A fixed asset inventory optimization management method based on RFID

Abstract

The application relates to an RFID-based fixed asset inventory optimization management method, and particularly relates to the field of asset management. The scheme synchronously captures original physical layer signal data strictly bound with the unique electronic code of an RFID electronic tag, generates normalized real-time radio frequency fingerprint vectors through common interference factor decoupling processing, further fuses the double reference difference degrees of the historical benchmark of the asset and the real-time state of the same group, calculates accurate state abnormality quantitative values, and drives the writing of structured state marker codes in the RFID electronic tag memory. The whole process changes the traditional inventory from passive static checking to an active asset health degree dynamic monitoring and decision support mechanism which is robust to environmental interference, can perceive the subtle changes of the asset physical state, can intelligently evaluate and warn, and realizes the synchronization of physical identification and digital system state.

Description

Technical Field

[0001] This invention relates to the field of asset management, and more specifically, to an RFID-based method for optimizing the inventory management of fixed assets. Background Technology

[0002] Archives, museums, laboratories, and specialized industrial warehouses preserve and manage a large number of assets of extremely high value or special nature, such as rare ancient books and documents, containers of volatile chemicals, precision optical instruments, and large composite material components. These assets are often extremely sensitive to environmental temperature and humidity, mechanical vibration, sealing, or structural integrity; even minor changes in their physical state can lead to irreversible damage, performance degradation, or safety risks. However, due to the inherent fragility of the assets, their special packaging, or limitations of the usage environment, directly installing various wired or independent sensor monitoring modules presents significant difficulties: on the one hand, adding external sensors can... It can damage the asset itself (such as ancient books and works of art) or change its inherent physical properties (such as the balance of precision instruments); on the other hand, in environments with many obstacles, strong interference, or hazardous materials storage, deploying complex sensor networks faces challenges such as high power supply and wiring costs and reliability issues; currently, managers mainly rely on regular manual inspections and static environmental monitoring, which is not only inefficient and has monitoring blind spots, but also cannot achieve real-time, contactless, and full-coverage perception of the physical state of individual assets (such as paper moisture content, liquid level in containers, and micro-deformation caused by internal structural stress), making preventive protection and risk warnings often lag behind the occurrence of actual damage.

[0003] Existing asset management systems based on UHF RFID technology, while capable of efficiently acquiring tag codes and associated static information through group reading for rapid asset location and inventory, suffer from fundamental limitations in their technological paradigm, restricting their application in deep asset status perception. Specifically, the core design objective of the baseband processing unit and communication protocol stack (e.g., conforming to the EPCglobal C1G2 standard) of current RFID systems is highly focused on reliably and quickly decoding the unique identifier of tags in complex channel environments. To achieve this, the signal processing algorithms at the reader end (such as channel estimation, equalization, and decoding algorithms) actively treat physical layer characteristics such as signal strength fluctuations, phase shifts, and spectral changes caused by multipath effects, path loss, environmental interference, and changes in electromagnetic coupling characteristics between the tag antenna and the attached object during radio frequency signal propagation in space as "noise" and "interference" that need to be suppressed or filtered. In other words, the entire system's signal processing link is optimized for "cleaning" the channel and extracting the digital bitstream; its hardware and firmware do not have corresponding channels and modules designed for communication with... The system collects, records, and outputs raw physical layer signal parameters generated in real time during communication, such as the instantaneous value sequence of received signal strength indication, the precise measurement value of carrier phase, or the frequency response characteristics of backscattered signals. However, it is precisely these physical layer characteristics, which are ignored by the existing technology stack, that contain key information reflecting changes in the physical state of the tagged object (i.e., the asset). For example, changes in the dielectric constant of paper caused by moisture in ancient books, changes in the equivalent load impedance of antennas caused by changes in the liquid level of chemical containers, or deformation of the antenna substrate caused by microcracks in the composite material structure will all modulate the antenna impedance of the attached RFID tag, thereby changing the backscattering characteristics of the tag to the incident radio frequency signal. This ultimately manifests as a systematic shift in the physical layer characteristics of the signal received by the reader. Existing technologies lack a complete technical solution from hardware acquisition to software processing to capture and utilize these characteristics, resulting in the sensing capabilities of RFID systems being limited to identification and rough positioning. The communication link itself cannot be built into a distributed, passive sensing network, thus losing the key technical approach for non-destructive, real-time, and indirect assessment of asset integrity and internal physical state. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing an RFID-based method for optimizing the inventory management of fixed assets, thereby resolving the issues raised in the background section.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is as follows: specifically, it includes the following steps: Step S1: When the RFID reader performs inventory communication with multiple RFID electronic tags, the baseband processing unit of the RFID reader synchronously captures and records the original physical layer signal data bound to the unique electronic code of each successfully responding RFID electronic tag. The original physical layer signal data includes at least the instantaneous value sequence of the received signal strength indication, the continuous sampling sequence of the carrier phase, and the instantaneous Doppler frequency offset sequence calculated based on the phase difference between adjacent sampling points. Step S2: For each set of original physical layer signal data bound with a unique electronic code in Step S1, estimate the common interference factor in the current environment using the original physical layer signal data of all RFID electronic tags read in the same inventory cycle; decouple the original physical layer signal data based on the common interference factor to remove common environmental interference, and extract the feature values ​​of the decoupled signal in multiple preset dimensions to generate a normalized real-time radio frequency fingerprint vector corresponding to the RFID electronic code. Step S3: Read the pre-stored standard state reference RF fingerprint vector corresponding to the RFID electronic code, and obtain the real-time RF fingerprint vector of the same type of reference tag set selected according to preset rules; calculate the first difference degree between the real-time RF fingerprint vector obtained in step S2 and the standard state reference RF fingerprint vector; at the same time, calculate the second difference degree between the real-time RF fingerprint vector and the real-time RF fingerprint vector of the same type of reference tag set; based on a state assessment model that integrates the first difference degree and the second difference degree, calculate a comprehensive state anomaly quantification value. Step S4: When the quantified value of the state anomaly calculated in step S3 exceeds the preset threshold, a tag instruction is generated; the RFID reader writes a state tag code into the memory of the RFID electronic tag according to the tag instruction. In a preferred embodiment, the specific process by which the baseband processing unit of the RFID reader synchronously captures and records the original physical layer signal data in step S1 is as follows: Inside the RFID reader's baseband processing unit, the physical layer signal sampling channel and the standard tag data decoding channel operate in parallel. When the RFID reader executes the inventory communication process, the physical layer signal sampling channel continuously captures in-phase and quadrature analog signals from the RF front-end, converting them into digitized in-phase and quadrature signal sequences, which are then input into the dedicated buffer memory of the physical layer signal sampling channel. Simultaneously, the standard tag data decoding channel performs digital protocol interaction with the RFID tag. When the standard tag data decoding channel successfully demodulates an RF... When the ID electronic tag replies with a unique electronic code, a decoding success event containing this unique electronic code and the corresponding communication time slot index is generated. Based on the decoding success event, the physical layer signal sampling channel accesses the buffer memory and, according to the start and end time points of the precise communication time slot corresponding to the communication time slot index in the decoding success event, extracts in-phase signal data segments and quadrature signal data segments that are perfectly aligned in time from the continuously buffered in-phase signal sequence and quadrature signal sequence, respectively. The in-phase signal data segments and quadrature signal data segments together constitute the original signal data block bound to the unique electronic code. Furthermore, the original signal data block is processed in real time to extract physical layer parameters: for each pair of corresponding in-phase signal data segment sampling point values ​​and orthogonal signal data segment sampling point values ​​in the original signal data block, the signal amplitude characterization value is calculated to form an instantaneous value sequence of received signal strength indication; its arctangent value is calculated to obtain the initial phase angle, and arranged to form an initial phase angle sequence; a phase dewinding operation is performed on the initial phase angle sequence to obtain a continuous carrier phase sampling sequence; based on the continuous carrier phase sampling sequence, the difference between the values ​​of the continuous carrier phase sampling sequence corresponding to adjacent sampling points is calculated to obtain the phase change, and then the phase change is divided by the product of the time interval between adjacent sampling points and twice the radius of pi, and the calculated value is used as the instantaneous Doppler frequency offset characterization value corresponding to the current sampling point, and arranged in the order of sampling points to form an instantaneous Doppler frequency offset sequence; Finally, the unique electronic code currently being processed, the corresponding communication time slot index, the timestamp corresponding to the original signal data block, the instantaneous value sequence of the received signal strength indication of length N, the carrier phase continuous sampling sequence, and the instantaneous Doppler frequency offset sequence are encapsulated together to form the original physical layer signal data.

