A warehouse data management method based on RFID tag data

Abstract

The application relates to the technical field of warehouse logistics informatization control, and discloses a warehouse data management method based on RFID label data, which comprises the following steps: a management server divides a radio frequency identification label event time sequence flow into discrete time windows, calculates a label event density factor, compares the factor with an inventory transaction confidence weight generated by a previous discrete time window to complete comparison and verification, and writes a virtual compensation event with a missing label mark in situ when the factor is lower than a set threshold to complete differential time sequence filling; when the number of interruptions reaches an upper limit of 4, the filling is suspended, a historical cross-section event change rate of adjacent reader nodes is retrieved, the weight is limited in a historical average range when there is no removal of a library transaction, and an asset logic management state attribute is adjusted to update an inventory database; the application effectively blocks the conduction of signal consumption parameters to a management transaction chain, and guarantees the logical consistency of inventory updating.

Description

Technical Field

[0001] This invention relates to a warehouse data management method based on RFID tag data, belonging to the field of warehouse logistics information control technology. Background Technology

[0002] Currently, automated warehouse management systems commonly use radio frequency identification (RFID) technology as a means of asset transfer tracking. The management server receives the RFID response stream captured by the reader, obtains the original event sequence containing device identification and timestamps, and converts it into inventory change records in the management ledger database to synchronously reproduce the flow status of physical assets and administrative transfer layers. This approach is based on the assumption that physical scanning and management transactions directly correspond, requiring the underlying acquisition unit to have high temporal integrity. In high-frequency concurrent traffic environments, the dense stacking of multiple layers of items causes spatial electromagnetic interference at the transfer site. Mediums such as metal and liquids cause wireless signal attenuation and shielding loss, and multipath refraction produces complex waveform scattering. Such interference causes the reader to produce transient missed reads and non-business-related misreads, resulting in a large number of lost events and stray redundancy mixed in with the original response stream. If the management server lacks a data verification shell, physical signal noise will be transmitted to the management transaction logic, causing the asset status time sequence to be reversed and the inventory change records to become logically disordered, resulting in a deviation between the management system's ledger data and the actual status of physical assets.

[0003] To improve accuracy, conventional techniques attempt to increase reader power or node density. However, these approaches not only induce out-of-bounds misreads and increase signal confusion in multipath paths, but also multiply costs and complexity. Hardware deployment limitations have led to a shift in focus towards software control methods. However, these control methods still have shortcomings. For example, Chinese invention patent CN103473574B discloses a non-deterministic RFID data processing method based on temporal relationships, which fills in missed location data based on object movement history and positional relationships. However, this data filling relies on the highly predictable group movement trajectories of items. When faced with high-frequency concurrent traffic and dense multi-layer stacking of items, asset flow is independent and discrete. The missed reads caused by physical shielding present transient bursts, causing the inference mechanism based on historical trajectory to be fundamentally mismatched. Moreover, the use of coarse-grained parent class positions to replace inconsistent positions leads to the deterioration of the precision of account and inventory management, which cannot meet the causal consistency requirements of real-time auditing. Another type of scheme that implements static filtering based on fixed time thresholds fails when faced with continuous response interruptions caused by long-term blocking, because it lacks the ability to verify the spatiotemporal correlation between adjacent nodes, and cannot rebuild the missing data chain.

[0004] Therefore, the technical problem to be solved by this invention is how to construct a discrete transaction orchestration method that can automatically smooth the original scanning noise, accurately resolve concurrent timing conflicts and satisfy the causal relationship of administrative audit, and ensure the consistency of inventory account data under high interference concurrent traffic peaks. Summary of the Invention

[0005] To address the problems in the background art, the technical solution of the present invention is as follows: A warehouse data management method based on RFID tag data, comprising the following steps:

[0006] Step S1: The management server obtains the RFID tag event time stream containing multiple sets of tag identifiers reported by the reader nodes in the warehouse data management system, and divides the RFID tag event time stream into discrete time windows according to the set time step.

[0007] Step S2: The management server calculates the tag event density factor of the RFID tag event time-series stream within the discrete time window, compares the tag event density factor with the confidence weight of the inventory transaction generated in the previous discrete time window to complete the data feedforward comparison verification, and writes a virtual compensation event with the lost tag identifier in place in the current time-series data stream when the tag event density factor is lower than the set sudden drop threshold to complete the differential time-series filling to build the self-calibration data feedforward verification logic relationship.

[0008] Step S3: When the number of consecutive interrupted windows of RFID tag events within the discrete time window reaches the upper limit of the sliding window (4), the management server suspends differential timing filling, retrieves the historical cross-sectional event change rate of adjacent reader nodes that are spatially adjacent to the current reader node within the same time span, and when neither adjacent reader node detects a lost tag and no inventory transfer transaction status transition is reported, the inventory transaction confidence weight is limited to the historical average range, and the asset logical management status attribute is adjusted according to the limited inventory transaction confidence weight, and the inventory database is updated.

