A RFID multi-tag identification optimization method and system

Abstract

The present invention discloses an RFID multi-tag identification optimization method and system, the method includes data collection, data serialization, time series feature prediction, collision avoidance and real-time feedback and adjustment. The present invention belongs to the field of wireless radio frequency identification technology, specifically refers to an RFID multi-tag identification optimization method and system, the scheme constructs a GRU network to identify the nonlinear relationship between time series data, predicts the type of tags output in the future time series and their response probability, dynamically adapts to the length of the input sequence, has fast operating efficiency, high real-time performance, and improves the flexibility of multi-tag identification; the collision avoidance strategy is designed to effectively reduce the collision probability, and by dynamically selecting the time slot division strategy and the binary tree strategy, the system can more effectively utilize the communication time and spectrum resources, and more accurately locate the tags, avoid unnecessary queries and communication conflicts, and help the system maintain good performance under different working conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of radio frequency identification, and in particular to an RFID multi-tag identification optimization method and system. Background Art

[0002] RFID is a technology used for wireless identification and tracking of tags. It is based on the principle of radio frequency communication and realizes data reading, writing and communication by transmitting wireless radio frequency signals between readers and tags. RFID technology has been widely used in automatic identification, logistics, supply chain management, inventory tracking and other fields. General readers directly identify the tag signal when receiving it, without considering the response characteristics of the tag and the timing characteristics of the tag distribution. There are often network delays in the communication process, which slows down the tag recognition speed and causes channel congestion. General multi-tag recognition will collide, especially in the case of high-speed or intensive movement. Due to the complexity of signal propagation and signal interference in the communication channel, there are errors in tag positioning, which leads to recognition errors or tag loss. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an RFID multi-tag identification optimization method and system. For general readers that directly identify when receiving tag signals without considering the response characteristics of the tags and the timing characteristics of the tag distribution, and there are often network delays in the communication process, which slows down the tag recognition speed and thus causes channel congestion. This solution constructs a GRU network to identify the nonlinear relationship between time series data, predicts the tag type and its response probability output in the future time series, dynamically adapts to the length of the input sequence, has high operating efficiency and high real-time performance, and improves the flexibility of multi-tag identification; for general multi-tag identification, collisions may occur, especially in the case of high-speed or intensive movement, and due to the complexity of signal propagation and signal interference in the communication channel, there are errors in tag positioning, which leads to identification errors or tag loss. This solution designs a collision avoidance strategy to effectively reduce the collision probability. By dynamically selecting the time slot division strategy and the binary tree strategy, the system can more effectively utilize the communication time and spectrum resources, and more accurately locate the tags, avoid unnecessary queries and communication conflicts, and help the system maintain good performance under different working conditions.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides an RFID multi-tag identification optimization method, which comprises the following steps:

[0005] Step S1: Data collection, using a central processor to construct a tag history response data set based on the tag information received by the reader, the tag history response data set includes the tag's timing response, signal strength, and response time. The reader is a device that sends an activation signal to the tag and receives tag information. The tag is an electronic device that stores and transmits information. The tag information consists of a tag ID and a user data area.

[0006] Step S2: Data serialization: convert the label history response data set into time series data in units of time steps to obtain the converted label history response data set. Randomly select 70% of the converted label history response data set as a training set and 30% as a test set.

[0007] Step S3: Time series feature prediction, building a GRU network, identifying the nonlinear relationship between time series data, and predicting the label type and response probability of the output in the future time series;

[0008] Step S4: collision avoidance, establishing a communication channel between the reader and the tag, designing a collision avoidance strategy, including a time slot division strategy and a binary tree strategy, using a collision detection algorithm to detect the response of the tag in the communication channel, and selecting a collision avoidance strategy based on the output and response of the GRU network;

[0009] Step S5: Real-time feedback and adjustment. The central processor continuously monitors the communication channel and dynamically adjusts the parameters of the GRU network and the parameters of the collision avoidance strategy according to the response situation.

