RFID anti-interference positioning method and system and electronic equipment

Abstract

The invention relates to the field of wireless communication and the Internet of Things, and discloses an RFID anti-interference positioning method and system and electronic equipment, and the method comprises the following steps: collecting and standardizing signal features through multiple reader-writers, and constructing a multi-dimensional feature structure; identifying an interference signal by using a machine learning model, and generating a frequency / power tuning parameter through a game strategy; when interference is detected, dynamically screening a base station subset, calculating a label position in combination with multi-base-station cooperative positioning and a triangulation algorithm, and synchronously generating a positioning error and a channel evaluation value; model parameters and tuning weights are updated in a closed-loop mode based on feedback indexes, and dynamically-optimized anti-interference positioning is achieved. According to the invention, interference signals are effectively suppressed, the positioning precision of a complex environment is improved, signal conflicts in a dense label environment are optimized, dynamic tuning and closed-loop optimization of parameters are realized, the adaptive capacity of the system is improved, co-frequency interference is reduced through cooperative gaming of multiple reader-writers, and long-term operation stability is guaranteed.

Description

Technical Field

[0001] The present invention relates to the fields of wireless communications and the Internet of Things, and in particular to an RFID anti-interference positioning method, system, and electronic equipment. Background Art

[0002] Radio frequency identification (RFID) technology uses wireless radio frequency signals to achieve contactless target recognition and tracking, playing a vital role in logistics warehousing, smart manufacturing, and smart retail. However, as application scenarios become more complex, the presence of multiple interference signals in the electromagnetic environment leads to a decrease in tag reading success rates. This is particularly true in industrial plants and densely packed shelves, where signal attenuation, multipath effects, and co-channel interference are prominent issues, severely limiting the positioning accuracy and reliability of RFID technology.

[0003] Existing RFID positioning methods generally rely on basic physical quantities such as signal strength (RSSI), time of arrival (ToA), or frequency characteristics to infer positioning. To cope with complex electromagnetic environments and interfering signals, traditional technologies generally adopt two approaches: one is to enhance signal transmission by increasing transmit power, changing antenna gain, or deploying relay equipment; the other is to introduce fixed frequency band planning or static frequency modulation schemes during the system design phase to reduce the probability of co-channel interference. In addition, some methods attempt to combine filtering algorithms or heuristic strategies to eliminate outliers and improve positioning accuracy. However, most of these methods rely on static configurations, making it difficult to respond to environmental changes in real time and fundamentally eliminate the impact of strong external interference signals on the RFID reader's recognition ability.

[0004] In practical applications, RFID positioning environments are often highly dynamic, with densely distributed tags, frequent electromagnetic interference, and significant multipath signal propagation. This leads to frequent signal overlap, conflicting interference, and missed tag reads during positioning. Existing technologies lack closed-loop feedback mechanisms for identifying interference signals and dynamically adapting to optimization, making it difficult to automatically adjust identification strategies and tuning parameters based on real-time positioning errors and channel state information, achieving continuous robust optimization. This is especially true in scenarios where multiple readers operate collaboratively, as devices lack effective coordination mechanisms, leading to frequent policy conflicts and further exacerbating channel resource waste and performance degradation.

[0005] Therefore, the present invention proposes an RFID anti-interference positioning method, system and electronic device to address the deficiencies of the prior art. Summary of the Invention

[0006] The purpose of the present invention is to provide an RFID anti-interference positioning method, system and electronic equipment to solve the problems of difficulty in effectively identifying interference signals, lack of adaptive tuning mechanism, and insufficient coordination ability in multi-reader scenarios.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: an RFID anti-interference positioning method, comprising the following steps: Multiple RFID readers are used to collect feature data including at least signal strength, arrival time, frequency offset, channel response, and device identification in real time and perform standardized processing; Construct a multi-dimensional expression structure that integrates spatiotemporal and device features based on the standardized feature data; Establishing an interference signal recognition model based on machine learning, using the multi-dimensional expression structure to extract frequency and device features, identifying interference signals and outputting recognition results; Based on the recognition results, the tuning parameters including the transmission frequency and power are generated through a game optimization strategy, and the RFID reader configuration is adaptively adjusted; When interference is detected, a set of undisturbed base stations is dynamically selected based on signal quality, and the tag position is determined by combining multi-base station collaborative positioning and triangulation algorithm. The positioning error and channel status are calculated to generate positioning error and channel status information. Based on the positioning error and channel status information, the interference signal identification model and tuning parameters are updated.

[0008] Preferably, the step of collecting characteristic data including at least signal strength, arrival time, frequency offset, channel response and device identification in real time through multiple RFID readers and writers and performing standardization processing includes: Utilize multiple distributed RFID readers to collect return signals from RFID tags and extract feature data including at least signal strength RSS, arrival time ToA, frequency offset Δf, channel response CSI, and device identification ID to form a multi-dimensional raw data set; For each eigenvector x i ∈{RSS,ToA,Δf,CSI,ID} is normalized using the following formula: Among them, x i is the original value of the i-th feature; μ i is the mean value of this type of feature in historical samples; σ i is the standard deviation of this type of feature in historical samples; x i ' is the standard value after normalization.