[0006] In a preferred embodiment, the specific process of performing a phase dewinding operation on the initial phase angle sequence to obtain a continuous carrier phase sampling sequence is as follows: The initial phase angle calculated from the first sampling point in the original signal data block is directly used as the first value of the carrier phase continuous sampling sequence obtained after unwinding. For each sampling point from the second point onwards in the initial phase angle sequence, perform the following recursive calculation operation in the order of the sampling points: First, the difference between the initial phase angle value of the current sampling point and the carrier phase continuous sampling sequence value calculated at the previous sampling point is calculated to obtain the original phase difference value. Next, the original phase difference value is corrected by adding the original phase difference value to the value of pi in radians. Then, the sum is moduloed by twice the value of pi in radians. The result of the modulo operation is a value between zero and twice the value of pi in radians. Finally, the value of pi in radians is subtracted from the result of the modulo operation to obtain the corrected phase difference value. The range of the corrected phase difference value is constrained to between negative and positive pi in radians. Afterward, the carrier phase continuous sampling sequence value of the previous sampling point is added to the corrected phase difference value corresponding to the current sampling point. The result is the carrier phase continuous sampling sequence value of the current sampling point. Following this recursive calculation operation, starting from the second sampling point, all sampling points in the original signal data block are sequentially traversed and processed, ultimately generating a continuous carrier phase sampling sequence.

[0007] In a preferred embodiment, step S2, which involves estimating the common interference factor in the current environment, specifically includes: From the raw physical layer signal data, the carrier phase continuous sampling sequences of all RFID tags are extracted, and each RFID tag's carrier phase continuous sampling sequence is treated as a row vector. All row vectors are arranged in rows to construct a phase observation matrix. For the constructed phase observation matrix, weighted robust principal component analysis is performed to estimate the common interference factor. The specific iterative calculation process is as follows: A weight is initialized for each row vector corresponding to an RFID tag in the phase observation matrix. This weight is positively correlated with the local smoothness of the carrier phase continuous sampling sequence represented by the current row vector, and also positively correlated with the overall similarity between the current row vector and all other row vectors in the phase observation matrix. A common interference subspace to be optimized is defined. In each iteration, an optimization problem is solved. The objective of this optimization problem is to find a common interference subspace such that the square of the L2 norm of the residual vector generated by projecting each row vector in the phase observation matrix onto this common interference subspace, multiplied by the current weight of that row vector, results in all RFID tags being unaffected. The weighted sum of the D electronic tags is minimized; after each iteration, the L2 norm of the projection residual vector of each row vector in the common interference subspace obtained in the current iteration is calculated, and the weight of the row vector in the next iteration is dynamically adjusted according to the magnitude of the calculated projection residual L2 norm; the iterative calculation and weight adjustment are repeated until the weight change of all row vectors is less than a preset convergence threshold, the iterative calculation process is terminated, and the converged common interference subspace is obtained; after the iteration converges, the first principal component vector is extracted from the converged common interference subspace, and the extracted first principal component vector is defined as the phase common interference factor vector; For all the instantaneous value sequences of received signal strength indicators extracted from the original physical layer signal data, a received signal strength indicator observation matrix is ​​constructed, and a weighted robust principal component analysis process that is exactly the same as that for the continuous sampling sequence of carrier phase is performed to obtain the received signal strength indicator common interference factor vector. The phase common interference factor vector and the received signal strength indicator common interference factor vector together constitute the estimated common interference factor in the current environment.

[0008] In a preferred embodiment, the specific operation of decoupling the original physical layer signal data based on the common interference factor is as follows: For each RFID tag, calculate the projection coefficient of its carrier phase continuous sampling sequence onto the phase common interference factor vector. Then, subtract the product of this projection coefficient and the phase common interference factor vector from the carrier phase continuous sampling sequence to obtain the decoupled phase residual sequence. For each RFID tag, calculate the projection coefficient of its instantaneous value sequence of received signal strength indication onto the common interference factor vector of received signal strength indication. Then, subtract the product of this projection coefficient and the common interference factor vector of received signal strength indication from the instantaneous value sequence of signal strength indication to obtain the decoupled signal strength residual sequence.

[0009] In a preferred embodiment, the process of extracting feature values ​​of the decoupled signal in multiple preset dimensions and generating a normalized real-time radio frequency fingerprint vector corresponding to the RFID electronic code specifically includes: For each RFID tag, perform the following operations: A second-order polynomial is fitted to the decoupled phase residual sequence, and the ratio of the absolute value of the quadratic term coefficient to the absolute value of the linear term coefficient is defined as the phase nonlinearity eigenvalue. C1. Perform a Discrete Fourier Transform on the decoupled phase residual sequence to obtain the power spectral density. Find the frequency corresponding to the largest power value in the obtained power spectral density and determine this frequency as the main peak frequency. Extend the frequency upwards and downwards by half a preset bandwidth, centered on the main peak frequency, to form a continuous frequency band. Calculate the integral value of the power spectral density in this frequency band as the local energy. Calculate the integral value of the power spectral density over the entire frequency band from zero frequency to Nyquist frequency as the total energy. Divide the local energy by the total energy and define the quotient as the frequency domain stability characteristic value. C2. Fit the decoupled phase residual sequence with a second-order polynomial and take the ratio of the absolute value of the quadratic term coefficient to the absolute value of the linear term coefficient as the characteristic value of phase nonlinearity. C3. Perform a Hilbert transform on the decoupled signal strength residual sequence to obtain the envelope signal, and calculate the normalized Shannon entropy of this envelope signal, which is defined as the eigenvalue of the residual signal envelope entropy. The unique electronic code of the RFID tag being processed, the calculated frequency domain stability characteristic value, phase nonlinearity characteristic value, residual signal envelope entropy characteristic value, projection coefficient on the phase common interference factor vector, and projection coefficient on the received signal strength indication common interference factor vector are assembled into an original fingerprint vector in this order. Finally, for the original fingerprint vector set of all RFID tags in the current inventory period, for each feature dimension except for the unique electronic code, calculate the arithmetic mean and the absolute value of the median of all feature values ​​in the current feature dimension; for each RFID tag, subtract the arithmetic mean from its feature value in the current feature dimension, and then divide by the sum of the absolute value of the median adjusted by the scaling factor and the smoothing constant to obtain the normalized feature value of that dimension. The normalized feature values ​​after normalizing all feature dimensions are reassembled with the unique electronic code to generate a normalized real-time radio frequency fingerprint vector corresponding to the code.

[0010] In a preferred embodiment, the specific process of calculating the first difference degree and the second difference degree in step S3 is as follows: First, for each feature dimension of the real-time RF fingerprint vector, calculate the diachronic stability weight and the synchronic stability weight: the diachronic weight is equal to 1 divided by the sum of the median absolute deviation of the feature value of this dimension under the historical normal state of the tag itself and the normal protection constant; the synchronic weight is equal to 1 divided by the sum of the median absolute deviation of the feature value of this dimension of the current set of similar reference tags and the protection constant. Next, the first difference is calculated: for each dimension, the square of the difference between the real-time RF fingerprint vector and the standard state baseline RF fingerprint vector in that dimension is calculated, multiplied by the corresponding epochal stability weight, and the square root is taken after summing the products over all dimensions. Then, the second difference is calculated: First, the robust center of the real-time RF fingerprint vector of the same reference tag set is calculated. The value of each dimension of the center is the median value of all vectors in the set in that dimension. Then, for each dimension, the square of the difference between the real-time RF fingerprint vector and the robust center in that dimension is calculated, multiplied by the corresponding synchronicity stability weight, and the square root is taken after summing the products of all dimensions.

[0011] In a preferred embodiment, the process of calculating a comprehensive state anomaly metric based on a state assessment model that integrates a first difference degree and a second difference degree specifically includes: Calculate the first relative difference: Divide the first difference by the sum of the third quartile of all first difference values ​​under the historical normal state of this tag and the normal protection constant; Calculate the second relative difference: Divide the second difference by the sum of the third quartile of all second difference values ​​in the current set of reference tags of the same type and the normal protection constant; The first relative difference and the second relative difference are respectively subjected to constant exponentiation greater than 1; Multiply the first relative difference after exponential calculation by the first fusion weight coefficient, multiply the second relative difference after exponential calculation by the second fusion weight coefficient, and then add the two products to obtain the median value. The hyperbolic tangent function is used to calculate this intermediate value, and the output value is the quantified value of the state anomaly.