[0010] Furthermore, in step S3, the time series feature prediction specifically includes the following steps:

[0011] Step S31: Network construction, constructing a GRU network to identify the nonlinear relationship between time series data. The GRU network consists of an update gate, a reset gate, and a hidden state. The GRU network contains multiple cell units. The hidden state of each cell unit is used as the input data of the next cell unit. The initial input data is time series data. The steps are as follows:

[0012] Step S311: Update gate construction, using the hidden state of the previous time step and the input data of the current time step to construct the update gate, the formula used is as follows:

[0013] ;

[0014] In the formula, represents the update gate, represents the sigmoid function, represents the current time step, represents the weight of the update gate, represents the bias value of the update gate, represents the hidden state at the previous time step, Represents input data;

[0015] Step S312: Reset gate construction, using the hidden state of the previous time step and the input data of the current time step to construct the reset gate, the formula used is as follows:

[0016] ;

[0017] In the formula, Represents the reset gate, represents the sigmoid function, Represents resetting the weight of the gate, Indicates the bias value of the reset gate;

[0018] Step S313: Candidate hidden state: The candidate hidden state is calculated based on the hidden state of the previous time step, the input data of the current time step and the reset gate. The formula used is as follows:

[0019] ;

[0020] In the formula, represents the candidate hidden state, represents the tanh function, represents the hidden weight, represents the hidden bias value, Represents element-wise multiplication;

[0021] Step S314: Hidden state update: Use the hidden state of the previous time step, the update gate and the candidate hidden state to update the hidden state. The formula used is as follows:

[0022] ;

[0023] In the formula, represents the updated hidden state;

[0024] Step S32: Add an output layer and construct a fully connected layer as the output layer. Add it after the last hidden state in the GRU network. Use the softmax function to convert the output result of the last hidden state into a category probability, and combine the category probability of each input data into a category probability vector. The formula used is as follows:

[0025] ;

[0026] ;

[0027] In the formula, represents the predicted value, represents the output vector of the last hidden state, represents the output weight, Indicates the output bias value, represents the softmax function, The output vector of the last hidden state elements, Indicates the number of categories of labels, Represents all elements in the output vector of the last hidden state The index and;

[0028] Step S33: Model training, inputting the training set into the GRU network for forward propagation, and calculating the predicted value of the GRU network;

[0029] Step S34: Loss function calculation, using the cross entropy loss function to calculate the difference between the predicted value and the actual value of the GRU network, the formula used is as follows:

[0030] CE y, y ̂ =-[y· log y ̂ + 1-y ·log(1- y ̂ )] ;

[0031] In the formula, represents the cross entropy loss function, is the actual value, is the predicted value;

[0032] Step S35: Back propagation, using the gradient descent algorithm to update the parameters of the GRU network, including the update gate, reset gate, hidden state, output layer weights and bias values. The formula used is as follows:

[0033] ;

[0034] In the formula, represents the updated parameters, represents the original parameters, represents the learning rate;

[0035] Step S36: Model testing: input the test set into the GRU network and use the confusion matrix to calculate the classification accuracy of the GRU model. The formula used is as follows:

[0036] ;

[0037] ;

[0038] In the formula, They are true positive examples, false negative examples, false positive examples, and true negative examples in the classification result confusion matrix. Represents the model's ability to correctly predict positive examples, Indicates the model's ability to incorrectly predict positive examples.

[0039] Further, in step S4, the collision avoidance specifically includes the following steps:

[0040] Step S41: establishing a communication channel between the reader and the tag, the communication channel being used to transmit the query request and activation signal sent by the reader and the response and tag information sent by the tag;

[0041] Step S42: Preliminary strategy selection: a collision avoidance strategy is preliminarily selected according to a threshold set based on the number of tags predicted by the GRU network. The collision avoidance strategy includes a time slot partitioning strategy and a binary tree strategy. If the number of tags is less than the threshold, the time slot partitioning strategy is used; if the number of tags is more than the threshold, the binary tree strategy is used.

[0042] Step S43: Time slot division strategy, divide the time into fixed time slots, the reader and the tag communicate in each time slot, perform collision detection after the time slot ends, set the maximum number of retransmissions, and retransmit if a collision is detected. If a collision is still detected when the maximum number of retransmissions is reached, use the binary tree strategy to query. The steps are as follows:

[0043] Step S431: The time slot length is determined according to the transmission time of the tag and other overheads in the communication process, and the time slot length is adjusted according to the utilization of the system. The formula used is as follows:

[0044] ;

[0045] ;

[0046] In the formula, Indicates the time slot length, Indicates the data length, Indicates the transmission rate of the tag, Indicates other overhead in the communication process, Indicates the utilization rate of the system. represents the probability of tag sending, Indicates the number of tags;

[0047] Step S432: counter initialization, designing a time slot counter, and initializing the time slot counter to 0;

[0048] Step S433: Data transmission. At the beginning of each time slot, the reader and the tag communicate. The reader sends a query request to nearby tags to check whether there is a tag that wants to transmit data. If a tag senses the query request and is ready to transmit data, it sends a response to the reader. After receiving the response, the reader sends an activation signal to the tag to activate the tag. After the tag is activated, it transmits the tag information to the reader through the communication channel.

