Large-scale antenna for RFID tag checking

Abstract

The invention discloses a large-scale antenna for RFID (Radio Frequency Identification) tag inventory, which relates to the technical field of wireless communication, extracts channel state information by using a communication link, extracts main characteristics from a sparse channel through a compressed sensing technology, constructs a channel model, effectively captures a signal path, a multi-path attenuation characteristic and noise intensity, and improves the performance of the antenna in a dynamic environment. According to the signal strength and return delay of the tag, dynamically dividing the inventory area, and adjusting the phase and amplitude distribution of the antenna array through real-time feedback, so that the beam coverage is matched with the channel characteristics, and the signal focusing capability of the target area is enhanced; a recurrent neural network (RNN) is introduced to analyze CSI information and feedback signals, multi-path interference and effective signals are separated by learning time sequence features, independent responses of tags are extracted, and the influence of reflected signal interference on tag identification is reduced; the identification success rate of the weak signal label is improved by preferentially processing the label with the weak signal or the shielded label and combining a dynamic recording and alarming mechanism.

Description

Technical Field

[0001] The invention relates to the technical field of wireless communications, and in particular to a large-scale antenna for RFID tag inventory counting. Background Art

[0002] In RFID inventory, large-scale antenna technology can focus signal energy on the target area through beamforming, thereby covering multiple tags at the same time and improving inventory efficiency. However, most RFID inventory environments contain a large number of reflective surfaces, such as shelves, walls or equipment. These reflective surfaces will cause multipath effects during signal propagation. The multipath effect causes the signal to reach the tag not only with a direct path, but also with reflected signals from other directions, interfering with the normal communication of the tag.

[0003] Some traditional antenna transmission methods enhance the directionality of signals by optimizing beamforming. However, when the multipath effect is severe, the directional optimization of beamforming is difficult to completely isolate the interference of reflected signals, which will cause the tag response to be fuzzy or fail. At the same time, the tags in RFID inventory scenarios are often randomly distributed and subject to dynamic environmental influences, which further limits the effectiveness of traditional solutions in suppressing multipath interference.

[0004] In inventory tasks where tags are densely packed and dynamically changing, problems such as fast channel changes, weak tag signals, and complex interference sources are more prominent. Traditional methods are difficult to meet the requirements of RFID inventory for efficiency and reliability; therefore, a targeted large-scale antenna for RFID tag inventory is urgently needed to solve such problems. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a large-scale antenna for RFID tag inventory to solve the problem that during the RFID inventory process, tags are densely distributed and dynamic, and there is multipath interference in the environment, which makes it difficult to stably identify the signals of some tags.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] An embodiment of the present invention provides a large-scale antenna for RFID tag inventory, which includes:

[0009] The communication link is used to establish the initial connection between the RFID reader and the tag, transmit the pilot signal, synchronize and initialize the communication link, and extract the channel state information CSI;

[0010] The array configuration adjustment unit dynamically divides the inventory area according to the signal strength and return delay of the tag, and adjusts the phase and amplitude distribution of the antenna array in each area;

[0011] A beamforming module adjusts the beam direction, power allocation and coverage according to the feedback signal adjusted by the array configuration adjustment unit;

[0012] Multipath signal separation module, used to analyze CSI information and feedback signals, separate multipath interference signals from valid signals, and extract independent responses of each tag;

[0013] The tag priority processing unit is used to give priority to tags with weak signals or blocked tags. After the processing is completed, it automatically resumes the normal inventory mode and dynamically records the tag processing status. For tags that have failed to be counted multiple times, an alarm is triggered to notify manual intervention;

[0014] The performance monitoring module is used to continuously monitor the performance indicators of large-scale antenna operations, including inventory success rate, energy consumption and processing delay, and collect and analyze real-time performance indicators.

