Preparation method and device of RFID electronic tag

Abstract

The invention relates to the technical field of RFID electronic tags, and discloses an RFID electronic tag preparation method and device, and the method comprises the steps: carrying out the signal conditioning and synchronous digital processing of a multi-channel RFID tag identification space receiving device, and obtaining a multi-channel digital data stream; performing singular value decomposition, low-order polynomial fitting and gradient-based binary search optimization on the RFID tag response signal to obtain a beam forming weight vector; performing particle swarm iterative optimization on the RFID tag identification parameters to obtain a target working parameter set; according to the method, sector division is carried out on a multi-tag identification space, RFID tag anti-collision identification is carried out in combination with dynamic time slot distribution, a target tag identification result is obtained, the target tag identification result is used for preparation parameter optimization of the RFID electronic tags, power consumption is minimized on the premise that the identification accuracy is guaranteed, the system operation efficiency is improved, and the system reliability is improved. The service life of equipment is prolonged, and the maintenance cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of RFID electronic tags, and in particular to a method and device for preparing an RFID electronic tag. Background Art

[0002] Traditional RFID reader antennas are bulky and difficult to integrate. Furthermore, they suffer from low recognition accuracy and efficiency in densely populated environments, making them unable to meet the growing demand for efficient identification. In particular, in complex electromagnetic environments, traditional RFID systems face challenges such as short recognition distances, low accuracy, and high energy consumption. The complex interaction model between tags and readers makes it difficult to optimize recognition parameters, making it difficult for the system to adapt to different environments.

[0003] Current research on RFID tag recognition and detection methods primarily focuses on improving reader antenna performance and optimizing recognition algorithms. While existing technologies have achieved some progress in antenna miniaturization, this often comes with performance degradation. In particular, input impedance matching is difficult, requiring additional matching networks and increasing system complexity and power consumption. Regarding algorithm optimization, existing semidefinite programming methods suffer from high computational complexity and struggle to meet real-time processing requirements. Furthermore, methods for estimating recognition accuracy are limited in applicability and lack an effective numerical solution for the accuracy, hindering further improvements in system performance. Summary of the Invention

[0004] The present invention provides a method and device for preparing an RFID electronic tag, which minimizes power consumption while ensuring recognition accuracy, improves system operation efficiency, extends equipment service life, and reduces maintenance costs.

[0005] In a first aspect, the present invention provides a method for preparing an RFID electronic tag, the method comprising: Connecting the multi-channel RFID tag identification space receiving device to the multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream; Performing singular value decomposition, low-order polynomial fitting, and gradient-based binary search optimization on the RFID tag response signal according to the multi-channel digital data stream to obtain a beamforming weight vector; Performing particle swarm iterative optimization on RFID tag recognition parameters based on the beamforming weight vector to obtain a target working parameter set; The multi-tag identification space is sectored according to the target working parameter set and RFID tag anti-collision identification is performed in combination with dynamic time slot allocation to obtain a target tag identification result, which is used for optimizing the preparation parameters of the RFID electronic tag.

[0006] In a second aspect, the present invention provides a device for preparing an RFID electronic tag, the device comprising: A digital processing module is used to connect the multi-channel RFID tag identification space receiving device to the multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream; A fitting module is used to perform singular value decomposition, low-order polynomial fitting, and gradient-based binary search optimization on the RFID tag response signal according to the multi-channel digital data stream to obtain a beamforming weight vector; an iterative optimization module, configured to perform particle swarm iterative optimization on RFID tag recognition parameters based on the beamforming weight vector to obtain a target working parameter set; The identification module is used to sectorize the multi-tag identification space according to the target working parameter set and perform RFID tag anti-collision identification in combination with dynamic time slot allocation to obtain a target tag identification result, which is used to optimize the preparation parameters of the RFID electronic tag.

[0007] In the technical solution provided by the present invention, a miniaturized RFID reader antenna designed using zigzag line technology maintains excellent radiation performance while reducing its size. It directly connects to the receiving circuit through conjugate impedance matching, eliminating the need for an external matching network. This simplifies system design, reduces power consumption, and improves antenna performance and integration. High-precision digital processing and precision clock synchronization ensure the complete preservation of the time and frequency domain characteristics of multi-channel signals. FPGA real-time preprocessing and high-speed cache design ensure the continuity and real-time nature of data processing, providing high-quality input data for subsequent algorithms. A low-order model based on singular value decomposition and polynomial fitting accurately describes the interaction characteristics between the reader and the tag, significantly reducing computational complexity and enabling real-time adaptive adjustments in complex environments, thereby improving the robustness and adaptability of the recognition system. The gradient-based binary search optimization method proposed in the present invention solves directly in the original beamforming domain, achieving real-time beamforming and significantly improving tag direction recognition accuracy and signal-to-interference-noise ratio performance. Particle swarm iterative optimization technology is used to globally optimize system parameters, minimizing power consumption while ensuring recognition accuracy, improving system efficiency, extending equipment life, and reducing maintenance costs. By combining spatial partitioning with the improved dynamic time-slot ALOHA algorithm, efficient recognition is achieved in a dense multi-tag environment. The fairness guarantee mechanism solves the tag starvation problem. Doppler feature analysis enhances the tracking and recognition capabilities of mobile tags, significantly improving the system's recognition performance in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 Schematic diagram of the steps of the method for preparing an RFID electronic tag according to an embodiment of the present invention; Figure 2 Schematic diagram of the structure of the RFID electronic tag preparation device in an embodiment of the present invention. DETAILED DESCRIPTION

[0010] The embodiments of the present invention provide a method and apparatus for preparing an RFID electronic tag. The terms "first," "second," "third," "fourth," and so on (if any) in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the numbers used in this way are interchangeable where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatuses.

[0011] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 An embodiment of a method for preparing an RFID electronic tag according to an embodiment of the present invention includes: Step S1: Connect the multi-channel RFID tag identification space receiving device to the multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream; It is understandable that the execution subject of the present invention may be a device for preparing RFID electronic tags, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0012] Specifically, the zigzag line structure of the RFID tag antenna's radiating element was calculated using electromagnetic simulation software to achieve size compression and directionality control while maintaining the frequency characteristics unchanged. In the specific design, a structure with five zigzags was adopted, the width of the zigzag line was set to 0.8 mm, and the adjacent spacing was set to 1.2 mm. The two-dimensional planar topology model of the antenna unit was established through the optimized geometric parameters, and the model was printed on the flexible substrate material Arlon CuClad 250LX by high-precision laser etching. The substrate has the characteristics of a dielectric constant of 2.5, a loss tangent of 0.0018, and a thickness of 0.5 mm, which is conducive to achieving low-loss transmission in the high-frequency band and stability of the antenna's electrical performance, thereby producing a unit antenna with a complete structure and excellent electrical performance. Based on beamforming theory, the antenna elements are arrayed with half-wavelength spacing, ensuring phase synchronization and pattern reconstruction capabilities between array elements near the operating frequency of 915 MHz. After this arrangement, each antenna element is independently connected to its corresponding RF front-end circuit module, which includes a low-noise amplifier, bandpass filter, and mixer, forming a structurally independent and functionally complementary multi-channel receiving path. On this basis, phase adjustment circuits and digital control logic are integrated, using an FPGA as the control core to achieve amplitude and phase adjustment, on-off control, and signal calibration for each receiving channel, thus constructing a steerable antenna array with beam steering capabilities. To improve the system's environmental adaptability and deployment stability, a comprehensive packaging solution is applied to the antenna array and the circuit structure above it. The sealing structure is designed using packaging materials with excellent electromagnetic wave transmission properties. Without affecting electrical performance, it meets the IP65 waterproof and dustproof requirements, forming a multi-channel RFID tag identification spatial receiving device. The receiving device is connected to the multi-channel signal acquisition circuit through a standard RF signal interface. The acquisition circuit integrates a high-linearity, low-noise front end, a synchronous sampling analog-to-digital converter, and a digital phase-locked module. After receiving the analog signal from each channel, it sequentially completes signal amplification, noise filtering, spectrum interception, and high-speed synchronous sampling. It also maintains the synchronization accuracy of the multi-channel signals at the nanosecond level through a time alignment mechanism and outputs a multi-channel digital data stream.