[0009] Preferably, the method further includes a multipath scattering misreading elimination step: Step S4, the management server extracts the response events reported by the first reader node and the second reader node that are spatially adjacent within the same discrete time window, and calculates the event overlap factor between the first reader node and the second reader node; Step S5, when the event overlap factor exceeds the cross-determination threshold of 0.65, the management server compares the tag event density factor of the first reader node in the current discrete time window with the tag event density factor of the second reader node in the current discrete time window, and removes the overlapping events with the smaller tag event density factor from the current time-series data stream.

[0010] Preferably, the system also includes a data conflict arbitration step: Step S6, when multiple reader nodes output conflicting state transition requests, the management server invokes the shortest topological path constraint for asset transfer; Step S7, the management server inputs the conflicting discrete transaction sequence into the topological adjacency matrix to calculate the semantic matching degree, and adjudicates the conflicting state transition requests based on the semantic matching degree.

[0011] Preferably, the data feedforward comparison verification in step S2 includes the following sub-steps: Step S21, the management server compares the tag event density factor in the discrete time window with the confidence weight of the inventory transaction generated in the previous discrete time window to complete the causal logic verification; Step S22, when the tag event density factor is lower than the set sudden drop threshold, the management server writes a virtual compensation event with the lost tag identifier in place in the time series data stream of the current discrete time window.

[0012] Preferably, after limiting the confidence weight of inventory transactions to the historical average range in step S3, the following sub-steps are also included: Step S31, the management server changes the current position status determination to a static retention calibration and limits the confidence weight of inventory transactions to the historical average range; Step S32, when the confidence weight of inventory transactions is within the historical average range, the management server suspends the book inventory status update process.

[0013] Preferably, the system also includes an inventory business status node update step: Step S8, the management server calculates the inventory transaction confidence weight of the current inventory business status node based on the time-series event flow after filling and removing items, and the update rule is that the inventory transaction confidence weight of the current discrete time window is equal to the algebraic average of the inventory transaction confidence weight of the previous discrete time window and the tag event density factor; Step S9, when the inventory transaction confidence weight crosses the state transition critical point of 0.80, the management server triggers the state transition decision of the state machine to change the asset logic management state attribute.

[0014] Preferably, the length of the discrete time window is negatively correlated with the material turnover rate, and the management server uses time-series state machine logic and multi-source transaction arbitration scheduling to adjust the state data in the inventory database.

[0015] Preferably, the discrete time window includes a sliding discrete time window, the moving step size of which is positively correlated with the administrative decision-making delay within the warehouse, and the management server updates the inventory database by reorganizing the time sequence flow of RFID tag events using time sequence operators.

[0016] Preferably, the management server constructs a causal logic verification structure and sends a data freeze command to the inventory database when the confidence weight of the inventory transaction is within the historical average range.

[0017] Compared with the prior art, the beneficial effects of the present invention are:

[0018] 1. In warehouse data management of RFID tag data, by dividing the time-series of tag events into discrete time windows, and combining the event density factor within the time window with the confidence weight of the preceding window, when the number of responses to a specific tag drops abnormally, implicit shielding loss is determined through data feedforward decoupling verification. Virtual compensation events with associated tags are written in situ, constructing a data feedforward verification structure with self-calibration capabilities. When the continuous interruption of response events reaches the upper limit, the event change rate of adjacent nodes in the retrieval space is subject to spatiotemporal correlation feedback constraints, freezing the confidence weight at a historical stable value. This effectively blocks the error transmission of physical layer information dissipation to the upper-level administrative management transaction chain, ensuring the logical consistency of inventory data updates in a multi-layered, high-interference environment.

[0019] 2. The management server extracts response events reported by the first and second spatially adjacent reader nodes within the same time window, calculates the event overlap factor that characterizes the multipath refraction features of the signal, and removes overlapping data of nodes with lower response counts based on the signal strength trend when the cross-judgment threshold is exceeded. When different reader nodes output conflicting state transition requests, the shortest topological path constraint for asset transfer is invoked, and the conflicting transaction sequences are input in parallel into the topological adjacency matrix to calculate the semantic matching degree to implement hard truncation arbitration. Through multi-mechanism collaboration, multipath scattering misreading is eliminated, avoiding data conflicts and decision lag in the inventory database. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the self-calibration and update process of RFID tag data in warehouse storage according to the present invention.

[0021] Figure 2 This is a logical framework diagram for warehouse data management of RFID tag data according to the present invention.

[0022] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0024] A warehouse data management method based on RFID tag data includes the following steps:

[0025] Step S1: The management server obtains the RFID tag event time stream containing multiple sets of tag identifiers reported by the reader nodes in the warehouse data management system, and divides the RFID tag event time stream into discrete time windows according to the set time step.