[0015] As a preferred solution of the large-scale antenna for RFID tag inventory in the present invention, the transmission mode of the large-scale antenna is:

[0016] Step S1, using the pilot signal to establish a communication link between the RFID reader and the tag, extracting the channel state information CSI from the sparse channel through the compressed sensing technology, and constructing a channel model;

[0017] The CSI information includes signal path, multipath attenuation characteristics and interference strength;

[0018] The channel model is used to describe the channel state between the tag and the reader;

[0019] Step S2, dividing the inventory area according to the signal strength and return delay of the tag, and adjusting the antenna array phase and amplitude distribution of each sub-area based on the channel model;

[0020] Step S3, evaluating the signal quality of each area in combination with the feedback signal of the inventory area adjusted in step S2, adjusting the beamforming parameters, including beam direction, power allocation and coverage, setting a priority mechanism, and giving priority to optimizing the low signal-to-noise ratio area;

[0021] The feedback signal includes signal strength, signal-to-noise ratio and bit error rate;

[0022] Step S4, constructing a multipath signal separation model based on a recurrent neural network (RNN), taking CSI information and feedback signals as input, analyzing multipath interference characteristics, separating effective signals from interference paths, and extracting independent responses for each label;

[0023] Step S5, continuously monitoring the performance indicators during the large-scale antenna transmission process, and optimizing the antenna activation strategy, feedback frequency and processing priority using a Q-learning reinforcement learning algorithm based on the performance indicators;

[0024] The performance indicators include inventory success rate, energy consumption and processing delay.

[0025] As a preferred solution of the large-scale antenna for RFID tag inventory described in the present invention, the steps of extracting channel state information CSI from sparse channels by compressed sensing technology and constructing a channel model are as follows:

[0026] After the communication link is established, channel transmission is performed and the sparse channel model of the received signal is:

[0027] y1=Φ1h1+n1,

[0028] Among them, y1 is the received signal vector,

[0029] Φ1 is the pilot signal matrix, h1 is the sparse channel gain vector, n1 is the Gaussian noise vector,

[0030]

[0031] Among them, P1 is the number of paths, α p is the gain of the pth path, δ is the pulse function, f p is the frequency of the pth path, and f is the frequency variable;

[0032] The sparse channel gain h1 is reconstructed using compressed sensing technology, and the optimization objective is defined as:

[0033]

[0034] in, is the reconstructed channel gain vector, |h1|1 is the sparsity of h1, and represents the l1-norm of the channel gain,

[0035] is the reconstruction error, ∈1 is the allowable error range;

[0036] According to the reconstructed channel gain vector Extract channel state information CSI, including: signal path gain α p 、Multipath attenuation characteristics |α p |, and noise intensity |n1|2, |n1|2 represents the l2-norm of the noise;

[0037] The channel model is constructed based on the extracted CSI, and the model is expressed as:

[0038]

[0039] Among them, h ab is the channel gain, which represents the channel characteristics between device a and device b, Q1 is the total number of paths in the channel, α q is the gain coefficient of the qth path, f q is the frequency of the qth path, d ab is the distance between devices a and b, is the phase change of the signal on path q, and j is an imaginary unit.

[0040] As a preferred solution of the large-scale antenna for RFID tag inventory described in the present invention, in which: in step S2, when the feedback signal change exceeds a preset threshold, the area division and antenna configuration are updated.

[0041] As a preferred solution of the large-scale antenna for RFID tag inventorying described in the present invention, the step of dividing the inventory area according to the signal strength and return delay of the tag is as follows:

[0042] According to the signal strength S m and return delay T m Divide the inventory area, the division formula is:

[0043] R m = {x|S(x)∈[S m-1 ,S m ],T(x)∈[T m-1 ,T m ]},

[0044] Among them, R m is the mth inventory area, S(x) is the signal strength of tag x, S m-1 ,S m are the upper and lower bounds of the signal strength in the mth region, T(x) is the return delay of tag x, and T m-1 ,T m are the upper and lower bounds of the return delay for the mth region,

[0045] If the feedback signal of the divided area changes beyond the threshold, the area is re-divided, and the re-dividing formula is:

[0046]

[0047] Among them, R′ m is the adjusted regional division, |S(R m )-S avg | is the deviation between the current area signal strength and the average signal strength, S avg is the average of all tag signal intensities.

[0048] As a preferred solution of the large-scale antenna for RFID tag inventory in the present invention, the step of adjusting the antenna array phase and amplitude distribution of each sub-area based on the channel model is as follows:

[0049] For the antenna array in the mth area, the phase is adjusted, and the adjustment formula is:

[0050]

[0051] Among them, φ m is the optimal phase of the mth region, h xy is the channel gain between tag x and antenna y, φ is the phase angle, indicating the beam direction of the antenna;

[0052] Optimize the amplitude distribution of the antenna array and adjust the formula as follows:

[0053]

[0054] Among them, A m is the amplitude distribution of the mth region, A is the amplitude distribution vector, y m is the received signal of region m, λ1|A|1 is the regularization term used to control the sparsity of the amplitude distribution, where λ1 is the regularization coefficient, which controls the weight of the regularization term in the objective function, and |A|1 is the l1-norm of A, which indicates the sparsity of vector A.