[0013] For the antenna analog signals output by the multi-channel RFID receiver, a set of highly linear amplifiers and bandpass filters with programmable gain are configured for each channel. Preset control logic adjusts the amplifier gain range to adapt to varying incident signal strengths, ensuring high dynamic range even in the presence of strong interference or long-distance identification. The center frequency of the bandpass filter is strictly limited to the RFID operating frequency band of 915MHz, with a bandwidth of 100MHz to effectively suppress out-of-band interference and harmonic components. This constructs a signal conditioning circuit with adaptive amplification and spectrum shaping capabilities. The output signal of the signal conditioning circuit is input into a 16-bit high-precision analog-to-digital converter for 10MSPS sampling. This ensures wide dynamic response while achieving fine-grained quantization of weak signals, allowing the analog waveform of each channel to be accurately characterized as a digital signal sequence with high time resolution. The sampled raw digital signal is input to the FPGA logic control platform, where a multi-stage digital signal preprocessing process is performed. The digital down-conversion module shifts the signal from the center frequency to the zero-band, freeing up system processing bandwidth and enhancing frequency-domain efficiency. The decimation filter module then uses a combination of CIC and FIR filters to reduce redundant sampling points while preserving the primary frequency component, thereby reducing subsequent data throughput. The time-domain alignment module then compensates for phase and time offsets in each channel's data to achieve sample-level synchronization between multi-channel sampling points. After the preprocessing process is complete, the processed data is written in real time to a 256MB DDR3 memory cache. This cache supports high-bandwidth, low-latency access and can accommodate data streams across multiple sampling periods, ensuring continuous and stable subsequent data scheduling. Furthermore, a separately constructed clock distribution network, using a global master clock as a reference source, calibrates and realigns each cached data block through phase synchronization circuitry and clock correction logic, resulting in a time-aligned, synchronized data stream. The synchronized data stream is transmitted to the digital signal processor at the core of the system through a high-bandwidth, low-latency high-speed interface (such as AXI or PCIe bus), and unified structured data integration is completed within the processor to form a multi-channel digital data stream.

[0014] Step S2: performing singular value decomposition, low-order polynomial fitting, and gradient-based binary search optimization on the RFID tag response signal according to the multi-channel digital data stream to obtain a beamforming weight vector; Specifically, complex vector analysis is performed on the multi-channel digital data stream. This involves extracting the amplitude envelope, instantaneous phase, and frequency offset of the signal received by each antenna channel in the time domain. This constructs an M×N complex signal matrix, where M represents the number of receiving channels in the array and N represents the number of time sampling points corresponding to each channel. This matrix describes the signal distribution characteristics of the spatial receiving device in both time and space. Singular value decomposition is then performed on the M×N complex matrix, expressing it in the form of UΣV^H. U and V are unit orthogonal matrices, and Σ is a diagonal matrix of singular values. The decomposition results reveal the distribution characteristics of the dominant and noise subspaces in the signal. By analyzing the number of dominant singular values ​​in Σ and their energy contribution, the low-dimensional feature space of the signal is extracted. Based on the feature space, the fitting error and model complexity ratio are calculated for different order models. The optimal model order k is determined using the minimum error-complexity ratio as the criterion. The k value represents the minimum dimension required for low-order modeling and controls the degrees of freedom in subsequent polynomial modeling. Based on the optimal model order k, a set of three-dimensional polynomial basis functions is constructed based on three physical variables strongly correlated with the tag response characteristics: distance, frequency, and angle of incidence. This set of basis functions is used to model the changing trends in the low-dimensional feature space, constructing an intermediate model containing a three-variable polynomial expression. Response data from known tag positions in the system are collected as calibration samples, and the polynomial model is fitted using the least squares method to obtain an initial fitting coefficient vector. These coefficients are iteratively optimized using the recursive least squares method. By continuously introducing new data samples to adjust the coefficient values, the model achieves good real-time and adaptive capabilities, resulting in a low-order polynomial fitting model. Based on the constructed fitting model, it is used as the response evaluation function for weight vector optimization. Combined with array signal beamforming theory, a gradient descent-based binary search strategy is introduced with the goals of enhancing directivity and suppressing interference. A progressive search is performed in the complex space of the weight vector. At each step, the search direction is guided by the objective function gradient calculation, the feasible domain of the weight vector is narrowed through the bisection strategy, and the optimal solution is gradually converged while ensuring the constraints (such as amplitude normalization and phase continuity). Finally, a set of beamforming weight vectors that match the current environment and the tag response characteristics is obtained.

[0015] In this embodiment, a low-order fitting model is used as the mathematical basis for expressing the signal directional response characteristics. This model is then used as the core to construct a joint optimization objective function. This objective function focuses on minimizing energy leakage in non-target directions and maximizing signal gain in the target direction. It comprehensively considers the array's mainlobe directivity, sidelobe suppression capability, and signal interference environment. By introducing a joint control mechanism for directional angle, frequency, and distance variables into the low-order model, a multi-dimensional optimization cost function is established. This function reduces the risk of signal coupling and mutual interference while satisfying directional enhancement requirements. To improve optimization efficiency and reduce the computational complexity of global search in high-dimensional space, a binary search-based framework is employed to structurally transform the joint optimization objective function, transforming the continuous variable optimization problem into a set of decidable feasibility checks. The cost function threshold in the minimization problem is represented as a parameter λ, and a determination is made as to whether a set of beamforming weight vectors exists within this λ such that the cost function value does not exceed λ. If so, the λ is feasible; otherwise, it is not. By continuously adjusting the λ value range and narrowing the search range, the lower bound of the optimal objective function value is gradually approached. During each feasibility determination, multiple initial beamforming weights are randomly generated, and the complex weight distribution of the multi-channel antenna array is initialized using random vectors to obtain an initial set of beamforming vectors. Starting from the initial beamforming vectors, a descent update is performed on each vector in the gradient direction based on the objective function. The gradient vector of the objective function with respect to the current weight is calculated, and iterative optimization is performed in the opposite direction of the gradient with a set step size. The updated vectors gradually approach the local optimal solution. Because the optimization process must be completed within a set of vector spaces that satisfy electromagnetic and physical constraints, after each gradient update, an orthogonal projection operation is performed on the optimized weight vectors onto a constraint set. This constraint set includes conditions such as weight normalization constraints, maximum sidelobe level limits, and fixed beam mainlobe directions. The projection operation ensures that all weights remain within the feasible region, thereby preventing model divergence and failure during the solution process. After completing multiple rounds of projection updates, all iterated projection weight vectors are input into the global optimization module. Combined with heuristic strategies such as simulated annealing, perturbation re-optimization, local optimal integration and other mechanisms, multiple local optimal results are performance evaluated and cross-combined to find the weight solution with the best global performance in the entire weight vector solution space, and finally the beamforming weight vector is obtained.

[0016] Step S3: performing particle swarm iterative optimization on the RFID tag recognition parameters based on the beamforming weight vector to obtain a target working parameter set; Specifically, based on the beamforming weight vector and a low-order fitting model, a multidimensional optimization variable set containing key RFID control elements is constructed. This set includes the two basic electrical parameters of the RFID system: transmit power and operating frequency. It also includes the beamforming weight vector itself used for spatial pointing control, as well as multiple algorithmic control variables involved in the signal processing algorithm, such as the detection threshold, filtering parameters, and response time constant. This forms a complete and interdependent parameter joint optimization problem. Based on this optimization problem, a reasonable search range and initial value are set for each dimension in the above optimization variable set. For example, the transmit power range is set to 0 to 30dBm with a step size of 0.5dBm, the operating frequency is limited to 860 to 960MHz in the UHF band, the beam weight dimension sets the upper and lower limits of the real and imaginary parts of the complex vector according to the size of the antenna array, and the signal processing parameters are set with threshold ranges based on algorithm stability, resulting in an initial parameter state covering the entire system control domain. Based on the initial parameter state, the particle swarm algorithm (PSO) uses its swarm intelligence search mechanism to generate a swarm of M particles. Each particle represents a candidate parameter combination. Each particle is assigned a velocity vector of consistent dimension. The initial value of this velocity vector is generated using a normal or uniform distribution, ensuring sufficient exploration in the parameter space and forming the initial particle distribution. At the beginning of the optimization phase, each parameter combination represented by a particle undergoes a fitness function evaluation. This evaluation function uses a label response prediction surface constructed using a low-order fitting model and a weighted scoring mechanism based on performance metrics such as recognition success rate, response time, energy efficiency, and recognition distance. Each particle is assigned a comprehensive performance score. Based on the fitness function output, the algorithm records the global optimal particle and the historical optimal position of each individual particle. In the next iteration, the algorithm updates the velocity and position of each particle according to the classical PSO update formula, aiming to maintain individual learning memory while moving it towards the swarm optimal position. An inertia weight is introduced in the velocity update process to control the convergence trend of the search step size. Local and global learning factors are also set to guide the balance between exploration and exploitation. After the generation of a new particle swarm, a new fitness function evaluation and iterative update are performed. After multiple rounds of iterative optimization, the particle swarm gradually converges and stabilizes in the optimal region within the global solution space. The system then extracts the particle position with the highest global score as the final optimization result. The resulting globally optimal particle corresponds to a set of target operating parameters, including the optimal transmit power, optimal operating frequency, most appropriate beamforming weight vector, and the signal processing algorithm parameters with the highest recognition efficiency in the current electromagnetic environment, selected through the particle swarm's intelligent search. These parameters will be used in tag recognition scheduling, beam steering strategy execution, and interference suppression mechanism deployment during subsequent system operation.