[0009] Preferably, the step of constructing a multi-dimensional expression structure integrating spatiotemporal and device features based on the standardized feature data includes: Spatiotemporal feature extraction: Group the standardized data by time window T and reader spatial coordinates (x, y) to generate a spatiotemporal feature matrix Among them, T is the preset time window; S is the number of readers; D is the single-device single-time feature dimension; Device feature encoding: Map the device ID to a device feature vector through the embedding layer Where E is the embedding dimension, forming the device feature matrix Multi-dimensional structure fusion: The spatiotemporal feature matrix and the device feature matrix are tensor-joined according to the reader-writer dimension to generate a fused multi-dimensional expression structure. Its elements are calculated as: in, Represents vector concatenation operation; Represents the fused feature vector under time window t and reader s; M d [s,:] represents the device feature matrix M d The device feature vector corresponding to the s-th reader / writer in .

[0010] Preferably, the steps of establishing an interference signal recognition model based on machine learning, extracting frequency and device features using the multidimensional expression structure, identifying interference signals and outputting recognition results include: The multidimensional expression structure Input to the interference signal recognition model based on deep neural network, and extract in parallel through the feature separation layer: Among them, * represents the convolution operation; W f ,W d is the trainable weight matrix; b f ,b d is the bias term; The fusion feature F = [F f ||F d ] Input the classifier and calculate the probability of interference signal: Among them, || represents feature splicing; σ is the Sigmoid function; is the trainable weight vector; When p ≥ θ, it is determined to be an interference signal and the binary recognition result y∈{0,1} is output, where θ∈[0.5,1) is the preset judgment threshold.

[0011] Preferably, the step of generating tuning parameters including transmission frequency and power through a game optimization strategy based on the recognition result and adaptively adjusting the RFID reader configuration includes: When the recognition result y=1, initialize the reader set The tuning parameter candidate set {(f s ,P s )}; Among them, f s ∈[f min ,f max ] is the adjustable frequency range; P s ∈[P min ,P max ] is the adjustable power range; Constructing a non-cooperative game model Define the utility function: U s (f s ,P s )=αlog(1+SINR s )-βP s -γ∑ i≠s |f s -f j |; Among them, αlog(1+SINR s ) is the communication quality item, Reflects the signal transmission quality requirements; βP s is the energy consumption cost item; γ∑|f s -f j | is the spectrum conflict term; α, β, γ are weight coefficients, and α+β+γ=1; Update the strategy by iteration: in, Indicates the reader r in the nth iteration s strategy; Indicates division by r s The strategy combination of all other readers in round n except U s (f s ,P s |·) indicates that under the condition that other node strategies are known, the reader r s The conditional utility function of By maximizing its own utility U in each round s , each reader continuously updates its transmission frequency and power, converging to the Nash equilibrium point: The obtained parameter pair That is the optimized tuning parameter; The tuning parameters are sent to each RFID reader through the control channel to perform configuration updates.

[0012] Preferably, when interference is detected, dynamically selecting a set of undisturbed base stations based on signal quality, combining multi-base station collaborative positioning with a triangulation algorithm to determine the tag position, and calculating the positioning error and evaluating the channel state, the steps of generating positioning error and evaluating channel state information include: When the interference identification result is y=1, from the set of all base stations In the process, based on the signal quality, the undisturbed base stations that meet the signal-to-noise ratio threshold condition are dynamically selected to form a subset. Specific satisfaction: Among them, P t Indicates the tag transmission power; G m Indicates tag to base station b m The channel gain; N0 represents the ambient noise power; I m Indicates base station b m The received interference power; δ represents the signal-to-noise ratio decision threshold; Base station collection Each base station b m , using the received signal strength or arrival time difference to estimate the distance value d between it and the tag m , combined with the known location of the base station (x m ,y m ), establish a two-dimensional geometric positioning equation group: Among them, (x t ,y t ) is the label coordinate to be solved; d m From tag to base station b m The estimated distance of Solve the above equations through triangulation algorithm to get the estimated coordinates of the label Compare the estimated position with the true reference position of the label Compare and calculate the positioning error: And the non-interfered base station set The received signal quality of each base station is evaluated and a channel state quantization model is constructed: Q m =ω1·SNR m +ω2·(1-PLR m )-ω3·I m ; Among them, PLR m Base station b m The data packet loss rate; ω1, ω2, ω3 are channel assessment weights, and they satisfy ω1+ω2+ω3=1; The final output positioning error Eloc and estimated channel state information

[0013] Preferably, the step of updating the interference signal identification model and tuning parameters according to the positioning error and the evaluated channel state information includes: Positioning error E loc and estimated channel state information Construct a system performance feedback index Ψ, which is used to jointly reflect the positioning accuracy and channel quality level and is defined as: Among them, E max is the preset maximum allowable positioning error; Q m Indicates base station b m The channel quality evaluation value; λ1,λ2∈[0,1] are weighting coefficients, and λ1+λ2=1; The performance feedback index Ψ is used to update the feature threshold vector θ=[θ1,θ2,...,θ K ], where each parameter θ k Corresponding to a type of feature discrimination condition in the interference signal recognition model, it is updated through the following iterative formula: in, is the value of the kth parameter in the nth round; η is the update step size of the interference signal recognition model; Ψ ref is the benchmark value for ideal system performance; At the same time, the indicator Ψ is introduced into the game-based tuning parameter optimization process to dynamically adjust the weighted parameter set {α, β, γ} in the utility function. The specific adjustment method is: α=α0+κ1·Ψ, β=β0-κ2·Ψ, γ=1-α-β; Among them, α0, β0 are the initially set weight values; κ1, κ2 are weight adjustment coefficients; α, β, γ∈(0,1) and satisfy α+β+γ=1, corresponding to the proportion of transmission frequency, power and interference avoidance strategy in the utility function respectively; The updated weights are introduced into the tuning parameters, and for each reader r s The strategy in round n (f s ,P s ) to find the optimal response: Among them, f s 、P s Respectively represent the transmission frequency and power of the reader; is the strategy of other readers in round n; U s(·) is the comprehensive utility function.