[0049] Step S434: updating the time slot counter. When a time slot ends and data transmission is completed, the time slot counter is incremented by 1 to point to the next time slot.

[0050] Step S435: Collision detection. After each time slot ends, if the reader recognizes that there are multiple tag information, it is determined that a collision occurs. The reader sends a collision notification to the tag. After receiving the collision notification, the tag retransmits in the next time slot. After reaching the maximum number of retransmissions, the binary tree strategy is used to query the tag ID.

[0051] Step S436: loop, repeating steps S433 to S435 until no tag responds or the maximum number of retransmissions is reached and the loop ends;

[0052] Step S44: Binary tree strategy, construct a binary tree to divide the tag set into small subsets, and use a step-by-step query method to locate the tag position. The steps are as follows:

[0053] Step S441: The root node is determined, a binary bit is selected from the ID of the tag as the ID specific bit, and the tag set is divided into two subsets according to the ID specific bit;

[0054] Step S442: internal node division, reselecting a binary bit as the ID specific bit in the divided subset, dividing the subset into two smaller subsets according to the ID specific bit, and further dividing each subset into smaller subsets step by step until only one binary bit is left in the final subset, thus forming a binary tree;

[0055] Step S443: Leaf node identification. After the binary tree is constructed, each leaf node is uniquely identified by a label ID;

[0056] Step S444: Determine the tag position. After a collision occurs during communication between the reader and the tag, the reader traverses the binary tree according to the tag probability predicted by the GRU network, reads the tag ID from the leaf node, and determines the tag position.

[0057] Step S445: Data reading, after determining the location of the tag, the reader directly reads the user data area from the tag.

[0058] The present invention provides an RFID multi-tag identification optimization system, comprising a data collection module, a data serialization module, a time series feature prediction module, a collision avoidance module and a real-time feedback and adjustment module;

[0059] The data collection module uses the central processor to construct a tag history response data set based on the tag information received by the reader. The tag history response data set includes the tag's timing response, signal strength, and response time. The reader refers to a device that sends an activation signal to the tag and receives tag information. The tag is an electronic device that stores and transmits information. The tag information consists of a tag ID and a user data area. The tag history response data set is sent to the data serialization module;

[0060] The data serialization module receives the label history response data set from the data collection module, converts the label history response data set into time series data in units of time steps, obtains the converted label history response data set, randomly selects 70% of the converted label history response data set as a training set and 30% as a test set, and sends the training set and the test set to the time series feature prediction module;

[0061] The time series feature prediction module receives the training set and the test set from the data serialization module, constructs a GRU network, identifies the nonlinear relationship between the time series data, predicts the label type and its response probability output in the future time series, and sends the output result of the GRU network to the collision avoidance module and the real-time feedback and adjustment module;

[0062] The collision avoidance module receives the output result of the GRU network from the timing feature prediction module, establishes a communication channel between the reader and the tag, designs a time slot division strategy and a binary tree strategy, uses a collision detection algorithm to detect the response of the tag in the communication channel, selects a collision avoidance strategy according to the output and response of the GRU network, and sends the collision avoidance strategy to the real-time feedback and adjustment module;

[0063] The real-time feedback and adjustment module receives the output results of the GRU network from the timing feature prediction module and the collision avoidance strategy from the collision avoidance module. The central processing unit continuously monitors the communication channel and dynamically adjusts the parameters of the GRU network and the parameters of the collision avoidance strategy according to the response situation.

[0064] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0065] (1) General readers directly identify tags when they receive signals, without considering the response characteristics of tags and the timing characteristics of tag distribution. In addition, there are often network delays during the communication process, which slows down the tag recognition speed and causes channel congestion. This scheme constructs a GRU network to identify the nonlinear relationship between time series data, predict the tag type and its response probability output in the future time series, dynamically adapt to the length of the input sequence, run efficiently, and have high real-time performance, thereby improving the flexibility of multi-tag recognition.

