A high-frequency RFID tag identification method based on a hybrid branch tree

Abstract

This invention discloses a high-frequency RFID tag identification method based on a hybrid branch tree, comprising: Step 1, obtaining the cardinality of tags to be identified in the current radio frequency field; Step 2, dynamically selecting the optimal branch query configuration and execution sequence based on the estimated tag cardinality and a preset decision model; Step 3, sending an inventory command according to the configuration determined in Step 2; Step 4, receiving tag responses and judging the time slot status; if a conflict occurs, recursively executing Step 2 to Step 3 for the child nodes generated in the conflicting time slot until all tags are successfully identified; This method, through a hybrid branch tree architecture, cardinality estimation algorithm, and adaptive switching mechanism, can not only significantly reduce the total time slot overhead and identification time while adapting to various dynamic tag environments from extremely sparse to extremely dense, but also is fully compatible with the ISO 15693 standard protocol and can be directly deployed in commercial RFID systems without modifying existing tag hardware.

Description

Technical Field

[0001] This invention relates to the field of radio frequency identification (RFID) technology, and in particular to a high-frequency RFID tag identification method based on a hybrid branch tree. Background Technology

[0002] Radio Frequency Identification (RFID) technology, as a core means of the Internet of Things (IoT) sensing layer, has been widely applied in key areas such as smart warehousing, medical supply management, library inventory, and unmanned retail. Among these, high-frequency RFID systems operating in the 13.56 MHz band, with their advantages of precise magnetic field boundary control and strong penetration into liquids and biological tissues, occupy an important position in near-field dense target identification scenarios. Current commercial high-frequency RFID systems generally comply with the ISO 15693 international standard. In this context, efficient RFID tag identification methods are crucial. RFID tag identification refers to the process by which a reader uses anti-collision protocols to identify all tags within the radio frequency field and obtain their unique identifiers (UIDs) and user data. This data is the foundation for object authentication, asset tracking, and information management. Large-scale RFID systems mean that numerous RFID tags will simultaneously compete for limited channel resources, leading to increased channel collisions and transmission delays, severely impacting information collection efficiency and consequently failing to guarantee the real-time performance of upper-layer applications.

[0003] Tag identification is the most basic and core operation in RFID systems. Efficient tag identification methods can quickly complete the inventory and verification of massive amounts of goods, ensuring the real-time and accuracy of data in applications such as logistics, retail, and healthcare, thereby significantly improving supply chain management efficiency and reducing operating costs. However, existing commercial high-frequency RFID systems still have significant limitations in addressing these challenges, mainly in the following aspects:

[0004] Existing standard protocols suffer from rigid mechanisms, making it difficult to adapt to dynamically changing tag densities. The Media Access Control (MAC) layer of the ISO 1569 standard employs a deterministic tree-structured anti-collision mechanism, strictly limiting the reader to a fixed time slot configuration during inventory checks. This static configuration cannot adapt to the dynamic changes in the number of tags in real-world scenarios. In densely tagged scenarios, a single-time-slot configuration results in frequent signal collisions within the channel due to the narrow search branches, leading to a large number of time slots being in a conflict state, causing a sharp decrease in algorithm convergence speed and limiting system throughput. In sparsely tagged scenarios, a 16-time-slot configuration requires the reader to scan all 16 time slots completely. When there are few tags, this results in a large number of time slots being in an unresponsive "idle state," causing a serious waste of valuable time and bandwidth resources.

[0005] Although numerous optimization algorithms have been proposed in the field of RFID collision avoidance by academia and industry, the vast majority of research results focus on the UHF band. UHF collision avoidance algorithms heavily rely on the unique physical layer mechanisms of UHF protocols and cannot be directly applied to the high-frequency domain. In fact, high-frequency RFID primarily relies on near-field magnetic induction coupling, possessing physical advantages unmatched by UHF, such as precise magnetic field boundary control and strong penetration of liquids and biological tissues. It is the only reliable solution in scenarios rich in liquids or with highly dense physical spaces, such as medical supply management and library inventory management. However, existing commercial high-frequency RFID tags generally follow the ISO 15693 international standard, whose media access control layer strictly stipulates that only a fixed 1-slot or 16-slot configuration can be used, and does not support low-level bit-level collision detection. The method provided in this invention overcomes the rigid technical bias of high-frequency standard protocol mechanisms. Without changing the existing high-frequency RFID underlying physical mechanism or requiring customization or modification of the tag chip hardware design, it cleverly utilizes standard instruction sequences (such as the Application Family Identifier (AFI) mechanism and mask extension) to achieve seamless integration of hybrid branch tree logic in high-frequency scenarios, filling the technical gap in dynamic anti-collision scheduling for high-frequency RFID. Summary of the Invention

[0006] Purpose of the invention: The technical problem to be solved by the present invention is to provide a high-frequency RFID tag identification method based on a hybrid branch tree, which addresses the shortcomings of the existing technology.

[0007] To address the aforementioned technical problems, this invention discloses a high-frequency RFID tag identification method based on a hybrid branch tree, comprising the following steps:

[0008] Step 1: Obtain the cardinality of tags to be identified in the current radio frequency field. For the initial environment where the number of tags is unknown, a cardinality estimation algorithm based on a single round of 16-slot frame observation is used. The tag density is inverted by statistically analyzing the number of idle time slots in the observation frame to obtain the estimated tag cardinality value.