[0014] The present invention also provides an RFID anti-interference positioning system, comprising: A data acquisition module is used to collect characteristic data including at least signal strength, arrival time, frequency offset, channel response and device identification in real time through multiple RFID readers; A standardization processing module, used for performing standardization processing on the feature data; The feature fusion module is used to construct a multi-dimensional expression structure that integrates time, space and device features based on the standardized feature data; the interference identification module establishes an interference signal identification model based on machine learning, uses the multi-dimensional expression structure to extract frequency and device features, identifies interference signals and outputs the identification results; The parameter tuning module is used to generate tuning parameters including transmission frequency and power through game optimization strategy and adaptively adjust the RFID reader configuration; The positioning optimization and evaluation module is used to dynamically select a set of undisturbed base stations based on signal quality when interference is detected, determine the tag position by combining multi-base station collaborative positioning and triangulation algorithms, calculate the positioning error and evaluate the channel state, and generate positioning error and evaluation channel state information; The model updating module is used to update the interference signal identification model and tuning parameters according to the positioning error and the evaluated channel state information.

[0015] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method when executing the computer program.

[0016] The present invention also provides a storage medium storing a computer program, wherein the computer program implements the above method when executed by a processor.

[0017] In summary, the present invention includes at least one of the following beneficial technical effects: 1. By integrating multidimensional spatiotemporal features with device characteristics and combining them with a deep neural network's joint interference signal discrimination mechanism, this design effectively distinguishes between environmental noise and true tag signals. This design leverages the inherent correlation of multidimensional data to enhance interference identification accuracy, ensuring stable operation of RFID anti-interference positioning in strong electromagnetic interference scenarios.

[0018] 2. A parameter tuning strategy based on a non-cooperative game model enables each reader to dynamically adjust its transmission frequency and power, minimizing co-channel interference while ensuring communication quality. This mechanism achieves efficient allocation of spectrum resources through distributed optimization, significantly reducing the probability of signal overlap in dense deployment scenarios.

[0019] 3. A signal-to-noise ratio-driven dynamic base station selection mechanism, combined with a multi-base station collaborative positioning algorithm, rapidly builds a highly reliable positioning network when interference occurs. This approach combines real-time channel assessment with geometric positioning to enhance the robustness of RFID tag position estimation in complex environments.

[0020] 4. Based on positioning error and channel quality feedback, the interference identification model threshold and game tuning weight parameters are synchronously updated, enabling the RFID anti-interference positioning method to continuously track environmental changes and self-optimize, ensuring stability and response speed in long-term operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the method flow of the present invention; Figure 2 Schematic diagram of the system architecture of the present invention; Figure 3 It is a schematic diagram of the structure of the electronic device of the present invention. DETAILED DESCRIPTION

[0022] The following is combined with Figure 1 -Attached Figure 3 , the present invention is described in further detail.

[0023] An embodiment of the present invention provides an RFID anti-interference positioning method, comprising the following steps: S1. Collecting feature data including at least signal strength, arrival time, frequency offset, channel response, and device identification in real time through multiple RFID readers and writers, and performing standardization processing; S2. Construct a multi-dimensional expression structure integrating spatiotemporal and device features based on the standardized feature data; S3. Establish an interference signal recognition model based on machine learning, use the multi-dimensional expression structure to extract frequency and device characteristics, identify interference signals and output recognition results; S4. Based on the recognition results, a game optimization strategy is used to generate tuning parameters including transmission frequency and power, and the RFID reader configuration is adaptively adjusted. S5. When interference is detected, a set of undisturbed base stations is dynamically selected based on signal quality, and the tag position is determined by combining multi-base station collaborative positioning and triangulation algorithms. The positioning error and estimated channel state are calculated to generate positioning error and estimated channel state information. S6. Update the interference signal identification model and tuning parameters according to the positioning error and the estimated channel state information.

[0024] For step S1, in this embodiment, the specific implementation is as follows: A spatial sensing network is constructed using multiple RFID readers deployed in a distributed manner. Each reader is preferably located at a different spatial coordinate within the positioning area. Multi-node collaboration enables full coverage of signal characteristics. The readers preferably support the ultra-high frequency (UHF) communication protocol, operating in the 860MHz to 960MHz frequency band, and are capable of collecting multi-dimensional signal parameters.

[0025] After receiving the return signal from the RFID tag, each reader / writer synchronously extracts at least the following five types of raw feature data: Signal strength (RSS): Measured by the received signal strength indicator (RSSI) circuit, it reflects the path loss between the tag and the reader, and is measured in dBm. Time of Arrival (ToA): Based on the reader's internal high-precision clock, it calculates the propagation time of the signal from the tag to the reader in nanoseconds (ns). It uses a timestamp differential algorithm to eliminate clock deviations between devices. Frequency offset (Δf): The instantaneous offset of the carrier frequency is detected by a phase-locked loop (PLL). It is measured in Hertz (Hz) and is used to identify Doppler effects or ambient frequency interference. Channel Response Information (CSI): This analyzes channel state information using orthogonal frequency division multiplexing (OFDM) technology to obtain multi-subcarrier response data containing amplitude and phase characteristics, characterizing the signal propagation path characteristics. Device Identification (ID): A unique identifier in the decoded tag's EPC code, used to distinguish different tags and associated device attributes.

