of the overlapping area covered by the i-th and j-th antennas. This means that the objective function only focuses on the region where signals overlap, rather than the entire space, which precisely corresponds to the requirement of "minimizing the overlapping area".

[0083] Quantizing the overlap strength, the expression inside the integral (P) i Gi (θ,φ)·P j G j (θ,φ) is the effective radiated power density of the i-th antenna in the direction (θ,φ) (proportional to P). i G i The effective radiated power density of the j-th antenna in that direction (proportional to P) j G j The product of ).

[0084] P i G i (θ,φ) represents the signal strength of antenna i in directions θ,φ (more accurately, a measure of power density or squared field strength).

[0085] P j G j (θ,φ) represents the signal strength of antenna j in directions θ and φ.

[0086] Their product P i G i (θ,φ)·P j G j (θ,φ) measures the degree to which two antenna signals "exist simultaneously" and both have a certain strength at a specific point (θ,φ) in space. If at a point, the signal from antenna i or antenna j is very weak (P... i G i or P j G j If the signal strength is close to zero, then their product will be very small. The product is only large when the signals from both antennas are relatively strong at the same point. Therefore, this product term effectively quantifies the "strength" or "poorness" of signal overlap at a point in space.

[0087] At any point of overlap, the total field strength of the reader antenna radiation field is proportional to the linear superposition of the field strengths of each antenna component. The integrand P in the objective function... i G i (θ,φ)·P j G j (θ,φ) directly characterizes the electromagnetic coupling strength (energy density product) of the two antennas in three-dimensional space, which precisely reflects the physical essence of crosstalk phenomenon.

[0088] Integrating this product term over the overlapping region volume V sums up the overlap intensity at each point within the overlapping region, resulting in a quantized value representing the total degree of overlap across the entire region. A larger integral value indicates a wider range and higher intensity of simultaneous strong signals from both antennas within the overlapping region, signifying more severe overlap.

[0089] The objective function is based on the antenna's dynamic power P.i and P j The objective function is a variable. This means that the value of the objective function can be changed by adjusting these powers. Since the objective function reasonably quantifies the degree of overlap, minimizing it means finding a set of power configurations that minimizes the simultaneous signal strength of all antenna pairs within the overlapping region, thereby minimizing the signal radiation overlap area.

[0090] This objective function constructs a mathematical model that can effectively measure and guide how to minimize signal radiation overlap by precisely defining the overlapping region, reasonably quantifying the overlap intensity of each point within the overlapping region (through the product of signal strength), and accumulating these intensities within the overlapping region.

[0091] S302, using an iterative optimization algorithm, calculate the gradient of the objective function with respect to the transmit power of each antenna port.

[0092] In this step, the RFID reader needs to determine how to adjust the transmit power of each antenna to most effectively reduce the value of the objective function, thereby reducing signal overlap. This is typically achieved by calculating the partial derivative of the objective function with respect to the transmit power of each antenna, i.e., the gradient. The gradient is a vector whose direction points to the direction in which the objective function value increases the most; the opposite direction of the gradient points to the direction in which the objective function value decreases the most. Therefore, adjusting the power along the opposite direction of the gradient can gradually approach the minimum value of the objective function.

[0093] In traditional optimization methods, gradients may be calculated analytically based entirely on the mathematical expression of the objective function. However, considering that the objective function has been adjusted or parameterized according to environmental changes and may contain terms that are difficult to differentiate precisely (e.g., the part related to tag collision rate), RFID readers employ gradient calculation or evaluation methods based on environmental feedback.

[0094] One feasible approach is the numerical gradient method: the reader makes a small positive or negative adjustment to the transmit power of a particular antenna, then immediately monitors environmental changes and assesses the corresponding changes in the objective function value. By comparing the change in the objective function value before and after the adjustment with the power adjustment amount, the reader can estimate the degree and direction of the impact of the antenna power change on the objective function. This essentially involves "measuring" the partial derivative of the antenna power in the actual environment. The reader performs similar small adjustments and environmental monitoring for each antenna port sequentially, thereby estimating the gradient vector of the objective function with respect to the power of all antennas.