[0026] Step S2: The management server calculates the tag event density factor of the RFID tag event time-series stream within the discrete time window, compares the tag event density factor with the confidence weight of the inventory transaction generated in the previous discrete time window to complete the data feedforward comparison verification, and writes a virtual compensation event with the lost tag identifier in place in the current time-series data stream when the tag event density factor is lower than the set sudden drop threshold to complete the differential time-series filling to build the self-calibration data feedforward verification logic relationship.

[0027] Step S3: When the number of consecutive interrupted windows of RFID tag events within the discrete time window reaches the upper limit of the sliding window (4), the management server suspends differential timing filling, retrieves the historical cross-sectional event change rate of adjacent reader nodes that are spatially adjacent to the current reader node within the same time span, and when neither adjacent reader node detects a lost tag and no inventory transfer transaction status transition is reported, the inventory transaction confidence weight is limited to the historical average range, and the asset logical management status attribute is adjusted according to the limited inventory transaction confidence weight, and the inventory database is updated.

[0028] Preferably, the method further includes a multipath scattering misreading elimination step: Step S4, the management server extracts the response events reported by the first reader node and the second reader node that are spatially adjacent within the same discrete time window, and calculates the event overlap factor between the first reader node and the second reader node; Step S5, when the event overlap factor exceeds the cross-determination threshold of 0.65, the management server compares the tag event density factor of the first reader node in the current discrete time window with the tag event density factor of the second reader node in the current discrete time window, and removes the overlapping events with the smaller tag event density factor from the current time-series data stream.

[0029] Preferably, the system also includes a data conflict arbitration step: Step S6, when multiple reader nodes output conflicting state transition requests, the management server invokes the shortest topological path constraint for asset transfer; Step S7, the management server inputs the conflicting discrete transaction sequence into the topological adjacency matrix to calculate the semantic matching degree, and adjudicates the conflicting state transition requests based on the semantic matching degree.

[0030] Preferably, the data feedforward comparison verification in step S2 includes the following sub-steps: Step S21, the management server compares the tag event density factor in the discrete time window with the confidence weight of the inventory transaction generated in the previous discrete time window to complete the causal logic verification; Step S22, when the tag event density factor is lower than the set sudden drop threshold, the management server writes a virtual compensation event with the lost tag identifier in place in the time series data stream of the current discrete time window.

[0031] Preferably, after limiting the confidence weight of inventory transactions to the historical average range in step S3, the following sub-steps are also included: Step S31, the management server changes the current position status determination to a static retention calibration and limits the confidence weight of inventory transactions to the historical average range; Step S32, when the confidence weight of inventory transactions is within the historical average range, the management server suspends the book inventory status update process.

[0032] Preferably, the system also includes an inventory business status node update step: Step S8, the management server calculates the inventory transaction confidence weight of the current inventory business status node based on the time-series event flow after filling and removing items, and the update rule is that the inventory transaction confidence weight of the current discrete time window is equal to the algebraic average of the inventory transaction confidence weight of the previous discrete time window and the tag event density factor; Step S9, when the inventory transaction confidence weight crosses the state transition critical point of 0.80, the management server triggers the state transition decision of the state machine to change the asset logic management state attribute.

[0033] Preferably, the length of the discrete time window is negatively correlated with the material turnover rate, and the management server uses time-series state machine logic and multi-source transaction arbitration scheduling to adjust the state data in the inventory database.

[0034] Preferably, the discrete time window includes a sliding discrete time window, the moving step size of which is positively correlated with the administrative decision-making delay within the warehouse, and the management server updates the inventory database by reorganizing the time sequence flow of RFID tag events using time sequence operators.

[0035] Preferably, the management server constructs a causal logic verification structure and sends a data freeze command to the inventory database when the confidence weight of the inventory transaction is within the historical average range.

[0036] Example 1: The current warehouse management information system operates within a high-frequency, dynamically changing physical warehouse environment. Multiple sets of non-electrical variable scanning units are deployed. The management server continuously receives the raw tag event time-series stream collected by the RFID link. This raw tag event time-series stream contains a set of tag RFID response events with timestamp noise. The management server performs windowing processing according to a preset discrete time period, dividing the raw tag event time-series stream into multiple equally long discrete time windows along the time axis. The time window length is set to 50ms. For each discrete time window, the management server extracts the RFID response characteristics of a specific tag and calculates the tag event density factor within the current window. And retrieve the confidence weight of the business status node in the preceding window. Tag event density factor Satisfy the following definition: ,in, The density factor of the labeled events within the i-th time window; The number of radio frequency responses for a specific tag within the i-th time window; The duration of the i-th time window is set to 50ms in this embodiment, and the number of radio frequency responses of a specific tag within the current time window is also represented. The confidence weight of the business status node in the preceding window is reduced to 0. When the value exceeds the stability safety threshold of 0.75, the management server determines that the current signal has experienced implicit shielding and missed reading due to dense stacking. After detecting the above state, the management server writes a virtual compensation event with the current service-related tag in situ into the time-series data stream of the current discrete time window, and performs differential timing padding to compensate for the missed reading loss of the physical layer. To eliminate non-service misreading interference caused by radio frequency multipath scattering, the management server extracts the radio frequency response events reported by the first and second spatially adjacent reader nodes for the same tag within the same time window and calculates the event overlap factor between the nodes. Event overlap factor The following relationship must be satisfied: ,in, The event overlap factor between the first and second spatially adjacent reader nodes; This represents the number of responses from the first reader node within the current time window. This represents the number of responses from the second reader node within the current time window. This represents the number of shared, overlapping responses collected by the two reader nodes within the current time window.