[0055] Φ2 is the regional pilot signal matrix.

[0056] As a preferred solution of the large-scale antenna for RFID tag inventory in the present invention, the step of adjusting the beamforming parameters is as follows:

[0057] For the beam direction of the mth region, the optimization goal is to maximize the signal gain, and the adjustment formula is:

[0058]

[0059] Among them, w m is the beam direction vector of the mth region, h xy is the channel gain between tag x and antenna y, w is the antenna array weight vector;

[0060] For the transmission power of area m, the optimization goal is to maximize the signal strength, and the adjustment formula is:

[0061]

[0062] Among them, P m is the transmission power allocation vector of region m, P is the power allocation vector, |h xy | 2is the square of the channel gain strength, used to measure channel quality;

[0063] Adjust the area coverage range ρ based on the channel feedback bit error rate BER m , the adjustment formula is:

[0064]

[0065] Among them, ρ m is the coverage of the mth region, ρ is the coverage, I is the indicator function used to determine whether the bit error rate condition is met, BER(x) is the bit error rate of label x, BER th is the preset bit error rate threshold.

[0066] As a preferred solution of the large-scale antenna for RFID tag inventory described in the present invention, wherein: in step S4, based on the independent response of each tag, the beam angle and power distribution of the corresponding area are preferentially adjusted for tags with weak signals or blocked, and after the processing is completed, the normal inventory mode is automatically restored;

[0067] In step S4, the processing status of the tag is dynamically recorded, and an alarm is triggered for tags that have failed to be counted multiple times, notifying manual intervention.

[0068] As a preferred solution of the large-scale antenna for RFID tag inventory described in the present invention, the steps of constructing a multipath signal separation model based on a recursive neural network RNN, taking CSI information and feedback signals as inputs, analyzing multipath interference characteristics, and separating effective signals from interference paths are as follows:

[0069] The extracted CSI information h ab and the feedback signal f ab Constructed as an input sequence, represented as:

[0070] X ab ={[h ab (t),f ab (t)]|t=1,2,…,T1},

[0071] Among them, X ab is the input feature sequence between device a and device b, h ab (t) is the channel gain at time t, f ab (t) is the feedback signal at time t, T1 is the time step of the input sequence,

[0072] Construct a recursive neural network RNN ​​model for multipath signal separation, and the update equation is:

[0073] h t+1 =σ(W h h t+W x X t +b h ),

[0074] Among them, h t is a hidden state, indicating the network state at step t, W h is the hidden layer weight matrix, W x is the input layer weight matrix, b h is the bias term, σ is the activation function;

[0075] Extract the effective signal s through the output layer ab and interference signal i ab , the extraction process is expressed as:

[0076] [s ab ,i ab ]=softmax(W o h T +b o ),

[0077] Among them, s ab is the extracted effective signal, i ab is the extracted interference signal, W o is the output layer weight matrix, h T is the hidden state of the last time step, b o is the output layer bias,

[0078] Will s ab and i ab Further separation into path response, the separation formula is:

[0079] r pq =s ab ·δ(ff q ),

[0080] Among them, r pq is the independent response of the pth path, δ is the impulse function used to match the path frequency f q .

[0081] As a preferred solution of the large-scale antenna for RFID tag inventory described in the present invention, the step of optimizing the antenna activation strategy, feedback frequency and processing priority by using the Q-learning reinforcement learning algorithm based on the performance index is as follows:

[0082] Define the state space, including inventory success rate, energy consumption and processing delay,

[0083] Define the action space. The adjustment strategies of the action space include: dynamically adjusting the feedback frequency according to the channel conditions, giving priority to tags with low signal-to-noise ratio or blocked signals,

[0084] Define the reward function and set the weights to balance the success rate of inventory counting, energy consumption, and processing delay;

[0085] The Q-learning update strategy is adopted to learn the rewards of each state-action combination and update the action strategy. After each action is executed, the gap between the current state and the target performance is calculated to update the action priority. Through continuous learning and adjustment by Q-learning, the antenna can adapt to real-time changing channel conditions, dynamically optimize activation strategies, feedback frequency and processing priority, and ultimately achieve continuous improvement of performance indicators.