[0095] Another feasible approach is to leverage machine learning or reinforcement learning techniques to "learn" the relationship between power adjustments and changes in the objective function through trial and error and environmental feedback, thereby indirectly evaluating the gradient. For example, a reader can try different power combinations, observe changes in environmental metrics, and adjust its internal model based on this feedback to predict the objective function values ​​and trends under different power configurations, thus inferring the effective gradient direction.

[0096] This gradient calculation or evaluation method based on environmental changes enables the reader to overcome model uncertainties caused by complex environments, ensuring that the calculated gradient is effective for the current actual environment and can guide the power to be adjusted in a way that truly improves reading performance.

[0097] S303, based on the direction of the gradient, iteratively update the transmit power of each antenna port until the objective function converges to a minimum value to obtain the real-time transmit power.

[0098] The opposite direction of the gradient indicates the direction in which the objective function value decreases the fastest. Therefore, adjusting the power along this direction can gradually reduce the value of the objective function, thereby reducing the degree of signal overlap.

[0099] In each iteration, the reader updates the transmit power of all antennas based on the current gradient information, using a certain step size (learning rate). The updated power configuration takes effect immediately. The reader then monitors environmental changes again, calculates the new objective function value and gradient, and performs the next round of power updates. This process is repeated until the objective function converges to a minimum value. "Convergence" here not only means that the change in the objective function value itself becomes very small, but more importantly, the convergence criterion must include the aforementioned environmental changes.

[0100] Traditional optimization might stop iterating when the objective function value changes less than a certain threshold. However, in this embodiment, even if the objective function value doesn't change significantly numerically, if environmental changes indicate that tag reading performance remains unsatisfactory (e.g., a high collision rate or cross-reading still exists), or if the environment suddenly changes significantly (e.g., a new metallic object is introduced), the reader will not consider it to have converged to the optimal state. True convergence means that under the current environmental conditions, through power adjustments, the tag's overall reading performance (e.g., read rate, cross-reading rate, collision rate, read stability, etc.) has reached a relatively stable and acceptable level, and further power adjustments cannot bring significant performance improvements, and may even lead to performance degradation.

[0101] The reader continuously monitors these environmental metrics and uses them as crucial criteria for determining convergence. Only when the environmental metrics stabilize and reach the optimization target will the reader stop iterating, set the current power configuration as the "real-time transmit power," and maintain this configuration for a period until environmental changes become significant enough to trigger a new round of dynamic power adjustments. This approach, which incorporates environmental changes into the convergence determination, ensures that the optimization process not only converges mathematically but also achieves performance optimization goals in practical applications.

[0102] S304, after transmitting the detection signal at the real-time transmission power at the antenna port, the backscatter signal returned by the read tag is obtained through each of the antenna ports.

[0103] This step can be referred to the description in the previous embodiments, and will not be repeated here.

[0104] S305, calculate the phase difference of the backscattered signal between each of the antenna ports based on the phase, and calculate the phase distance difference based on the wavelength and the phase difference.

[0105] For backscattered signals emitted by the same tag, the phase accumulated during propagation will differ due to the different path lengths to different antenna ports of the reader. The reader precisely measures the phase value of the same backscattered signal as it arrives at each of its connected antenna ports.

[0106] Then, using the phase of a reference antenna (e.g., antenna 1) as a reference, the phase difference between the signal reaching each of the other antenna ports and the reference antenna is calculated. This phase difference reflects the phase delay caused by the signal propagating along different paths. Since the phase is periodic (with a period of 2π), the reader needs to handle phase ambiguity, typically by using unwinding algorithms or combining other information to determine the true phase difference.

[0107] Once the accurate phase difference is obtained, the reader converts it into a distance difference using the known wavelength of the radio frequency signal. Since a phase difference of 2π corresponds to one wavelength, the phase-distance difference can be calculated by multiplying the phase difference by the wavelength and then dividing by 2π. This phase-distance difference roughly reflects the difference in path length between the tag and different antenna ports, and is one of the important bases for determining the tag's spatial location and its affiliation.

[0108] S306, determine the propagation delay difference of the backscattered signal to each of the antenna ports, and calculate the delay distance difference based on the propagation delay difference.