[0017] Step S4: sectorize the multi-tag identification space according to the target working parameter set and perform RFID tag anti-collision identification in combination with dynamic time slot allocation to obtain a target tag identification result, which is used to optimize the preparation parameters of the RFID electronic tag.

[0018] Specifically, an anti-collision recognition mechanism is constructed using the target operating parameter set derived from particle swarm optimization as input. This mechanism establishes a tag response control protocol based on an initial number of time slots of 16. Based on this, a dynamically updateable Q-value adjustment strategy is designed. By providing real-time feedback on the current tag recognition success rate, number of collisions, and response distribution, dynamic Q-value adjustment parameters are set, enabling the system to improve resolution in densely identified areas and enhance response efficiency in sparsely identified areas. Supported by a beamforming antenna array, the recognition space is finely divided into multiple sectors based on the beamforming weight vectors in the target operating parameter set. The entire space is divided into several angular regions, each assigned independent RF activation directions and detection control instructions. Combined with dynamic Q-value adjustment parameters, joint anti-collision control in the time and space domains is achieved. Doppler shift features are extracted and analyzed for the tag reflection signals received within each sector. A short-time Fourier transform is then performed on the tag's motion status and historical response records to generate predicted trajectory data for tags with small frequency drifts. This data is used to dynamically estimate the tag's potential spatial position and response priority, providing motion trend support for subsequent scheduling mechanisms. Based on tag trajectory prediction information, a fairness guarantee mechanism is introduced to assign higher response priority to tags that have not been successfully identified within multiple consecutive identification cycles. Specifically, the status of unidentified tags is incorporated into the priority scheduling function. This, combined with the response probability of the traditional Q algorithm, dynamically increases the activation probability and channel resource allocation weight of such tags in each round of time slot allocation. This avoids the "starvation effect" where some tags lose contact for extended periods due to path obstruction, signal attenuation, or multi-tag collisions, thereby improving the overall identification coverage integrity and response fairness of the system. After completing dynamic sector division, collision resolution, and trajectory prediction, the system performs feature analysis and screening on the tag reflection signals successfully received within multiple identification cycles. A feature fusion decision algorithm is executed based on the tag's signal strength, response delay, Doppler shift, phase change, and historical trajectory characteristics to extract the tag's identity information, relative position, and operating status. The final data is then structured through multi-channel data fusion and a tag decoding module. The resulting target tag recognition result includes a high-confidence tag identity code, an estimated three-dimensional position in physical space, motion trend analysis, and operational status indicators such as tag stability and collision frequency. The target tag identification results are fed back to the RFID electronic tag preparation system as reverse adjustment input to dynamically optimize the tag antenna parameters, matching impedance structure and packaging form, so that the subsequently prepared RFID tags have better recognition adaptability, response consistency and signal stability in the corresponding spatial sector and specific environment.

[0019] In this embodiment, a parameter mapping relationship associated with antenna performance is constructed based on the target tag recognition results. The recognition results include key performance indicators such as recognition strength, response stability, and bit error rate under different directions, positions, and signal environments. By correlating and analyzing the changing trends between performance indicators and antenna structural parameters, the positive and negative effects of adjustments to multiple variables, such as meander line width, length, spacing, antenna overall size, substrate thickness, and dielectric constant, on the recognition results are attributed statistically. The degree of influence of each design parameter on the tag performance change is quantified, and the sensitivity weight of each structural parameter is obtained, reflecting the importance and control priority of each parameter in overall performance optimization. Based on the target design parameters and their sensitivity weights, a multi-objective optimization model is constructed that comprehensively considers tag size, read distance, and cost. This model simultaneously considers three core objectives: tag size reduction, read distance maximization, and manufacturing cost control. The model structure is configured as a combination of objective functions with weighted coefficients. The size function is measured by the total antenna area, the read performance is measured by the minimum recognition power at a standard test distance, and the cost is quantified by the material unit price, process complexity, and assembly difficulty. By incorporating sensitivity weights to constrain the influence of design variables in each objective function, the optimization model not only seeks optimal performance but also implements an efficient and controllable sample generation strategy within the design space. Based on this model, Latin hypercube sampling or genetic algorithms are used to extensively explore the high-dimensional parameter space, generating a set of candidate parameter combinations containing a large number of feasible solutions. Each combination corresponds to a specific antenna structure design. Using these candidate parameter combinations as input, RFID tag samples are batch-produced in an experimental environment. The sample preparation process involves laser etching, die attach, impedance matching layer design, and packaging process control based on the optimized structural parameters, ensuring that each sample is produced within controlled process deviations. Following preparation, the sample tags are performance-tested under standardized test conditions. Test criteria include typical read distance, recognition angle range, frequency response bandwidth, extreme recognition sensitivity, and tag signal consistency. By accumulating test data from multiple batches and environments, a tag performance validation dataset is formed. Statistical analysis is then conducted on the average and extreme performance of each parameter combination. All tag performance verification data is fed into a comprehensive evaluation model, which conducts a multi-dimensional evaluation based on structural parameters, performance, and manufacturing costs. The model compares the performance, cost, and size trade-offs of various parameter combinations. By incorporating a Pareto-optimal frontier-based decision-making mechanism, the model selects a set of parameter configurations with the best performance or the highest overall score across all evaluation dimensions. This selection then determines the final RFID tag manufacturing parameters. The optimal manufacturing parameters encompass antenna geometry design, impedance matching methods (such as conjugate matching embedding strategies), packaging material selection, and structural layout.

[0020] In an embodiment of the present invention, a miniaturized RFID reader antenna designed using zigzag technology maintains excellent radiation performance while reducing its size. It directly connects to the receiving circuit through conjugate impedance matching, eliminating the need for an external matching network. This simplifies system design, reduces power consumption, and improves antenna performance and integration. High-precision digital processing and precision clock synchronization ensure the complete preservation of the time and frequency domain characteristics of multi-channel signals. FPGA real-time preprocessing and high-speed cache design ensure the continuity and real-time nature of data processing, providing high-quality input data for subsequent algorithms. A low-order model based on singular value decomposition and polynomial fitting accurately describes the interaction characteristics between the reader and tag, significantly reducing computational complexity and enabling real-time adaptive adjustments in complex environments, thereby improving the robustness and adaptability of the recognition system. The gradient-based binary search optimization method proposed in this invention solves directly in the original beamforming domain, achieving real-time beamforming and significantly improving tag direction recognition accuracy and signal-to-interference-noise ratio performance. Particle swarm iterative optimization is used to globally optimize system parameters, minimizing power consumption while ensuring recognition accuracy, improving system efficiency, extending equipment life, and reducing maintenance costs. By combining spatial partitioning with the improved dynamic time-slot ALOHA algorithm, efficient recognition is achieved in a dense multi-tag environment. The fairness guarantee mechanism solves the tag starvation problem. Doppler feature analysis enhances the tracking and recognition capabilities of mobile tags, significantly improving the system's recognition performance in complex environments.