[0037] When the calculated event overlap factor When the cross-referencing threshold of 0.65 is exceeded, the management server determines that a multipath misread has occurred. Based on the dynamic evolution characteristics of the signal strength gradient, the management server removes the corresponding overlapping events of nodes with fewer responses from the current time-series data stream, retaining the business events of the high-confidence main path. The management server combines the time-series event stream purified by filling and deleting to calculate the confidence weight of the current business status node. It represents the stage of certainty in the current logic of warehousing administration, with confidence weights. The following relationship must be satisfied: ,in, The confidence weight is the business status corresponding to the current discrete time window; The confidence weight for the previous discrete time window; To eliminate the dimensional conflict between the reciprocal time dimension of the density factor and the dimensionless probability attribute of the confidence weights when calculating the algebraic average of the tag event density factor within the current discrete time window, the management server performs a normalization transformation mapping at the underlying level. Specifically, the calculation process multiplies the current tag event density factor by a preset standard time base constant of 50ms, thereby converting it into a dimensionless scalar scaling factor reflecting the actual scan completeness within the current window, converging its value range between 0 and 1 to align with the dimensions of the confidence weights. Simultaneously, for scenarios where the time window dynamically scales from 50ms to 15ms or 100ms, the management server simultaneously adjusts the denominator of the time span to the current actual window width when calculating the density factor, and uses the reciprocal of the window scaling factor to perform linear gain compensation on the calculated scalar scaling factor. This ensures that, with a constant number of tag responses, the normalized density factor value participating in the algebraic average calculation remains stable, eliminating false weight spikes or drops caused by sudden changes in the denominator, and ensuring the stability of the state machine decision-making. When the confidence weights... When the state transition threshold of 0.80 is crossed, the system triggers the state machine's state transition decision, changes the current logical management state attribute of the asset, and the management server, based on the final convergence confidence weight and time series relationship, reorganizes the out-of-order or delayed event stream into a discrete transaction sequence that conforms to the consistency of warehousing logic, and outputs the final asset allocation control instruction, driving the downstream warehousing execution agency to update the corresponding state data field in the book inventory database.

[0038] When the number of consecutive interruption windows of a specific tag's RF response event within a discrete time window reaches four, the system initiates spatiotemporal correlation feedback constraints. The determination of the upper limit of the sliding window (four) and the set drop threshold uses the following basic parameter test data under the test environment as the quantitative basis: In a static shelf environment, the management server continuously collects the baseline RF response counts of a specific tag within 100 discrete time windows and calculates the average value of the baseline RF response counts; 30% of the average value is fixed as the set drop threshold to identify transient signal amplitude drops caused by dense obstruction; for the upper limit of the sliding window, a 50mm thick metal shielding plate is set on the RF identification link to simulate physical shielding blockage, and the number of continuous windows of tag RF response interruption is collected. Test results show that when the number of consecutive interruption windows reaches four, the probability of signal recovery drops below 2%, while adjacent read / write operations... Since the historical cross-sectional event change rate of the reader node remains stable and there are no changes in the state of data transfer, the upper limit of the sliding window is determined to be 4, which serves as the quantization trigger boundary for suspending differential time-series filling and initiating spatiotemporal correlation feedback constraints. The management server actively suspends the aforementioned differential time-series filling step and instead retrieves the historical cross-sectional event change rate of adjacent reader nodes that are spatially topologically adjacent to the current reader node within the same time span. If none of the adjacent reader nodes have detected the tag and there is no state transition report for data transfer transactions, the asset is determined to be in a static retention state. The management server changes the current state determination to a static retention calibration, truncates the time-series filling operator, forcibly freezes the current confidence weight within the historical stationary mean range, sends a data freeze command to the inventory database, suspends the book inventory status update process, and calculates the overall information entropy of the number of active tag responses within the current time window. ,when When the rate of change exceeds the preset stepped threshold, the management server dynamically scales the current time window length, compressing it from 50ms to 15ms to improve the resolution of the time stream. When the job is in a rest period, the window length is adjusted to 100ms. During this control process, the management server achieves refined updates to the inventory database by reorganizing the time stream of RFID tag events into time-series process operators. A time-series process operator is an independent logical execution unit that encapsulates specific data processing behaviors such as data cleaning, event merging, feedforward verification, and status judgment. The underlying data structure of the operator includes operator type identifier, timestamp range boundaries, and input / output event queue pointers. A group refers to the management server dynamically rearranging the serial parallel execution sequence and time axis call frequency of each operator based on the moving step size of the sliding discrete time window, which is positively adjusted by the administrative decision delay. When the delay increases, causing the step size to increase, the management server merges the original short-cycle, high-frequency operator chain into a merge deduplication operator containing a larger time span data set, and increases the parallel processing ratio of the batch verification operator, thereby filtering out micro-signal disturbances during the step size increase and directly outputting a large-granularity transaction flow to update the inventory database. When the first reader and the second reader output conflicting state transition requests for the same tag within the same time span, the management server calls the preset shortest asset transfer time. Topological path constraints are applied by inputting conflicting transaction sequences in parallel into the topological adjacency matrix to calculate semantic matching degree. Hard truncation arbitration is then performed based on the semantic matching degree to eliminate abnormal state transitions. In practice, the input of the aforementioned conflicting discrete transaction sequences and the topological adjacency matrix, along with the semantic matching degree calculation, is implemented through the following plain text logic steps: The management server extracts the identifiers of multiple conflicting reader nodes and retrieves the system's pre-stored warehouse space topological adjacency matrix. The rows and columns of this matrix represent the various physical reader nodes within the warehouse; a matrix element of 1 indicates direct connectivity between nodes, and 0 indicates disconnection. Next, the timestamped discrete transaction sequences reported by each reader are transformed into... For node transfer path chains, the connectivity state corresponding to the path chain is found in the topological adjacency matrix. The semantic matching degree is calculated by counting the number of legal transfer steps in the transaction sequence that meet the connectivity constraints of the topological adjacency matrix, dividing it by the total number of transfer steps in the sequence, and outputting a semantic matching degree scalar value between 0 and 1. The management server compares the semantic matching degrees of each conflicting sequence, decides to retain the request with the highest matching degree and the shortest path, and hard-truncates and removes conflicting requests that do not conform to the real path of the spatial topology. After the asset allocation control command drives the inventory database to execute the status update, the corresponding book inventory data and the asset entity status in the physical storage environment are logically consistent.