[0109] In addition to phase information, RFID readers also utilize the propagation delay information of backscattered signals reaching different antenna ports. While phase provides periodic information about the distance, the delay provides absolute propagation time information, which can be used to resolve phase ambiguity and provide more accurate distance measurements.

[0110] The reader uses high-precision time synchronization and measurement technology to determine the exact time when the same backscattered signal arrives at each antenna port after being emitted from the tag. By comparing the arrival times of the signal at different antenna ports, the reader can calculate the propagation delay difference of the signal along different paths. For example, the time difference between the signal arriving at antenna 2 and arriving at antenna 1 is the propagation delay difference. This delay difference directly reflects the difference in the propagation time of the signal along different paths.

[0111] The reader then uses the speed of electromagnetic wave propagation in air (approximately the speed of light) to convert the propagation delay difference into a distance difference. The time-delay distance difference equals the propagation delay difference multiplied by the speed of electromagnetic wave propagation. Compared to the phase distance difference, the time-delay distance difference typically has higher accuracy and is non-periodic, more accurately reflecting the absolute path length differences from the tag to different antenna ports. Combining the phase distance difference and the time-delay distance difference allows for a more reliable determination of the tag's spatial location information, which is crucial for subsequent tag attribution.

[0112] S307, Generate a comprehensive distance difference based on the phase distance difference and the time delay distance difference.

[0113] While phase distance difference is sensitive to small distance changes, it is subject to periodic ambiguity; while time delay distance difference is generally more accurate and non-periodic, it may be affected in certain situations (e.g., multipath environments or clock synchronization errors).

[0114] There are several methods for generating the combined distance difference. One feasible approach is to take a weighted average of the two measurements, with the weights determined based on their respective measurement accuracy or reliability. For example, if the time delay measurement accuracy is high, the time delay distance difference can be given a larger weight.

[0115] Another feasible approach is to utilize data fusion techniques such as Kalman filtering or particle filtering to combine phase and time delay information, thereby estimating a more accurate distance difference in real time. Time delay information can also be used to resolve phase ambiguity; that is, the integer number of periods of the phase difference can be determined by using the time delay distance difference to obtain the true phase difference after unwinding. A more accurate phase distance difference can then be calculated, and finally, the unwound phase distance difference is combined with the time delay distance difference.

[0116] The purpose of generating the integrated distance difference is to obtain a more accurate and reliable estimate of the distance difference between the tag and different antenna ports than using phase or time delay alone, thereby providing a more solid spatial location information basis for subsequent tag attribution determination.

[0117] S308, Based on the integrated distance difference, the signal strength, and the phase, the key feature is constructed.

[0118] After calculating the comprehensive distance difference that reflects the tag's spatial location information, the RFID reader combines this information with other important backscattered signal features—signal strength and phase—to construct key features for determining tag ownership.

[0119] Signal strength (RSSI) reflects the attenuation of a signal during propagation and is closely related to the distance from the tag to the antenna and losses in the environment. Generally, the closer the distance, the higher the signal strength. Phase information, even after calculating the phase distance difference, still contains information about the signal propagation path and environmental characteristics in its original or unwound form.

[0120] Therefore, the reader combines the combined range difference, signal strength, and raw or processed phase values ​​(e.g., phase relative to a reference antenna) into a multi-dimensional feature vector. This feature vector contains key information extracted from the backscattered signal that best distinguishes different tags or signals emitted by the same tag at different locations. For example, a key feature might be a triple: [combined range difference vector, signal strength vector, phase vector], where each element of the vector corresponds to an antenna port.

[0121] This key feature vector integrates spatial location information (through distance difference), signal attenuation information (through signal strength), and signal propagation characteristics information (through phase) to form a unique "fingerprint" used to identify and distinguish different backscattered signals, and ultimately determine which tag they come from and which antenna coverage area the tag may currently be located in, thereby achieving tag attribution determination.

[0122] S309, Based on the key features, calculate the posterior probability of the tag belonging to each of the antenna ports using a Bayesian probability framework.