[0012] In a preferred embodiment, in step S4, the preset threshold on which the marking instruction is generated is a dynamic threshold; the process of calculating the dynamic threshold includes: Calculate the current environmental uncertainty factor, which is the sum of the squares of the elements of the phase common interference factor vector and the sum of the squares of the elements of the received signal strength indication common interference factor vector; The confidence factor for the current state assessment is calculated as the result of an exponential function with the natural logarithm as the base, in the following negative value: This negative value is obtained by subtracting one from the ratio of the first relative difference to the sum of the second relative difference and the protection constant, taking the absolute value, and then dividing by the scale parameter. The dynamic threshold is calculated as follows: the static base threshold plus the product of the first positive coefficient and the environmental uncertainty factor, minus the product of the second positive coefficient and the state assessment confidence factor. When the quantified value of the state anomaly calculated in step S3 exceeds this dynamic threshold, a marking instruction is generated.

[0013] In a preferred embodiment, the specific process by which the RFID reader writes a status tag code into the memory of the RFID electronic tag according to the tagging instruction is as follows: First, a status tag code with a multi-dimensional information structure is constructed. The status tag code consists of four fields in sequence: the first field is a two-bit status tag field, used to encode the overall status level of the asset to which the RFID electronic tag is attached. Different binary codes correspond to four predefined statuses: normal, watch out, warning, and severe. The second field is a four-bit level field, whose value is obtained by mapping the status anomaly quantification value calculated in step S3 to different levels, and is used to store the quantification level corresponding to the quantification value. The third field is a four-bit dominant dimension field, whose value is determined by analyzing the contribution ratio of the weighted square difference of each feature dimension of the real-time RFID fingerprint vector to the total difference when calculating the first and second difference in step S3. The index value of the feature dimension with the largest contribution ratio is filled into this field. The fourth field is a timestamp field. After constructing the status tag code, the RFID reader will only perform the operation of writing the newly constructed status tag code to the tag memory if all of the following conditions are met: The first condition is that the quantified value of the state anomaly calculated in step S3 exceeds the dynamic threshold. The second condition is that the status flag field of the old tag code read from the RFID electronic tag memory is different from the status flag field of the newly constructed status tag code; or, the level field of the old tag code is different from the level field of the newly constructed status tag code; or, the difference between the current system time and the time recorded in the timestamp field of the old tag code is greater than a preset retention time. Immediately after writing, read and verify. If there is a discrepancy, record the failure and arrange a retry; if there is a discrepancy, update the status and timestamp information of the digital asset record corresponding to the tag in the backend system.

[0014] The beneficial effects of this invention are as follows: This solution synchronously captures the original physical layer signal data that is strictly bound to the unique electronic code of the RFID electronic tag, generates a normalized real-time radio frequency fingerprint vector through decoupling processing of common interference factors, and then integrates the dual reference difference degree of the asset's own historical benchmark and the real-time status of similar groups to calculate an accurate quantified value of the state anomaly. This value is then used to drive the writing of structured status marker codes into the RFID electronic tag's memory. The entire process transforms traditional inventory from passive static verification into a proactive dynamic monitoring and decision support mechanism for asset health that is robust to environmental interference, can sense subtle changes in the asset's own physical status, can intelligently assess and warn, and achieves synchronization between physical identification and digital system status. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0018] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0019] Example 1

[0020] This embodiment provides, for example Figure 1 The method for optimizing fixed asset inventory management based on RFID, as shown, specifically includes the following steps: Step S1: When the RFID reader performs inventory communication with multiple RFID electronic tags, the baseband processing unit of the RFID reader synchronously captures and records the original physical layer signal data that is strictly bound to the unique electronic code of each successfully responding RFID electronic tag. The original physical layer signal data includes at least the instantaneous value sequence of the received signal strength indication, the continuous sampling sequence of the carrier phase, and the instantaneous Doppler frequency offset sequence calculated based on the phase difference between adjacent sampling points. Step S2: For each set of original physical layer signal data bound with a unique electronic code in Step S1, estimate the common interference factor in the current environment using the original physical layer signal data of all RFID electronic tags read in the same inventory cycle; decouple the original physical layer signal data based on the common interference factor to remove common environmental interference, and extract the feature values ​​of the decoupled signal in multiple preset dimensions to generate a normalized real-time radio frequency fingerprint vector corresponding to the RFID electronic code. Step S3: Read the pre-stored standard state reference RF fingerprint vector corresponding to the RFID electronic code, and obtain the real-time RF fingerprint vector of the same type of reference tag set selected according to preset rules; calculate the first difference degree between the real-time RF fingerprint vector obtained in step S2 and the standard state reference RF fingerprint vector; at the same time, calculate the second difference degree between the real-time RF fingerprint vector and the real-time RF fingerprint vector of the same type of reference tag set; based on a state assessment model that integrates the first difference degree and the second difference degree, calculate a comprehensive state anomaly quantification value. Step S4: When the quantified value of the state anomaly calculated in step S3 exceeds the preset threshold, a tag instruction is generated; the RFID reader writes a state tag code into the memory of the RFID electronic tag according to the tag instruction.

[0021] In this embodiment, it is specifically necessary to explain the process in step S1 where the baseband processing unit of the RFID reader synchronously captures and records the original physical layer signal data as follows: Inside the RFID reader's baseband processing unit, a physical layer signal sampling channel and a standard tag data decoding channel are configured to operate in parallel and simultaneously. The physical layer signal sampling channel independently includes a high-speed analog-to-digital converter and a dedicated buffer memory. Its sampling rate is configured to be several times higher than the RFID communication carrier frequency to ensure accurate capture of phase changes. When the RFID reader executes the inventory communication process, the physical layer signal sampling channel continuously captures in-phase and quadrature analog signals from the RF front-end, converting them into digitized in-phase and quadrature signal sequences, which are then input to the dedicated buffer memory of the physical layer signal sampling channel. The buffer memory uses a circular buffer structure, configured to store the original signal data for at least several consecutive complete inventory cycles across all time slots, ensuring that the signal segment corresponding to any successful decoding event can be retrieved retrospectively. Simultaneously, the standard tag data decoding channel performs communication with the RFID electronic... The tag's digital protocol interaction: When the standard tag data decoding channel successfully demodulates the unique electronic code of an RFID tag's response, a decoding success event is generated. The decoding success event contains this unique electronic code and its corresponding communication time slot index. Based on the decoding success event, the physical layer signal sampling channel accesses the buffer memory and, according to the start and end times of the precise communication time slot corresponding to the communication time slot index in the decoding success event, extracts time-aligned in-phase signal data segments and quadrature signal data segments from the continuously buffered in-phase and quadrature signal sequences in the buffer memory. The in-phase and quadrature signal data segments together constitute the original signal data block bound to the unique electronic code. The start and end times of the precise communication time slot are precisely given by the time slot controller of the reader's baseband processing unit according to the communication protocol timing, thereby ensuring that the extracted signal segments and the tag response are perfectly aligned in time and eliminating adjacent channel interference. Furthermore, the original signal data block is processed in real time to extract physical layer parameters. The extraction process includes: for each pair of corresponding in-phase signal data segment sampling point values ​​and quadrature signal data segment sampling point values ​​in the original signal data block, firstly, the squares of the in-phase signal data segment sampling point values ​​and the quadrature signal data segment sampling point values ​​are added together; then, the logarithm of the sum is taken to the base 10; finally, the result is multiplied by 10, and the calculated value is used as the signal amplitude characterization value at the current sampling point. All signal amplitude characterization values ​​are arranged in the order of sampling points to form an instantaneous value sequence of received signal strength indication; this sequence reflects the micro-fluctuations of signal amplitude and provides a basis for subsequent analysis of signal stability and disturbances; for the same pair of sampling point values, the arctangent value of the quadrature signal data segment sampling point value is calculated by dividing it by the arctangent value of the in-phase signal data segment sampling point value to obtain the initial phase angle of the current sampling point, whose value range is within the negative pi. Between radians and positive pi radians, all initial phase angles are arranged in the order of sampling points to form an initial phase angle sequence. A phase unwinding operation is then performed on this initial phase angle sequence to obtain a continuous carrier phase sampling sequence. This operation eliminates periodic phase jumps, resulting in continuous phase information that accurately reflects changes in the electromagnetic wave propagation path. This is a crucial prerequisite for subsequent calculation of Doppler frequency offset and high-precision fingerprint comparison. Based on the continuous carrier phase sampling sequence, the difference between the values ​​of the continuous carrier phase sampling sequences corresponding to adjacent sampling points is calculated to obtain the phase change. This phase change is then divided by the product of the time interval between adjacent sampling points and twice the number of pi radians. The calculated value serves as the instantaneous Doppler frequency offset characterization value for the current sampling point. These values ​​are arranged in the order of sampling points to form an instantaneous Doppler frequency offset sequence. This instantaneous Doppler frequency offset sequence can effectively sense the subtle movements of tags or environmental reflectors, providing a direct basis for determining whether an asset is in a state of vibration or movement. Finally, the unique electronic code currently being processed, the corresponding communication time slot index, the timestamp corresponding to the original signal data block, and the instantaneous value sequence of the received signal strength indication, the carrier phase continuous sampling sequence, and the instantaneous Doppler frequency offset sequence, all of which are of length N, are encapsulated together to form the original physical layer signal data that is strictly bound to the unique electronic code as defined in step S1. The encapsulated original physical layer signal data is organized in the form of a tag physical layer feature data packet, which is the output of step S1, where N is the total number of sampling points contained in the original signal data block. The tag physical layer feature data packet has a unified and standardized data structure, which ensures that step S2 can receive and process the original sensing data from different tags and different time slots in a standardized manner. The specific process of performing a phase dewinding operation on the initial phase angle sequence to obtain a continuous carrier phase sampling sequence is as follows: The initial phase angle calculated from the first sampling point in the original signal data block is directly used as the first value of the carrier phase continuous sampling sequence obtained after unwinding. For each sampling point from the second point onwards in the initial phase angle sequence, perform the following recursive calculation operation in the order of the sampling points: First, the difference between the initial phase angle value of the current sampling point and the carrier phase continuous sampling sequence value calculated at the previous sampling point is calculated to obtain the original phase difference value. Next, the original phase difference value is corrected by adding the original phase difference value to the value of pi in radians. Then, the sum is modulo 2 (twice the value of pi in radians). The result of the modulo operation is a value between zero and twice the value of pi in radians. Finally, the value of pi in radians is subtracted from the result of the modulo operation to obtain the final phase difference value. The corrected phase difference value is constrained to a range between negative and positive pi radians. The core of this correction step is to detect and correct phase jumps that cross the boundaries of positive and negative pi. Through mathematical mapping, potentially large jumps (close to twice pi) are converted into actual small-amplitude continuous changes. Then, the carrier phase continuous sampling sequence value of the previous sampling point is added to the corrected phase difference value corresponding to the current sampling point. The result is the carrier phase continuous sampling sequence value of the current sampling point. Following this recursive calculation operation, starting from the second sampling point, all sampling points in the original signal data block are sequentially traversed and processed, ultimately generating a continuous carrier phase sampling sequence that eliminates periodic jumps. This algorithm ensures that even when the signal has a large Doppler frequency shift or rapid phase changes, it can still accurately recover the continuous phase trajectory, laying a solid foundation for the accuracy of subsequent feature extraction.