[0066] (2) In general, multi-tag identification will cause collisions, especially in high-speed or intensive movement. Due to the complexity of signal propagation and signal interference in the communication channel, there are errors in tag positioning, which leads to recognition errors or tag loss. The collision avoidance strategy designed in this scheme effectively reduces the probability of collision. By dynamically selecting the time slot division strategy and the binary tree strategy, the system can more effectively utilize communication time and spectrum resources, and more accurately locate tags, avoiding unnecessary queries and communication conflicts, which helps the system maintain good performance under different working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A schematic diagram of a flow chart of an RFID multi-tag identification optimization method proposed by the present invention;

[0068] Figure 2 A schematic diagram of an RFID multi-tag identification optimization system proposed by the present invention;

[0069] Figure 3 is a schematic flow chart of step S3;

[0070] Figure 4 is a flow chart of step S31;

[0071] Figure 5 is a schematic flow chart of step S4;

[0072] Figure 6 is a flow chart of step S43;

[0073] Figure 7 is a flow chart of step S44;

[0074] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0075] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0076] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0077] Example 1, see Figure 1 The present invention provides an RFID multi-tag identification optimization method, which comprises the following steps:

[0078] Step S1: Data collection, using a central processor to construct a tag history response data set based on the tag information received by the reader, the tag history response data set includes the tag's timing response, signal strength, and response time. The reader is a device that sends an activation signal to the tag and receives tag information. The tag is an electronic device that stores and transmits information. The tag information consists of a tag ID and a user data area.

[0079] Step S2: Data serialization: convert the label history response data set into time series data in units of time steps to obtain the converted label history response data set. Randomly select 70% of the converted label history response data set as a training set and 30% as a test set.

[0080] Step S3: Time series feature prediction, building a GRU network, identifying the nonlinear relationship between time series data, and predicting the label type and response probability of the output in the future time series;

[0081] Step S4: collision avoidance, establishing a communication channel between the reader and the tag, designing a collision avoidance strategy, including a time slot division strategy and a binary tree strategy, using a collision detection algorithm to detect the response of the tag in the communication channel, and selecting a collision avoidance strategy based on the output and response of the GRU network;

[0082] Step S5: Real-time feedback and adjustment. The central processor continuously monitors the communication channel and dynamically adjusts the parameters of the GRU network and the parameters of the collision avoidance strategy according to the response situation.

[0083] Example 2, see Figure 1 , Figure 3 and Figure 4 This embodiment is based on the above embodiment. In step S3, the time series feature prediction specifically includes the following steps:

[0084] Step S31: Network construction, constructing a GRU network to identify the nonlinear relationship between time series data. The GRU network consists of an update gate, a reset gate, and a hidden state. The GRU network contains multiple cell units. The hidden state of each cell unit is used as the input data of the next cell unit. The initial input data is time series data. The steps are as follows:

[0085] Step S311: Update gate construction, using the hidden state of the previous time step and the input data of the current time step to construct the update gate, the formula used is as follows:

[0086] ;

[0087] In the formula, represents the update gate, represents the sigmoid function, represents the current time step, represents the weight of the update gate, represents the bias value of the update gate, represents the hidden state at the previous time step, Represents input data;

[0088] Step S312: Reset gate construction, using the hidden state of the previous time step and the input data of the current time step to construct the reset gate, the formula used is as follows:

[0089] ;

[0090] In the formula, Represents the reset gate, represents the sigmoid function, Represents resetting the weight of the gate, Indicates the bias value of the reset gate;

[0091] Step S313: Candidate hidden state: The candidate hidden state is calculated based on the hidden state of the previous time step, the input data of the current time step and the reset gate. The formula used is as follows:

[0092] ;

[0093] In the formula, represents the candidate hidden state, represents the tanh function, represents the hidden weight, represents the hidden bias value, Represents element-wise multiplication;

[0094] Step S314: Hidden state update: Use the hidden state of the previous time step, the update gate and the candidate hidden state to update the hidden state. The formula used is as follows:

[0095] ;

[0096] In the formula, represents the updated hidden state;

[0097] Step S32: Add an output layer and construct a fully connected layer as the output layer. Add it after the last hidden state in the GRU network. Use the softmax function to convert the output result of the last hidden state into a category probability, and combine the category probability of each input data into a category probability vector. The formula used is as follows:

[0098] ;