[0026] The above feature data constitutes the original multidimensional data set, whose dimension is N×5 (N is the number of sampling points), and each data point corresponds to a complete feature vector x collected by a reader at a specific moment. i ∈{RSS,ToA,Δf,CSI,ID}.

[0027] To eliminate the impact of different feature dimension differences on subsequent model training, Z-score normalization is performed on each feature dimension: Among them, x i ′ is the standard value after normalization; x i is the original value of the i-th feature; μ i is the mean value of this type of feature in historical samples; σ i is the standard deviation of this type of feature in historical samples. i and σ i All are obtained through offline statistical calculations. In specific implementation: For continuous features (RSS, ToA, Δf, CSI): calculate the mean and standard deviation of each feature dimension to ensure that the normalized data conforms to the zero mean and unit variance distribution; For discrete features (ID): use One-Hot Encoding or Embedding to convert them into numerical vectors, and then perform normalization; The historical data preferably includes sampling data under a variety of typical environmental scenarios (such as no interference, multipath interference, and co-channel interference) to ensure the generalization of statistical parameters.

[0028] The reader array coordinates its operations through a multiplexing mechanism to avoid mutual interference. The feature collection frequency is set based on real-time requirements. Raw data is transmitted to the central processing unit via a communication link and time-synchronized. During normalization, if the variance of a feature dimension approaches a minimum threshold, a feature filtering mechanism is triggered to eliminate redundant dimensions. The normalized feature vector serves as input for subsequent multidimensional feature fusion.

[0029] Distributed multi-reader collaborative data collection overcomes single-node signal coverage blind spots and improves data spatial resolution. Multi-dimensional feature extraction comprehensively reflects the physical characteristics of signal propagation, environmental interference characteristics, and device association characteristics, providing a comprehensive information foundation for interference identification.

[0030] For step S2, in this embodiment, the specific implementation is as follows: First, for the standardized feature data, the data is grouped by time window T and reader spatial coordinates (x, y) through spatiotemporal feature extraction. The time window T is used to define the time interval for data collection and is usually set according to actual application requirements. Through this grouping method, the data can be structured by time and space to generate a spatiotemporal feature matrix. Among them, T represents the size of the time window; S represents the number of readers participating in data collection; D represents the feature dimension of a single device at a single moment. The feature matrix M t,s Organized as a three-dimensional tensor, its first dimension is the time window index, the second dimension is the reader number, and the third dimension is the feature dimension of the data read by the device at that point in time. In practical applications, this matrix can be constructed by sliding windows on the original normalized data, ensuring that both temporal and spatial features are clearly expressed in the matrix, facilitating subsequent deep modeling.

[0031] Next, in order to further introduce the identity semantic information of the device and enhance the model's ability to identify different reader characteristics, this embodiment adopts a device feature encoding mechanism to map the device identification (i.e., ID) of each reader into a fixed-dimensional vector representation. Specifically, the discrete device ID is encoded into a continuous low-dimensional vector through the embedding layer (embedding#layer), and the device feature vector is obtained. Where E is the embedding dimension. The embedding vectors of all readers and writers constitute the device feature matrix The Sth row in this matrix corresponds to the embedding representation of the Sth reader / writer.

[0032] This construction allows the semantic information of each device to be independently expressed in vector space and modeled uniformly with spatiotemporal features. Furthermore, using an embedded encoding method to encode device identifiers offers greater expressive efficiency and semantic compactness than traditional one-hot encoding.

[0033] In order to form the final fusion structure for subsequent anti-interference modeling, in this embodiment, the tensor splicing method is used to combine the above-mentioned spatiotemporal feature matrix M t,s and the device feature matrix M d Specifically, in each time window t and each reader s, the feature vector M corresponding to the time and space point t,s [s,:] and the vector representation M of the sth reader / writer in the device feature matrix d [s,:] performs splicing operations on the feature dimension to generate a fused multi-dimensional expression structure: in, Represents vector concatenation operation; Represents the fused feature vector under the time window t and the reader s; finally, the fused multi-dimensional expression structure is generated This multidimensional expression structure simultaneously includes the original environmental characteristics of each time window, device signal characteristics, and identity semantic information of each reader / writer, making it easier for the model to jointly model and distinguish interference patterns and positioning signals in subsequent processing.

[0034] Through the above design, it is possible to integrate multi-source information into a unified data structure, improve the system's ability to express signal interference patterns in complex environments, and provide a structured and semantically rich input basis for subsequent anti-interference positioning models.

[0035] Regarding step S3, in this embodiment, the specific implementation is as follows: First, the multi-dimensional expression structure that integrates time, space and device characteristics is constructed The input is fed into a pre-built interference signal recognition model based on a deep neural network. The interference signal recognition model preferably includes multiple parallel processing paths for extracting signal features and device semantic features. The feature separation mechanism extracts frequency-related features and device-related features separately, thereby enhancing the model's ability to identify different interference types.