[0009] Step 2: Based on the estimated tag cardinality and a pre-defined decision model, dynamically select the optimal branch query configuration and execution sequence. The decision model incorporates physical layer time parameters (including frame establishment time, idle time slots, singleton time slots, and conflict time slot durations) to construct an accurate time cost model, and derives the optimal dynamic switching threshold and performance range accordingly.

[0010] 1) When the number of tags is less than or equal to the switching threshold (low-density scenario), select the 2-branch configuration (corresponding to the 1-slot configuration of ISO15693 standard) to reduce idle slots by utilizing its depth-first feature and fast pruning capability;

[0011] 2) When the number of tags is greater than the switching threshold and is within the optimal performance range of the 16-branch configuration (medium-density scenario), select the 16-branch configuration (corresponding to the 16-slot configuration of the ISO 15693 standard) and use its breadth-first characteristic to quickly divert conflicting tags.

[0012] 3) When the number of tags exceeds the optimal performance range (high-density large-scale scenarios), a generalized delay strategy based on sequence optimization is adopted. Several layers of 2-branch configuration with 1 time slot are pre-distributed. After the subset size falls back to the optimal range, the configuration is switched back to 16-branch configuration for parallel processing.

[0013] Step 3: Send inventory instructions according to the configuration and sequence determined in Step 2. For 16-branch nodes, perform breadth-first search, mapping to 16 time-slot frames for parallel splitting; for 2-branch nodes, perform depth-first search, mapping to a 1-time-slot frame sequence implemented through mask expansion for binary search.

[0014] Step 4: Receive tag responses and determine time slot status. If a conflict occurs, generate a new child node for the conflicting time slot and recursively execute steps 2 to 3 for that child node until all tags are successfully identified.

[0015] The method for estimating the cardinality of labels in step 1 includes:

[0016] Step 1-1: Perform a standard 16-slot inventory as an observation frame; count the number of idle time slots n in this observation frame. e Number of conflicting time slots n c Number of successfully read time slots n s ;

[0017] Steps 1-2: Construct a mapping relationship between the number of idle time slots and the total number of tags using the method of moments estimation, through the formula... The estimated value of the label cardinality was calculated. ;

[0018] Where, if time slot n e = 16, determining the total number of tags. If the value is 0, the inventory count ends; if time slot n e = 0, indicating system saturation, and adopting a generalized delay strategy based on sequence optimization to pre-split the entire tag set.

[0019] Step 2 specifically involves: based on the label cardinality n (i.e., the calculated estimated value) The system dynamically selects between a 16-branch query configuration and a 2-branch query configuration using a preset decision model. The 16-branch query configuration corresponds to the 16-slot configuration in the ISO 15693 standard, and the 2-branch query configuration corresponds to the 1-slot configuration in the ISO 15693 standard.

[0020] The decision model described in step 2 specifically involves: theoretically calculating the critical point at which the total expected time of a 16-branch configuration is equal to the total expected time of a 2-branch configuration, i.e., the time switching threshold n. time The optimal branch query configuration and execution sequence are dynamically selected based on the time switching threshold and the pre-split strategy's splitting threshold.

[0021] The dynamic selection of the optimal branch query configuration and execution sequence specifically includes:

[0022] Strategy A: When n time When selecting a configuration branch, choose 2 branches.

[0023] Strategy B: When When setting the traffic splitting threshold, select the 16-branch configuration;

[0024] Strategy C: When When the traffic splitting threshold is reached, a generalized delay strategy is adopted.

[0025] The generalized delay strategy first forces the execution of a k-layer 2-branch configuration for pre-distribution, where k = (0, Using a k-layer 2-branch structure as a dilution layer, after the subset size falls back to within the diversion threshold, switch to the instant 16-branch configuration in strategy B.

[0026] The optimal delay layer k for pre-splitting is based on the formula... (n) = max(0, ) Sure.

[0027] The decision model described in step 2 includes a correction strategy based on physical layer time overhead: introducing physical layer time parameters, including frame establishment time t. f Singleton time slot duration t s Idle time slot duration t e and the duration t of the conflict slot c Construct the time gain function.

[0028] We need to accurately estimate the identification overhead of the 2-branch configuration under different numbers of labels. To this end, this study defines an expected time slot function S(n, d), which indicates the expected number of time slots required to resolve n labels under the constraint of the maximum remaining search depth d. This function not only provides a theoretical benchmark for analyzing the query efficiency of the 2-branch configuration but also serves as a core mathematical tool for determining the advantage interval of the 16-branch strategy. We use a recursive analysis method to rigorously define S(n, d), first establishing the boundary conditions of the recursion:

[0029] 1) When d=0, it means that the mask length has reached the maximum number of bits of the tag UID, and the system cannot continue to issue more refined query instructions. At this time, no matter how many tags exist in the coverage area, the reader's ability to distinguish at the physical level has reached its limit, and the recursive process will be forcibly terminated. Therefore, S(n, 0)=0 is defined.

[0030] 2) Single-label recognition condition (n=1): When a query node contains only a single-label response (i.e., n=1), it means that the conflict has been completely resolved, and the label has been successfully recognized. Since no additional split query operation is required, the additional time slot overhead is zero, so S(1, d)=0 is defined.

[0031] 3) Idle slot processing rule (n=0): If there is no tag response under the current mask (i.e., n=0), it indicates that the branch is empty. The system will skip this branch directly without building an additional subtree, so its subsequent overhead is zero, i.e., S(0,d)=0.