[0022] In this embodiment, it is specifically necessary to explain that the process of estimating the common interference factor in the current environment in step S2 is as follows: For all RFID tags read during the current inventory period, the carrier phase continuous sampling sequence of each RFID tag is extracted from the raw physical layer signal data (i.e., tag physical layer feature data packet) output in step S1. Each RFID tag's carrier phase continuous sampling sequence is treated as a row vector, and all row vectors are arranged in rows to construct a phase observation matrix. For the constructed phase observation matrix, a weighted robust principal component analysis is performed to estimate the common interference factor. The specific iterative calculation process of the weighted robust principal component analysis is as follows: A weight is initialized for each row vector corresponding to an RFID tag in the phase observation matrix. This weight is positively correlated with the local smoothness of the continuous carrier phase sampling sequence represented by the current row vector. Local smoothness can be quantified by calculating the root mean square value of the difference between adjacent sampling points in the sequence; the smaller the value, the smoother the sequence. It is also positively correlated with the overall similarity between the current row vector and all other row vectors in the phase observation matrix. Overall similarity can be measured by calculating the cosine similarity or correlation coefficient between this row vector and the average vector of all other row vectors in the matrix. A common interference subspace to be optimized is defined. In each iteration, an optimization problem is solved. The goal is to find a common interference subspace such that the sum of the weighted results of all RFID tags is minimized when the square of the L2 norm of the residual vector generated by projecting each row vector in the phase observation matrix onto this common interference subspace is multiplied by the current weight of that row vector. After each iteration, the L2 norm of the projected residual vector of each row vector in the common interference subspace obtained in the current iteration is calculated, and the weight of that row vector in the next iteration is dynamically adjusted based on the calculated L2 norm of the projected residual vector. The weight adjustment rule is: the larger the L2 norm of the projected residual vector, the greater the reduction in its weight. This can be further refined using Huber. Alternatively, a Tukey weight function can be used. For example, the new weights can be inversely proportional to a robust function value with respect to the projection residuals to suppress the excessive influence of outlier labels on the estimation of common disturbance factors. Iterative calculations and weight adjustments are repeated until the weight changes of all row vectors are less than a preset convergence threshold, such as one ten-thousandth. The iterative calculation process terminates, resulting in a converged common disturbance subspace. After convergence, the first principal component vector representing the most significant common change direction is extracted from this converged common disturbance subspace. This extracted first principal component vector is defined as the phase common disturbance factor vector. This vector captures the common disturbance mode that has the most significant impact on phase in the current environment. The formula includes multipath field changes caused by large-scale opening and closing of doors and windows or movement of people; for all instantaneous value sequences of received signal strength indications extracted from the original physical layer signal data (i.e., tag physical layer feature data packets), a received signal strength indication observation matrix is ​​constructed, and the same weighted robust principal component analysis process as for the carrier phase continuous sampling sequence is performed to obtain the received signal strength indication common interference factor vector; this vector captures the common influence of the environment on the signal amplitude, such as changes in overall path loss; the phase common interference factor vector and the received signal strength indication common interference factor vector together constitute the common interference factor in the current environment estimated in step S2; The specific steps for decoupling the original physical layer signal data based on the common interference factor are as follows: For each RFID tag, perform the following two steps in sequence to calculate its phase residual sequence: A1. Calculate the projection coefficient of the carrier phase continuous sampling sequence of this RFID electronic tag on the phase common interference factor vector. The calculation process of the projection coefficient is as follows: perform a dot product operation between the carrier phase continuous sampling sequence and the phase common interference factor vector, and then divide the result of the dot product operation by the result of the phase common interference factor vector and itself. This projection coefficient quantifies the relative intensity of the common phase interference currently experienced by the tag. A2. Perform a subtraction operation to subtract the product of the calculated projection coefficient and the phase common interference factor vector from the carrier phase continuous sampling sequence, resulting in the decoupled phase residual sequence. The specific execution process of the subtraction operation is as follows: subtract the product of the projection coefficient and the phase common interference factor vector at the corresponding position from each value in the carrier phase continuous sampling sequence to obtain the value at the corresponding position in the phase residual sequence. This operation removes the estimated common phase interference component from the original phase signal. The resulting phase residual sequence theoretically mainly contains phase disturbances caused by changes in the physical state of the tag itself and random noise that cannot be explained by the common pattern, laying the foundation for subsequent extraction of individual sensitive features. For each RFID tag, its signal strength residual sequence is calculated using the same logic: B1. Calculate the projection coefficient of the instantaneous value sequence of the received signal strength indication of this RFID electronic tag onto the received signal strength indication common interference factor vector; this coefficient quantifies the relative intensity of the common amplitude interference currently experienced by the tag. B2. Perform a subtraction operation by subtracting the product of the calculated projection coefficient and the common interference factor vector of the received signal strength indication from the instantaneous value sequence of the received signal strength indication, to obtain the decoupled signal strength residual sequence. This operation removes the common amplitude variation trend from the original signal strength indication signal, and the resulting signal strength residual sequence better reflects the signal amplitude modulation caused by the tag's own physical state (such as deformation and changes in attachments). The process of extracting feature values ​​of the decoupled signal across multiple preset dimensions to generate a normalized real-time radio frequency fingerprint vector corresponding to the RFID electronic code is as follows: For each RFID tag, perform the following operations: C1. Perform a Discrete Fourier Transform on the decoupled phase residual sequence to obtain the power spectral density of the decoupled phase residual sequence. Find the frequency corresponding to the largest power value in the obtained power spectral density and determine this frequency as the main peak frequency. With the main peak frequency as the center, extend upwards and downwards by half a preset bandwidth, for example, by ±5 Hz, to form a continuous frequency band. Calculate the integral value of the power spectral density in this frequency band as the local energy. Calculate the integral value of the power spectral density over the entire frequency band from zero frequency to Nyquist frequency as the total energy. Divide the local energy by the total energy and define the quotient as the frequency domain stability characteristic value. The closer this characteristic value is to 1, the more concentrated the energy of the phase residual sequence is in the frequency domain, the more stable the signal, and the more stable the asset state may be. The smaller the value, the more dispersed the energy, and the more likely there is abnormal vibration or complex modulation. C2. Using the time sequence number of the sampling points as the independent variable and the numerical value of the decoupled phase residual sequence as the dependent variable, a second-order polynomial curve fitting is performed to obtain a fitting function containing quadratic coefficients, linear coefficients, and a constant term. The absolute value of the quadratic coefficient is taken to obtain the first value. The absolute value of the linear coefficient is taken and added to a very small positive protection constant, for example, the protection constant is taken to be on the order of ten to the power of negative ten to prevent division by zero error, to obtain the second value. The quotient obtained by dividing the first value by the second value is defined as the phase nonlinearity characteristic value. This characteristic value reflects the nonlinear acceleration of phase change and is extremely sensitive to detecting nonlinear phase changes caused by structural deformation, material aging, etc. C3. Perform a Hilbert transform on the decoupled signal strength residual sequence to construct the analytic signal of the decoupled signal strength residual sequence; calculate the magnitude of this analytic signal to obtain the envelope signal of the decoupled signal strength residual sequence; normalize all values ​​of the envelope signal to the interval between zero and one, and divide it into multiple equally spaced sub-intervals, for example, sixteen or thirty-two sub-intervals; statistically analyze the probability that the normalized envelope signal value falls into each sub-interval; calculate the Shannon entropy of the envelope signal based on the statistically obtained probability; then divide the calculated Shannon entropy by the logarithm with the number of sub-intervals as the base, and normalize it. The result is defined as the residual signal envelope entropy characteristic value; this characteristic value characterizes the randomness and uncertainty of the signal strength envelope, and its increase may indicate that the asset state tends to be unstable or there is abnormal disturbance; The unique electronic code of the currently processed RFID electronic tag, the calculated frequency domain stability feature value, phase nonlinearity feature value, residual signal envelope entropy feature value, projection coefficient on the phase common interference factor vector, and projection coefficient on the received signal strength indication common interference factor vector are assembled into an original fingerprint vector in this order. This original fingerprint vector is an ordered list. The first element of the list is the unique electronic code, the second element is the frequency domain stability feature value, the third element is the phase nonlinearity feature value, the fourth element is the residual signal envelope entropy feature value, and the subsequent elements are the projection coefficients on the phase common interference factor vector and the projection coefficients on the received signal strength indication common interference factor vector, respectively. Finally, for the set of original fingerprint vectors of all RFID tags within the current inventory period, the following normalization operations are performed on each feature dimension of the original fingerprint vectors, excluding the unique electronic code: Calculate the arithmetic mean of the eigenvalues ​​corresponding to all original fingerprint vectors in the current feature dimension; calculate the absolute value of the median of the eigenvalues ​​corresponding to all original fingerprint vectors in the current feature dimension; using the absolute value of the median instead of the standard deviation enhances the robustness of the normalization process to outliers that may exist in the data; for each RFID tag, subtract the calculated arithmetic mean from the eigenvalue of the original fingerprint vector of this RFID tag in the current feature dimension, and use the difference as the numerator; [The text abruptly ends here, likely due to an incomplete sentence or missing information.] The calculated absolute value of the median is multiplied by a preset scaling factor, such as 1.4826, so that for normally distributed data, the scaled absolute value of the median is approximately equal to the standard deviation. The product is then added to a smoothing constant, such as 10 to the power of negative 8, to prevent the denominator from being zero and to ensure numerical stability. The result is used as the denominator. The numerator is divided by the denominator, and the quotient is the normalized feature value of this RFID tag in the current feature dimension. This normalization operation transforms feature values ​​of different dimensions and orders of magnitude to a similar scale, eliminating the dominant differences between features, so that the subsequent state assessment model can make equal and effective use of information from each feature dimension. The normalized feature values, after normalizing all feature dimensions, are arranged in their original order in the original fingerprint vector except for the unique electronic code, and then reassembled with the unique electronic code into a new vector. The newly assembled vector is the normalized real-time radio frequency fingerprint vector generated in step S2, corresponding to this RFID electronic code. This real-time radio frequency fingerprint vector is a standardized, multi-dimensional state feature representation of the asset at the current moment, and is a direct input for assessing the state anomaly degree.