[0099] ;

[0100] In the formula, represents the predicted value, represents the output vector of the last hidden state, represents the output weight, Indicates the output bias value, represents the softmax function, The output vector of the last hidden state elements, Indicates the number of categories of labels, Represents all elements in the output vector of the last hidden state The index and;

[0101] Step S33: Model training, inputting the training set into the GRU network for forward propagation, and calculating the predicted value of the GRU network;

[0102] Step S34: Loss function calculation, using the cross entropy loss function to calculate the difference between the predicted value and the actual value of the GRU network, the formula used is as follows:

[0103] CE y, y ̂ =-[y· log y ̂ + 1-y ·log(1- y ̂ )] ;

[0104] In the formula, represents the cross entropy loss function, is the actual value, is the predicted value;

[0105] Step S35: Back propagation, using the gradient descent algorithm to update the parameters of the GRU network, including the update gate, reset gate, hidden state, output layer weights and bias values. The formula used is as follows:

[0106] ;

[0107] In the formula, represents the updated parameters, represents the original parameters, represents the learning rate;

[0108] Step S36: Model testing: input the test set into the GRU network and use the confusion matrix to calculate the classification accuracy of the GRU model. The formula used is as follows:

[0109] ;

[0110] ;

[0111] In the formula, They are true positive examples, false negative examples, false positive examples, and true negative examples in the classification result confusion matrix. Indicates the model's ability to correctly predict positive examples, Indicates the model's ability to incorrectly predict positive examples.

[0112] By executing the operations, a general reader directly identifies the tag signal when it receives it, without considering the response characteristics of the tag and the timing characteristics of the tag distribution, and there is often a network delay in the communication process, which slows down the tag recognition speed and causes channel congestion. This solution constructs a GRU network to identify the nonlinear relationship between time series data, predict the tag type and its response probability output in the future time series, dynamically adapt to the length of the input sequence, have fast operating efficiency, high real-time performance, and improve the flexibility of multi-tag recognition.

[0113] Example 3, see Figure 1 , Figure 5 , Figure 6 and Figure 7 This embodiment is based on the above embodiment. In step S4, collision avoidance specifically includes the following steps:

[0114] Step S41: establishing a communication channel between the reader and the tag, the communication channel being used to transmit the query request and activation signal sent by the reader and the response and tag information sent by the tag;

[0115] Step S42: Preliminary strategy selection: a collision avoidance strategy is preliminarily selected according to a threshold set based on the number of tags predicted by the GRU network. The collision avoidance strategy includes a time slot partitioning strategy and a binary tree strategy. If the number of tags is less than the threshold, the time slot partitioning strategy is used; if the number of tags is more than the threshold, the binary tree strategy is used.

[0116] Step S43: Time slot division strategy, divide the time into fixed time slots, the reader and the tag communicate in each time slot, perform collision detection after the time slot ends, set the maximum number of retransmissions, and retransmit if a collision is detected. If a collision is still detected when the maximum number of retransmissions is reached, use the binary tree strategy to query. The steps are as follows:

[0117] Step S431: The time slot length is determined according to the transmission time of the tag and other overheads in the communication process, and the time slot length is adjusted according to the utilization of the system. The formula used is as follows:

[0118] ;

[0119] ;

[0120] In the formula, Indicates the time slot length, Indicates the data length, Indicates the transmission rate of the tag, Indicates other overhead in the communication process, Indicates the utilization rate of the system. represents the probability of tag sending, Indicates the number of tags;

[0121] Step S432: counter initialization, designing a time slot counter, and initializing the time slot counter to 0;

[0122] Step S433: Data transmission. At the beginning of each time slot, the reader and the tag communicate. The reader sends a query request to nearby tags to check whether there is a tag that wants to transmit data. If a tag senses the query request and is ready to transmit data, it sends a response to the reader. After receiving the response, the reader sends an activation signal to the tag to activate the tag. After the tag is activated, it transmits the tag information to the reader through the communication channel.

[0123] Step S434: updating the time slot counter. When a time slot ends and data transmission is completed, the time slot counter is incremented by 1 to point to the next time slot.

[0124] Step S435: Collision detection. After each time slot ends, if the reader recognizes that there are multiple tag information, it is determined that a collision occurs. The reader sends a collision notification to the tag. After receiving the collision notification, the tag retransmits in the next time slot. After reaching the maximum number of retransmissions, the binary tree strategy is used to query the tag ID.