[0086] The beneficial effects of the present invention are as follows: the present invention utilizes the communication link to extract channel state information, extracts the main features from the sparse channel through the compressed sensing technology, constructs a channel model, effectively captures the signal path, multipath attenuation characteristics and noise intensity, and dynamically divides the inventory area according to the signal strength and return delay of the tag in a dynamic environment, and adjusts the phase and amplitude distribution of the antenna array through real-time feedback to match the beam coverage with the channel characteristics, thereby enhancing the signal focusing capability on the target area; introduces a recursive neural network RNN ​​to analyze CSI information and feedback signals, separates multipath interference from effective signals by learning time series features, extracts the independent response of the tag, and reduces the influence of reflected signal interference on tag recognition; on this basis, by giving priority to processing tags with weak or blocked signals, combined with dynamic recording and alarm mechanisms, the recognition success rate of weak signal tags is improved; uses a reinforcement learning model to optimize antenna activation strategies, feedback frequencies and processing priorities, and adaptively optimizes dynamic resource allocation through continuous learning driven by performance indicators.

[0087] In summary, the present invention can significantly improve the efficiency and reliability of RFID inventory counting while optimizing energy consumption and processing delay. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0089] Figure 1 It is a schematic diagram of the framework of the large-scale antenna of the present invention.

[0090] Figure 2 It is a schematic diagram of the information transmission process of the large-scale antenna of the present invention. DETAILED DESCRIPTION

[0091] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0092] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0093] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0094] Example 1, reference Figure 1 and Figure 2 , this embodiment provides a large-scale antenna for RFID tag inventory, including:

[0095] The communication link is used to establish the initial connection between the RFID reader and the tag, transmit the pilot signal, synchronize and initialize the communication link, and extract the channel state information CSI;

[0096] The array configuration adjustment unit dynamically divides the inventory area according to the signal strength and return delay of the tag, and adjusts the phase and amplitude distribution of the antenna array in each area;

[0097] A beamforming module adjusts the beam direction, power allocation and coverage according to the feedback signal adjusted by the array configuration adjustment unit;

[0098] Multipath signal separation module, used to analyze CSI information and feedback signals, separate multipath interference signals from valid signals, and extract independent responses of each tag;

[0099] The tag priority processing unit is used to give priority to tags with weak signals or blocked tags. After the processing is completed, it automatically resumes the normal inventory mode and dynamically records the tag processing status. For tags that have failed to be counted multiple times, an alarm is triggered to notify manual intervention;

[0100] The performance monitoring module is used to continuously monitor the performance indicators of large-scale antenna operations, including inventory success rate, energy consumption and processing delay, and collect and analyze real-time performance indicators.

[0101] This embodiment also provides a large-scale antenna for RFID tag inventory, including:

[0102] Step S1, using the pilot signal to establish a communication link between the RFID reader and the tag, extracting the channel state information CSI from the sparse channel through the compressed sensing technology, and constructing a channel model;

[0103] CSI information includes signal path, multipath attenuation characteristics, and interference strength;

[0104] The channel model is used to describe the channel status between the tag and the reader;

[0105] The compressed sensing technology is used to extract the channel state information CSI from the sparse channel. The steps to construct the channel model are as follows:

[0106] After the communication link is established, channel transmission is performed and the sparse channel model of the received signal is:

[0107] y1=Φ1h1+n1,

[0108] Among them, y1 is the received signal vector,

[0109] Φ1 is the pilot signal matrix, h1 is the sparse channel gain vector, n1 is the Gaussian noise vector,

[0110]

[0111] Among them, P1 is the number of paths, α p is the gain of the pth path, δ is the pulse function, f p is the frequency of the pth path, and f is the frequency variable;

[0112] The sparse channel gain h1 is reconstructed using compressed sensing technology, and the optimization objective is defined as:

[0113]

[0114] in, is the reconstructed channel gain vector, |h1|1 is the sparsity of h1, and represents the l1-norm of the channel gain,

[0115] is the reconstruction error, ∈1 is the allowable error range;

[0116] According to the reconstructed channel gain vector Extract channel state information CSI, including: signal path gain α p 、Multipath attenuation characteristics |α p |, and noise intensity |n1|2, |n1|2 represents the l2-norm of the noise;

[0117] The channel model is constructed based on the extracted CSI, and the model is expressed as:

[0118]

[0119] Among them, h ab is the channel gain, which represents the channel characteristics between device a and device b, Q1 is the total number of paths in the channel, α q is the gain coefficient of the qth path, f q is the frequency of the qth path, d ab is the distance between devices a and b, is the phase change of the signal on path q, j is the imaginary unit;

[0120] Specifically, step S1 uses compressed sensing technology to extract channel state information CSI from the received sparse signal, constructs an optimization problem, reconstructs the sparse channel using l1-norm minimization, and extracts the signal path, attenuation characteristics and noise.