[0039] Example 2: The current management server relies on a multi-layered concurrent warehouse management environment. It manages warehouse status data by implementing refined time-slice windowing and data feedforward verification on the RFID tag event time-series stream. During the experimental verification process, the discrete time window length was set to 50ms. The experimental sample group contained 500 tags, and inbound simulation operations were performed at flow rates of 100 tags / s, 500 tags / s, and 1000 tags / s, respectively. A specific physical shielding condition caused by high-density goods obstruction was also set to verify the stability of the management server's status determination under signal loss conditions. This experiment constructed three-dimensional control groups for performance evaluation. Control group A was the missing time-series filling sample group, which only performed the original RFID event scan without triggering the differential time-series filling step. Control group B was the out-of-range filling frequency sample group, with a set time window of 250ms, and forced execution of compensation event writing even when the tag response count did not suddenly drop. The sample group of this invention executed complete feedforward comparison verification and in-situ differential time-series filling logic. The management server based on the tag event density factor... Confidence weight of business status nodes The evolution relationship is determined by the number of responses to a specific tag. During a specific time slot where the threshold suddenly drops to 0, the system decouples business logic in real time using the following relational expression: ,in, The confidence weight is the business status corresponding to the current discrete time window; The confidence weight for the previous discrete time window; The label event density factor is calculated within the current discrete time window, and ,in, The number of responses to a specific tag within the current time window; The duration of the current time window is set to 50ms in this embodiment.

[0040] Test data shows that when Greater than the stability safety threshold of 0.75 and detected When the value is 0, the inventory data breakage rate of control group A under high throughput conditions is 14.5%, leading to inconsistencies between the records and the actual inventory. In contrast, the sample group of this invention successfully filled the logical gaps caused by physical shielding in the information space by writing virtual compensation events with associated tags in situ in the time-series data stream. At a high throughput rate of 1000 items / s, its inventory data breakage rate was reduced to 0.82%, effectively ensuring the integrity of the data chain. For parameter determinism, the management server introduces a multipath scattering misreading elimination mechanism to calculate the event overlap factor between the first and second spatially adjacent reader nodes in real time. ,in, This is the event overlap factor between spatially adjacent nodes; This represents the number of responses from the first node. This refers to the number of responses from the second node. This represents the number of shared, intersecting responses collected by the two nodes within the current time window.