[0125] Step S436: loop, repeating steps S433 to S435 until no tag responds or the maximum number of retransmissions is reached and the loop ends;

[0126] Step S44: Binary tree strategy, construct a binary tree to divide the tag set into small subsets, and use a step-by-step query method to locate the tag position. The steps are as follows:

[0127] Step S441: The root node is determined, a binary bit is selected from the ID of the tag as the ID specific bit, and the tag set is divided into two subsets according to the ID specific bit;

[0128] Step S442: internal node division, reselecting a binary bit as the ID specific bit in the divided subset, dividing the subset into two smaller subsets according to the ID specific bit, and further dividing each subset into smaller subsets step by step until only one binary bit is left in the final subset, thus forming a binary tree;

[0129] Step S443: Leaf node identification. After the binary tree is constructed, each leaf node is uniquely identified by a label ID;

[0130] Step S444: Determine the tag position. After a collision occurs during communication between the reader and the tag, the reader traverses the binary tree according to the tag probability predicted by the GRU network, reads the tag ID from the leaf node, and determines the tag position.

[0131] Step S445: Data reading, after determining the location of the tag, the reader directly reads the user data area from the tag.

[0132] By executing the operations, collisions may occur in general multi-tag identification, especially in high-speed or intensive movement, and due to the complexity of signal propagation and signal interference in the communication channel, there are errors in tag positioning, which leads to identification errors or tag loss. The collision avoidance strategy designed in this solution effectively reduces the probability of collision. By dynamically selecting time slot division strategy and binary tree strategy, the system can more effectively utilize communication time and spectrum resources, and more accurately locate tags, avoiding unnecessary queries and communication conflicts, which helps the system maintain better performance under different working conditions.

[0133] Example 4, see Figure 2, this embodiment is based on the above embodiment, and the present invention provides an RFID multi-tag identification optimization system, including a data collection module, a data serialization module, a time series feature prediction module, a collision avoidance module and a real-time feedback and adjustment module;

[0134] The data collection module uses the central processor to construct a tag history response data set based on the tag information received by the reader. The tag history response data set includes the tag's timing response, signal strength, and response time. The reader refers to a device that sends an activation signal to the tag and receives tag information. The tag is an electronic device that stores and transmits information. The tag information consists of a tag ID and a user data area. The tag history response data set is sent to the data serialization module;

[0135] The data serialization module receives the label history response data set from the data collection module, converts the label history response data set into time series data in units of time steps, obtains the converted label history response data set, randomly selects 70% of the converted label history response data set as a training set and 30% as a test set, and sends the training set and the test set to the time series feature prediction module;

[0136] The time series feature prediction module receives the training set and the test set from the data serialization module, constructs a GRU network, identifies the nonlinear relationship between the time series data, predicts the label type and its response probability output in the future time series, and sends the output result of the GRU network to the collision avoidance module and the real-time feedback and adjustment module;

[0137] The collision avoidance module receives the output result of the GRU network from the timing feature prediction module, establishes a communication channel between the reader and the tag, designs a collision avoidance strategy, including a time slot division strategy and a binary tree strategy, uses a collision detection algorithm to detect the response of the tag in the communication channel, selects a collision avoidance strategy according to the output and response of the GRU network, and sends the collision avoidance strategy to the real-time feedback and adjustment module;

[0138] The real-time feedback and adjustment module receives the output results of the GRU network from the timing feature prediction module and the collision avoidance strategy from the collision avoidance module. The central processing unit continuously monitors the communication channel and dynamically adjusts the parameters of the GRU network and the parameters of the collision avoidance strategy according to the response situation.