[0121] Step S2, dividing the inventory area according to the signal strength and return delay of the tag, and adjusting the antenna array phase and amplitude distribution of each sub-area based on the channel model;

[0122] In step S2, when the feedback signal changes beyond a preset threshold, the area division and antenna configuration are updated;

[0123] The steps to divide the inventory area according to the signal strength and return delay of the tag are as follows:

[0124] According to the signal strength S m and return delay T m Divide the inventory area, the division formula is:

[0125] R m = {x|S(x)∈[S m-1 ,S m ],T(x)∈[T m-1 ,T m ]},

[0126] Among them, R m is the mth inventory area, S(x) is the signal strength of tag x, S m-1 ,S m are the upper and lower bounds of the signal strength in the mth region, T(x) is the return delay of tag x, and T m-1 ,T m are the upper and lower bounds of the return delay for the mth region,

[0127] If the feedback signal of the divided area changes beyond the threshold, the area is re-divided, and the re-dividing formula is:

[0128]

[0129] Among them, R′ m is the adjusted regional division, |S(Rm )-S avg | is the deviation between the current area signal strength and the average signal strength, S avg is the average signal intensity of all tags,

[0130] The steps for adjusting the antenna array phase and amplitude distribution of each sub-area based on the channel model are:

[0131] For the antenna array in the mth area, the phase is adjusted, and the adjustment formula is:

[0132]

[0133] Among them, φ m is the optimal phase of the mth region, h xy is the channel gain between tag x and antenna y, φ is the phase angle, indicating the beam direction of the antenna;

[0134] Optimize the amplitude distribution of the antenna array and adjust the formula as follows:

[0135]

[0136] Among them, A m is the amplitude distribution of the mth region, A is the amplitude distribution vector, y m is the received signal of region m, λ1|A|1 is the regularization term used to control the sparsity of the amplitude distribution, where λ1 is the regularization coefficient, which controls the weight of the regularization term in the objective function, and |A|1 is the l1-norm of A, which indicates the sparsity of vector A.

[0137] Φ2 is the regional pilot signal matrix;

[0138] Specifically, step S2 dynamically divides the inventory area by signal strength and return delay, adjusts the phase and amplitude distribution of the antenna array based on the channel model, updates the area division using the feedback signal, matches the antenna configuration to the channel characteristics, and performs dynamic area optimization and antenna signal enhancement.

[0139] Step S3, evaluating the signal quality of each area in combination with the feedback signal of the inventory area adjusted in step S2, adjusting the beamforming parameters, including beam direction, power allocation and coverage, setting a priority mechanism, and giving priority to optimizing the low signal-to-noise ratio area;

[0140] Feedback signals include signal strength, signal-to-noise ratio, and bit error rate;

[0141] The steps for adjusting the beamforming parameters are:

[0142] For the beam direction of the mth region, the optimization goal is to maximize the signal gain, and the adjustment formula is:

[0143]

[0144] Among them, w m is the beam direction vector of the mth region, h xy is the channel gain between tag x and antenna y, w is the antenna array weight vector;

[0145] For the transmission power of area m, the optimization goal is to maximize the signal strength, and the adjustment formula is:

[0146]

[0147] Among them, P m is the transmission power allocation vector of region m, P is the power allocation vector, |h xy | 2 is the square of the channel gain strength, used to measure channel quality;

[0148] Adjust the area coverage range ρ based on the channel feedback bit error rate BER m , the adjustment formula is:

[0149]

[0150] Among them, ρ m is the coverage of the mth region, ρ is the coverage, I is the indicator function used to determine whether the bit error rate condition is met, BER(x) is the bit error rate of label x, BER th is a preset bit error rate threshold;

[0151] Specifically, step S3 adjusts the beamforming parameters, including beam direction, power allocation and coverage, by enhancing the signal quality in low signal-to-noise ratio areas, improving the ability of large-scale antennas to identify weak signal tags, and dynamically optimizing the coverage to improve overall communication efficiency.