[0032] For multi-tag collision scenarios, the system needs to initiate a recursive splitting mechanism to resolve conflicts. The reader precisely divides the n colliding tags into two mutually exclusive subsets by applying an additional one-bit mask to the most significant bit: a left subset with the most significant bit being bit '0' and a right subset with the most significant bit being bit '1'. Under the ideal channel assumption, we assume that the bits of the tag UID follow a uniform distribution, meaning that each tag independently and equally (with a probability of 0.5) chooses either the '0' branch or the '1' branch. This random allocation process statistically strictly follows a binomial distribution model with parameters (n, 0.5). Specifically, the probability distribution p(k) that exactly k tags fall into the '0' prefix group (left subtree) among the n colliding tags can be expressed as follows:

[0033]

[0034] During each 2-branch configuration operation, regardless of how the tags are assigned, the system must consume two basic time slots to identify the two tag subsets, corresponding to two inventory operations with send masks of "0..." and "1...", respectively. This inherent overhead is represented by a fixed constant term—'2'—in the recursive formula. After the initial partitioning is completed, the system will recursively process the left subset containing k tags and the right subset containing nk tags under the constraint of the reduced search depth d-1. Based on the law of total probability and the linear property of expectation, the recursive relationship of the expectation function S(n, d) is shown in the following formula:

[0035]

[0036] The formula precisely describes the recursive nature of the binary conflict resolution process. The summation term in the formula represents the weighted sum of the expected future costs of the left and right subtrees for all possible label assignments, from 0 to n, while the '2' on the right represents the splitting cost at the current level.

[0037] After obtaining the actual execution time for each type of time slot, we use it in a theoretical model to calculate the overall execution time more accurately, rather than simply comparing the number of time slots required.

[0038]

[0039] The execution time is calculated as shown in the formula above, where, Expected execution time configured for 2 branches, The parameter represents the cumulative cost of frame establishment. , and These represent the cumulative expected number of singleton time slots, idle time slots, and conflict time slots during the entire binary tree traversal process. Expected execution time configured for 16 branches, parameters The parameter represents the expected number of singleton time slots. This represents the expected number of idle time slots, a parameter. This represents the expected number of conflict slots. = ,and According to the formula, when the number of tags exceeds a certain number, the time required to search forward 4 bits in a single time slot exceeds the time required for a complete round of inventory consisting of 16 time slots, which is the time switching threshold and the diversion threshold.

[0040] Step 3 specifically involves constructing the corresponding ISO 15693 directive based on the strategy determined in Step 2, performing the inventory count, and handling conflicts.

[0041] Step 3-1: If the 16-branch configuration is selected, the reader sends an inventory command, sets the Slot flag in the flag field to 0, and attaches the current mask. The tag is mapped to the 0-15 time slot response according to the UID hash.

[0042] Step 3-2: If the 2-branch configuration is selected, the reader sends an inventory command, sets the Slot flag in the flag field to 1, and sends two commands consecutively by appending '0' and '1' to the current mask respectively, logically simulating a binary tree split.

[0043] Step 3-3: For the time slot where a conflict occurs, mark it as a new child node to be processed. For the set of tags in the child node, recursively execute steps 2 to 3 until all branches are resolved to singleton or idle.

[0044] The method is fully compatible with the ISO 15693 communication protocol and achieves rapid reading of all tags solely through media access control layer scheduling at the reader end.

[0045] Beneficial effects:

[0046] This method, through a hybrid branch tree architecture, cardinality estimation algorithm, and adaptive switching mechanism, can not only significantly reduce total time slot overhead and identification time while adapting to various dynamic tag environments ranging from extremely sparse to extremely dense, but also is fully compatible with the ISO 15693 standard protocol and can be directly deployed in commercial RFID systems without modifying existing tag hardware. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the hardware deployment of a high-frequency RFID system according to an embodiment of the present invention.

[0048] Figure 2 This is an overall flowchart of the label recognition method based on a hybrid branch tree provided in an embodiment of the present invention.

[0049] Figure 3 This is a schematic diagram of the tag base estimation process based on a single round of 16 time slot observations in an embodiment of the present invention.

[0050] Figure 4 This is a hybrid branch tree strategy based on the physical layer time-series decision model in this embodiment of the invention.

[0051] Figure 5 This is a schematic diagram comparing the slot overhead of three strategies for large-scale tag scenarios (e.g., n = 32) in an embodiment of the present invention.

[0052] Figure 6 This is a schematic diagram of the inventory instruction execution and conflict recursive processing flow in an embodiment of the present invention.

[0053] Figure 7 This is a comparison chart of tag reading times under different strategies in a real library scenario.