[0021] In a specific embodiment, the process of executing step S1 may specifically include the following steps: The meander line structure of the antenna radiating element is calculated to obtain the structural parameters of the radiating element. The structural parameters of the radiating element are then applied to the flexible substrate Arlon CuClad 250LX for printing to obtain a unit antenna. Arrange the unit antennas at half-wavelength intervals to form an antenna array, and connect each unit antenna of the antenna array to an independent RF front-end circuit to obtain a multi-channel receiving path; Phase adjustment circuits and digital control logic are configured for the multi-channel receiving path to obtain a controllable antenna array, and the controllable antenna array is designed to be waterproof and dustproof to obtain a multi-channel RFID tag recognition space receiving device; The multi-channel RFID tag identification space receiving device is connected to the multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream.

[0022] Specifically, a zigzag line structure is used to tune the electromagnetic performance of the antenna. The core purpose of adopting a zigzag structure for the antenna radiating unit is to significantly reduce the size of the antenna without sacrificing effective radiation efficiency. Based on the target operating frequency band (860MHz to 960MHz in the UHF band), a two-dimensional planar structural model is constructed, and the initial antenna geometric framework is established through electromagnetic simulation software. The shape of the zigzag line is set as a periodic return line segment. By relying on geometric compression, the effective electrical length is increased within a limited space, so that the antenna obtains optimized standing wave ratio characteristics and controllable radiation pattern near 915MHz. In this process, multiple structural parameters are set as solution variables, including the line width, line spacing, number of returns, number of periods, total unfolded length and coverage area width of the zigzag line, to optimize its input impedance, electromagnetic standing wave, gain directivity and current distribution uniformity at the target frequency point. After multiple rounds of simulation optimization, the researchers determined a meander line width of 0.8mm, a spacing of 1.2mm, a total fold length of approximately 42mm, a width of 15mm, and a fold count of five, resulting in an optimal radiating element structure with both performance and size. These radiating element structural parameters were then applied to the Arlon CuClad 250LX flexible dielectric substrate for physical printing. This substrate material exhibits excellent dielectric properties (dielectric constant εr = 2.5), extremely low dielectric loss (loss tangent tanδ = 0.0018), and high high-frequency signal transparency. In actual fabrication, a high-precision laser etching process was used to etch the 0.035mm thick copper foil onto the 0.5mm thick CuClad 250LX substrate according to the designed pattern. This ensured sharp etched line edges and a line width error of less than ±0.05mm, preventing standing wave reflections or energy leakage during high-frequency signal transmission. This resulted in the fabrication of a single, high-performance flexible antenna unit. After the unit antennas are fabricated, their spatial arrangement is designed based on the array signal processing requirements. They are arranged in a linear array with half-wavelength spacing. Half-wavelength spacing refers to the physical distance formed by half a period of electromagnetic wave propagation in free space at an operating frequency of 915 MHz. This spacing prevents strong coupling between units, which can cause directional pattern distortion, and ensures the array spatial resolution required by the beamforming algorithm. During the arrangement process, all antenna units are aligned in their orientation and centers, and each is connected to an independent RF front-end module via coaxial feed lines of equal length. Each front-end module contains key functional circuits such as a low-noise amplifier, bandpass filter, and mixer, forming a multi-channel receiving path with spatial diversity characteristics. These paths receive RFID tag response signals in parallel from different directions and distances. Phase adjustment circuits and digital control logic are added to the multi-channel receiving path.The phase adjustment module uses programmable phase shifters to perform high-precision phase adjustment on the input signal of each channel, enabling coherent combination of the signals in the digital domain to support subsequent beamforming processing. The digital control logic, implemented in an FPGA architecture, dynamically configures and controls the phase shifters and performs operations such as channel enabling, amplitude equalization, and signal synchronization, ensuring flexible and reliable adjustment even when identifying densely populated tags or in environments with strong interference. The entire steerable antenna array must also meet the environmental adaptability requirements of actual deployment scenarios. Therefore, mechanical and process packaging is performed. The entire antenna array is sealed and encapsulated with a high-transmittance, low-water-absorption polymer protective material, such as modified PC or PES. The IP65 standard design provides dust and water resistance, ensuring industrial-grade durability and long-term operation in high humidity, dust, and outdoor rain and snow. This constitutes a multi-channel RFID tag identification spatial receiver with directional control capabilities and environmental adaptability. After the antenna array and control circuitry are packaged, the receiver connects to the backend signal acquisition circuitry via a standard RF signal channel to establish a data acquisition link. The acquisition circuitry comprises multiple parallel channels, each corresponding to an antenna path. It amplifies and bandpass filters the analog RF signal, then performs high-rate analog-to-digital conversion with 16-bit precision and a 10MSPS sampling rate, converting each channel's signal into a digital format. The acquisition system integrates a unified clock source and synchronous calibration mechanism to ensure that multi-channel sampling results are aligned in the time domain with nanosecond accuracy, preventing the accumulation of phase errors during subsequent processing. After analog-to-digital conversion, all digital signals are synchronously input into the FPGA platform, where digital down-conversion, decimation filtering, amplitude and phase calibration, and time alignment are performed to form a multi-channel digital data stream.

[0023] In a specific embodiment, the process of connecting the multi-channel RFID tag identification space receiving device to the multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream may specifically include the following steps: An amplifier and a bandpass filter with a preset programmable gain range are configured for each antenna channel to obtain a signal conditioning circuit; The output signal of the signal conditioning circuit is input into a 16-bit high-precision analog-to-digital converter for 10MSPS sampling to obtain a digital signal; The digital signal is processed by digital down-conversion, decimation filtering, and time-domain alignment through FPGA to obtain a pre-processed data stream, which is then temporarily stored in a 256MB DDR3 cache to obtain cached data. A clock distribution network is used to perform time synchronization processing on the cached data to obtain a synchronous data stream, and the synchronous data stream is transmitted to the digital signal processor through a high-speed interface to obtain a multi-channel digital data stream.

[0024] Specifically, a signal amplification unit with programmable gain and a bandpass filter responsive to the center frequency are configured for each antenna channel to construct a high-performance signal conditioning circuit. During this process, the antenna output signal connected to each channel passes through a low-noise amplifier. This amplifier supports multi-level gain control and implements dynamic gain adjustment through programmable logic, allowing the system to flexibly set the gain value according to the input signal strength under different environmental conditions. The gain setting range is between 0 and 30dB, and it has linear response characteristics and low phase distortion capabilities, which can effectively enhance the amplitude of weak signals without introducing excessive noise. When the amplifier signal enters the bandpass filter module, the filter center frequency is fixed at 915MHz and the bandwidth is set to 100MHz to accurately match the operating frequency band of the RFID system. The filter also has excellent out-of-band suppression characteristics and extremely low insertion loss, which can effectively filter out interference signals far away from the frequency band and retain the required spectrum portion, thereby completing the signal's frequency domain shaping and noise isolation. At the conditioned signal output, each channel's analog RF signal is fed into a high-precision analog-to-digital converter (ADC). Designed with 16-bit quantization accuracy and a high-speed sampling rate of 10MSPS, this ADC offers high linearity, high dynamic range, and low nonlinear distortion, enabling high-fidelity digitization of transient RF signals. Sampling is performed synchronously, with independent ADCs configured for each channel. This ensures consistent sampling time and signal correlation within the multi-channel parallel processing architecture, resulting in high temporal and amplitude resolution of the output digital signal. The digitized signals are then fed into an FPGA logic processing platform, where digital downconversion is performed, shifting the 915MHz RF signal down to zero intermediate frequency (IF) or near baseband. This reduces the computational bandwidth requirements of subsequent processing algorithms and enhances the time-domain processability of the signal representation. The downconversion process, including quadrature mixing, low-pass filtering, and resampling, is performed simultaneously across all channels to ensure consistent processing timing. The decimation filter module then enters the data center. The decimation filter utilizes an FIR structure, coupled with a CIC cascade architecture. This reduces the sampling rate while retaining key frequency components and effectively suppressing image interference and high-frequency residuals, thereby reducing data bandwidth and processing load. Time-domain alignment is then performed on the digital signals of all channels. Interpolation delay compensation and frame boundary synchronization mechanisms eliminate sample offsets introduced by hardware delay differences between channels, ensuring that all channel signals at the same moment represent consistent spatial response states and outputting stable, aligned, and synchronized pre-processed data streams. The pre-processed data stream is written to a 256MB DDR3 cache module, which supports high-speed parallel writes and multi-channel parallel reads. Address mapping and timing control are used internally to achieve conflict-free access to multiple data channels, maintaining data integrity while ensuring system stability and efficiency.The data cache stage coordinates with the system master clock for unified scheduling, introduces a dedicated clock distribution network, and globally time-stamps and sequentially integrates the cached data to ensure that all data blocks have consistent timestamps and continuous structural order before being read by the processor. After completing time synchronization and cache stabilization, the cache controller sends the synchronized multi-channel digital data streams in batches to the back-end digital signal processor module through a high-speed interface (such as AXI, LVDS or PCIe). This module is composed of a high-performance floating-point DSP or SoC platform, and internally executes complex algorithms including beamforming, singular value decomposition, spectrum analysis, data fusion and label recognition, and performs dynamic parameter estimation and target judgment on each digital signal. The final output multi-channel digital data stream is a basic signal set with unified structure, time synchronization and signal enhancement, with complete time-frequency-amplitude information and channel mapping.