[0152] Step S4, constructing a multipath signal separation model based on a recurrent neural network (RNN), taking CSI information and feedback signals as input, analyzing multipath interference characteristics, separating effective signals from interference paths, and extracting independent responses for each label;

[0153] In step S4, based on the independent response of each tag, the beam angle and power distribution of the corresponding area are preferentially adjusted for tags with weak signals or blocked. After the processing is completed, the normal inventory mode is automatically restored;

[0154] In step S4, the processing status of the tag is dynamically recorded, and an alarm is triggered for tags that have failed to be counted multiple times, notifying manual intervention;

[0155] A multipath signal separation model based on recurrent neural network (RNN) is constructed. The CSI information and feedback signal are used as input to analyze the multipath interference characteristics. The steps to separate the effective signal and the interference path are as follows:

[0156] The extracted CSI information h ab and the feedback signal f ab Constructed as an input sequence, represented as:

[0157] X ab ={[h ab (t),f ab (t)]|t=1,2,…,T1},

[0158] Among them, X ab is the input feature sequence between device a and device b, h ab (t) is the channel gain at time t, f ab (t) is the feedback signal at time t, T1 is the time step of the input sequence,

[0159] Construct a recursive neural network RNN ​​model for multipath signal separation, and the update equation is:

[0160] h t+1 =σ(W h h t +W x X t +b h ),

[0161] Among them, h t is a hidden state, indicating the network state at step t, W h is the hidden layer weight matrix, W x is the input layer weight matrix, b h is the bias term, σ is the activation function;

[0162] Extract the effective signal s through the output layer ab and interference signal i ab , the extraction process is expressed as:

[0163] [s ab ,i ab ]=softmax(W o h T +b o ),

[0164] Among them, s ab is the extracted effective signal, i ab is the extracted interference signal, W o is the output layer weight matrix, h T is the hidden state of the last time step, b ois the output layer bias,

[0165] Will s ab and i ab Further separation into path response, the separation formula is:

[0166] r pq =s ab ·δ(ff q ),

[0167] Among them, r pq is the independent response of the pth path, δ is the impulse function used to match the path frequency f q ;

[0168] Specifically, step S4 uses a recursive neural network to analyze the CSI and feedback signal features in the time series, dynamically learns the distribution characteristics of the multipath signal through the hidden state, separates the effective signal from the interference path, and outputs the response of the independent path.

[0169] Step S5, continuously monitoring the performance indicators during the large-scale antenna transmission process, and optimizing the antenna activation strategy, feedback frequency and processing priority using a Q-learning reinforcement learning algorithm based on the performance indicators;

[0170] Performance indicators, including inventory success rate, energy consumption, and processing latency;

[0171] The steps of using Q-learning reinforcement learning algorithm to optimize antenna activation strategy, feedback frequency and processing priority based on performance indicators are as follows:

[0172] Define the state space, including inventory success rate, energy consumption and processing delay,

[0173] Define the action space. The adjustment strategies of the action space include: dynamically adjusting the feedback frequency according to the channel conditions, giving priority to tags with low signal-to-noise ratio or blocked signals,

[0174] Define the reward function and set the weights to balance the success rate of inventory counting, energy consumption, and processing delay;

[0175] The Q-learning update strategy is used to learn the rewards of each state-action combination and update the action strategy. After each action is executed, the gap between the current state and the target performance is calculated to update the action priority. Through continuous learning and adjustment by Q-learning, the antenna can adapt to the real-time changing channel conditions, dynamically optimize the activation strategy, feedback frequency and processing priority, and ultimately achieve continuous improvement of performance indicators.

[0176] Specifically, step S5 uses the Q-learning reinforcement learning method to dynamically optimize antenna activation, feedback frequency, and processing priority. Through real-time learning and updating, it can automatically adjust in complex environments, which is very suitable for RFID scenarios that require real-time response.

[0177] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A large-scale antenna for RFID tag inventory, characterized by: include, The communication link is used to establish the initial connection between the RFID reader and the tag, transmit the pilot signal, synchronize and initialize the communication link, and extract the channel state information CSI; The array configuration adjustment unit dynamically divides the inventory area according to the signal strength and return delay of the tag, and adjusts the phase and amplitude distribution of the antenna array in each area; A beamforming module adjusts the beam direction, power allocation and coverage according to the feedback signal adjusted by the array configuration adjustment unit; Multipath signal separation module, used to analyze CSI information and feedback signals, separate multipath interference signals from valid signals, and extract independent responses of each tag; The tag priority processing unit is used to give priority to tags with weak signals or blocked tags. After the processing is completed, it automatically resumes the normal inventory mode and dynamically records the tag processing status. For tags that have failed to be counted multiple times, an alarm is triggered to notify manual intervention; The performance monitoring module is used to continuously monitor the performance indicators of large-scale antenna operations, including inventory success rate, energy consumption and processing delay, and collect and analyze real-time performance indicators.