[0054] Figure 8 This is a comparison chart of the execution time overhead of different strategies under different tag sizes in embodiments of the present invention. Detailed Implementation

[0055] This invention constructs a complete prototype high-frequency passive RFID system based on a commercial hardware platform to verify the actual performance of the proposed hybrid branch tree architecture. The integrated architecture of the hardware platform and software system is shown in Figure 1. The experimental platform uses two industrial-grade reader devices: the FEIG ELECTRONIC LRM 5400 and LRM 2500-B series reader modules, both of which strictly comply with the ISO 15693 standard and have wide applicability in commercial high-frequency RFID deployments. The tags used are ICODE SLIX series high-frequency passive tags manufactured by NXP Semiconductors, with an operating frequency of 13.56 MHz, comprising a configuration area and a storage area. The configuration area stores an 8-byte unique identifier, a 1-byte Application Family Identifier (AFI), and a 1-byte Data Storage Format Identifier (DSFID). The storage area contains several bytes of user data blocks, which can be written with specified data. All read and write operations are performed using a single-loop antenna (model: FEIG ANT 310) with adjustable gain (0-30dBm). The antenna reading distance is precisely calibrated within the range of 0.5-1.0 meters depending on the application scenario. The core logic of the protocol is implemented in C# 10.0 and developed using the .NET 6.0 framework, ensuring compatibility with existing commercial high-frequency RFID middleware, avoiding the need for underlying hardware modifications, and providing a scalable technical path for practical deployment. This invention discloses a high-frequency RFID tag identification method based on a hybrid branch tree, the method comprising: a high-frequency RFID tag identification method based on a hybrid branch tree, such as... Figure 2 As shown:

[0056] Step S1: Use a cardinality estimation algorithm based on single-round 16-slot frame observation to estimate the size of the tags to be identified in the current radio frequency field and obtain the estimated tag cardinality value;

[0057] Step S2: Based on the estimated cardinality of the tags, a hybrid branch tree strategy based on a physical layer time series decision model is used to dynamically select the optimal query configuration (16 branches or 2 branches) and execution sequence.

[0058] Step S3: Send an inventory command according to the determined configuration, perform inventory and conflict handling on the target tags, and recursively execute the above steps until all tags are identified.

[0059] Tag identification specifically involves the reader interacting with the tags to parse out the unique identifier (UID) of all tags within the radio frequency field. For simplicity, this method is based on the ISO 15693 standard protocol, where the 16-branch configuration corresponds to the 16-slot configuration in the standard, and the 2-branch configuration corresponds to the 1-slot configuration (implemented through mask bit extension). Detection scenarios are mainly divided into two categories: static scenarios with a known number of tags and dynamic scenarios with an unknown number of tags. This invention focuses on solving the efficiency optimization problem in the initial state where the number of tags is unknown (i.e., non-preset number scenarios). This method first performs step S1 to estimate the tag size within the radio frequency field, specifically by sending a round of "observation frames" to count the number of idle time slots and deduce the total number of tags. The core logic of the cardinality estimation algorithm is based on a probability mapping model of idle time slots. In the 16-slot inventory of the ISO 15693 protocol, it is assumed that the probability of a tag selecting any time slot is equal. As the total number of tags n increases, the probability of any time slot being selected increases, and the probability of remaining "idle" increases. This then decays exponentially. Therefore, the number of idle time slots, n e There is a strict monotonically decreasing relationship between the number of tags and the total number of tags, n. This method utilizes the concept of moment estimation, treating the observed number of idle time slots as the theoretical expected value. Through inverse function derivation, the tag density in the current radio frequency field can be inverted in real time with extremely low computational complexity O(1), providing a quantitative basis for subsequent branch switching. For example... Figure 3 As shown, the cardinality estimation algorithm based on single-round 16-slot frame observations in step S1 includes:

[0060] Step S11: The reader sends a standard inventory command, sets the time slot number flag to 16 time slots, but does not need to parse the specific UID data; it is only used to observe the channel status.

[0061] Step S12: Count the number of idle time slots that appear in the observation frame, and denote it as n. e An idle time slot refers to a time slot in which no subcarrier signal is detected within a specified waiting period.

[0062] Step S13: Calculate the cardinality estimate of the labels using the mapping model constructed by the method of moments. At the same time, the local collision density is obtained.

[0063] The specific formula for estimating the label technology in step S13 is as follows: If n e =16, judge = 0, the reader directly ends the inventory process; if n e= 0, indicating that the system is in a saturated state (usually n > 95), and a pre-splitting strategy for large-scale tag sets is directly adopted.

[0064] After estimating the tag size, the next step is to dynamically select the optimal query configuration in step S2. This process primarily utilizes the physical layer time overhead model to balance "protocol interaction overhead" and "time slot conflict overhead." The hybrid branch tree strategy based on the physical layer time-series decision model in step S2 is as follows: Figure 4 As shown:

[0065] Step S21: Obtain the physical layer time parameters under the current hardware environment, including the frame establishment time t. f Singleton slot time t s Conflict slot time t c and idle time slot t e .

[0066] In the actual deployment and data interaction of commercial RFID readers, the data acquisition process is typically divided into two modes: Host Mode and Buffer Mode. In Host Mode, the reader is completely controlled by the host computer and transmits the data acquired from the tags to the host system for processing in real time. However, for the FEIG series readers used in this experiment, the standard inventory command response in Host Mode often only contains basic information such as the tag UID, making it difficult to provide precise timing details of the physical layer communication process. In Buffer Mode, the reader temporarily caches tag data using its internal memory and, with the help of its built-in high-precision timing mechanism, fully records detailed metadata including tag response time, signal strength, and various time slot events. After the inventory cycle ends or the buffer threshold is triggered, these data packets containing timing parameters are uploaded to the host in batches. Given the need for precise time measurement in this invention, the FEIG reader's Buffer Mode is chosen as the experimental benchmark. This choice not only overcomes the functional limitations of the standard Host Mode in timestamp acquisition but also fully utilizes the high throughput advantage brought by its batch processing mechanism, thereby significantly improving the completeness of experimental data and the usability of timing analysis.