[0025] In a specific embodiment, the process of executing step S2 may specifically include the following steps: Perform complex vector analysis on the multi-channel digital data stream to extract the signal amplitude, phase and frequency characteristics, and obtain an M×N dimensional complex matrix, where M represents the number of antenna array channels and N represents the number of sampling points; Perform singular value decomposition on the M×N dimensional complex matrix to obtain a low-dimensional feature space, and calculate the fitting error and complexity ratio of the increasing order model based on the low-dimensional feature space to obtain the optimal model order k value; A polynomial basis function representation is constructed based on the optimal model order k value, and a polynomial model containing three variables: the distance between the reader and the tag, the operating frequency, and the incident angle is obtained; The calibration data collected under known tag positions are used for least squares calculation to obtain the fitting coefficients of the polynomial model, and the fitting coefficients are recursively calculated using least squares to obtain a low-order fitting model. A low-order fitting model is used to perform gradient-based binary search optimization on the multi-channel antenna array weights to obtain the beamforming weight vector.

[0026] Specifically, complex vector analysis is performed on multi-channel digital data streams. In a multi-channel RFID receiving system, each receiving antenna channel receives a complex signal containing amplitude, phase, and frequency drift at each time sampling point. Complex vector analysis maps the original discrete signal into a structured complex matrix. The specific operation involves calculating the real and imaginary parts of each channel at each sampling moment and combining them into a complex-valued representation. This results in an M×N complex matrix, where M represents the number of receiving antenna channels and N represents the number of time sampling points. In this matrix, each row corresponds to the time evolution information of a channel, while each column represents the complex signal distribution in all spatial directions at a given moment. This constructs a signal representation structure with triple coupling characteristics of space, time, and phase. Singular value decomposition is performed on the M×N complex matrix, decomposing it into the product of three matrices, namely, UΣV^H, where U and V are unit orthogonal matrices, representing the spatial and temporal eigenvectors of the signal, respectively, and Σ is a diagonal matrix containing singular values ​​arranged in descending order of energy. These singular values ​​correspond to the basis for distinguishing the dominant components of the signal from the noise components. Only the characteristic subspaces corresponding to the first few main singular values ​​are retained, covering the vast majority of the signal energy, to obtain a low-dimensional feature space. Based on the low-dimensional feature space, the fitting error and complexity ratio of the increasing order model are calculated. For each increasing order k, the fitting error and model complexity of the model constructed based on this k-order subspace are calculated separately. Then, by constructing the "error-complexity ratio" indicator to determine the minimum value, the optimal model order k value that minimizes this ratio is obtained. This k value reflects the minimum sufficient expression dimension of the signal structure under the current system environment. Based on the optimal model order k value, a low-order polynomial model suitable for beam pointing control is constructed. Three key physical factors influencing signal response—the distance between the reader and the tag, the system's current operating frequency, and the tag's relative incident angle—are considered independent variables. A three-dimensional joint expression structure is established in the form of polynomial basis functions. By constructing a set of polynomial basis functions, a function mapping model in k-dimensional space is formed. Each principal singular value or principal eigenvector is approximated by a nonlinear combination of these three variables, establishing a mathematical connection between the tag's spatial position and the signal characteristic expression. To obtain the specific coefficients in this polynomial model, a batch of calibration data with known tag positions is introduced. This data consists of the actual spatial coordinates of the tags recorded in an experimental environment and their corresponding actual response matrices. The model parameters are fitted using the least squares method to minimize the sum of squared errors between the signal characteristics output by the fitted model and the actual sampled data, thereby determining the coefficient values ​​for each polynomial basis function.Considering the potential for new location changes or interference conditions during actual tag deployment, the initial fitting coefficients are continuously updated using a recursive least squares method. This means that the fitting parameters are updated in real time as new tag identification data arrives. This allows the model to maintain historical memory while also enabling adaptive updates, resulting in a low-order fitting model. This low-order fitting model is then applied to the optimization of the beamforming weight vector. The fitting model serves as a signal directional gain estimation function, and a target optimization function is established. This function aims to improve the main beam gain, suppress sidelobe leakage, and enhance the signal-to-noise ratio and interference rejection. The sum of the squared amplitudes of the beamforming vector is constrained to be a constant, forming a typical nonlinear constrained optimization problem. A binary search method based on gradient descent is used to solve the problem. A set of candidate initial weight vector values ​​is constructed, and their corresponding directional gains or objective function values ​​are calculated based on the fitting model. A cost threshold λ is then constructed to transform the optimization problem into determining whether a feasible solution exists that satisfies the constraints and has a cost less than λ. In each search, the search bounds are adjusted based on the feasibility judgment of the current λ value, and the search interval is continuously narrowed until convergence is met. In each iteration, the weight vector is updated based on the gradient direction of the cost function and normalized to ensure it always falls within the set of constraints. After several rounds of iterative optimization, the system converges to a set of beamforming weight vectors with optimal performance.

[0027] In a specific embodiment, the step of performing gradient-based binary search optimization on the multi-channel antenna array weights using a low-order fitting model to obtain a beamforming weight vector may specifically include the following steps: A joint optimization objective function for minimizing RFID tag orientation recognition is constructed based on a low-order fitting model; A binary search framework is used to transform the joint optimization objective function to obtain a feasibility check problem. Based on the feasibility check problem, a random weight vector is initialized to obtain an initial beamforming vector. Calculate the objective function gradient for the initial beamforming vector and update it in the opposite direction of the gradient to obtain an iteratively optimized weight vector, and project the iteratively optimized weight vector onto the constraint set to obtain a projected weight vector; The projection weight vector is globally optimized to obtain the beamforming weight vector.

[0028] Specifically, based on a low-order fitting model, a comprehensive objective function is established for the RFID tag directional recognition problem, encompassing directional enhancement, interference suppression, and recognition stability control. This objective function describes the mapping relationship between received signal characteristics in different directions and beamforming weights. Its core goal is to maximize the signal response in the direction of the target tag while simultaneously reducing interference responses in non-target directions, thereby improving tag recognition accuracy and directional resolution. The low-order fitting model provides a mapping mechanism for predicting signal strength in any direction. The joint optimization objective function, based on this, integrates multiple recognition performance metrics and performs a weighted combination of the performance across all possible directions to measure the overall recognition performance of the current beamforming vector. To address the nonlinear, multivariate, and multi-constrained nature of this optimization objective function, a feasibility problem reconstruction method based on a binary search framework is introduced. This transforms the original performance optimization problem into a problem of determining whether a solution exists under a certain performance threshold. This process no longer directly seeks the exact minimum or maximum value, but instead transforms the optimization process into multiple rounds of feasibility testing. Each round of search sets a performance boundary and determines whether a specific beamweight combination can achieve the desired overall system performance. If a solution exists, the performance requirement is further narrowed and the search continues. If not, the lower bound is raised and a new range is explored. This approach uses a stepwise approximation strategy to reduce computational redundancy and convergence difficulties associated with blind searches in complex non-convex spaces. During each feasibility check, a random initialization method is used to generate several beamforming vectors to initialize the possible solution space. All initial vectors satisfy multiple constraints, including amplitude normalization, power limits, and phase ranges. These vectors are randomly distributed within the multidimensional complex space to ensure global search coverage. For each randomly initialized vector, its performance under the current objective function is evaluated to determine whether it falls within the feasible solution domain. If so, further local optimization is initiated. A gradient-based information update is performed on each initial vector. By evaluating the current vector's trend along the performance curve, a downward direction is determined, indicating the direction in which system performance is most easily improved. Minor adjustments are made to the vector in this direction to gradually improve recognition performance. The resulting vector is projected during each update to ensure it remains within the legal space defined by all constraints. Projection operations are performed within system parameters such as amplitude constraints, total power constraints, and phase continuity, ensuring that the optimization process remains within design boundaries. After completing multiple rounds of gradient updates and constrained projection, a set of local solutions with superior performance is obtained. A global optimization search is then performed across the entire weight space. This process utilizes various strategies, such as simulated annealing, perturbation-perturbation rearrangement, genetic algorithms, or particle swarm optimization algorithms. By combining, mutating, and reselecting multiple local solutions, a set of beamforming weight vectors with the most stable performance, the most accurate direction, and the strongest interference resistance can be more effectively identified.