[0041] when When the cross-referencing threshold of 0.65 is exceeded, the management server determines that a multipath misread has occurred. Experimental observations show that after exceeding 0.65, the data state migration misjudgment rate of the sample group of this invention decreased from 12.3% to 0.45%. In control group B, due to the excessively long time window, the displacement of the tag between different positions was mistakenly counted as the same event, causing false drift of the asset position coordinates in the book inventory, with a drift rate of 9.8%. When the number of consecutive interruption windows reaches 4, the system triggers the spatiotemporal correlation feedback constraint. The management server determines the asset status by retrieving the historical cross-sectional event change rate of adjacent reader nodes. If the adjacent readers If no node detects the tag and no data transfer transaction is reported, the management server changes the current state determination to a static retention flag, truncates the time-series filling operator, freezes the current confidence weight within the historical stable average range, and sends a data freeze command to the inventory database, suspending the inventory status update process. This mechanism, when simulating radio frequency link interruptions exceeding 200ms, can reduce the write probability of error compensation events from 28.5% to below 1.2%, avoiding the accumulation of false inventory records under high interference. The management server calculates the overall information entropy of the number of active tag responses within the current time window. The calculation of the overall information entropy does not rely on complex mathematical formulas, but is executed in the background through pure text logic steps: The management server counts the proportion of radio frequency response times of each active tag within the current time window to the total number of responses of all tags within that window, and uses this proportion as the probability of that tag appearing in the current window. For the probability of each tag's appearance, the logarithm to the base 2 is calculated and multiplied by the probability itself. Finally, the calculation results of all tags are negatively summed to quantify the disorder and concurrent activity of the current data stream. The preset stepped thresholds include a first threshold of 15% and a second threshold of 30%. When the rate of change of information entropy is greater than 15% and less than or equal to 30%, the system is determined to have entered a moderate traffic fluctuation state, triggering the initial compression of the window width. When the rate of change exceeds 30%, the traffic concurrency is determined to have reached a preset peak, and the management server initiates a window compression command, adjusting the window length from 50ms to 15ms. When the rate of change exceeds the preset step threshold, the management server dynamically scales the current time window length, compressing it from 50ms to 15ms to improve the resolution of the time-series stream. When the job is in the metabolic rest period, the window length is stretched to 100ms. When the first reader and the second reader output conflicting state transition requests for the same tag within the same time span, the management server calls the preset shortest topology path constraint for asset transfer, inputs the conflicting transaction sequences in parallel into the topology adjacency matrix to calculate the semantic matching degree, and performs hard truncation arbitration based on the semantic matching degree to eliminate abnormal state transitions. After the asset allocation control command drives the inventory database to perform state updates, the corresponding book inventory data and the asset entity status in the physical storage environment are logically consistent.

[0042] Example 3: When the current warehouse asset management system faces combined interference from signal strength attenuation due to equipment aging and dynamic impedance fluctuations in the environment, it manages asset status through boundary early warning and adaptive compensation logic. The system monitors the real-time received signal strength indication value of the RFID link. and the average noise level within the preset time window Calculate the signal-to-noise ratio boundary value of the current link. Signal-to-noise ratio boundary value Satisfy the following formula: ,in, This represents the signal-to-noise ratio boundary value at the current moment. To receive real-time signal strength indication values, The average noise level within a preset time window, when the signal-to-noise ratio boundary value When the signal drops to the preset asset loss risk threshold of 12dB, the management server determines that the current radio frequency signal is in the defense warning range of asset misreading. The management server automatically starts the defensive compensation procedure, dynamically increases the query beam gain of the reader node by 3dB, and adjusts the avoidance time window of tag collision arbitration from 5ms to 2ms to improve the data retransmission success rate in low signal-to-noise ratio environments.

[0043] The management server introduces a second-order loss compensation operator to correct the response sequence deviation caused by high-frequency time delay lag. For physical gap delays generated in the signal transmission path, the management server calculates the response time difference compensation amount. By aligning the timestamp logic to correct the logical location map of the RFID tag, the compensation amount is increased. Satisfy the following formula: ,in, To compensate for the time difference in response, This refers to the impedance fluctuation adaptation coefficient calibrated based on the physical properties of the reader antenna array. The set reference signal-to-noise ratio value, The impedance fluctuation adaptation coefficient, which is the boundary value of the current signal-to-noise ratio, is determined by a standard RF network analyzer during the initial deployment of the system. In actual engineering implementation, this coefficient is used to characterize the modulation sensitivity of the antenna return loss to the signal propagation delay, and its engineering value range is limited to 0.1ms / dB to 0.5ms / dB. The specific calibration process is as follows: under standard factory test conditions, the impedance of the metal shielding medium around the reader is gradually changed, and the physical response time delay corresponding to each 1dB decrease in the signal-to-noise ratio from the reference value is recorded. Linear fitting is performed using the least squares method, and the absolute value of the slope of the fitted line is extracted as the inherent coefficient constant of the antenna array. When the environmental impedance fluctuates drastically and causes the coefficient to fall below 0.1ms / dB, the system will experience positional errors due to insufficient compensation. When it is above 0.5ms / dB, it will cause overcompensation, resulting in timing waveform distortion. In this embodiment, for the specific physical properties of the multi-antenna array, the coefficient is precisely assigned a value of 0.25ms / dB.