[0123] Specifically, the calculation formulas in the Bayesian probability framework include: Among them, P(A) i |x) is the posterior probability, representing the probability that, given a feature vector x, the label belongs to antenna A. iThe probability of tag attribution. Given that a specific eigenvector x of the backscattered signal has been observed, the probability that the signal (and the tag emitting the signal) actually belongs to antenna Ai. This is the most desirable probability, reflecting the latest and most accurate judgment of tag attribution after considering new observational evidence (eigenvector x). The higher this probability value, the more likely the tag is to be covered or read by antenna Ai based on the current signal characteristics x.

[0124] P(x∣A i ) is the likelihood probability, representing the probability that when antenna A... i The probability of observing feature vector x when the label is covered. Specifically, it means the probability of observing feature vector x assuming the label indeed belongs to antenna A. i (i.e., the tag is located at antenna A) i If antenna A is within the effective coverage area, then what is the probability of observing a specific backscattered signal feature vector x? This probability measures how well the currently observed signal feature conforms to a specific attribution hypothesis. If feature vector x is within the effective coverage area of ​​antenna A, then what is the probability of observing a specific backscattered signal feature vector x? i The patterns that frequently occur when covering tags, then P(A) i |x) will be relatively high. Conversely, if the eigenvector x is at antenna A i This rarely occurs when the label is overridden, so P(A) i |x) will be relatively low.

[0125] P(A i ) is the prior probability, indicating that the tag belongs to antenna A. i The initial probability; before observing any backscattered signal eigenvector x, the tag is assigned to antenna A. i The initial probability estimate reflects prior knowledge or assumptions about tag attribution. For example, if historical data or environmental layout indicates that a tag is more likely to appear in the coverage area of ​​a certain antenna, then the prior probability for that antenna will be set higher. Without specific prior information, the prior probabilities of all antennas are typically set to be equal, meaning that before a signal is observed, the tag has an equal probability of belonging to any antenna. Prior probabilities introduce existing knowledge into the Bayesian framework and are updated upon receiving new data.

[0126] P(x∣A j ) is another likelihood probability, representing the probability when antenna A j The probability of observing feature vector x when the label is covered. Its significance is: assuming the label indeed belongs to antenna A. j (i.e., the tag is located at antenna A) jIf the antenna is within the effective coverage area, then what is the probability of observing a specific backscattered signal eigenvector x? This term appears in the summation part of the denominator and is used to calculate the total probability of observing eigenvector x when any antenna covers the tag.

[0127] P(A j ) is another prior probability, indicating that the tag belongs to antenna A. j The initial probability. Its significance is: before observing any backscattered signal eigenvector x, the initial probability of the tag belonging to antenna A. j The initial probability estimate. This term also appears in the summation part of the denominator.

[0128] Σ[P(x∣A j )*P(A j The denominator is the summation term. It represents the total probability of observing a specific backscattered signal eigenvector x, after considering all possible antenna assignments (from antenna 1 to antenna N) and their respective prior probabilities. This denominator serves a normalization function, ensuring that the calculated posterior probability P(A) is consistent with the summation term in the denominator. i |x) is a valid probability value, meaning the sum of the posterior probabilities of all antennas equals 1. It represents the overall probability that the signal could originate from any one of the antennas given the currently observed eigenvector x.

[0129] N is the total number of antennas. It represents the total number of antenna ports connected to the RFID reader in the system. The summation operation in the denominator iterates through these N antennas, calculates the joint probability (likelihood probability multiplied by prior probability) for each possible assignment, and sums them.

[0130] In this embodiment, the role and significance of the Bayesian probabilistic framework in RFID tag attribution determination lies in providing a probabilistic method that integrates prior knowledge and observational data to make decisions. Through the aforementioned calculation formula, this framework can quantify the probability (i.e., posterior probability) that a tag belongs to each antenna after receiving a specific backscattered signal feature vector. This probabilistic output is more flexible and robust than simple hard decisions (e.g., judging solely based on signal strength), especially when the signal is affected by environmental interference, multipath effects, or the tag is located at the edge of multiple antenna coverage areas, providing a more detailed distribution of attribution probability. Its significance lies in enabling RFID readers to consider multiple features (represented by the feature vector x) and prior knowledge of the tag's location (represented by the prior probability) rather than relying solely on a single or simple signal feature, thereby improving the accuracy and reliability of tag attribution determination. By calculating the posterior probability of each antenna, the reader can select the antenna with the highest probability as the tag's attribution result, or utilize this probabilistic information for more complex applications, such as probability-based tag tracking or area positioning. Essentially, the Bayesian framework transforms the label assignment problem into a probabilistic inference problem, using observational data to continuously update the belief in label location, making the decision-making process more consistent with the uncertainties in real-world applications.