[0029] In a specific embodiment, the process of executing step S3 may specifically include the following steps: Constructing an optimization variable set including transmit power, operating frequency, beamforming weights, and signal processing parameters based on the beamforming weight vector and the low-order fitting model; Setting a search range and initial values ​​for the optimization variable set to obtain an initial state of parameters including a transmit power range and an operating frequency range; Based on the initial state of the parameters, a group of M particles is generated and random velocities are assigned to obtain the initial particle distribution. The fitness function is then evaluated for each particle in the initial particle distribution to obtain a score value. The particle positions and velocities are updated according to the scoring values ​​to obtain a new generation of particle swarms that move towards the global optimal solution and the individual optimal solution; The new generation of particle swarm is iteratively calculated and the global optimal solution is selected to obtain the target operating parameter set including the optimal transmit power, operating frequency, beamforming weights and signal processing parameters.

[0030] Specifically, a multidimensional parameter optimization set is constructed based on the beamforming weight vector and low-order fitting model. This set includes two basic control variables at the RF hardware configuration level: transmit power and operating frequency. It also includes the antenna array beamforming weight vector obtained from the previous beam optimization calculation. It also integrates various parameter settings for decision, filtering, synchronization, and decoding in the signal processing chain, including key digital signal processing parameters such as detection threshold, matched filter length, synchronization window size, decision delay compensation, and time drift tolerance. This optimized variable set is a multidimensional control vector in the system's operating state space. The value of each variable directly affects the spatial reception capability, directional sensitivity, signal-to-noise ratio control capability, and recognition response speed of the identified signal. Reasonable search ranges and initial values ​​are set for each set of optimization variables. The transmit power range is set between 0 and 30 dBm based on the capabilities of the power amplifier module, and the operating frequency range is controlled between 860 MHz and 960 MHz in the UHF band. The initial values ​​of the beam weight parameters are based on previously optimized complex vectors, and boundary conditions are set to allow for small perturbations. Signal processing parameters are defined with upper and lower bounds based on algorithm performance requirements and computational resource constraints, and initial default values ​​are set. The initial set of all variables forms a central point in the search space and serves as a reference region for initializing each particle in subsequent optimization. The initial state is set to take into account practical hardware limitations, system stability requirements, and the flexible resource allocation requirements of the target recognition task. Based on this multi-dimensional parameter initial state, the particle swarm optimization algorithm enters the initialization phase, generating a swarm of M particles. Each particle is a vector structure composed of multiple optimization variables. Each particle is randomly assigned a velocity vector, which represents the trend and amplitude of the particle's change in the parameter space in each dimension. To maintain diversity in the initial population, the initial position and velocity of each particle are randomly generated using Gaussian perturbations or uniform distributions. This ensures global exploration and avoids falling into local optima throughout the search process. The initial particle population generated in this stage serves as the system's first search sample and is projected into a multidimensional parameter space for evaluation using a fitness function. The fitness function's design considers the overall RFID system performance, including tag recognition rate, recognition speed, system energy consumption, response stability, and beam direction performance. This function integrates metrics such as read success rate, recognition latency, spatial resolution, and energy efficiency, performing a weighted combination to output a comprehensive performance score for each particle under the corresponding parameter combination. The fitness evaluation process is performed on all initial particles one by one. Based on the scoring results, the individual optimal position achieved by each particle is recorded. The highest score among all particles is then compared to update the current global optimal position, completing the first round of feedback recording.Based on the fundamental principles of particle swarm optimization, the particle's historical optimal position and the current global optimal position are used as references. The velocity of each particle is updated using a set inertia factor, individual learning factor, and global learning factor. The current velocity value is then used to correct each particle's position in the search space, forming a new generation of particle swarms. The updated particles are distributed closer to the global optimal solution, gradually converging near the optimal value. After each round of updates, a new round of fitness evaluation and individual and global optimal state updates is performed. This iterative process repeats until a set termination criterion is met, such as reaching an upper limit on the number of iterations or the fitness function reaching stability and no further improvement within a certain number of rounds. When the optimization algorithm converges, the particle with the highest fitness score is selected from all particles. The set of parameters corresponding to its position represents the optimal operating parameter configuration for the current system, given the target performance weights. These parameters include the optimal transmit power, optimal operating frequency, a re-adjusted and optimized beamforming weight vector, and signal processing algorithm parameters adapted to the current wireless environment and channel conditions.

[0031] In a specific embodiment, the process of executing step S4 may specifically include the following steps: Based on the target working parameter set, a label conflict handling mechanism with an initial time slot number of 16 is designed. The Q value of the label conflict handling mechanism is adaptively adjusted according to the rules, and the Q value dynamic adjustment parameter is obtained. The recognition space is divided into multiple sectors using the beamforming weight vector and the Q value dynamic adjustment parameter to obtain recognition sectors. The Doppler characteristics of the tag signal in each recognition sector are analyzed to obtain trajectory prediction data. Based on trajectory prediction data, labels that have not been identified for multiple times are given priority to obtain a fairness guarantee mechanism; All identified tag signals are subjected to feature analysis and screening to obtain target tag recognition results containing tag identity, location, and status information. The target tag recognition results are used to optimize the preparation parameters of RFID electronic tags.

[0032] Specifically, based on the target working parameter set, an anti-collision scheduling strategy is constructed to adapt to the current identification environment and tag density. The target working parameter set includes a series of controllable factors such as transmission power, operating frequency, beamforming weights, and signal processing parameters, providing the necessary physical layer support and control space for system scheduling. In the tag identification scheduling link, the initial structure of the time slot scheduling strategy is set, where the initial value of the total number of time slots is set to 16. This value is derived from the typical tag initial response window design in the EPC Gen2 protocol, and is dynamically adjusted in combination with the current tag density and identification rate requirements, laying the foundation for building an efficient time slot anti-collision mechanism. At the same time, in order to avoid frequent conflicts or resource waste caused by static Q value settings, a Q value adaptive adjustment mechanism is introduced. The Q value represents the range of the number of tags that can be accommodated in the current identification round, and its size directly affects the granularity of each round of time slot scheduling. The adaptive rule integrates dynamic metrics such as current slot utilization, number of tag responses, collision ratio, and idle ratio to make decisions. When numerous collisions occur during the recognition process, the system automatically adjusts the Q value upward to increase the number of slots and reduce the probability of overlapping responses. When most slots are idle, the Q value is automatically reduced, improving the system's recognition rate and resource utilization. To enhance the responsiveness and decision stability of the adjustment mechanism, a smoothing factor and a memory window mechanism are introduced to ensure a gradual change in the Q value adjustment and avoid frequent oscillations. This results in a set of dynamic Q value adjustment parameters that adapt to environmental conditions. Based on this, the spatial directional weight vectors derived from antenna array beamforming are combined to perform fine-grained sectorization of the entire recognition space. This partitioning strategy, based on spatial angle, partitions the recognition area according to the main lobe direction of the beam. Each sector is controlled by a set of weight vectors that control the antenna array's receiving direction, thereby physically achieving directional awareness of tags in different spatial regions. Combining the dynamic Q value adjustment parameters with the sector control strategy, conflict control logic is independently executed within each sector, achieving decoupled scheduling in both spatial and temporal domains. During each recognition cycle, the system activates different spatial sectors in turn, identifying and processing tags within each area in turn. To improve the ability to identify moving tags in dynamic scenarios, Doppler signature analysis is performed on tag signals received in each sector. By detecting the frequency drift, phase change, and amplitude fluctuations of the tag response signal within adjacent time windows, the tag's speed, direction, and relative motion trend are inferred. Combined with time-domain continuity, Doppler trajectory feature vectors are generated. A motion model is then used to predict the tag's likely position in future rounds, enabling short-term trajectory prediction. This trajectory prediction data is used to determine whether a tag has moved out of the recognition area and to assist the system in prioritizing tags that are about to leave the recognition boundary to avoid missed or delayed recognition. Based on these trajectory prediction results, a dynamic tag priority adjustment mechanism is introduced to establish a fairness guarantee strategy.If a tag fails to respond successfully in multiple consecutive identification rounds, or if its sector experiences frequent conflicts and identification opportunities are scarce, the system assigns a higher recognition priority to that tag and prioritizes resource allocation in the next round of time slot allocation. This prevents some tags from becoming "starved" due to prolonged periods of unrecognized activity, improving overall recognition fairness and tag response integrity. A weighting factor and cumulative priority adjustment mechanism are used during the priority adjustment process to prevent high-priority tags from overcommitting and depriving other tags of recognition opportunities, thereby creating a more balanced tag scheduling structure. After the system completes multiple identification rounds and collects response data from all tags, it enters the comprehensive feature analysis phase of the tag signal. This phase performs structured decoding and attribute analysis on each successfully identified tag signal, extracting multi-dimensional attributes such as the tag's unique identification code, response frequency band, signal strength, estimated relative position, movement trajectory information, and its sector and time slot number. Combining this information, a global distribution map and status table of tags in the current identification space are reconstructed, annotating each tag's recognition quality, response stability, and potential recognition anomaly risk. The output target tag recognition result has identity information, and also carries the tag's relative position, relative speed, movement trend and multi-round recognition trajectory in physical space.

