A method and system for identifying the location of file boxes based on RFID read / write timing and shelf position feedback
By adopting a dynamic identification method based on RFID reading and writing timing and shelf position feedback, the problems of missing tag reading and response window conflict under the variable speed movement of electric mobile shelving were solved, realizing the high efficiency and reliability of in-situ identification of file boxes and improving the identification completeness rate of mobile inventory.
Patent Information
- Application Number
- CN202611543974.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-08
- Publication Date
- 2026-11-10
AI Technical Summary
When electric mobile shelving moves at variable speeds and there is jitter and delay in the feedback of the shelving's position, existing technologies cannot effectively match the tags into the reading and writing field, resulting in missed tag readings and response window conflicts, which affect the integrity and reliability of the in-situ identification of file boxes.
A file box in-situ identification method based on RFID read/write timing and shelf position feedback is adopted. Through logistic probability boundary calibration, adaptive Kalman filtering and delay compensation, kinematic extrapolation and window overlap pruning, dynamic frame length anti-collision and spatial time slot frequency joint scheduling, and recursive Bayes confidence determination, the read/write timing and tag storage are dynamically adjusted.
It reduces label misreading and response window conflicts, improves the completeness and reliability of in-situ file box identification, and ensures accurate file box identification in the variable speed movement environment of electric mobile shelving.
Smart Images

Figure CN123070148A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial data acquisition and processing technology, and in particular to a method and system for identifying the location of file boxes based on RFID read / write timing and shelf position feedback. Background Technology
[0002] Currently, in the field of electric mobile shelving file management, to achieve batch inventory of file box tags during shelving movement, fixed RFID readers are typically installed at fixed positions in the shelving's operating aisle, and position sensors are configured on the shelving. The existing processing path is generally as follows: the reader sends an inventory command according to a fixed polling cycle or a fixed reading window; when a file box tag passes through the reader's effective area with the shelving, the tag responds and forms a reading event record; simultaneously, the position sensor feeds back the shelving position reading at a fixed cycle, and the backend correlates the tag reading status with the shelving position, using a standard anti-collision algorithm based on frame time slots to handle multi-tag response conflicts. For the boundary of the effective RFID reading area, current practices often use a unified empirical range or boundary parameters based on static calibration; for jitter and delay in position feedback, simple averaging or threshold filtering is usually performed, or the original position reading is directly used as the trigger for reading and writing.
[0003] For example, during the start-up, braking, and speed change processes of electric mobile shelving units, the speed, acceleration, and jerk of the shelving unit continuously change, causing the time it takes for a tag to pass through the effective RF reading area to become constant. Fixed polling cycles or fixed reading window lengths cannot match the process of tags entering the read / write field. Some tags may leave the effective reading area before completing anti-collision inventory, resulting in missed reads. If there is overlap or gap between adjacent reading windows, tags in nearby areas may be activated by incorrect windows, causing overlapping response windows, time slot conflicts, and increased collisions. Simultaneously, the shelving unit position sensor itself has jitter and communication delays, meaning position feedback cannot be directly used for window triggering. Static calibration or unified empirical boundaries cannot reflect the actual entry and exit characteristics of different tags in the RF field, causing a misalignment between the read / write start position and the actual tag entry time. When reading tags in batches from multiple layers and columns of file boxes, the high tag density and concentrated relative positions lead to a high degree of temporal overlap in tag responses. Under fixed timing, response conflicts are further amplified, resulting in incomplete reading records after a single inventory check, and frequent missed checks or pending verifications in subsequent file box location determinations.
[0004] Therefore, there is an urgent need for a file box location identification method that can still utilize real-time location feedback of the shelving unit and coordinated control of RFID reading and writing timing to reduce tag omissions and response window conflicts in scenarios where electric mobile shelving moves at variable speeds and the shelving unit position feedback has jitter and delay, so as to improve the one-time identification completeness rate and location determination reliability of mobile inventory. Summary of the Invention
[0005] To address the aforementioned technical shortcomings, the present invention aims to propose a file box location identification method based on RFID reading / writing timing and shelf position feedback. This method aims to solve the technical problem in the prior art where fixed polling cannot match tags to enter the reading / writing field under the conditions of variable speed movement of electric mobile shelving and position feedback jitter delay. In particular, it is difficult to reduce missed readings and conflicts under the condition of overlapping response windows or missed readings leading to incomplete inventory data during dynamic inventory.