[0131] In some embodiments, a feature noise covariance matrix Γ can be introduced into the Bayesian probabilistic framework to quantify observation uncertainty. Specifically, this includes the following steps: a) Construct the main diagonal elements γ of the characteristic noise covariance matrix Γ kk , SNR k Let be the signal-to-noise ratio of the k-th dimension feature, used to quantify independent measurement noise; b) Construct the off-diagonal elements γ of the Γ matrix kl Where ρ is the noise correlation coefficient between features, which is obtained through historical data statistics and is used to characterize feature coupling noise caused by multipath interference; c) Add Γ to the prior covariance matrix ∑ i This forms a noise-adaptive likelihood function: Where μ i It is the expected value (mean vector) of the feature vector x when the antenna Ai covers the tag, which is obtained through statistical learning of historical data.

[0132] d) Accelerate (∑) using Woodbury matrix identities i +Γ) -1 The inverse operation is performed, and weakly correlated terms with |ρ|<0.3 are sparsified to reduce computational complexity.

[0133] The core significance of introducing the feature noise covariance matrix Γ into the Bayesian probabilistic framework in this embodiment lies in: by dynamically quantifying the uncertainty of feature observation caused by environmental disturbances such as multipath interference and thermal noise (including phase shift, RSSI fluctuation and their coupling effects), the traditional static Gaussian model is upgraded to a noise adaptive likelihood function, so that the posterior probability calculation can reflect the changes in environmental reliability in real time (e.g. when the signal-to-noise ratio drops to 6dB, the main diagonal element γ_kk of Γ automatically increases by 3 times to expand the probability distribution range), thereby reducing the feature mismatch rate by 78% under strong interference. At the same time, by explicitly decoupling the false correlations between features caused by multipath interference through off-diagonal elements (e.g., the false correlation between RSSI and phase in metal reflection), the confidence of the attribution determination is ultimately improved.

[0134] S310, the antenna port with the largest median posterior probability is determined as the home port corresponding to the tag.

[0135] This step can be referred to the description in the previous embodiments, and will not be repeated here.

[0136] S311, Receive truth-value port information associated with the actual location of the tag being read.

[0137] To evaluate and optimize the performance of a Bayesian probabilistic framework, RFID readers need to obtain a "standard answer," namely, the truth-value port information of the read tag. This truth-value port information is typically obtained during system deployment and training phases through manual annotation or by utilizing other high-precision positioning systems. For example, during model training, an operator might place a tag within a clearly defined area covered by a specific antenna and record the tag's ID and its actual antenna port number. Alternatively, in an environment with a known, precise layout, it can be predetermined which antenna a tag read at a certain location should belong to.

[0138] After reading a tag and making an initial attribution determination, the RFID reader will look up pre-recorded truth-based attribution port information associated with the tag ID or the time of reading. This truth-based information is an objective fact independent of the Bayesian framework calculation results and is used to measure the accuracy of the model's determination.

[0139] The implementation method for receiving truth-based home port information depends on the specific system design. It may be obtained by interacting with an external database or file that stores tag IDs read in different test scenarios and their corresponding actual home antenna ports. Obtaining this truth-based information is fundamental for subsequent calculation of the loss function and optimization of model parameters, as it provides a benchmark for evaluating model performance.

[0140] S312, compare the truth value attribution port information with the attribution port to generate a loss function value.

[0141] After obtaining the true home port information of the tag being read and the home port determination result calculated through the Bayesian framework, the RFID reader will compare them to quantify the difference between the model determination result and the actual situation, and represent this difference as a loss function value.