[0033] In a specific embodiment, the method for preparing an RFID electronic tag further includes the following steps: Perform sensitivity analysis on antenna structure parameters based on target tag recognition results to obtain target design parameters and their sensitivity weights; A multi-objective optimization model that comprehensively considers tag size, reading distance, and cost is constructed based on the target design parameters and their sensitivity weights. The parameter space of the multi-objective optimization model is explored to obtain a set of candidate parameter combinations. Prepare multiple sets of sample labels based on the candidate parameter combination set and perform performance testing and verification to obtain label performance verification data; A comprehensive evaluation of the tag performance verification data is conducted and the optimal preparation parameters of the RFID electronic tag, including antenna design, impedance matching and packaging structure, are selected.

[0034] Specifically, a systematic sensitivity analysis of antenna structural parameters was conducted based on key performance indicators (KPIs) such as signal strength, read distance, directional stability, signal-to-noise ratio variation, recognition angle range, and recognition stability, as reflected in the target tag recognition results. This analysis revealed the extent to which small changes in antenna geometry affect recognition performance in real-world environments, establishing a mapping between structural parameters and performance output. Several key design parameters were selected as primary variables, such as antenna meander width, line spacing, number of turns, total length, conductor thickness, and the dielectric constant and thickness of the dielectric substrate. Single-factor perturbation simulations or experimental perturbations were performed on each variable dimension, and the rate of change in recognition performance before and after the perturbation was recorded. The sensitivity coefficient of each design parameter was then calculated based on the relative change ratio. Through normalization, all sensitivity coefficients were converted to relative weights, resulting in sensitivity weights representing the contribution of each parameter to tag recognition performance. A multi-objective optimization model was constructed based on the target design parameters and their sensitivity weights, comprehensively considering tag size, read distance, and cost. Tag size is related to meander line length, turn density, and substrate area; read distance is related to antenna gain, impedance matching, and pattern distribution; and cost is influenced by conductor usage, substrate material grade, and processing accuracy. The optimization model is constructed using a weighted objective function combination approach, expressing the three performance objectives as cost functions. Sensitivity weights are then used to apply control factors to different parameter dimensions, thereby establishing a control structure that retains multi-objective optimization capabilities while being adaptive to changes in key parameter sensitivities. After modeling, the model's parameter space is systematically explored. To ensure global coverage while balancing solution density and diversity, candidate solutions are generated using Latin hypercube sampling, multi-point uniform random distribution, orthogonal experimental methods, or evolutionary optimization methods (such as NSGA-II). This generates a comprehensive set of candidate parameter combinations. Each combination represents a specific set of antenna structure parameter configurations, including geometry, conductor width, substrate properties, and matching interface dimensions. Sample preparation experiments are conducted based on these candidate parameter combinations, and each parameter configuration is converted into a physical tag using standardized RFID tag processing techniques. During the production process, it is necessary to ensure that processing accuracy is within a repeatable range. For example, the line width error of laser etching does not exceed ±0.05mm, the position deviation of the mounted chip is controlled within ±0.1mm, and the packaging material thickness is uniform and has consistent electromagnetic wave transmission performance. After sample preparation, performance verification experiments are carried out under standard test conditions. Test items include static reading distance testing, multi-angle recognition rate testing, recognition stability under changing tag posture, signal integrity testing under reflection and interference conditions, and response consistency at different frequency points. By recording and comparing each set of samples across multiple test dimensions, systematic tag performance verification data is generated. This tag performance verification data is then comprehensively evaluated.A weighted summary analysis of recognition performance, structural indicators, and cost metrics corresponding to each candidate parameter configuration is performed. The evaluation process uses multi-attribute decision-making methods (such as TOPSIS, AHP, or hierarchical fuzzy comprehensive evaluation) to normalize different performance dimensions. This process, combined with the weight distribution of key parameters in the aforementioned sensitivity weighted reinforcement model, aims to establish a balance between performance score, cost factors, and structural feasibility. The highest-scoring parameter combination in the evaluation results represents the optimal tag fabrication solution under the current constraints. The optimal parameter combination ultimately selected is summarized into a complete fabrication parameter set, encompassing antenna geometry design parameters, matching impedance configuration strategies, and tag packaging materials and structural design methods. The antenna design defines the specific zigzag structure's shape, linewidth distribution, overall dimensions, and grounding structure. The impedance matching component determines whether to integrate a matching network or achieve implicit matching through structural conjugation based on the chip's impedance. The packaging structure is configured hierarchically based on signal transmission efficiency, protection level, material adhesion, and tag flexibility requirements.

[0035] The above describes the method for preparing the RFID electronic tag in the embodiment of the present invention. The following describes the device for preparing the RFID electronic tag in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, a device for preparing an RFID electronic tag includes: A digital processing module is used to connect the multi-channel RFID tag identification space receiving device to the multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream; A fitting module is used to perform singular value decomposition, low-order polynomial fitting, and gradient-based binary search optimization on the RFID tag response signal according to the multi-channel digital data stream to obtain a beamforming weight vector; Iterative optimization module, used to perform particle swarm iterative optimization on RFID tag recognition parameters based on beamforming weight vector to obtain the target working parameter set; The identification module is used to sectorize the multi-tag identification space according to the target working parameter set and perform RFID tag anti-collision identification in combination with dynamic time slot allocation to obtain the target tag identification result, which is used to optimize the preparation parameters of the RFID electronic tag.

[0036] Through the synergistic cooperation of the above-mentioned components, the present invention uses a miniaturized RFID reading antenna designed with zigzag line technology to reduce the volume while maintaining good radiation performance, and is directly connected to the receiving circuit through conjugate impedance matching. No external matching network is required, which simplifies the system design, reduces power consumption, and improves the antenna performance and integration. High-precision digital processing and precision clock synchronization technology are used to ensure that the time domain and frequency domain characteristics of multi-channel signals are fully preserved. Through FPGA real-time preprocessing and high-speed cache design, the continuity and real-time performance of data processing are guaranteed, providing high-quality input data for subsequent algorithms. The low-order model based on singular value decomposition and polynomial fitting accurately describes the interaction characteristics between the reader and the tag, significantly reduces the computational complexity, enables the system to perform real-time adaptive adjustments in complex environments, and improves the robustness and adaptability of the recognition system. The gradient-based binary search optimization method proposed in the present invention is directly solved in the original beamforming domain, realizes real-time beamforming, and greatly improves the tag direction recognition accuracy and signal-to-interference-noise ratio performance. Particle swarm iterative optimization technology is used to globally optimize system parameters, minimizing power consumption while ensuring recognition accuracy. This improves system efficiency, extends equipment life, and reduces maintenance costs. By combining spatial partitioning with an improved dynamic time-slot ALOHA algorithm, efficient recognition is achieved in densely populated environments with multiple tags. A fairness guarantee mechanism addresses tag starvation, and Doppler signature analysis enhances the tracking and recognition capabilities of mobile tags, significantly improving the system's recognition performance in complex environments.