[0067] In the physical layer time parameter model established in this invention, we assume that the 1-slot configuration and the 16-slot configuration have the same frame establishment time t. fThis assumption is based on the physical layer specification of the ISO 15693 standard: the inventory command initiated by the reader has a fixed frame structure (including frame header, flag bits, command code, mask, and check bits). The switching of the number of time slots is determined only by the state of the 5th bit in the flag field (0 for 16 time slots, 1 for 1 time slot). Since the state flipping of the flag bit does not change the total bit length of the command frame, the physical transmission time of the command in the radio frequency channel is strictly constant. Although there may be slight differences in the firmware preprocessing time inside the reader under different configurations, compared with the millisecond-level over-the-air transmission overhead, this difference is a high-order small quantity and can be ignored in engineering calculations. Therefore, t f Treating it as a system fixed invariant independent of time slot configuration not only simplifies the model complexity but also conforms to the actual characteristics of physical layer communication.

[0068] To accurately obtain the execution time parameters of the physical layer, we designed a systematic measurement scheme on the LRM 5400 commercial reader. Because this device uses a frame-level encapsulation mechanism in a 16-slot configuration, all slots execute continuously at the physical layer but only return complete frame data via API, making it impossible to directly separate the time overhead of individual slots through the software interface. Therefore, we adopted a differential measurement method: by systematically controlling the number and distribution of tags, we constructed a controlled scenario containing different combinations of slot types, and used the difference in the execution time of the complete frame to establish a linear constraint equation, thereby calculating the precise values ​​of the time parameters.

[0069] The specific measurement process is as follows: First, five tags with known UIDs are deployed for a 16-slot inventory. Based on the known tag UIDs, the distribution of each slot type can be determined. The experiment consists of two rounds of continuous inventory cycles: In the first round, three tags were successfully identified, and a tag conflict was detected in one slot. That is, the 16 slots in the first round included three singleton slots, one conflicting slot, and 12 idle slots, and the completion time was T. round1 The second round successfully identified the remaining 2 tags without any conflicts, meaning it included 2 singleton time slots, 0 conflicting time slots, and 14 idle time slots, completing at time T. round2 The time difference between the two rounds of 16-slot execution T = T round2 - T round1The time difference between singleton time slots and idle time slots can be separated, and the total execution time of the frame is finally measured to be approximately 210 ms. Subsequently, keeping the distribution of time slot types in the first round of inventory unchanged, the number of singleton time slots in the second round of inventory is increased by adjusting the label configuration, i.e., the number of idle time slots is reduced accordingly. Experiments show that for each additional singleton time slot, the execution time of a 16-slot frame increases by an average of 43 ms, from which t can be obtained. s – t e = 43 ms. Using a similar method, the experiment kept the number of inter-frame time slots constant and found that when idle time slots were replaced with conflicting time slots, the total frame time did not change significantly. This indicates that in the processing logic of this device, the duration of conflicting time slots and idle time slots is basically the same, that is, it satisfies t c – t e 0.

[0070] To further obtain the frame establishment time t f We designed a single-tag response experiment. By deploying tags with specific UIDs, we enabled the 1-slot inventory process to execute along a preset path. We measured the response times of tags tag1 and tag2 under different scenarios, denoted as T. tag1 and T tag2 And calculate its time interval. T = T tag1 - T tag2 Experiments revealed a fixed average time difference of approximately 15 ms between adjacent scenes, stemming from the additional overhead required for frame establishment. To ensure measurement accuracy, all experiments were repeated 100 times, and the average value was taken as the final result. Based on the above measurement results, we established a complete physical layer time parameter model, as shown in the following set of equations:

[0071]

[0072] Finally, after system calibration, the specific time slot parameters of the LRM 5400 reader used in this experiment can be approximated as: frame establishment time t f = 8 ms, Singleton slot duration t s = 50 ms, duration of conflict slot t c = 7 ms, idle time slot duration t e = 7 ms. These precisely measured parameters provide crucial input for subsequent time efficiency analysis, ensuring a high degree of consistency between the theoretical model and the actual system performance evaluation. Substitute these parameters into the following formula for calculation:

[0073]

[0074] It should be noted that the physical layer time parameters listed in this embodiment (such as frame establishment time t) f 8ms, Singleton slot time t s The 50 ms (etc.) values ​​are merely practical examples based on specific commercial hardware (such as the FEIG LRM 5400 reader) and are not intended to limit the numerical value of the technical solution of this invention. In actual industrial applications, differences in the hardware architecture or underlying firmware versions of RFID readers from different manufacturers will inevitably lead to fluctuations in the internal processing latency of the device. To address this, this invention provides a widely applicable differential measurement method to dynamically acquire the aforementioned parameters: by controlling the response of known tags in specific time slots, test frames with different combinations of idle, singleton, and conflict time slots are constructed, and the overall physical duration of each test frame is recorded; subsequently, based on the time difference of these different state frames, a system of linear equations is constructed and solved, thereby separating and calculating the precise time parameters (t) adapted to the current hardware environment without needing to crack the reader's underlying firmware. f t s t c t e This measurement method is independent of specific equipment models, and any high-frequency RFID system that conforms to the ISO 15693 standard can use this method to complete the self-calibration of physical layer parameters.