[0036] In the specific implementation, a set of trainable convolution kernels are used to align the input tensor Perform convolution operation to extract frequency features. The convolution operation captures the pattern of frequency change and spatial distribution in the local time window by jointly sliding calculation with the time dimension and feature dimension in the tensor, thereby generating a frequency feature tensor F f The process is expressed as follows: Among them, * represents the convolution operation; W f is the trainable convolution kernel weight matrix of the frequency channel; b f is the corresponding bias term; ReLU is the linear rectification activation function, which is used to introduce nonlinear expression capabilities. This partial convolution operation can effectively extract the time-correlated interference signal characteristics from the fused tensor, reflecting the perturbation pattern of the signal in the time domain.

[0037] In parallel, the model also has a device feature extraction channel that uses fully connected transformations to extract feature representations related to device semantics from the fusion structure, enhancing the model's ability to learn about the interference tendencies of specific devices or readers. This part is calculated using the following expression: Among them, W d is the fully connected weight matrix of the device feature path; b d is the corresponding bias term. This process converts the three-dimensional tensor Flatten along the spatial dimension and feature dimension, and generate a structured device feature representation F through multi-layer mapping d , thereby exploring the response differences of different readers in interference environments.

[0038] Then, the above frequency characteristics F f With device feature F d Splice to form a fusion feature representation F, that is: F = [F f ||F d ]; Here, || represents the concatenation operation along the feature dimension. This fused feature not only contains local time-frequency pattern information but also preserves the semantic characteristics of the device embedding representation, improving the model's comprehensive ability to discriminate against interference signals.

[0039] Finally, the fusion feature F is input into the subsequent binary classification, and the Sigmoid activation function is used to calculate the probability value p that the current input is an interference signal. The calculation formula is as follows: Where σ is the Sigmoid function, which is used to map the output to the probability value interval [0,1]; is a trainable weight vector; b is a bias term. The probability value p represents the model’s confidence in whether the current input has interference.

[0040] To output the final binary recognition result, the model compares the probability value with the preset threshold θ. If p≥θ is satisfied, the output is an interference signal (y=1); otherwise, the output is a non-interference signal (y=0). The decision process is formally expressed as: y∈{0,1}, when p≥θ; Among them, θ∈[0.5,1) is a preset classification threshold, which is preferably set according to the system application environment and is used to control the sensitivity of the interference signal recognition model to interference signal judgment.

[0041] The overall design of the above-mentioned interference signal recognition model ensures the ability of structured expression and high-order feature extraction, while fully modeling the fusion features through the deep network structure, realizing adaptive recognition of complex interference patterns.

[0042] Regarding step S4, in this embodiment, the specific implementation is as follows: When the recognition result output by the interference recognition model is y=1, that is, it is determined that there is an interference signal in the current environment, the system activates the interference response mechanism and enters the parameter adaptive optimization stage. The core goal of this stage is to generate a set of optimal tuning parameter pairs (f s ,P s ) to simultaneously optimize communication quality, power consumption, and spectrum conflicts, and improve the interference robustness of the overall system.

[0043] First, initialize the current reader-writer set to Each reader r S The operating frequency and transmission power can be adjusted within a certain range. s Construct the tuning parameter candidate set {(f s ,P s )}; Wherein, the frequency f s ∈[f min ,f max ] indicates the adjustable working frequency range of the reader, power P s ∈[P min ,P max ] indicates its adjustable transmission power range.

[0044] During the parameter optimization process, a non-cooperative game model is constructed in, Represents the game participants, that is, all RFID readers in the system; Indicates reader / writer s The optional strategy set, corresponding to its optional (fs ,P s )Yes; U s Indicates reader / writer s The utility function under a given strategy is used to evaluate the comprehensive benefits of a certain parameter combination on its communication performance and resource utilization.

[0045] The utility function is defined as follows: U s (f s ,P s )=αlog(1+SINR s )-βP s -γ∑ i≠s |f s -f j |; Among them, αlog(1+SINR s ) is the communication quality item, which is used to measure the signal transmission quality of the reader under the current parameter configuration. Reflects the signal transmission quality requirements; among them, G s represents the channel gain; N0 is the background noise power; ∑ k≠s P k |h k,s | 2 Indicates that other readers are s interference effects.

[0046] βP s is the energy cost item, which penalizes the reader's transmission power to avoid ineffective high-power transmission; γ∑|f s -f j | is the spectrum conflict term, which encourages different readers to select non-overlapping frequencies to reduce co-channel interference; α, β, and γ are weight coefficients, and α+β+γ=1 is satisfied, which is used to balance the weight distribution between different targets.

[0047] During the optimization process, each reader updates its strategy independently based on its local information and the current system state. Specifically, it uses an iterative game optimization method to update its tuning parameter pair (f s ,P s ), so that the utility function reaches the optimal value: The above formula indicates that in the nth round of iteration, each reader r s Based on the current parameter combination of other readers Find the current optimal (f s ,P s ), to maximize its conditional utility function U s .

[0048] in, Indicates the reader r in the nth iteration s strategy; Indicates division by r s The strategy combination of all other readers in round n except U s (f s ,P s |·) indicates that under the condition that other node strategies are known, the reader r s The conditional utility function of Through the above-mentioned strategy update mechanism, each reader / writer in each round of the game adaptively adjusts its parameter combination according to the current system state and searches for a stable optimal response strategy in continuous iteration. Ultimately, the system converges to the Nash equilibrium point under the conditions met, which is expressed as: This is the tuning parameter pair after game optimization, which is used to guide subsequent configuration update operations.