[0044] When an asset moves dynamically in the edge area, the management server monitors the slope change of the received signal strength indicator in real time. If the absolute value of the signal slope exceeds 0.5dB / ms, the asset is determined to be in a critical state of leaving the coverage area. The management server sends a status storage instruction with a temporary location lock identifier to the inventory database. This instruction triggers the system to ignore subsequent radio frequency silence events of the asset within 500ms, retaining its position coordinates in the high-confidence time window of the last frame until a handover confirmation signal or silent retention calibration feedback is received from the reader in the adjacent spatial topology area, thus realizing the cross-regional asset transfer process. Continuous trajectory tracking and the management server recording the response confidence weight evolution curve of each tag, within the defense range of 12dB to 15dB signal-to-noise ratio, through dynamic gain adjustment, the asset identification accuracy of the sample group of this invention remains above 99.1%. Under the defense condition without gain adjustment, the identification accuracy drops to 85.3%, and multiple false alarm events of inventory status are generated due to false signals. As the signal-to-noise ratio continues to decrease to below 10dB, the sample group of this invention triggers the location locking mark, effectively avoiding systemic accounting disorder caused by signal interruption and maintaining the logical stability of warehouse data under edge conditions.

[0045] Example 4: Under the condition that the radio frequency link characteristics of the current warehouse asset management system have shifted due to long-term operation of equipment, the system performs parameter calibration on the asset status judgment criteria by executing the initial calibration procedure. The management server obtains the radio frequency response benchmark data of the reader in the empty warehouse resting state and calculates the benchmark value of the average tag arrival rate per unit time. This calibration process is performed during the initial deployment of the system. It involves continuously collecting radio frequency events over 100 discrete time windows, eliminating abnormal interference sampling points, and... The baseline value is calculated, where, Let M be the label arrival rate of the j-th calibration window, and M be the total number of calibration windows.

[0046] During normal operation, the management server monitors the average arrival rate in real time within the i-th discrete time window. Standard deviation of event arrival rate And generate fluctuation suppression coefficients based on calibration parameters. Fluctuation suppression coefficient The sensitivity used for dynamically adjusting state transition decisions is calculated using the following formula: ,in, This is the fluctuation suppression coefficient for the current window, reflecting the degree of deviation of the current data stream dispersion from the baseline operating condition; The average arrival rate within the current discrete time window. The average arrival rate of the labels obtained through calibration is the baseline value. This represents the standard deviation of the event arrival rate within the current window.

[0047] Management server based on fluctuation suppression coefficient Discretize the inventory status of assets and set critical thresholds for state transitions. The value is 0.65, when the weighted product calculated for the label in three consecutive scanning cycles exceeds... When the system triggers an inventory update, the parameter is set based on a physical attenuation simulation experiment of the RF link in a high-density environment. This ensures that transient random interference caused by fluctuations in the equipment environment is filtered out while maintaining logical accuracy. The fluctuation suppression coefficient... When the value is within the predetermined range of 0.85 to 1.15, the system maintains the normal asset status update frequency; when... When the value is below 0.85, it indicates that the data dispersion is increased due to the influence of environmental noise. The system automatically changes the trigger control of asset status updates in the inventory database from the hardware counting threshold to a smooth judgment mechanism based on weighted product, thereby suppressing the logic jitter caused by noise by reducing the interaction frequency.

[0048] Example 5: When the current warehouse asset management system performs automatic inventory record updates, it configures standardized offline parameter construction procedures to ensure data confidence under high concurrency environments. Before system deployment, the management server calls a standard RFID test bench to collect multiple sets of tag response data under different inventory densities. The experiment selects 50 standard RFID tags as test subjects and performs a full scan in a static shelf environment with a step size of 200ms. Noise data with a signal strength lower than 30dBm is removed, and the average arrival rate benchmark value of the tags per unit time is statistically obtained. This baseline value is written to the non-volatile storage area of ​​the management server as a standardized reference for measuring the intensity of interaction in the current environment in subsequent business logic judgments.

[0049] The management server implements data deviation compensation logic within a discrete time window based on a standardized benchmark reference system. The system truncates the original event time stream with a sampling period of 50ms and calculates the average tag arrival rate within the current window. With arrival rate standard deviation The system calculates the fluctuation suppression coefficient for the current window using the following formula. ,in, The fluctuation suppression coefficient reflects the ratio of the current data interaction intensity to the degree of environmental disturbance. The average arrival rate within the current discrete time window. The average arrival rate of the labels obtained through calibration is the baseline value. This represents the standard deviation of the event arrival rate within the current window.

[0050] The management server will set the fluctuation suppression coefficient. Applied to inventory transaction state machine transition determination, the system sets a confidence threshold for asset in-stock state transitions. The threshold is 0.65. If the weighted product calculated over three consecutive scan cycles exceeds this threshold, an atomic update of the account is triggered. Within a predetermined confidence interval of 0.85 to 1.15, the system performs state judgment with a sampling period of 50ms; when When the value is below 0.85, it indicates that environmental noise has increased data dispersion. The system automatically extends the sampling period to 100ms and increases the trigger weight judgment threshold to suppress random interference. This parameter calibration logic eliminates the dependence on hardware sensitivity changes during the account update process by transforming the physical properties of data fluctuations into logical decision thresholds, thus ensuring the accuracy of data consistency auditing in a multi-layered warehousing environment.