[0139] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0140] Although 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 the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0141] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A RFID multi-tag identification optimization method, characterized in that: The method comprises the following steps: Step S1: Data collection, using a central processor to construct a tag history response data set based on the tag information received by the reader, the tag history response data set includes the tag's timing response, signal strength, and response time. The reader is a device that sends an activation signal to the tag and receives tag information. The tag is an electronic device that stores and transmits information. The tag information consists of a tag ID and a user data area. Step S2: Data serialization: convert the label history response data set into time series data in units of time steps to obtain the converted label history response data set. Randomly select 70% of the converted label history response data set as a training set and 30% as a test set. Step S3: Time series feature prediction, building a GRU network, identifying the nonlinear relationship between time series data, and predicting the label type and response probability of the output in the future time series; Step S4: collision avoidance, establishing a communication channel between the reader and the tag, designing a collision avoidance strategy, including a time slot division strategy and a binary tree strategy, using a collision detection algorithm to detect the response of the tag in the communication channel, and selecting a collision avoidance strategy based on the output and response of the GRU network; Step S5: Real-time feedback and adjustment, the central processor continuously monitors the communication channel and dynamically adjusts the parameters of the GRU network and the parameters of the collision avoidance strategy according to the response situation; In step S3, the time series feature prediction specifically includes the following steps: Step S31: Network construction, constructing a GRU network to identify the nonlinear relationship between time series data. The GRU network consists of an update gate, a reset gate, and a hidden state. The GRU network contains multiple cell units. The hidden state of each cell unit is used as the input data of the next cell unit. The initial input data is time series data. The steps are as follows: Step S311: Update gate construction, using the hidden state of the previous time step and the input data of the current time step to construct the update gate, the formula used is as follows: ； In the formula, represents the update gate, represents the sigmoid function, represents the current time step, represents the weight of the update gate, represents the bias value of the update gate, represents the hidden state at the previous time step, Represents input data; Step S312: Reset gate construction, using the hidden state of the previous time step and the input data of the current time step to construct the reset gate, the formula used is as follows: ； In the formula, Represents the reset gate, represents the sigmoid function, Represents resetting the weight of the gate, Indicates the bias value of the reset gate; Step S313: Candidate hidden state: The candidate hidden state is calculated based on the hidden state of the previous time step, the input data of the current time step and the reset gate. The formula used is as follows: ； In the formula, represents the candidate hidden state, represents the tanh function, represents the hidden weight, represents the hidden bias value, Represents element-wise multiplication; Step S314: Hidden state update: Use the hidden state of the previous time step, the update gate and the candidate hidden state to update the hidden state. The formula used is as follows: ； In the formula, Represents the updated hidden state; Step S32: Add an output layer and construct a fully connected layer as the output layer. Add it after the last hidden state in the GRU network. Use the softmax function to convert the output result of the last hidden state into a category probability, and combine the category probability of each input data into a category probability vector. The formula used is as follows: ； ； In the formula, represents the predicted value, represents the output vector of the last hidden state, represents the output weight, Indicates the output bias value, represents the softmax function, The output vector of the last hidden state elements, Indicates the number of categories of labels, Represents all elements in the output vector of the last hidden state The index and; Step S33: Model training, inputting the training set into the GRU network for forward propagation, and calculating the predicted value of the GRU network; Step S34: Loss function calculation, using the cross entropy loss function to calculate the difference between the predicted value and the actual value of the GRU network, the formula used is as follows: ； In the formula, represents the cross entropy loss function, is the actual value, is the predicted value; Step S35: Back propagation, using the gradient descent algorithm to update the parameters of the GRU network, including the update gate, reset gate, hidden state, output layer weights and bias values. The formula used is as follows: ； In the formula, represents the updated parameters, represents the original parameters, represents the learning rate; Step S36: Model testing: input the test set into the GRU network and use the confusion matrix to calculate the classification accuracy of the GRU model. The formula used is as follows: ； ； In the formula, They are true positive examples, false negative examples, false positive examples, and true negative examples in the classification result confusion matrix. Indicates the model's ability to correctly predict positive examples, Indicates the model's ability to incorrectly predict positive examples; In step S4, the collision avoidance specifically includes the following steps: Step S41: establishing a communication channel between the reader and the tag, the communication channel being used to transmit the query request and activation signal sent by the reader and the response and tag information sent by the tag; Step S42: Preliminary strategy selection: a collision avoidance strategy is preliminarily selected according to a threshold set based on the number of tags predicted by the GRU network. The collision avoidance strategy includes a time slot partitioning strategy and a binary tree strategy. If the number of tags is less than the threshold, the time slot partitioning strategy is used; if the number of tags is more than the threshold, the binary tree strategy is used. Step S43: time slot division strategy, divide the time into fixed time slots, the reader and the tag communicate in each time slot, perform collision detection after the time slot ends, set the maximum number of retransmissions, and retransmit if a collision is detected. If a collision is still detected when the maximum number of retransmissions is reached, use the binary tree strategy to query; Step S44: Binary tree strategy, constructing a binary tree to divide the tag set into small subsets, and using a step-by-step query method to locate the tag position.