[0049] After the optimization is completed, the system will set the final tuning parameters of each reader. The parameter-level configuration update is completed by sending it to the corresponding device through the control channel. This configuration process uses the control signaling mechanism supported by the standard RFID communication protocol to achieve remote parameter loading and immediate effect.

[0050] Through the above-mentioned parameter tuning mechanism based on non-cooperative game, the reader can achieve adaptive frequency and power adjustment after identifying interference. It does not rely on a central coordinator, has good distributed deployment capabilities, and is suitable for RFID system applications in complex dynamic interference environments.

[0051] Regarding step S5, in this embodiment, the specific implementation is as follows: When the output result is y=1, that is, it is confirmed that there is an interference signal in the current environment, the system starts the anti-interference positioning mechanism. Filter out the subset of base stations that are not interfered with The screening process is based on the signal-to-noise ratio (SNR) of the received signal of each base station. The specific screening conditions are: Among them, P t Indicates the tag transmission power; G m Indicates tag to base station b m The channel gain; N0 represents the ambient noise power; I m Indicates base station b m The received interference power; δ represents the signal-to-noise ratio decision threshold. This threshold is preferably set based on the system deployment environment and positioning accuracy requirements to ensure that the signal quality of the selected base station meets the positioning calculation requirements.

[0052] For the filtered base station subset The system uses the received signal strength (RSS) or time difference of arrival (TDoA) information of each base station to estimate the distance between the tag and each base station b m The distance value d between m . Combined with the known location coordinates of each base station (x m ,y m ), construct the two-dimensional plane positioning equations: Among them, (x t ,y t ) is the label coordinate to be solved; d m From tag to base station b m When the number of base stations When , the above nonlinear equations are solved by triangulation positioning algorithm. Specifically, the least squares method is used to transform the equations into linear form, and the estimated value of the tag position is obtained through matrix operation. This algorithm can effectively handle ranging errors and improve positioning robustness in interference environments.

[0053] To quantify the accuracy of the positioning results, the system will estimate the coordinates The actual reference position of the label For comparison, calculate the positioning error in the form of Euclidean distance: Among them, the reference position It can be obtained through a high-precision positioning system or preset calibration points and is used to evaluate the degree of deviation of the current positioning result.

[0054] At the same time, the base station set participating in positioning The channel status of each node in the network is comprehensively evaluated and a channel status quantization model is constructed to reflect the signal reception quality. Calculate its channel state score Q m :Q m =ω1·SNR m +ω2·(1-PLR m )-ω3·I m ; Among them, PLR m Base station b m The data packet loss rate reflects the reliability of the communication link; ω1, ω2, ω3 are channel evaluation weights and coefficients, corresponding to the evaluation weights of the signal-to-noise ratio, transmission reliability, and interference suppression capability, respectively, and satisfy ω1+ω2+ω3=1.

[0055] Finally, the system outputs the positioning error E loc and channel state assessment information set Serves as the basis for subsequent model updates and parameter tuning.

[0056] Regarding step S6, in this embodiment, the specific implementation is as follows: First, based on the positioning error E loc and channel state assessment information set Construct a system performance feedback index Ψ to comprehensively reflect the positioning accuracy and channel availability in the current environment. This index is defined as follows through a weighted fusion method: Among them, E max is the preset maximum allowable positioning error; Q m Indicates base station b m The channel quality evaluation value reflects the signal reception quality; λ1,λ2∈[0,1] are weighting coefficients, and λ1+λ2=1, which respectively control the contribution ratio of positioning accuracy and channel quality in the comprehensive index.

[0057] The value range of this indicator Ψ is [0,1]. The closer its value is to 1, the better the overall performance of the system is, providing a quantitative basis for subsequent model and parameter updates.

[0058] For the interference signal recognition model, it contains a set of feature discrimination threshold parameters θ=[θ1,θ2,...,θ K ], each θ k The threshold value of a certain interference judgment condition in the corresponding model. According to the performance feedback indicator Ψ, the threshold parameter is dynamically adjusted using the following iterative formula: in, is the value of the kth parameter in the nth round; η is the update step size of the interference signal recognition model; Ψ ref This is the baseline value for ideal system performance.

[0059] Through the above update mechanism, when the actual performance Ψ is lower than the benchmark Ψ ref When k , enhance the model's sensitivity to potential interference signals; conversely, increase the threshold to reduce misjudgment.

[0060] At the same time, the performance feedback index Ψ is introduced into the tuning parameters to dynamically adjust the weight parameter set {α, β, γ} of the utility function in the non-cooperative game model. The specific adjustment formula is: α=α0+κ1·Ψ, β=β0-κ2·Ψ, γ=1-α-β; Among them, α0, β0 are the initially set weight values; κ1, κ2 are weight adjustment coefficients; α, β, γ∈(0,1) and satisfy α+β+γ=1, corresponding to the proportion of transmission frequency, power and interference avoidance strategy in the utility function respectively.

[0061] After the weights are adjusted, each reader will recalculate the optimal strategy based on the updated utility function during the game optimization process. s , its strategy update formula is: Among them, the utility function U s The weight parameters α, β, and γ in (·) have been adjusted according to the current Ψ value. Through iterative optimization, the system eventually converges to a new Nash equilibrium point and generates a tuning parameter combination that adapts to the latest environmental state.

[0062] Among them, f s 、P s Respectively represent the transmission frequency and power of the reader; is the strategy of other readers in round n; U s (·) is the comprehensive utility function.