[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a method for identifying the presence of file boxes based on RFID reading and writing timing and shelf position feedback.
[0007] The method for identifying the location of file boxes based on RFID read / write timing and shelf position feedback includes:
[0008] Step S10: Obtain file box tag reading event data, and perform radio frequency effective reading boundary calibration task based on the file box tag reading event data using the logistic probability boundary calibration mechanism, and output archive calibration data packet;
[0009] Step S20: Based on the archived calibration data package, an adaptive Kalman filter and delay compensation mechanism are used to perform a position feedback filtering task, and a dynamic identification parameter set is output;
[0010] Step S30: Based on the dynamic identification parameter set, a kinematic extrapolation and window overlap pruning mechanism is used to perform read / write timing calculation tasks, and output the dynamic read window parameter set;
[0011] Step S40: Based on the dynamic reading window parameter set, a dynamic frame length anti-collision and spatial, time slot and frequency joint three-dimensional scheduling mechanism is used to execute the tag inventory task and output the tag inventory reading record;
[0012] Step S50: Based on the tag inventory reading record, the file box presence determination task is performed using a dual determination mechanism of recursive Bayes confidence and continuous missed reading, and the file box presence identification result is output.
[0013] Preferably, step S10, which involves acquiring file box tag reading event data, performing radio frequency effective read boundary calibration tasks based on the file box tag reading event data using a logistic probability boundary calibration mechanism, and outputting an archive calibration data packet, specifically includes:
[0014] Step S101: Control the electric mobile shelving to pass through the effective reading area of the fixed RFID reader at a preset calibration speed, and simultaneously record the tag identification, number of reads, reading position, received signal strength indication value and the corresponding shelf position sensor reading at the time of each tag, and mark the reading event with the received signal strength indication value not lower than the preset effective threshold as a valid reading event, forming the file box tag reading event data;
[0015] Step S102: For each tag, calculate the ratio of valid reads to total queries according to the shelf position to obtain the read probability of the corresponding position, and use a four-parameter logistic model to fit the curve of read probability changing with shelf position to obtain the reading probability outside the reading area, the reading probability saturation at the center of the reading area, the midpoint of the reading probability, and the steepness of the reading probability change.
[0016] Step S103: Substitute the preset entry probability threshold and the preset exit probability threshold into the entry segment fitting curve and the exit segment fitting curve, respectively, to determine the entry and exit boundaries of each tag; establish the correspondence between the file box identifier, the tag identifier, and the nominal position, and associate the entry boundary, the exit boundary, the reading center saturation probability, the reading probability change steepness, the preset effective threshold for received signal strength indication, and the standard deviation of the calibration residual to form the archive calibration data packet.
[0017] Preferably, step S20, which involves performing a position feedback filtering task based on the archived calibration data packet using an adaptive Kalman filter and a delay compensation mechanism, and outputting a dynamic identification parameter set, specifically includes:
[0018] Step S201: Obtain the time-series readings of the frame position sensor, construct a state vector consisting of frame position, frame velocity and frame acceleration, determine the frame jerk based on the sampling period of the position sensor and the difference in frame acceleration between adjacent times, use the frame jerk to predict the motion state of the state vector, and obtain the state prediction value and prediction covariance matrix.
[0019] Step S202: Obtain the filtered velocity, filtered acceleration, and rate of change of filtered acceleration at the previous moment, and determine the stationary mode, uniform speed mode, acceleration mode, deceleration mode, or impact transition mode based on the comparison results between the filtered velocity, filtered acceleration, and rate of change of filtered acceleration at the previous moment and the corresponding preset motion state threshold; adaptively determine the process noise covariance and measurement noise variance based on the determined motion mode and the difference between the filtered acceleration, the filtered velocity, and the filtered acceleration at adjacent moments.