[0037] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0038] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0039] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for preparing an RFID electronic tag, characterized in that: include: Connecting the multi-channel RFID tag identification space receiving device to the multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream; Performing singular value decomposition, low-order polynomial fitting, and gradient-based binary search optimization on the RFID tag response signal according to the multi-channel digital data stream to obtain a beamforming weight vector; Performing particle swarm iterative optimization on RFID tag recognition parameters based on the beamforming weight vector to obtain a target working parameter set; The multi-tag identification space is sectored according to the target working parameter set and RFID tag anti-collision identification is performed in combination with dynamic time slot allocation to obtain a target tag identification result, which is used for optimizing the preparation parameters of the RFID electronic tag.

2. The method for preparing an RFID electronic tag according to claim 1, wherein: The multi-channel RFID tag identification space receiving device is connected to a multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream, including: Performing meander line structure calculation on the antenna radiating element to obtain radiating element structural parameters, and applying the radiating element structural parameters to the flexible substrate Arlon CuClad 250LX for printing to obtain a unit antenna; Arranging the unit antennas at half-wavelength intervals to obtain an antenna array, and connecting each unit antenna of the antenna array to an independent radio frequency front-end circuit to obtain a multi-channel receiving path; The multi-channel receiving path is configured with a phase adjustment circuit and digital control logic to obtain an adjustable antenna array, and the adjustable antenna array is designed to be waterproof and dustproof to obtain a multi-channel RFID tag identification space receiving device; The multi-channel RFID tag identification space receiving device is connected to a multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream.

3. The method for preparing an RFID electronic tag according to claim 2, wherein: The multi-channel RFID tag identification space receiving device is connected to a multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream, including: An amplifier and a bandpass filter with a preset programmable gain range are configured for each antenna channel to obtain a signal conditioning circuit; The output signal of the signal conditioning circuit is input into a 16-bit high-precision analog-to-digital converter for 10MSPS sampling to obtain a digital signal; Performing digital down-conversion, decimation filtering, and time-domain alignment processing on the digital signal through FPGA to obtain a pre-processed data stream, and storing the pre-processed data stream in a 256MB DDR3 cache for temporary storage to obtain cached data; A clock distribution network is used to perform time synchronization processing on the cached data to obtain a synchronous data stream, and the synchronous data stream is transmitted to a digital signal processor through a high-speed interface to obtain a multi-channel digital data stream.

4. The method for preparing an RFID electronic tag according to claim 1, wherein: The step of performing singular value decomposition, low-order polynomial fitting, and gradient-based binary search optimization on the RFID tag response signal according to the multi-channel digital data stream to obtain a beamforming weight vector includes: Performing complex vector analysis on the multi-channel digital data stream to extract signal amplitude, phase, and frequency characteristics to obtain an M×N dimensional complex matrix, where M represents the number of antenna array channels and N represents the number of sampling points; Performing singular value decomposition on the M×N dimensional complex matrix to obtain a low-dimensional feature space, and calculating the fitting error and complexity ratio of the increasing order model based on the low-dimensional feature space to obtain the optimal model order k value; Constructing a polynomial basis function representation according to the optimal model order k value to obtain a polynomial model including three variables: the distance between the reader and the tag, the operating frequency, and the incident angle; The calibration data is collected under known tag positions and a least squares calculation is performed to obtain fitting coefficients of the polynomial model, and a recursive least squares calculation is performed on the fitting coefficients to obtain a low-order fitting model; The low-order fitting model is used to perform gradient-based binary search optimization on the multi-channel antenna array weights to obtain a beamforming weight vector.

5. The method for preparing an RFID electronic tag according to claim 4, wherein: The method of performing gradient-based binary search optimization on the multi-channel antenna array weights using the low-order fitting model to obtain a beamforming weight vector includes: constructing a joint optimization objective function for minimizing RFID tag orientation recognition based on the low-order fitting model; Transforming the joint optimization objective function using a binary search framework to obtain a feasibility check problem, and initializing a random weight vector based on the feasibility check problem to obtain an initial beamforming vector; Calculating the objective function gradient for the initial beamforming vector and updating it in the reverse direction of the gradient to obtain an iteratively optimized weight vector, and projecting the iteratively optimized weight vector onto a constraint set to obtain a projected weight vector; A global optimization process is performed on the projection weight vector to obtain a beamforming weight vector.

6. The method for preparing an RFID electronic tag according to claim 5, wherein: The particle swarm iterative optimization of the RFID tag recognition parameters based on the beamforming weight vector is performed to obtain a target working parameter set, including: Constructing an optimization variable set including transmit power, operating frequency, beamforming weights, and signal processing parameters according to the beamforming weight vector and the low-order fitting model; Setting a search range and initial values ​​for the optimization variable set to obtain an initial parameter state including a transmission power range and an operating frequency range; Based on the initial state of the parameters, a group of M particles is generated and random velocities are assigned to obtain an initial particle distribution, and a fitness function is performed on each particle in the initial particle distribution to obtain a score value; The particle positions and velocities are updated according to the score values ​​to obtain a new generation of particle swarms moving toward the global optimal solution and the individual optimal solution; Iterative calculation is performed on the new generation particle swarm and a global optimal solution is selected to obtain a target operating parameter set including optimal transmit power, operating frequency, beamforming weights and signal processing parameters.

7. The method for preparing an RFID electronic tag according to claim 1, wherein: The multi-tag identification space is sectorized according to the target working parameter set and RFID tag anti-collision identification is performed in combination with dynamic time slot allocation to obtain a target tag identification result, which is used for optimizing the preparation parameters of the RFID electronic tag, including: Designing a label conflict handling mechanism with an initial value of 16 time slots based on the target working parameter set, and setting an adaptive adjustment rule for the Q value of the label conflict handling mechanism to obtain a Q value dynamic adjustment parameter; Dividing the recognition space into multiple sectors using the beamforming weight vector and the Q value dynamic adjustment parameter to obtain recognition sectors, and performing Doppler characteristic analysis on the tag signal in each recognition sector to obtain trajectory prediction data; Prioritizing tags that have not been identified for multiple consecutive times based on the trajectory prediction data to obtain a fairness guarantee mechanism; All identified tag signals are subjected to feature analysis and screening to obtain target tag recognition results containing tag identity, location, and status information. The target tag recognition results are used to optimize the preparation parameters of RFID electronic tags.

8. The method for preparing an RFID electronic tag according to claim 7, wherein: The method for preparing the RFID electronic tag further includes: Performing sensitivity analysis on antenna structure parameters according to the target tag recognition result to obtain target design parameters and their sensitivity weights; Constructing a multi-objective optimization model that comprehensively considers tag size, reading distance, and cost based on the target design parameters and their sensitivity weights, and performing parameter space exploration on the multi-objective optimization model to obtain a set of candidate parameter combinations; Prepare multiple groups of sample labels based on the candidate parameter combination set and perform performance testing and verification to obtain label performance verification data; A comprehensive evaluation is conducted on the tag performance verification data and the optimal preparation parameters of the RFID electronic tag including antenna design, impedance matching and packaging structure are selected.

9. A device for preparing an RFID electronic tag, characterized in that: The method for preparing an RFID electronic tag according to any one of claims 1 to 8 is used, wherein the device for preparing the RFID electronic tag comprises: A digital processing module is used to connect the multi-channel RFID tag identification space receiving device to the multi-channel signal acquisition circuit for signal conditioning and synchronous digital processing to obtain a multi-channel digital data stream; A fitting module is used to perform singular value decomposition, low-order polynomial fitting, and gradient-based binary search optimization on the RFID tag response signal according to the multi-channel digital data stream to obtain a beamforming weight vector; an iterative optimization module, configured to perform particle swarm iterative optimization on RFID tag recognition parameters based on the beamforming weight vector to obtain a target working parameter set; The identification module is used to sectorize the multi-tag identification space according to the target working parameter set and perform RFID tag anti-collision identification in combination with dynamic time slot allocation to obtain a target tag identification result, which is used to optimize the preparation parameters of the RFID electronic tag.

Images