2. A large-scale antenna for RFID tag inventory as claimed in claim 1, characterized in that: The transmission mode of the large-scale antenna is: Step S1, using the pilot signal to establish a communication link between the RFID reader and the tag, extracting the channel state information CSI from the sparse channel through the compressed sensing technology, and constructing a channel model; The CSI information includes signal path, multipath attenuation characteristics and interference strength; The channel model is used to describe the channel state between the tag and the reader; Step S2, dividing the inventory area according to the signal strength and return delay of the tag, and adjusting the antenna array phase and amplitude distribution of each sub-area based on the channel model; Step S3, evaluating the signal quality of each area in combination with the feedback signal of the inventory area adjusted in step S2, adjusting the beamforming parameters, including beam direction, power allocation and coverage, setting a priority mechanism, and giving priority to optimizing the low signal-to-noise ratio area; The feedback signal includes signal strength, signal-to-noise ratio and bit error rate; Step S4, constructing a multipath signal separation model based on a recurrent neural network (RNN), taking CSI information and feedback signals as input, analyzing multipath interference characteristics, separating effective signals from interference paths, and extracting independent responses for each label; Step S5, continuously monitoring the performance indicators during the large-scale antenna transmission process, and optimizing the antenna activation strategy, feedback frequency and processing priority using a Q-learning reinforcement learning algorithm based on the performance indicators; The performance indicators include inventory success rate, energy consumption and processing delay.

3. A large-scale antenna for RFID tag inventory as claimed in claim 2, characterized in that: The steps of extracting channel state information CSI from sparse channels by compressed sensing technology and constructing a channel model are as follows: After the communication link is established, channel transmission is performed and the sparse channel model of the received signal is: y1=Φ1h1+n1, Among them, y1 is the received signal vector, Φ1 is the pilot signal matrix, h1 is the sparse channel gain vector, n1 is the Gaussian noise vector, Among them, P1 is the number of paths, α p is the gain of the pth path, δ is the pulse function, f p is the frequency of the pth path, and f is the frequency variable; The sparse channel gain h1 is reconstructed using compressed sensing technology, and the optimization objective is defined as: in, is the reconstructed channel gain vector, |h1|1 is the sparsity of h1, and represents the l1-norm of the channel gain, is the reconstruction error, ∈1 is the allowable error range; According to the reconstructed channel gain vector Extract channel state information CSI, including: signal path gain α p 、Multipath attenuation characteristics |α p |, and noise intensity |n1|2, |n1|2 represents the l2-norm of the noise; The channel model is constructed based on the extracted CSI, and the model is expressed as: Among them, h ab is the channel gain, which represents the channel characteristics between device a and device b, Q1 is the total number of paths in the channel, α q is the gain coefficient of the qth path, f q is the frequency of the qth path, d ab is the distance between devices a and b, is the phase change of the signal on path q, and j is an imaginary unit.

4. A large-scale antenna for RFID tag inventory as claimed in claim 3, characterized in that: In step S2, when the feedback signal change exceeds a preset threshold, the area division and antenna configuration are updated.

5. A large-scale antenna for RFID tag inventory as claimed in claim 4, characterized in that: The step of dividing the inventory area according to the signal strength and return delay of the tag is: According to the signal strength S m and return delay T m Divide the inventory area, the division formula is: R m ={x∣S(x)∈[S m-1 ,S m ],T(x)∈[T m-1 ,T m ]}, Among them, R m is the mth inventory area, S(x) is the signal strength of tag x, S m-1 ,S m are the upper and lower bounds of the signal strength in the mth region, T(x) is the return delay of tag x, and T m-1 ,T m are the upper and lower bounds of the return delay for the mth region, If the feedback signal of the divided area changes beyond the threshold, the area is re-divided, and the re-dividing formula is: Among them, R′ m is the adjusted regional division, |S(R m )-S avg | is the deviation between the current area signal strength and the average signal strength, S avg is the average of all tag signal intensities.