[0075] Step S22: Theoretically calculate the critical point at which the total expected time of the 16-branch configuration is equal to the total expected time of the 2-branch configuration, i.e., the time switching threshold n. time , and the pre-splitting strategy's splitting threshold. Its calculation method is as follows: (1) When the time switching threshold is reached, it means that the two configurations have the same expected time for backward counting of 4 bits. That is... =0, substituting the calculated time parameters into the formula, the theoretical threshold for time switching can be calculated as 4. (2) When the pre-splitting strategy's splitting threshold is reached, it means that the expected time of directly using the 16-branch configuration is the same as the expected time of first using the 2-branch configuration and then executing the 16-branch configuration (i.e., delaying the 16-branch configuration by one layer). This is equivalent to the two strategies saving the same amount of time compared to executing the 5-layer 2-branch configuration, i.e. Similarly, substituting the calculated time parameters into the formula, the theoretical pre-shunting threshold can be calculated to be 18. Then, we substituted the actual duration data into tag inventory scenarios of different scales to obtain the optimal inventory switching threshold. The theoretical results are consistent with the actual verification, and finally the time switching strategy on the typical commercial hardware FEIG LRM 5400 is obtained: pre-shunting threshold of 18 and time switching threshold of 4.

[0076] Step S23: Based on the relationship between the estimated tag size, the threshold, and the sequence optimization boundary, select different strategies to conduct tag inventory.

[0077] Strategy A (Low-Density Fast Convergence): When the estimated cardinality of the labels... At time 4, select the 2-branch configuration (1-slot configuration). In this case, the 1-slot configuration has the characteristics of no additional frame establishment overhead and can quickly skip idle nodes, which is the most efficient.

[0078] Strategy B (Medium-Density Parallel Processing): When 4 At 18:00, select the 16-branch configuration (16-timeslot configuration and several rounds of 1-timeslot configuration). At this time, the parallel splitting capability of 16 time slots can be utilized to minimize the probability of collisions.

[0079] Strategy C (High-Density Sequence Optimization): When the tag cardinality estimate is... > At 18:00, a "generalized delay strategy" is adopted. This means that a pre-distribution of the traffic is first enforced using a k-layer 2-branch configuration, where k = (0, ... A k-level 2-branch structure is used as a "dilution layer." Because the 2-branch structure divides the set in two, each recursive level halves the expected number of tags under a child node. The system pre-divides the large-scale tag set into several appropriately sized subsets by calculating the optimal delay layer k. Once the subset size falls below 18, the system switches to the immediate 16-branch configuration in strategy B. Figure 5 As shown, in a large-scale scenario with n = 32 tags, this method can significantly reduce slot overhead compared to a single strategy.

[0080] Step S3 will construct the corresponding ISO 15693 directive based on the strategy determined in step S2, perform the inventory count, and handle conflicts, such as... Figure 6 As shown:

[0081] Step S31: If the 16-branch configuration is selected, the reader sends an inventory command, setting the Slot flag in the flag field to 0 (representing 16 slots), and attaching the current mask. The tag is mapped to slot 0-15 response based on the UID hash.

[0082] Step S32: If the 2-branch configuration is selected, the reader sends an inventory command, sets the Slot flag in the flag field to 1 (representing 1 slot), and sends two commands consecutively by appending '0' and '1' to the current mask respectively, logically simulating a binary tree split.

[0083] Step S33: For a time slot where a conflict occurs, the system marks it as a new child node to be processed. For the tag set within this child node, steps S2 to S3 are executed recursively until all branches are resolved to singletons or idle.

[0084] To verify the effectiveness of this method in a real physical environment, a typical "single-layer bookshelf in a library" experimental scenario was constructed. The experimental hardware platform used an industrial-grade FEIG LRM 5400 reader and 100 NXP ICODESLIX2 tags. To ensure the statistical robustness and scenario generalization ability of the experimental results, a random sampling method with replacement was employed. Different numbers of tags (n = 8, 12, 16) were randomly selected from the total sample pool of the 100 tags to generate tag sets of different sizes, which were then deployed on standard book carriers to simulate the tag distribution characteristics of a library bookshelf under different borrowing states. This sampling strategy effectively preserves the original distribution characteristics of the tag UIDs, while accurately reproducing the distribution characteristics of densely labeled areas in the real scenario through repeated sampling. Furthermore, this invention uses the fixed 16-slot and 1-slot configurations in the current ISO 15693 standard as benchmarks to verify the performance of this method in scenarios with different tag sizes.

[0085] This embodiment does not address a static inventory scenario where the number of tags is known in advance, but rather a blind identification scenario that conforms to the ISO 15693 standard, where the number of tags is unknown before inventory and changes dynamically during borrowing, returning, or relocation. In this scenario, while existing fixed 1-slot or fixed 16-slot inventory methods can be directly deployed in commercial systems, they struggle to simultaneously achieve efficiency: when tags are sparse, the 16-slot method generates a large number of idle slots; when tags are dense, the 1-slot method causes numerous conflicts and increases recursive query latency. On the other hand, while some existing optimization methods can improve recognition performance, they often rely on underlying physical layer signals, dedicated tag status mechanisms, or non-ISO 15693 standard capabilities, making them difficult to directly apply to closed-source commercial off-the-shelf (COTS) readers.

[0086] To address the aforementioned technical problem of "difficulty in balancing deployability and recognition efficiency," this invention proposes an adaptive blind inventory method for commercial ISO 15693 systems. First, the current tag size is estimated using a single round of 16-slot observation frames. Then, a physical layer time overhead model is constructed by combining frame establishment time, idle time slot time, single-instance time slot time, and conflict time slot time. Based on this model, an adaptive selection is made between 2-branch query, 16-branch query, and delayed pre-splitting sequence, thereby achieving a better total recognition time under unknown tag size conditions.