[0023] In this embodiment, the specific process of calculating the first difference degree and the second difference degree in step S3 is as follows: First, for each feature dimension of the real-time radio frequency fingerprint vector of the RFID electronic tag to be evaluated, a diachronic stability weight for evaluating the first degree of difference and a synchronic stability weight for evaluating the second degree of difference are calculated respectively. When calculating the historical stability weight, the dispersion of the RFID tag's value on this feature dimension under historical normal conditions is taken into account. This dispersion is characterized by calculating the median absolute deviation of historical data on this dimension. Using the median absolute deviation instead of the standard deviation can effectively eliminate the influence of random outliers in historical data on the dispersion estimation, making the weight calculation more robust. The calculation process of the historical stability weight is as follows: take the number 1 as the numerator, add the value of the historical median absolute deviation to a very small normal protection constant, use the sum as the denominator, and calculate the quotient obtained by dividing the numerator by the denominator, which is the historical stability weight. This historical stability weight design ensures that feature dimensions with low historical volatility (high stability) have larger corresponding weight values. This means that even if there is a small change in the value on this dimension, it will be given a higher weight in the difference calculation, thereby enhancing the model's sensitivity to changes in stable features. Conversely, for features with large historical volatility, their changes will be partially suppressed, thereby reducing the possibility of false alarms. When calculating the synchronicity stability weight, the dispersion of the real-time radio frequency fingerprint vectors of all RFID tags in the same reference tag set within the current inventory period is used as the basis for the calculation of the median absolute deviation of the feature values ​​of all RFID tags in the set in this dimension. The calculation process of the synchronicity stability weight is as follows: take the number 1 as the numerator, add the current median absolute deviation value to a very small normal protection constant, use the sum as the denominator, and calculate the quotient obtained by dividing the numerator by the denominator, which is the synchronicity stability weight. This synchronicity stability weight reflects the degree of consistency of the same asset group in a certain feature dimension under the current environment. The higher the consistency of the group, the greater the weight of the dimension, which means that the behavior that deviates from the group center in this dimension is regarded as a more significant abnormal signal. Next, the first dissimilarity is calculated. Its value is the square of the difference between the standard state baseline RF fingerprint vector and the real-time RF fingerprint vector across each feature dimension, multiplied by the corresponding historical stability weight, summed, and then the square root of the sum is taken. The calculation process is as follows: For each feature dimension, the difference between the real-time RF fingerprint vector value and the standard state baseline RF fingerprint vector value in that dimension is calculated, the square of this difference is calculated, and then the squared result is multiplied by the historical stability weight corresponding to that dimension. The products of all feature dimensions are summed, and finally the square root of the sum is taken. The result is the first dissimilarity. This first dissimilarity is a weighted Euclidean distance that measures the overall deviation of the current state from its historical health baseline, and this deviation has been intelligently weighted according to the historical stability of each feature. "Standard state baseline RF fingerprint vector" The establishment method is as follows: When the system is initially deployed or the asset is known to be in a normal and intact state, a baseline acquisition is performed. For each RFID electronic tag, under a standard environment (such as constant temperature and humidity, no strong interference, and the asset is stationary), the tag is read multiple times (e.g., dozens to hundreds of times) by an RFID reader. For each read, the steps S1 and S2 of this invention are executed to generate a real-time radio frequency fingerprint vector of the tag in the standard state. After collecting the real-time radio frequency fingerprint vectors generated by all reads, the statistical average (or median) of these vectors is calculated for each feature dimension. The vector composed of the statistical average (or median) of all feature dimensions in sequence is associated with the unique electronic code of the tag and stored as the "standard state baseline radio frequency fingerprint vector". This baseline vector represents the multi-dimensional radio frequency feature "fingerprint" of the asset in a healthy state. Then, the robust center of the real-time RF fingerprint vectors of the same reference label set in the feature space is calculated. The calculation process is as follows: for each feature dimension, the median of the values ​​of all real-time RF fingerprint vectors in the same reference label set in that dimension is found. The medians of all feature dimensions are arranged in dimensional order, and the resulting vector is the robust center vector. Using the median instead of the mean to calculate the population center can effectively resist the interference of a few abnormal labels in the population on the center position, ensuring the robustness of the center estimation. The second dissimilarity value is the square of the difference between the robust center vector and the real-time RF fingerprint vector in each feature dimension, multiplied by... The value is obtained by summing the corresponding synchronicity stability weights and then taking the square root of the sum. The calculation process is as follows: For each feature dimension, calculate the difference between the value of the real-time RF fingerprint vector in that dimension and the value of the robust center vector in that dimension, calculate the square of the difference, multiply the squared result by the synchronicity stability weight corresponding to that dimension, sum the product results of all feature dimensions, and finally take the square root of the sum. The result is the second dissimilarity. This second dissimilarity measures the degree of deviation of the current label state from the current state of its peer group, that is, its "outlier" degree. This metric also considers the inherent consistency of the group in each feature dimension. Based on a state assessment model that integrates the first and second dissimilarity measures, the process of calculating a comprehensive state anomaly metric is as follows: First, the first degree of difference is normalized to obtain the first relative degree of difference. This is done by using the first degree of difference value as the numerator and adding the third and fourth quartiles of all first degree of difference values ​​calculated for the currently evaluated RFID tags under historical normal conditions to a normal protection constant as the denominator. The quotient obtained by dividing the numerator by the denominator is the first relative degree of difference. Using the third and fourth quartiles of historical first degree of difference as the normalized denominator is equivalent to using the upper limit of the historical normal fluctuation range as a reference benchmark. This ensures that the first relative degree of difference represents the multiple of the current degree of difference relative to its historical normal fluctuation level, achieving cross-timescale representation. Comparability: The second dissimilarity is normalized to obtain the second relative dissimilarity. The processing method is to use the value of the second dissimilarity as the numerator, and add the third quartile of the second dissimilarity values ​​calculated from all RFID electronic tags in the same reference tag set within the current inventory period to a normal protection constant as the denominator. The quotient obtained by dividing the numerator by the denominator is the second relative dissimilarity. The third quartile of the second dissimilarity of the current group is used as the normalized denominator, so that the second relative dissimilarity represents the outlier degree of the current tag relative to the dispersion degree of the entire group, realizing fair comparison between different tags within the same inventory period. Next, the obtained first relative difference is exponentially calculated using a pre-defined constant greater than one; for example, this exponent can be set to 1.5. The second relative difference is then subjected to the same exponential calculation, again using a constant greater than one. This exponential calculation of the relative difference is a non-linear amplification operation, which allows larger relative difference values ​​to grow more rapidly, thereby enhancing the sensitivity of the state assessment model to significant anomalous signals and helping to distinguish between ordinary fluctuations and true anomalies. Then, the exponentially calculated first relative difference is multiplied by an adjustable first fusion weight coefficient, which can be obtained, for example, through training with historical data, and its default value can be set to 0.6. The exponentially calculated second relative difference is then multiplied by an adjustable second fusion weight coefficient, and the two products are summed to obtain an intermediate value. By adjusting the first and second fusion weight coefficients, the assessment model can be flexibly configured to focus more on the asset's own historical deviations or its relative deviations. To address the outlier characteristics of the group and adapt to different asset management strategies and application scenarios, the hyperbolic tangent function is used to calculate this intermediate value. The output value of the hyperbolic tangent function ranges from zero to one, and this output value is defined as the final comprehensive state anomaly metric. The saturation characteristic of the hyperbolic tangent function can compress arbitrarily large intermediate values ​​into a finite interval between zero and one, ensuring that the range of the output value is fixed and meaningful. The closer the value is to one, the higher the degree of anomaly, and zero indicates complete normality. This provides convenience for setting a unified warning threshold. The state anomaly metric is a value between zero and one, and its magnitude directly reflects the degree of anomaly in the asset state. The preset threshold of the state anomaly metric can be adjusted according to the actual operation after system deployment. A typical setting method is to collect a large number of state anomaly metric values ​​under normal conditions during the system trial operation phase, calculate their statistical distribution, and set the preset threshold at the 95th percentile or higher, such as 0.85, to effectively capture high-risk abnormal states while controlling the false alarm rate.