[0142] The loss function is a concept in machine learning and optimization used to measure the degree of error in a model's predictions (or decisions). In this scenario, it compares whether the model's predicted port (i.e., the port with the highest posterior probability) matches the actual true port. If they match, the model's decision is correct, and the loss function value should be small (ideally zero). If they don't match, the model's decision is incorrect, and the loss function value should be large; the magnitude of the loss reflects the severity of the error.

[0143] Commonly used loss functions include the 0-1 loss function (a loss of 1 if a decision is wrong, and 0 otherwise) or the cross-entropy loss function (more suitable for probabilistic outputs). For the probabilistic outputs of a Bayesian framework, the cross-entropy loss function can be used, which measures the difference between the probability distribution predicted by the model and the true attribution (which can be represented as a one-hot encoded vector, with the truth port being 1 and the others being 0).

[0144] The loss function provides a clear optimization objective: by adjusting the model parameters, the loss function value is minimized, thereby making the model's judgment result closer to the real situation.

[0145] S313, Based on the loss function value, backpropagate and update the model parameters in the Bayesian probabilistic framework.

[0146] After calculating the loss function value, the RFID reader can use this loss value to guide the adjustment of model parameters through an optimization algorithm running locally or on a connected backend system, thereby reducing future decision errors. This process typically involves backpropagation and parameter updates. Although the Bayesian probabilistic framework itself is not a typical deep learning network, it may contain some parameters that need to be learned and tuned, such as the prior probability P(Ai) or the model parameters used to calculate the likelihood probability P(x|Ai) (if the likelihood probability is estimated through a parameterized model, such as a Gaussian mixture model).

[0147] Backpropagation can be understood here as calculating the gradient of the loss function with respect to these model parameters. The gradient indicates the direction in which the parameters should be adjusted to reduce the loss function. Once the gradient is calculated, optimization algorithms (such as gradient descent or its variants) can be used to update the model parameters.

[0148] The rule for parameter updates is typically: New parameter = Current parameter - Learning rate * Gradient. The learning rate is a hyperparameter that controls the step size of parameter updates. By continuously receiving new label data, obtaining ground truth, calculating loss, and performing backpropagation and parameter updates, the model parameters of the Bayesian probabilistic framework are gradually optimized. This allows the model to more accurately learn the signal feature distribution corresponding to different antenna coverage areas, thereby improving the accuracy of label attribution. This process is an iterative optimization process, usually carried out during offline or online training, with the aim of enabling the Bayesian model to better adapt to real-world application environments.

[0149] In RFID reading and writing technology, multipath interference is a common problem. It refers to the fact that radio frequency signals encounter obstacles during propagation and are reflected, refracted or scattered, forming multiple propagation paths to the receiving end. The superposition of these signals from different paths may lead to signal distortion, fading or even failure to read tags normally.

[0150] To effectively address multipath interference, this embodiment also employs a three-dimensional decoupling mechanism involving time, frequency, and space. Specifically, it combines three technologies: TDMA (Time Division Multiple Access), frequency hopping anti-interference, and spatial filtering. This process treats the signal from three dimensions—time, frequency, and space—in order to suppress the adverse effects of multipath interference to the greatest extent possible.

[0151] First, TDMA decouples signals in the time dimension. In multi-antenna systems, if multiple antennas operate simultaneously, their emitted signals and received tag response signals may interfere with each other, exacerbating the complexity of multipath effects. TDMA technology ensures that only one antenna is transmitting or receiving at any given time by assigning different antennas to different time slots for alternating operation. This effectively avoids mutual interference between antennas, simplifies signal processing complexity, and makes it easier to identify and process the corresponding multipath signals when a single antenna is operating. Through temporal isolation, TDMA creates a more favorable environment for subsequent frequency and spatial domain processing, reducing temporal aliasing of signals from different antennas.

[0152] Secondly, frequency hopping (FHH) decoupling decouples interference across the frequency dimension. Multipath effects are typically frequency-selective, meaning that the impact of multipath interference varies at different frequencies. Frequency hopping disperses signal energy across a wider bandwidth by rapidly and randomly changing the operating frequency during communication. This has the advantage that even with severe multipath fading or narrowband interference at a particular frequency, some signal will always avoid the most unfavorable frequency due to the signal "hopping" across different frequencies, thus improving communication robustness. In RFID applications, FHH helps readers attempt to communicate with tags at multiple frequency points, increasing the probability of successful reading and reducing signal distortion caused by multipath at specific frequencies. By dispersing across the frequency dimension, FHH helps mitigate the impact of multipath at specific frequencies on overall signal quality.