[0087] Furthermore, given that commercially available off-the-shelf reader firmware is closed and typically does not expose the underlying interface for custom hybrid query logic, this embodiment utilizes the Application Family Identifier (AFI) filtering mechanism already present in the ISO 15693 standard to logically divide the subset of tags participating in the current roulette wheel point. This allows the adaptive hybrid branch tree strategy to be deployed without modifying existing tags and the reader's underlying hardware / firmware. Therefore, the technical contribution of this invention lies not in the simple threshold switching itself, but in proposing a high-time-efficiency tag recognition scheme that can be directly implemented to address the problem of "coexistence of dynamic tag density, real physical layer overhead, and commercial closed-source hardware constraints" in the ISO 15693 blind recognition scenario.

[0088] Specifically, the experiment restricts the reader to inventory only a subset of tags with specific AFI values ​​by enabling the reader's AFI filtering mask function, thereby logically dividing the group of tags to be identified into different processing groups. Based on theoretical derivation and simulation results, for the tag size set in this experiment, the optimal tag inventory strategy should include an initial 16-slot configuration and several rounds of 1-slot configurations. Before the experiment, we pre-configured the tags with different AFI values: the AFI of the tags used to perform the initial 16-slot inventory was set to 0x01, and the AFI of the tags used to perform subsequent 1-slot inventory was set to 0x02. Through this logical mapping, the present invention successfully achieved precise control of different branch strategies at the physical level without modifying the reader firmware. To build a rigorous comparison benchmark, the present invention performed standard 16-slot inventory and 1-slot inventory on the same group of tags without applying the AFI filtering mask function (i.e., all tags kept the default AFI value, 0x00). For three different label sizes (n = 8, 12, 16), each experiment was independently repeated 50 times, and the average recognition time was taken as the final performance index to eliminate random errors and ensure the statistical significance of the data.

[0089] Experimental results are as follows Figure 7As shown, the measured data demonstrates that the present invention maintains a significant performance advantage across all test scales. In the scenario with n = 8, the average recognition time for hybrid queries is 574 ms; in the scenario with n = 12, the average recognition time for hybrid queries is 846 ms. Specifically, in the scenario with a full bookshelf and n = 16, the average recognition time for hybrid queries is 1150 ms, which is more time-efficient than the standard 16-slot inventory (1388 ms) and 1-slot inventory (1290 ms). Horizontal comparisons show that compared to the traditional 1-slot baseline scheme, the hybrid branch tree protocol improves time efficiency by 16.08%, 15.57%, and 17.15% at the three scales, respectively; compared to the 16-slot baseline scheme, the time efficiency improves by 8.89%, 9.81%, and 10.85%, respectively. This experiment not only verified the effectiveness of the hybrid branch tree protocol in a controlled laboratory environment, but also directly proved its excellent ability to handle high-concurrency tags and its practical engineering value in real high-frequency RFID application scenarios such as library book inventory.

[0090] To evaluate the theoretical upper limit and scalability of this method in a large-scale tagging environment, this embodiment underwent full-range testing on a simulation platform, with tag sizes covering a continuous range from n = 10 to n = 300, to ensure the universality and reliability of the evaluation results. Figure 8 As shown in the figure. Specifically, compared with the traditional 1-slot and 16-slot configurations, our method demonstrates significant advantages in various scenarios. For example, when the number of tags is 100, the execution times for 1-slot, 16-slot, and our proposed hybrid branch tree tag recognition are 8.57 s, 8.30 s, and 7.20 s, respectively; when the number of tags increases to 200, the corresponding execution times are 17.22 s, 17.05 s, and 13.91 s, respectively. Overall, compared with the baseline scheme, our protocol achieves average time savings of 17.64% and 17.23% at different tag sizes. The above results confirm that the method has excellent scalability: that is, it can achieve good recognition efficiency at any tag size, thus establishing its time performance advantage over the fixed-slot configuration.

[0091] It should be noted that the time switching threshold (e.g., 4) and sequence optimization boundary (e.g., 30) in this embodiment are optimal values ​​derived from physical layer parameters measured on typical commercial hardware (e.g., FEIG readers and NXP tags). In practical applications, although the underlying processing latency varies between different devices, the physical layer time overhead model constructed in this invention has complete universality. Any high-frequency RFID system conforming to the ISO 15693 standard can dynamically calculate the optimal threshold adapted to the current environment by pre-measuring its own hardware time parameters and substituting them into this model, without being bound to a specific hardware model. In addition, the derivation and application of the above thresholds bring two core advantages to the system:

[0092] 1) Micro-level time optimization: The time switching threshold (4) is the strict mathematical intersection of the expected time function of 16 time slots and 1 time slot configuration. It ensures that the system can unconditionally choose the evolution path with the shortest actual time consumption at each branch decision, fundamentally eliminating the game loss caused by the waste of idle time slots and the overhead of establishing redundant frames.

[0093] 2) Macro-level congestion prevention: The boundary threshold (30) precisely defines the optimal performance range of the 16-slot configuration. When the tag size is greater than 30, the collision probability of the 16-slot configuration increases significantly. This invention uses this value as a hard line to trigger the generalized delay strategy, eliminating channel congestion and protocol deadlock that are very likely to occur during large-scale intensive inventory checks, and ensuring that the system maintains efficient convergence even under extremely high load.