[0024] In this embodiment, it should be specifically noted that in step S4, the preset threshold on which the marking instruction is generated is a dynamic threshold; the process of calculating the dynamic threshold includes: First, calculate the current environmental uncertainty factor, which is the sum of the squares of the elements of the phase common interference factor vector obtained from step S2, and the sum of the squares of the elements of the received signal strength indication common interference factor vector. The phase common interference factor vector and the received signal strength indication common interference factor vector respectively characterize the influence mode of the environment on the signal phase and amplitude. The larger the energy (sum of squares), the stronger the electromagnetic disturbance of the current environment, the lower the "signal-to-noise ratio" of the state assessment, and the higher the uncertainty of the assessment conclusion. Next, the confidence factor for the current state assessment is calculated. The calculation process is as follows: using the first relative difference obtained from step S3 as the numerator and the sum of the second relative difference and a positive protection constant as the denominator, the ratio of the two is calculated; the absolute value of the difference between this ratio and one is calculated; this absolute value is divided by a scaling parameter and the result is negative; finally, the result of the exponential function with the natural logarithm as the base is calculated on this negative value, and this result is the confidence factor for the current state assessment. This confidence factor quantifies whether the current state anomaly is mainly caused by its own factors or affected by the public environment. When the values ​​of the first relative difference and the second relative difference are close (the ratio is close to 1), the confidence factor approaches zero, indicating that it is difficult to distinguish whether it is an anomaly of its own or a change in the public environment, and the assessment confidence is low; when the difference between the two is significant, the confidence factor approaches one, indicating that the assessment conclusion is clear and the confidence is high. The scaling parameter is used to control the steepness of the transition from "uncertainty" to "certainty", for example, it can be set to 0.5. Next, the preset static base threshold, the product of the current environmental uncertainty factor and a first positive coefficient, and the product of the current state assessment confidence factor and a second positive coefficient are algebraically calculated using the following formula: the static base threshold is added to the product of the first positive coefficient and the current environmental uncertainty factor, and then the product of the second positive coefficient and the current state assessment confidence factor is subtracted. The result is the dynamic threshold. The static base threshold is the benchmark point for decision-making, for example, it can be set to 0.8. The first positive coefficient controls the upward adjustment strength of the threshold due to environmental uncertainty, for example, it can be set to 0.1. The stronger the environmental disturbance, the higher the threshold, and the more conservative the decision to prevent false alarms. The second positive coefficient controls the downward adjustment strength of the threshold due to assessment confidence, for example, it can be set to 0.05. The higher the assessment confidence, the lower the threshold can be appropriately reduced to improve the detection rate by using high-confidence anomaly judgment. This dynamic threshold mechanism enables the system to intelligently adapt to different environmental noise levels and assessment confidence, and achieve robust decision-making. When the quantified value of the state anomaly calculated in step S3 exceeds this dynamic threshold, a marking instruction is generated. The specific process by which an RFID reader writes a status tag code into the memory of an RFID electronic tag according to a tagging instruction is as follows: First, a status tag code with a multi-dimensional information structure is constructed. The status tag code consists of four fields in sequence: The first field is a two-bit status tag field, used to encode the overall status level of the asset to which the RFID electronic tag is attached. Different binary codes correspond to four predefined states: normal, watchful, warning, and severe. For example, 00 represents normal, 01 represents watchful (minor anomalies require observation), 10 represents warning (obvious anomalies require planned inspection), and 11 represents severe (requires immediate handling). The second field is a four-bit level field, whose value is obtained by mapping the status anomaly quantification value calculated in step S3 to a different level. This level mapping can be linear or non-linear, for example, mapping the range from zero to one. The data is evenly divided into sixteen levels, or the high-end anomaly area can be further subdivided. This field retains quantitative information on the degree of anomaly, facilitating subsequent trend analysis. The third field is a four-bit dominant dimension field. Its value is determined by the contribution ratio of the weighted squared difference of each feature dimension of the real-time RF fingerprint vector to the total difference when calculating the first and second difference in analysis step S3. The index value of the feature dimension with the largest contribution ratio is filled into this field. This field provides preliminary anomaly characteristics, such as indicating that the anomaly mainly originates from specific dimensions such as "frequency domain stability," "phase nonlinearity," or "envelope entropy," providing clues for on-site investigation. The fourth field is a timestamp field, used to record the time when this tag code was generated. This timestamp provides key information for subsequent analysis of state change timing and calculation of anomaly duration. After constructing the status tag code, the physical write operation is not performed immediately. Instead, the RFID reader only performs the operation of writing the newly constructed status tag code to the tag memory when all of the following conditions are met: The first condition is that the quantified value of the state anomaly calculated in step S3 exceeds the dynamic threshold. The second condition is that the status flag field of the old tag code read from the RFID electronic tag memory is different from the status flag field of the newly constructed status flag code; or, the level field of the old tag code is different from the level field of the newly constructed status flag code; or, the difference between the current system time and the time recorded in the timestamp field of the old tag code is greater than a preset retention time. The retention time is used to prevent the tag memory from being repeatedly written too frequently during the period of continuous abnormal asset status, thereby effectively extending the service life of RFID electronic tags based on technologies such as EEPROM. For example, the retention time can be set to thirty minutes or one hour. This selective writing strategy significantly optimizes hardware resource consumption while ensuring the timeliness of status information. After the write operation is completed, the RFID reader immediately initiates a dedicated read access to the RFID tag, reading the newly written tag code data from its user memory. The read tag code data is then compared field by field with the newly constructed status tag code intended for writing. If any field is inconsistent, a write failure event is recorded, and the write operation is retried for the tag in the next inventory cycle. This verification and retry mechanism ensures reliable writing of tag information even under unstable communication conditions, improving system robustness. If all fields are completely consistent, the status tag field and timestamp field of the digital asset record associated with the unique electronic code of this RFID tag are updated in the backend management system connected to the RFID reader to the corresponding field values ​​of the verified newly constructed status tag code, and the update time is recorded. This ensures that the asset status record in the backend management system is consistent with the physical tag status stored in the RFID tag's memory. This operation completes the full status synchronization from physical perception to digital recording, constructing a unified online and offline asset health profile, providing an accurate data foundation for achieving full lifecycle management based on digital twins.