[0063] The above-mentioned update mechanism forms a closed-loop feedback control, which enables the interference signal recognition model to adaptively optimize the discrimination ability according to the actual positioning performance and channel status. At the same time, the tuning parameter optimization can dynamically adjust the strategy weight.

[0064] The present invention also provides an RFID anti-interference positioning system, comprising: A data acquisition module is used to collect characteristic data including at least signal strength, arrival time, frequency offset, channel response and device identification in real time through multiple RFID readers; A standardization processing module, used for performing standardization processing on the feature data; The feature fusion module is used to construct a multi-dimensional expression structure that integrates time, space and device features based on the standardized feature data; the interference identification module establishes an interference signal identification model based on machine learning, uses the multi-dimensional expression structure to extract frequency and device features, identifies interference signals and outputs the identification results; The parameter tuning module is used to generate tuning parameters including transmission frequency and power through game optimization strategy and adaptively adjust the RFID reader configuration; The positioning optimization and evaluation module is used to dynamically select a set of undisturbed base stations based on signal quality when interference is detected, determine the tag position by combining multi-base station collaborative positioning and triangulation algorithms, calculate the positioning error and evaluate the channel state, and generate positioning error and evaluation channel state information; The model updating module is used to update the interference signal identification model and tuning parameters according to the positioning error and the evaluated channel state information.

[0065] The system of this embodiment can be used to execute the above method embodiments, and its principles and technical effects are similar, so they will not be repeated here.

[0066] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above method is implemented.

[0067] The present invention also provides a storage medium on which a computer program is stored. When the computer program is run by a processor, the above method is executed.

[0068] Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0069] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. RFID anti-interference positioning method, characterized in that: The following steps are involved: Multiple RFID readers are used to collect feature data including at least signal strength, arrival time, frequency offset, channel response, and device identification in real time and perform standardized processing; Construct a multi-dimensional expression structure that integrates spatiotemporal and device features based on the standardized feature data; Establishing an interference signal recognition model based on machine learning, using the multi-dimensional expression structure to extract frequency and device features, identifying interference signals and outputting recognition results; Based on the recognition results, the tuning parameters including the transmission frequency and power are generated through a game optimization strategy, and the RFID reader configuration is adaptively adjusted; When interference is detected, a set of undisturbed base stations is dynamically selected based on signal quality. The tag position is determined by combining multi-base station collaborative positioning with a triangulation algorithm. The positioning error and channel state are then calculated to generate positioning error and channel state evaluation information. The interference signal identification model and tuning parameters are updated according to the positioning error and the estimated channel state information.

2. The RFID anti-interference positioning method according to claim 1, characterized in that: The step of collecting characteristic data including at least signal strength, arrival time, frequency offset, channel response and device identification in real time through multiple RFID readers and writers and performing standardization processing includes: Utilize multiple distributed RFID readers to collect return signals from RFID tags and extract feature data including at least signal strength RSS, arrival time ToA, frequency offset Δf, channel response CSI, and device identification ID to form a multi-dimensional raw data set; For each eigenvector x i ∈{RSS,ToA,Δf,CSI,ID} is normalized using the following formula: Among them, x i is the original value of the i-th feature; μ i is the mean value of this type of feature in historical samples; σ i is the standard deviation of this type of feature in historical samples; x i ' is the standard value after normalization.

3. The RFID anti-interference positioning method according to claim 1, characterized in that: The step of constructing a multi-dimensional expression structure integrating spatiotemporal and device features based on the standardized feature data includes: Spatiotemporal feature extraction: Group the standardized data by time window T and reader spatial coordinates (x, y) to generate a spatiotemporal feature matrix Among them, T is the preset time window; S is the number of readers; D is the single-device single-time feature dimension; Device feature encoding: Map the device ID to a device feature vector through the embedding layer Where E is the embedding dimension, forming the device feature matrix Multi-dimensional structure fusion: The spatiotemporal feature matrix and the device feature matrix are tensor-joined according to the reader-writer dimension to generate a fused multi-dimensional expression structure. Its elements are calculated as: Among them, ⊕ represents the vector splicing operation; Represents the fused feature vector under time window t and reader s; M d [s,:] represents the device feature matrix M d The device feature vector corresponding to the s-th reader / writer in .

4. The RFID anti-interference positioning method according to claim 1, characterized in that: The steps of establishing an interference signal recognition model based on machine learning, extracting frequency and device features using the multi-dimensional expression structure, identifying interference signals and outputting recognition results include: The multidimensional expression structure Input to the interference signal recognition model based on deep neural network, and extract in parallel through the feature separation layer: Among them, * represents the convolution operation; W f ,W d is the trainable weight matrix; b f ,b d is the bias term; The fusion feature F = [F f ||F d ] Input the classifier and calculate the probability of interference signal: Among them, || represents feature splicing; σ is the Sigmoid function; is the trainable weight vector; When p ≥ θ, it is determined to be an interference signal and the binary recognition result y∈{0,1} is output, where θ∈[0.5,1) is the preset judgment threshold.