[0020] Step S203: Determine the position information based on the predicted state value and the current reading of the frame position sensor. When the position information exceeds the anomaly boundary determined by the theoretical information standard deviation and the preset anomaly judgment coefficient, introduce a fading factor. Update the predicted covariance matrix and the state vector using the process noise covariance, the measurement noise variance, and the fading factor. Correct the communication delay online based on the deviation between the entry reading area boundary in the archived calibration data packet and the filtered position when the tag is first effectively read. Perform kinematic compensation on the updated state vector according to the communication delay. Combine the compensated predicted position, predicted velocity, predicted acceleration, predicted jerk, position estimation covariance, velocity estimation covariance, filtered position error standard deviation, and online correction delay to form the dynamic identification parameter set.
[0021] Preferably, in step S202, when adaptively determining the process noise covariance based on the motion mode, the filtered acceleration, the filtered velocity, and the difference between filtered accelerations at adjacent times, the process noise covariance is:
[0022]
[0023] in, For the first Process noise covariance at time step This is the process noise reference matrix. , and These are the process noise matrices corresponding to the impact transition mode, acceleration mode, and deceleration mode, respectively. , and In the shock transition mode Acceleration Mode and deceleration mode The default mode weights enabled below. The rate of change of filtered acceleration. For filtered acceleration, The filter time number is used; the measurement noise variance is determined by the weighted increment of the absolute value of the frame acceleration, the square value of the frame velocity, and the square value of the difference between the frame accelerations at adjacent times relative to the measurement noise variance reference value.
[0024] Preferably, in step S30, the step of performing read / write timing calculation tasks based on the dynamic identification parameter set using kinematic extrapolation and window overlap pruning mechanisms, and outputting the dynamic read window parameter set, specifically includes:
[0025] Step S301: Determine the advance start time based on the RFID reader's RF turn-on time, frequency stabilization time, and phase-locked loop lock time. Extrapolate the frame movement within the advance start time using the predicted velocity, predicted acceleration, and predicted jerk from the dynamic identification parameter set to obtain the trigger position advance amount. Determine the safety advance amount based on the position estimation covariance, velocity estimation covariance, online correction delay, and the standard deviation of the calibration residual in the archived calibration data packet. Finally, determine the RFID read / write trigger position based on the boundary of each tag's entry into the reading area, the trigger position advance amount, and the safety advance amount.
[0026] Step S302: Determine the reading area length based on the entry and exit boundaries of each tag; when the current acceleration estimate is less than a preset minimum acceleration threshold, determine the physical passage time based on the ratio of the reading area length to the predicted velocity when entering the reading area; when the current acceleration estimate is not less than the preset minimum acceleration threshold, determine the physical passage time based on the kinematic relationship between the reading area length, the predicted velocity, and the current acceleration estimate; add the physical passage time, the anti-collision storage time, and the reader termination protection time to obtain the initial reading window length.
[0027] Step S303: Determine the prediction end position based on the RFID read / write trigger position of the current reading window, the predicted average speed, the predicted average acceleration, and the initial reading window length. Compare the prediction end position with the position obtained by subtracting the preset minimum protection gap from the RFID read / write trigger position of the next reading window. When the current reading window intrudes into the starting area of the next reading window, determine the maximum allowable window length based on the distance between adjacent RFID read / write trigger positions, the preset minimum protection gap, the predicted average speed, and the predicted average acceleration. Determine the smaller value between the initial reading window length and the maximum allowable window length as the final reading window length. Mark the final reading window shorter than the preset minimum inventory time as a high-conflict window. Combine the RFID read / write trigger position, the final reading window length, the predicted speed, the predicted acceleration, the standard deviation of the filter position error, the online correction delay, the standard deviation of the calibration residual, the nominal position, the entry boundary of the reading area, the exit boundary of the reading area, the filter acceleration, and the rate of change of the filter acceleration to form the dynamic reading window parameter set.