2. The RFID multi-tag identification optimization method according to claim 1, characterized in that: In step S43, the time slot division strategy includes the following steps: Step S431: The time slot length is determined according to the transmission time of the tag and other overheads in the communication process, and the time slot length is adjusted according to the utilization of the system. The formula used is as follows: ； ； In the formula, Indicates the time slot length, Indicates the data length, Indicates the transmission rate of the tag, Indicates other overhead in the communication process, Indicates the utilization rate of the system. represents the probability of tag sending, Indicates the number of tags; Step S432: counter initialization, designing a time slot counter, and initializing the time slot counter to 0; Step S433: Data transmission. At the beginning of each time slot, the reader and the tag communicate. The reader sends a query request to nearby tags to check whether there is a tag that wants to transmit data. If a tag senses the query request and is ready to transmit data, it sends a response to the reader. After receiving the response, the reader sends an activation signal to the tag to activate the tag. After the tag is activated, it transmits the tag information to the reader through the communication channel. Step S434: updating the time slot counter. When a time slot ends and data transmission is completed, the time slot counter is incremented by 1 to point to the next time slot. Step S435: Collision detection. After each time slot ends, if the reader recognizes that there are multiple tag information, it is determined that a collision occurs. The reader sends a collision notification to the tag. After receiving the collision notification, the tag retransmits in the next time slot. After reaching the maximum number of retransmissions, the binary tree strategy is used to query the tag ID. Step S436: loop, repeating steps S433 to S435 until no tag responds or the maximum number of retransmissions is reached, and the loop ends.

3. The RFID multi-tag identification optimization method according to claim 2, characterized in that: In step S44, the binary tree strategy includes the following steps: Step S441: The root node is determined, a binary bit is selected from the ID of the tag as the ID specific bit, and the tag set is divided into two subsets according to the ID specific bit; Step S442: internal node division, reselecting a binary bit as the ID specific bit in the divided subset, dividing the subset into two smaller subsets according to the ID specific bit, and further dividing each subset into smaller subsets step by step until only one binary bit is left in the final subset, thus forming a binary tree; Step S443: Leaf node identification. After the binary tree is constructed, each leaf node is uniquely identified by a label ID; Step S444: Determine the tag position. After a collision occurs during communication between the reader and the tag, the reader traverses the binary tree according to the tag probability predicted by the GRU network, reads the tag ID from the leaf node, and determines the tag position. Step S445: Data reading, after determining the location of the tag, the reader directly reads the user data area from the tag.

4. An RFID multi-tag identification optimization system, used to implement an RFID multi-tag identification optimization method as described in any one of claims 1 to 3, characterized in that: It includes data collection module, data serialization module, time series feature prediction module, collision avoidance module and real-time feedback and adjustment module; The data collection module uses the central processor to construct a tag history response data set based on the tag information received by the reader. The tag history response data set includes the tag's timing response, signal strength, and response time. The reader is a device that sends an activation signal to the tag and receives tag information. The tag is an electronic device that stores and transmits information. The tag information consists of a tag ID and a user data area. The tag history response data set is sent to the data serialization module; The data serialization module receives the label history response data set from the data collection module, converts the label history response data set into time series data in units of time steps, obtains the converted label history response data set, randomly selects 70% of the converted label history response data set as a training set and 30% as a test set, and sends the training set and the test set to the time series feature prediction module; The time series feature prediction module receives the training set and the test set from the data serialization module, constructs a GRU network, identifies the nonlinear relationship between the time series data, predicts the label type and its response probability output in the future time series, and sends the output result of the GRU network to the collision avoidance module and the real-time feedback and adjustment module; The collision avoidance module receives the output result of the GRU network from the timing feature prediction module, establishes a communication channel between the reader and the tag, designs a time slot division strategy and a binary tree strategy, uses a collision detection algorithm to detect the response of the tag in the communication channel, selects a collision avoidance strategy according to the output and response of the GRU network, and sends the collision avoidance strategy to the real-time feedback and adjustment module; The real-time feedback and adjustment module receives the output results of the GRU network from the timing feature prediction module and the collision avoidance strategy from the collision avoidance module. The central processing unit continuously monitors the communication channel and dynamically adjusts the parameters of the GRU network and the parameters of the collision avoidance strategy according to the response situation.

Images