[0094] This invention provides a high-frequency RFID tag identification method based on a hybrid branch tree. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment of the invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A high-frequency RFID tag identification method based on a hybrid branch tree, characterized in that, Includes the following steps: Step 1: Obtain the cardinality of the tags to be identified in the current radio frequency field; Step 2: Based on the estimated cardinality of the labels and the preset decision model, dynamically select the optimal branch query configuration and execution sequence; Step 3: Send the inventory command according to the configuration determined in Step 2; Step 4: Receive tag responses and determine time slot status; if a conflict occurs, recursively execute steps 2 to 3 for the child nodes generated by the conflicting time slot until all tags are successfully identified.

2. The high-frequency RFID tag identification method based on a hybrid branch tree according to claim 1, characterized in that, The method for estimating the cardinality of labels in step 1 includes: Step 1-1: Perform a standard 16-slot inventory as an observation frame; count the number of idle time slots n in this observation frame. e Number of conflicting time slots n c Number of successfully read time slots n s ; Steps 1-2: Construct a mapping relationship between the number of idle time slots and the total number of tags using the method of moments estimation, through the formula... The estimated value of the label cardinality was calculated. ; Where, if time slot n e = 16, determining the total number of tags. If the value is 0, the inventory count ends; if time slot n e = 0, indicating system saturation, and adopting a generalized delay strategy based on sequence optimization to pre-split the entire tag set.

3. The high-frequency RFID tag identification method based on a hybrid branch tree according to claim 1, characterized in that, Step 2 specifically involves: based on the label cardinality n, dynamically selecting between a 16-branch query configuration and a 2-branch query configuration using a preset decision model; the 16-branch query configuration corresponds to the 16-slot configuration in the ISO 15693 standard, and the 2-branch query configuration corresponds to the 1-slot configuration in the ISO 15693 standard.

4. The high-frequency RFID tag identification method based on a hybrid branch tree according to claim 3, characterized in that, The decision model described in step 2 specifically involves: theoretically calculating the critical point at which the total expected time of a 16-branch configuration is equal to the total expected time of a 2-branch configuration, i.e., the time switching threshold n. time The optimal branch query configuration and execution sequence are dynamically selected based on the time switching threshold and the pre-split strategy's splitting threshold.

5. The high-frequency RFID tag identification method based on a hybrid branch tree according to claim 4, characterized in that, The dynamic selection of the optimal branch query configuration and execution sequence specifically includes: Strategy A: When n time When selecting a configuration branch, choose 2 branches. Strategy B: When When setting the traffic splitting threshold, select the 16-branch configuration; Strategy C: When When the traffic splitting threshold is reached, a generalized delay strategy is adopted.

6. The high-frequency RFID tag identification method based on a hybrid branch tree according to claim 5, characterized in that, The generalized delay strategy first forces the execution of a k-layer 2-branch configuration for pre-distribution, where k = (0, Using a k-layer 2-branch structure as a dilution layer, after the subset size falls back to within the diversion threshold, switch to the instant 16-branch configuration in strategy B.

7. The high-frequency RFID tag identification method based on a hybrid branch tree according to claim 6, characterized in that, The optimal delay layer k for pre-splitting is based on the formula... (n) = max(0, ) Sure.

8. The high-frequency RFID tag identification method based on a hybrid branch tree according to claim 3, characterized in that, The decision model described in step 2 includes a correction strategy based on physical layer time overhead: introducing physical layer time parameters, including frame establishment time t. f Singleton time slot duration t s Idle time slot duration t e and the duration t of the conflict slot c Construct a time gain function; accurately calculate the overall execution time based on the durations of the various time slots mentioned above: in Expected execution time configured for 2 branches, The parameter represents the cumulative cost of frame establishment. , and These represent the cumulative expected number of singleton time slots, idle time slots, and conflicting time slots during the entire binary tree traversal process. Expected execution time configured for 16 branches, parameters The parameter represents the expected number of singleton time slots. This represents the expected number of idle time slots, parameter This represents the expected number of conflict slots. = ,and ; According to the formula, when the number of tags exceeds a certain number, the time required to search forward 4 bits in a single time slot exceeds the time required for a complete round of inventory consisting of 16 time slots, which is the time switching threshold and the diversion threshold.

9. The high-frequency RFID tag identification method based on a hybrid branch tree according to claim 1, characterized in that, Step 3 specifically involves constructing the corresponding ISO 15693 directive based on the strategy determined in Step 2, performing the inventory count, and handling conflicts. Step 3-1: If the 16-branch configuration is selected, the reader sends an inventory command, sets the Slot flag in the flag field to 0, and attaches the current mask. The tag is mapped to the 0-15 time slot response according to the UID hash. Step 3-2: If the 2-branch configuration is selected, the reader sends an inventory command, sets the Slot flag in the flag field to 1, and sends two commands consecutively by appending '0' and '1' to the current mask respectively, logically simulating a binary tree split. Step 3-3: For the time slot where a conflict occurs, mark it as a new child node to be processed. For the set of tags in the child node, recursively execute steps 2 to 3 until all branches are resolved to singleton or idle.

10. The high-frequency RFID tag identification method based on a hybrid branch tree according to claim 1, characterized in that, The method is fully compatible with the ISO 15693 communication protocol and achieves rapid reading of all tags solely through media access control layer scheduling at the reader end.

Images