[0153] Finally, spatial filtering decouples signals from a spatial dimension. Multipath signals typically originate from different spatial directions. Spatial filtering is usually achieved by using an antenna array (containing multiple antenna elements) and weighting and combining signals from different directions. By adjusting the phase and amplitude of the signals received by each antenna element in the array, a beam with specific directionality can be formed, enhancing the signal from the desired direction (e.g., the direction of the tag) while suppressing interference from other directions (e.g., multipath reflections). This is like using a "spatial filter" to selectively receive signals. In RFID readers, using antenna arrays and spatial filtering algorithms allows for more precise pointing towards the tag, reducing the energy of received multipath signals and thus improving the signal-to-noise ratio and reading performance. Spatial filtering directly utilizes the spatial separability of multipath signals and is a powerful means of suppressing multipath interference.

[0154] This embodiment employs a multi-pronged strategy of time-frequency-space three-dimensional joint decoupling. TDMA isolates the antenna in time, simplifying processing; frequency hopping disperses the signal in frequency, combating frequency-selective fading; and spatial filtering distinguishes the signal source in space, suppressing reflection paths. These three technologies complement each other, working together to reduce the impact of multipath interference from different dimensions, improving the reading performance and stability of the RFID system, especially in complex electromagnetic environments. Through this joint processing, the reader can receive signals from the target tag more clearly, thereby improving the accuracy and efficiency of tag identification.

[0155] The methods provided in the above embodiments can be executed by an RFID reader / writer. This reader / writer, in addition to being a conventional reader / writer, can also possess the following functional components, making it an electronic device capable of executing the methods of the above embodiments. The electronic device in the embodiments of the present invention is described below from a hardware processing perspective. Please refer to [link to relevant documentation]. Figure 5 This is a schematic diagram of the physical device structure of an electronic device in an embodiment of the present invention.

[0156] It should be noted that, Figure 5 The structure of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0157] like Figure 5 As shown, the electronic device includes a Central Processing Unit (CPU) 401, which can perform various appropriate actions and processes according to a program stored in Read-Only Memory (ROM) 402 or a program loaded from storage portion 408 into Random Access Memory (RAM) 403, such as performing the methods described in the above embodiments. The Random Access Memory (RAM) 403 also stores various programs and data required for system operation. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An Input / Output (I / O) interface 405 is also connected to the bus 404.

[0158] The following components are connected to the input / output (I / O) interface 405: an input section 406 including audio input devices, push-button switches, etc.; an output section 407 including displays, audio output devices, indicator lights, etc.; a storage section 408 including hard disks, etc.; and a communication section 409 including network interface cards such as LAN (Local Area Network) cards, modems, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output (I / O) interface 405 as needed. Removable media 411, such as disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on the drive 410 as needed so that computer programs read from them can be installed into the storage section 408 as needed.

[0159] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by the Central Processing Unit (CPU) 401, it performs the various functions defined in the present invention.

[0160] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0161] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0162] Specifically, the electronic device of this embodiment includes a processor and a memory. The memory is coupled to one or more processors and is used to store computer program code. The computer program code includes computer instructions. One or more processors call the computer instructions to cause the electronic device to perform the method provided in the above embodiment.

[0163] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The storage medium carries one or more computer programs that, when executed by a processor of the electronic device, cause the electronic device to implement the methods provided in the above embodiments.

[0164] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

[0165] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0166] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A multi-port-based tag anti-cross-reading identification method, characterized in that, include: With the goal of minimizing the signal radiation overlap area between antennas, the real-time transmit power of multiple antenna ports is adjusted; After transmitting the detection signal at the real-time transmission power at the antenna port, the backscattered signal returned by the read tag is obtained through each of the antenna ports; Based on the characteristic data of the backscattered signal, key features for determining the affiliation of the read tag are determined; Based on the aforementioned key features, the posterior probability of the tag belonging to each of the aforementioned antenna ports is calculated using a Bayesian probabilistic framework. The antenna port with the highest probability value among the posterior probabilities is determined as the home port corresponding to the tag.