[0025] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0026] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0027] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0028] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0029] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0030] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0031] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for optimizing the inventory management of fixed assets based on RFID, characterized in that, Specifically, the steps include the following: Step S1: When the RFID reader performs inventory communication with multiple RFID electronic tags, the baseband processing unit of the RFID reader synchronously captures and records the original physical layer signal data bound to the unique electronic code of each successfully responding RFID electronic tag. The original physical layer signal data includes at least the instantaneous value sequence of the received signal strength indication, the continuous sampling sequence of the carrier phase, and the instantaneous Doppler frequency offset sequence calculated based on the phase difference between adjacent sampling points. Step S2: For each set of original physical layer signal data bound with a unique electronic code in Step S1, estimate the common interference factor in the current environment using the original physical layer signal data of all RFID electronic tags read in the same inventory cycle; decouple the original physical layer signal data based on the common interference factor to remove common environmental interference, and extract the feature values ​​of the decoupled signal in multiple preset dimensions to generate a normalized real-time radio frequency fingerprint vector corresponding to the RFID electronic code. Step S3: Read the pre-stored standard status reference radio frequency fingerprint vector corresponding to the RFID electronic code, and obtain the real-time radio frequency fingerprint vector of the same type of reference tag set selected according to preset rules; Calculate the first difference between the real-time RF fingerprint vector obtained in step S2 and the standard state reference RF fingerprint vector; Simultaneously, a second degree of difference is calculated between the real-time radio frequency fingerprint vector and the real-time radio frequency fingerprint vector of the same reference tag set; based on a state assessment model that integrates the first degree of difference and the second degree of difference, a comprehensive state anomaly quantification is calculated. Step S4: When the quantified value of the state anomaly calculated in step S3 exceeds the preset threshold, a tag instruction is generated; the RFID reader writes a state tag code into the memory of the RFID electronic tag according to the tag instruction.

2. The fixed asset inventory optimization management method based on RFID according to claim 1, characterized in that: In step S1, the specific process by which the baseband processing unit of the RFID reader synchronously captures and records the original physical layer signal data is as follows: Inside the RFID reader baseband processing unit, the physical layer signal sampling channel and the standard tag data decoding channel work in parallel. When the RFID reader executes the inventory communication process, the physical layer signal sampling channel continuously captures in-phase and quadrature analog signals from the radio frequency front end, and converts the in-phase and quadrature analog signals into digital in-phase and quadrature signal sequences, which are then input into the dedicated buffer memory of the physical layer signal sampling channel. At the same time, the standard tag data decoding channel performs digital protocol interaction with the RFID electronic tag; When the standard tag data decoding channel successfully demodulates the unique electronic code returned by an RFID electronic tag, it generates a decoding success event containing this unique electronic code and the corresponding communication time slot index. Based on the decoding success event, the physical layer signal sampling channel accesses the buffer memory and, according to the start and end time points of the precise communication time slot corresponding to the communication time slot index in the decoding success event, extracts in-phase signal data segments and orthogonal signal data segments that are perfectly aligned in time from the continuously buffered in-phase signal sequence and orthogonal signal sequence, respectively. The in-phase signal data segments and orthogonal signal data segments together constitute the original signal data block bound to the unique electronic code. Then, the original signal data block is processed in real time to extract physical layer parameters: for each pair of corresponding in-phase signal data segment sampling point values ​​and quadrature signal data segment sampling point values ​​in the original signal data block, the signal amplitude characterization value is calculated to form an instantaneous value sequence of received signal strength indication; its arctangent value is calculated to obtain the initial phase angle, and arranged to form an initial phase angle sequence. A phase unwinding operation is performed on the initial phase angle sequence to obtain a continuous carrier phase sampling sequence. Based on the continuous carrier phase sampling sequence, the difference between the values ​​of the continuous carrier phase sampling sequences corresponding to adjacent sampling points is calculated to obtain the phase change. The phase change is then divided by the product of the time interval between adjacent sampling points and twice the radius of pi in radians. The calculated value is used as the instantaneous Doppler frequency offset characterization value corresponding to the current sampling point. The instantaneous Doppler frequency offset sequence is formed by arranging the sampling points in order. Finally, the unique electronic code currently being processed, the corresponding communication time slot index, the timestamp corresponding to the original signal data block, the instantaneous value sequence of the received signal strength indication of length N, the carrier phase continuous sampling sequence, and the instantaneous Doppler frequency offset sequence are encapsulated together to form the original physical layer signal data.

3. The fixed asset inventory optimization management method based on RFID according to claim 2, characterized in that: The specific process of performing phase dewinding operation on the initial phase angle sequence to obtain a continuous carrier phase sampling sequence is as follows: The initial phase angle calculated from the first sampling point in the original signal data block is directly used as the first value of the carrier phase continuous sampling sequence obtained after unwinding. For each sampling point from the second point onwards in the initial phase angle sequence, perform the following recursive calculation operation in the order of the sampling points: First, the difference between the initial phase angle value of the current sampling point and the carrier phase continuous sampling sequence value calculated at the previous sampling point is calculated to obtain the original phase difference value. Next, the original phase difference value is corrected by adding the original phase difference value to the value of pi in radians. Then, the sum is moduloed by twice the value of pi in radians. The result of the modulo operation is a value between zero and twice the value of pi in radians. Finally, the value of pi in radians is subtracted from the result of the modulo operation to obtain the corrected phase difference value. The range of the corrected phase difference value is constrained to between negative and positive pi in radians. Afterward, the carrier phase continuous sampling sequence value of the previous sampling point is added to the corrected phase difference value corresponding to the current sampling point. The result is the carrier phase continuous sampling sequence value of the current sampling point. Following this recursive calculation operation, starting from the second sampling point, all sampling points in the original signal data block are sequentially traversed and processed, ultimately generating a continuous carrier phase sampling sequence.

4. The fixed asset inventory optimization management method based on RFID according to claim 3, characterized in that: In step S2, the process of estimating the common interference factor under the current environment is as follows: From the raw physical layer signal data, the carrier phase continuous sampling sequences of all RFID tags are extracted, and each RFID tag's carrier phase continuous sampling sequence is treated as a row vector. All row vectors are arranged in rows to construct a phase observation matrix. For the constructed phase observation matrix, weighted robust principal component analysis is performed to estimate the common interference factor. The specific iterative calculation process is as follows: A weight is initialized for each row vector corresponding to an RFID tag in the phase observation matrix. This weight is positively correlated with the local smoothness of the carrier phase continuous sampling sequence represented by the current row vector, and also positively correlated with the overall similarity between the current row vector and all other row vectors in the phase observation matrix. A common interference subspace to be optimized is defined. In each iteration, an optimization problem is solved. The objective of this optimization problem is to find a common interference subspace such that the square of the L2 norm of the residual vector generated by projecting each row vector in the phase observation matrix onto this common interference subspace, multiplied by the current weight of that row vector, results in all RFID tags being unaffected. The weighted sum of the D electronic tags is minimized; after each iteration, the L2 norm of the projection residual vector of each row vector in the common interference subspace obtained in the current iteration is calculated, and the weight of the row vector in the next iteration is dynamically adjusted according to the magnitude of the calculated projection residual L2 norm; the iterative calculation and weight adjustment are repeated until the weight change of all row vectors is less than a preset convergence threshold, the iterative calculation process is terminated, and the converged common interference subspace is obtained; after the iteration converges, the first principal component vector is extracted from the converged common interference subspace, and the extracted first principal component vector is defined as the phase common interference factor vector; For all the instantaneous value sequences of received signal strength indicators extracted from the original physical layer signal data, a received signal strength indicator observation matrix is ​​constructed, and a weighted robust principal component analysis process that is exactly the same as that for the continuous sampling sequence of carrier phase is performed to obtain the received signal strength indicator common interference factor vector. The phase common interference factor vector and the received signal strength indicator common interference factor vector together constitute the estimated common interference factor in the current environment.