[0051] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A warehouse data management method based on RFID tag data, characterized in that, Includes the following steps: Step S1: The management server obtains the RFID tag event time stream containing multiple sets of tag identifiers reported by the reader nodes in the warehouse data management system, and divides the RFID tag event time stream into discrete time windows according to the set time step. Step S2: The management server calculates the tag event density factor of the RFID tag event time-series stream within the discrete time window, compares the tag event density factor with the confidence weight of the inventory transaction generated in the previous discrete time window to complete the data feedforward comparison verification, and writes a virtual compensation event with the lost tag identifier in place in the current time-series data stream when the tag event density factor is lower than the set sudden drop threshold to complete the differential time-series filling to build the self-calibration data feedforward verification logic relationship. Step S3: When the number of consecutive interrupted windows of RFID tag events within the discrete time window reaches the upper limit of the sliding window (4), the management server suspends differential timing filling, retrieves the historical cross-sectional event change rate of adjacent reader nodes that are spatially adjacent to the current reader node within the same time span, and when neither adjacent reader node detects a lost tag and no inventory transfer transaction status transition is reported, the inventory transaction confidence weight is limited to the historical average range, and the asset logical management status attribute is adjusted according to the limited inventory transaction confidence weight, and the inventory database is updated.

2. The warehouse data management method based on RFID tag data according to claim 1, characterized in that, It also includes a multipath scattering misreading elimination step: Step S4, the management server extracts the response events reported by the first reader node and the second reader node that are spatially adjacent within the same discrete time window, and calculates the event overlap factor between the first reader node and the second reader node. Step S5: When the event overlap factor exceeds the cross-determination threshold of 0.65, the management server compares the tag event density factor of the first reader node in the current discrete time window with the tag event density factor of the second reader node in the current discrete time window, and removes the overlapping events with the smaller tag event density factor from the current time-series data stream.

3. The warehouse data management method based on RFID tag data according to claim 1, characterized in that, It also includes a data conflict arbitration step: Step S6, when multiple reader nodes output conflicting state transition requests, the management server invokes the shortest topology path constraint for asset transfer; In step S7, the management server inputs the conflicting discrete transaction sequences into the topological adjacency matrix to calculate the semantic matching degree, and decides the conflicting state transition requests based on the semantic matching degree.

4. The warehouse data management method based on RFID tag data according to claim 1, characterized in that, The data feedforward comparison verification in step S2 includes the following sub-steps: Step S21, the management server compares the tag event density factor in the discrete time window with the confidence weight of the inventory transaction generated in the previous discrete time window to complete the causal logic verification; Step S22, when the tag event density factor is lower than the set sudden drop threshold, the management server writes a virtual compensation event with the lost tag identifier in place in the time series data stream of the current discrete time window.

5. A warehouse data management method based on RFID tag data according to claim 1, characterized in that, After limiting the confidence weight of inventory transactions to the historical average range in step S3, the following sub-steps are also included: Step S31, the management server changes the current position status determination to a static retention calibration and limits the confidence weight of inventory transactions to the historical average range; Step S32, when the confidence weight of inventory transactions is within the historical average range, the management server suspends the book inventory status update process.

6. A warehouse data management method based on RFID tag data according to claim 1, characterized in that, It also includes the inventory business status node update steps: Step S8, the management server calculates the inventory transaction confidence weight of the current inventory business status node according to the time-series event flow after the completion of filling and removal, and the update rule is that the inventory transaction confidence weight of the current discrete time window is equal to the algebraic average of the inventory transaction confidence weight of the previous discrete time window and the tag event density factor; Step S9, when the inventory transaction confidence weight crosses the state transition critical point of 0.80, the management server triggers the state transition decision of the state machine to change the asset logic management state attribute.

7. A warehouse data management method based on RFID tag data according to claim 1, characterized in that, The length of the discrete time window is negatively correlated with the material turnover rate. The management server uses time-series state machine logic and multi-source transaction arbitration scheduling to adjust the state data in the inventory database.

8. A warehouse data management method based on RFID tag data according to claim 1, characterized in that, Discrete time windows include sliding discrete time windows. The moving step size of the sliding discrete time window is positively correlated with the administrative decision-making delay in the warehouse. The management server updates the inventory database by reorganizing the time sequence flow of RFID tag events using time sequence operators.

9. A warehouse data management method based on RFID tag data according to claim 1, characterized in that, The management server constructs a causal logic verification structure and sends a data freeze command to the inventory database when the confidence weight of the inventory transaction is within the historical average range.

Images