5. The RFID anti-interference positioning method according to claim 1, characterized in that: The steps of generating tuning parameters including transmission frequency and power through a game optimization strategy based on the recognition result and adaptively adjusting the RFID reader configuration include: When the recognition result y=1, initialize the reader set The tuning parameter candidate set {(f s ,P s )}; Among them, f s ∈[f min ,f max ] is the adjustable frequency range; P s ∈[P min ,P max ] is the adjustable power range; Constructing a non-cooperative game model Define the utility function: U s (f s ,P s )=αlog(1+SINR s )-βP s -γ∑ i≠s |f s -f j |; Among them, αlog(1+SINR s ) is the communication quality item, Reflects the signal transmission quality requirements; βP s is the energy consumption cost item; γ∑|f s -f j | is the spectrum conflict term; α, β, γ are weight coefficients, and α+β+γ=1; Update the strategy by iteration: in, Indicates the reader r in the nth iteration s strategy; Indicates division by r s The strategy combination of all other readers in round n except U s (f s ,P s |·) indicates that under the condition that other node strategies are known, the reader r s The conditional utility function of By maximizing its own utility U in each round s , each reader continuously updates its transmission frequency and power, converging to the Nash equilibrium point: The obtained parameter pair That is the optimized tuning parameter; The tuning parameters are sent to each RFID reader through the control channel to perform configuration updates.

6. The RFID anti-interference positioning method according to claim 1, characterized in that: When interference is detected, dynamically selecting a set of undisturbed base stations based on signal quality, combining multi-base station collaborative positioning with a triangulation algorithm to determine the tag position, and calculating the positioning error and evaluating the channel state. The steps of generating positioning error and evaluating channel state information include: When the interference identification result is y=1, from the set of all base stations In the process, based on the signal quality, the undisturbed base stations that meet the signal-to-noise ratio threshold condition are dynamically selected to form a subset. Specific satisfaction: Among them, P t Indicates the tag transmission power; G m Indicates tag to base station b m The channel gain; N0 represents the ambient noise power; I m Indicates base station b m The received interference power; δ represents the signal-to-noise ratio decision threshold; Base station collection Each base station b m , using the received signal strength or arrival time difference to estimate the distance value d between it and the tag m , combined with the known location of the base station (x m ,y m ), establish a two-dimensional geometric positioning equation group: Among them, (x t ,y t ) is the label coordinate to be solved; d m From tag to base station b m The estimated distance of Solve the above equations through triangulation algorithm to get the estimated coordinates of the label Compare the estimated position with the true reference position of the label Compare and calculate the positioning error: And the non-interfered base station set The received signal quality of each base station is evaluated and a channel state quantization model is constructed: Q m =ω1·SNR m +ω2·(1-PLR m )-ω3·I m ; Among them, PLR m Base station b m The data packet loss rate; ω1, ω2, ω3 are channel assessment weights, and they satisfy ω1+ω2+ω3=1; The final output positioning error E loc and estimated channel state information 7. The RFID anti-interference positioning method according to claim 1, characterized in that: The step of updating the interference signal identification model and tuning parameters according to the positioning error and the estimated channel state information includes: Positioning error E loc and estimated channel state information Construct a system performance feedback index Ψ, which is used to jointly reflect the positioning accuracy and channel quality level and is defined as: Among them, E max is the preset maximum allowable positioning error; Q m Indicates base station b m The channel quality evaluation value; λ1,λ2∈[0,1] are weighting coefficients, and λ1+λ2=1; The performance feedback index Ψ is used to update the feature threshold vector θ=[θ1,θ2,...,θ K ], where each parameter θ k Corresponding to a type of feature discrimination condition in the interference signal recognition model, it is updated through the following iterative formula: in, is the value of the kth parameter in the nth round; η is the update step size of the interference signal recognition model; Ψ ref is the benchmark value for ideal system performance; At the same time, the indicator Ψ is introduced into the game-based tuning parameter optimization process to dynamically adjust the weighted parameter set {α, β, γ} in the utility function. The specific adjustment method is: α=α0+κ1·Ψ, β=β0-κ2·Ψ, γ=1-α-β; Among them, α0, β0 are the initially set weight values; κ1, κ2 are weight adjustment coefficients; α, β, γ∈(0,1) and satisfy α+β+γ=1, corresponding to the proportion of transmission frequency, power and interference avoidance strategy in the utility function respectively; The updated weights are introduced into the tuning parameters, and for each reader r s The strategy in round n (f s ,P s ) to find the optimal response: Among them, f s 、P s Respectively represent the transmission frequency and power of the reader; is the strategy of other readers in round n; U s (·) is the comprehensive utility function.

8. An RFID anti-interference positioning system, applied to the RFID anti-interference positioning method according to any one of claims 1 to 7, characterized in that: include: A data acquisition module is used to collect characteristic data including at least signal strength, arrival time, frequency offset, channel response and device identification in real time through multiple RFID readers; A standardization processing module, used for performing standardization processing on the feature data; Feature fusion module, used to build a multi-dimensional expression structure that integrates spatiotemporal and device features based on the standardized feature data; An interference identification module establishes an interference signal identification model based on machine learning, uses the multi-dimensional expression structure to extract frequency and device characteristics, identifies interference signals and outputs identification results; The parameter tuning module is used to generate tuning parameters including transmission frequency and power through game optimization strategy and adaptively adjust the RFID reader configuration; The positioning optimization and evaluation module is used to dynamically select a set of undisturbed base stations based on signal quality when interference is detected, determine the tag position by combining multi-base station collaborative positioning and triangulation algorithms, calculate the positioning error and evaluate the channel state, and generate positioning error and evaluation channel state information; The model updating module is used to update the interference signal identification model and tuning parameters according to the positioning error and the evaluated channel state information.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images