2. The method according to claim 1, characterized in that, The method of adjusting the real-time transmit power of multiple antenna ports with the goal of minimizing the signal radiation overlap area between antennas includes: Obtain an objective function with the transmit power of each antenna port as a variable. The objective function is used to minimize the integral of the field strength difference in the signal radiation overlap area formed by any two antennas in space. An iterative optimization algorithm is used to calculate the gradient of the objective function with respect to the transmit power of each antenna port; Based on the direction of the gradient, the transmit power of each antenna port is iteratively updated until the objective function converges to a minimum value to obtain the real-time transmit power.

3. The method according to claim 2, characterized in that, The objective function includes: Among them, P i Let G be the dynamic power of the i-th antenna, (θ,φ) be spherical coordinates, and G be the dynamic power of the ith antenna. i (θ,φ) is the antenna pattern function, P j G represents the dynamic power of the j-th antenna. j (θ,φ) is the antenna pattern function, V is the volume of the overlapping area covered by the i-th antenna and the j-th antenna, and N represents the total number of antennas.

4. The method according to claim 1, characterized in that, The feature data includes signal strength, phase, and wavelength; Based on the characteristic data of the backscattered signal, key features for determining the affiliation of the read tag are identified, including: The phase difference of the backscattered signal between each of the antenna ports is calculated based on the phase, and the phase distance difference is calculated based on the wavelength and the phase difference; Determine the propagation delay difference of the backscattered signal to each of the antenna ports, and calculate the delay distance difference based on the propagation delay difference; A comprehensive distance difference is generated based on the phase distance difference and the time delay distance difference; The key features are constructed based on the combined distance difference, the signal strength, and the phase.

5. The method according to claim 4, characterized in that, The relationship between the signal strength and the tag antenna distance is characterized by a modified free-space path loss model, which includes: RSSI=P TX +G TX +G RX -20log 10 (4πd / λ)–L env ; Where RSSI is the signal strength, P TX For transmission power, G TX For the transmit antenna gain, G RX Let L be the receiving antenna gain, λ be the signal wavelength, d be the tag antenna distance, and L be the distance to the receiving antenna. env Environmental depletion factor; The relationship between the phase and the tag antenna distance is characterized by a two-way path model, which includes: Where φ represents the phase angle of the backscattered signal, λ represents the signal wavelength, and φ offset This indicates a fixed phase offset of the system, and mod2π indicates a phase periodic constraint.

6. The method according to any one of claims 1-5, characterized in that, The calculation formulas in the Bayesian probability framework include: Among them, P(A) i |x) is the posterior probability, representing the probability that the label belongs to antenna A given feature vector x. i The probability of P(x|A) i ) and P(x|A j ) represent the likelihood probabilities, P(x|A) and P(x|A) are respectively. i ) indicates that when antenna A i When covering labels, the probability of observing feature vector x is P(x|A). j ) indicates that when antenna A j The probability of observing feature vector x when the label is overlaid; P(A i ) and P(A j P(A) represents the prior probability, and P(A) represents the prior probability. i This indicates that the tag belongs to antenna A. i The initial probability, P(A) j This indicates that the tag belongs to antenna A. j The initial probability; N is the total number of antennas.

7. The method according to claim 1, characterized in that, After determining the antenna port with the highest probability value among the posterior probabilities as the home port corresponding to the tag, the method further includes: Receive truth value port information associated with the actual location of the tag being read; The true value attribution port information is compared with the attribution port to generate a loss function value; Based on the loss function value, backpropagation is performed to update the model parameters in the Bayesian probabilistic framework.

8. An electronic device, characterized in that, Includes one or more processors and memory; The memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the electronic device to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium storing computer instructions, characterized in that, When the computer instructions are executed on the electronic device, the electronic device causes the electronic device to perform the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on an electronic device, it causes the electronic device to perform the method as described in any one of claims 1-7.

Images