6. A large-scale antenna for RFID tag inventory as claimed in claim 5, characterized in that: The step of adjusting the antenna array phase and amplitude distribution of each sub-area based on the channel model is: For the antenna array in the mth area, the phase is adjusted, and the adjustment formula is: Among them, φ m is the optimal phase of the mth region, h xy is the channel gain between tag x and antenna y, φ is the phase angle, indicating the beam direction of the antenna; Optimize the amplitude distribution of the antenna array and adjust the formula as follows: Among them, A m is the amplitude distribution of the mth region, A is the amplitude distribution vector, y m is the received signal of region m, λ1|A|1 is the regularization term used to control the sparsity of the amplitude distribution, where λ1 is the regularization coefficient, which controls the weight of the regularization term in the objective function, and |A|1 is the l1-norm of A, which indicates the sparsity of vector A. Φ2 is the regional pilot signal matrix.

7. A large-scale antenna for RFID tag inventory as claimed in claim 6, characterized in that: The step of adjusting the beamforming parameters is: For the beam direction of the mth region, the optimization goal is to maximize the signal gain, and the adjustment formula is: Among them, w m is the beam direction vector of the mth region, h xy is the channel gain between tag x and antenna y, w is the antenna array weight vector; For the transmission power of area m, the optimization goal is to maximize the signal strength, and the adjustment formula is: Among them, P m is the transmission power allocation vector of region m, P is the power allocation vector, |h xy | 2 is the square of the channel gain strength, used to measure channel quality; Adjust the area coverage range ρ based on the channel feedback bit error rate BER m , the adjustment formula is: Among them, ρ m is the coverage of the mth region, ρ is the coverage, I is the indicator function used to determine whether the bit error rate condition is met, BER(x) is the bit error rate of label x, BER th is the preset bit error rate threshold.

8. A large-scale antenna for RFID tag inventory as claimed in claim 7, characterized in that: In step S4, based on the independent response of each tag, the beam angle and power distribution of the corresponding area are preferentially adjusted for tags with weak signals or blocked. After the processing is completed, the normal inventory mode is automatically restored; In step S4, the processing status of the tag is dynamically recorded, and an alarm is triggered for tags that have failed to be counted multiple times, notifying manual intervention.

9. A large-scale antenna for RFID tag inventory as claimed in claim 8, characterized in that: The steps of constructing a multipath signal separation model based on a recurrent neural network RNN, taking CSI information and feedback signals as input, analyzing multipath interference characteristics, and separating effective signals from interference paths are as follows: The extracted CSI information h ab and the feedback signal f ab Constructed as an input sequence, represented as: X ab ={[h ab (t),f ab (t)]∣t=1,2,…,T1}, Among them, X ab is the input feature sequence between device a and device b, h ab (t) is the channel gain at time t, f ab (t) is the feedback signal at time t, T1 is the time step of the input sequence, Construct a recursive neural network RNN ​​model for multipath signal separation, and the update equation is: h t+1 =σ(W h h t +W x X t +b h ), Among them, h t is a hidden state, indicating the network state at step t, W h is the hidden layer weight matrix, W x is the input layer weight matrix, b h is the bias term, σ is the activation function; Extract the effective signal s through the output layer ab and interference signal i ab , the extraction process is expressed as: [s ab ,i ab ]=softmax(W o h T +b o ), Among them, s ab is the extracted effective signal, i ab is the extracted interference signal, W o is the output layer weight matrix, h T is the hidden state of the last time step, b o is the output layer bias, Will s ab and i ab Further separation into path response, the separation formula is: r pq =s ab ·δ(f-f q ), Among them, r pq is the independent response of the pth path, δ is the impulse function used to match the path frequency f q .

10. A large-scale antenna for RFID tag inventory as claimed in claim 9, characterized in that: The steps of optimizing antenna activation strategy, feedback frequency and processing priority by using Q-learning reinforcement learning algorithm based on performance indicators are: Define the state space, including inventory success rate, energy consumption and processing delay, Define the action space. The adjustment strategies of the action space include: dynamically adjusting the feedback frequency according to the channel conditions, giving priority to tags with low signal-to-noise ratio or blocked signals, Define the reward function and set the weights to balance the success rate of inventory counting, energy consumption, and processing delay; The Q-learning update strategy is used to learn the rewards of each state-action combination, update the action strategy, and calculate the gap between the current state and the target performance after each action is executed.

Images