5. The fixed asset inventory optimization management method based on RFID according to claim 4, characterized in that: The specific operation for decoupling the original physical layer signal data based on the common interference factor is as follows: For each RFID tag, calculate the projection coefficient of its carrier phase continuous sampling sequence onto the phase common interference factor vector. Then, subtract the product of this projection coefficient and the phase common interference factor vector from the carrier phase continuous sampling sequence to obtain the decoupled phase residual sequence. For each RFID tag, calculate the projection coefficient of its instantaneous value sequence of received signal strength indication onto the common interference factor vector of received signal strength indication. Then, subtract the product of this projection coefficient and the common interference factor vector of received signal strength indication from the instantaneous value sequence of signal strength indication to obtain the decoupled signal strength residual sequence.

6. The fixed asset inventory optimization management method based on RFID according to claim 5, characterized in that: The process of extracting feature values ​​of the decoupled signal in multiple preset dimensions and generating a normalized real-time radio frequency fingerprint vector corresponding to the RFID electronic code is as follows: For each RFID tag, perform the following operations: A second-order polynomial is fitted to the decoupled phase residual sequence, and the ratio of the absolute value of the quadratic term coefficient to the absolute value of the linear term coefficient is defined as the phase nonlinearity eigenvalue. C1. Perform a discrete Fourier transform on the decoupled phase residual sequence to obtain the power spectral density. Find the frequency corresponding to the power value with the largest value in the obtained power spectral density and determine this frequency as the main peak frequency. With the main peak frequency as the center, extend upward and downward by half of a preset bandwidth to form a continuous frequency band. The integral value of the power spectral density within this frequency band is calculated as the local energy; the integral value of the power spectral density over the entire frequency band from zero frequency to Nyquist frequency is calculated as the total energy; the quotient obtained by dividing the local energy by the total energy is defined as the frequency domain stability eigenvalue. C2. Fit the decoupled phase residual sequence with a second-order polynomial and take the ratio of the absolute value of the quadratic term coefficient to the absolute value of the linear term coefficient as the characteristic value of phase nonlinearity. C3. Perform a Hilbert transform on the decoupled signal strength residual sequence to obtain the envelope signal, and calculate the normalized Shannon entropy of this envelope signal, which is defined as the eigenvalue of the residual signal envelope entropy. The unique electronic code of the RFID tag being processed, the calculated frequency domain stability characteristic value, phase nonlinearity characteristic value, residual signal envelope entropy characteristic value, projection coefficient on the phase common interference factor vector, and projection coefficient on the received signal strength indication common interference factor vector are assembled into an original fingerprint vector in this order. Finally, for the original fingerprint vector set of all RFID tags in the current inventory period, for each feature dimension except for the unique electronic code, calculate the arithmetic mean and the absolute value of the median of all feature values ​​in the current feature dimension; for each RFID tag, subtract the arithmetic mean from its feature value in the current feature dimension, and then divide by the sum of the absolute value of the median adjusted by the scaling factor and the smoothing constant to obtain the normalized feature value of that dimension. The normalized feature values ​​after normalizing all feature dimensions are reassembled with the unique electronic code to generate a normalized real-time radio frequency fingerprint vector corresponding to the code.

7. The fixed asset inventory optimization management method based on RFID according to claim 6, characterized in that: In step S3, the specific process of calculating the first difference degree and the second difference degree is as follows: First, for each feature dimension of the real-time RF fingerprint vector, calculate the diachronic stability weight and the synchronic stability weight: the diachronic weight is equal to 1 divided by the sum of the median absolute deviation of the feature value of this dimension under the historical normal state of the tag itself and the normal protection constant; the synchronic weight is equal to 1 divided by the sum of the median absolute deviation of the feature value of this dimension of the current set of similar reference tags and the protection constant. Next, the first difference is calculated: for each dimension, the square of the difference between the real-time RF fingerprint vector and the standard state baseline RF fingerprint vector in that dimension is calculated, multiplied by the corresponding epochal stability weight, and the square root is taken after summing the products over all dimensions. Then, the second difference is calculated: First, the robust center of the real-time RF fingerprint vector of the same reference tag set is calculated. The value of each dimension of the center is the median value of all vectors in the set in that dimension. Then, for each dimension, the square of the difference between the real-time RF fingerprint vector and the robust center in that dimension is calculated, multiplied by the corresponding synchronicity stability weight, and the square root is taken after summing the products of all dimensions.

8. The fixed asset inventory optimization management method based on RFID according to claim 7, characterized in that: The process of calculating a comprehensive state anomaly metric based on a state assessment model that integrates the first and second differences is as follows: Calculate the first relative difference: Divide the first difference by the sum of the third quartile of all first difference values ​​under the historical normal state of this tag and the normal protection constant; Calculate the second relative difference: Divide the second difference by the sum of the third quartile of all second difference values ​​in the current set of reference tags of the same type and the normal protection constant; The first relative difference and the second relative difference are respectively subjected to constant exponentiation greater than 1; Multiply the first relative difference after exponential calculation by the first fusion weight coefficient, multiply the second relative difference after exponential calculation by the second fusion weight coefficient, and then add the two products to obtain the median value. The hyperbolic tangent function is used to calculate this intermediate value, and the output value is the quantified value of the state anomaly.

9. The fixed asset inventory optimization management method based on RFID according to claim 8, characterized in that: In step S4, the preset threshold on which the marking instruction is generated is a dynamic threshold. The process of calculating the dynamic threshold includes: Calculate the current environmental uncertainty factor, which is the sum of the squares of the elements of the phase common interference factor vector and the sum of the squares of the elements of the received signal strength indication common interference factor vector; The confidence factor for the current state assessment is calculated as the result of an exponential function with the natural logarithm as the base, in the following negative value: This negative value is obtained by subtracting one from the ratio of the first relative difference to the sum of the second relative difference and the protection constant, taking the absolute value, and then dividing by the scale parameter. The dynamic threshold is calculated as follows: the static base threshold plus the product of the first positive coefficient and the environmental uncertainty factor, minus the product of the second positive coefficient and the state assessment confidence factor. When the quantified value of the state anomaly calculated in step S3 exceeds this dynamic threshold, a marking instruction is generated.

10. The fixed asset inventory optimization management method based on RFID according to claim 9, characterized in that: The specific process by which the RFID reader writes a status tag code into the memory of the RFID electronic tag according to the tagging instruction is as follows: First, a status tag code with a multi-dimensional information structure is constructed. The status tag code consists of four fields in sequence: the first field is a two-bit status tag field, used to encode the overall status level of the asset to which the RFID electronic tag is attached. Different binary codes correspond to four predefined statuses: normal, watch out, warning, and severe. The second field is a four-bit level field, whose value is obtained by mapping the status anomaly quantification value calculated in step S3 to different levels, and is used to store the quantification level corresponding to the quantification value. The third field is a four-bit dominant dimension field, whose value is determined by analyzing the contribution ratio of the weighted square difference of each feature dimension of the real-time RFID fingerprint vector to the total difference when calculating the first and second difference in step S3. The index value of the feature dimension with the largest contribution ratio is filled into this field. The fourth field is a timestamp field. After constructing the status tag code, the RFID reader will only perform the operation of writing the newly constructed status tag code to the tag memory if all of the following conditions are met: The first condition is that the quantified value of the state anomaly calculated in step S3 exceeds the dynamic threshold. The second condition is that the status flag field of the old tag code read from the RFID electronic tag's memory is different from the status flag field of the newly constructed status flag code; Alternatively, the level field of the old tag code is different from the level field of the newly constructed status tag code; or, the difference between the current system time and the time recorded in the timestamp field of the old tag code is greater than a preset retention time. Immediately after writing, read and verify. If there is a discrepancy, record the failure and arrange a retry; if there is a discrepancy, update the status and timestamp information of the digital asset record corresponding to the tag in the backend system.

Images