[0028] Preferably, step S40, which involves executing the tag inventory task based on the dynamic reading window parameter set using a dynamic frame length anti-collision and a joint three-dimensional scheduling mechanism of space, time slot, and frequency, and outputting the tag inventory reading record, specifically includes:
[0029] Step S401: Determine the initial Q value based on the expected number of tags in the reading window, the preset target throughput, and the preset Q value range. The Q value is a frame length parameter used to determine the number of time slots in a single frame. Perform tag inventory in units of frames within the reading window. After each frame is completed, count the number of successfully read time slots, the number of collision time slots, and the number of idle time slots. Estimate the remaining number of unidentified tags based on the number of collision time slots. Determine the Q value of the next frame based on the remaining number of unidentified tags, the preset target throughput, the remaining time of the reading window, and the preset time slot time.
[0030] Step S402: Sort the tags in the reading window according to the nominal position, determine the number of position segments according to the preset maximum number of tags per segment, and determine the width of the position segments according to the prediction speed, the average time for each position segment to complete one sub-inventory, and the preset minimum segment width. Map the nominal position of each tag to the corresponding position segment index. Determine the adaptive protection band according to the standard deviation of the filtered position error, the product of the prediction speed and the online correction delay, the standard deviation of the calibration residual, and the tag density near the segment boundary. Add the tags whose nominal positions are located within the adaptive protection band of the segment boundary to the adjacent position segments.
[0031] Step S403: Activate the corresponding tags sequentially according to the location segmentation order for inventory; split the location segment when the collision rate of the location segment exceeds the preset first collision rate threshold; switch the RFID reader's operating frequency or adjust the transmission power within the preset transmission power range when the collision rate exceeds the preset second collision rate threshold within the continuous inventory period; and perform targeted inventory replenishment for unidentified tags in the high-collision window; record the single-round successful identification flag, reading position estimate, and single-round reading received signal strength indication mean of each tag; and associate the filtered position error standard deviation, the filtered acceleration, the filtered acceleration change rate, the nominal position, the preset received signal strength indication effective threshold, and the preset received signal strength indication saturation threshold with the records to form the tag inventory reading record.
[0032] Preferably, step S50, which involves performing the file box presence determination task based on the tag inventory read records using a dual determination mechanism of recursive Bayesian confidence and continuous missed reads, and outputting the file box presence identification result, specifically includes:
[0033] Step S501: For each tag, determine the single-round successful identification flag based on the comparison between the number of valid reads in a single round and the preset minimum number of valid reads; determine the position consistency weight based on the magnitude of the deviation between the estimated reading position and the corresponding nominal position relative to the standard deviation of the filter position error; and determine the reading quality weight based on the relative position of the average value of the received signal strength indication in a single round between the preset effective threshold and the preset saturation threshold of the received signal strength indication.
[0034] Step S502: Determine the adaptive learning rate based on the weighted normalized result of the filtering acceleration and the rate of change of filtering acceleration from the previous round relative to the preset base learning rate, and recursively determine the label. In the In-position reliability of the wheel:
[0035]
[0036] in, For tags In the The reliability of the wheel in position. For tags In the The reliability of the wheel in position. For the first The adaptive learning rate of the wheel, For tags In the The single-wheel successful identification mark of the wheel, For tags In the Wheel position consistency weight, For tags In the The read quality weight of the wheel, For label serial numbers, This is the inventory cycle number;
[0037] Step S503: Update the continuous missed reading count according to the single-round successful recognition flag of the current round. When the current round of recognition is successful, the continuous missed reading count is cleared to zero. When the current round of recognition fails, the continuous missed reading count of the previous round is incremented by one. The in-position confidence is compared with the preset in-position confidence threshold and the preset out-of-position confidence threshold, and the continuous missed reading count is compared with the preset maximum number of continuous missed readings. Based on the comparison results, the in-position recognition results of the file box are output as follows: in-shelf, in-shelf and triggered targeted review, out-of-shelf, or pending and triggered review and inventory.
[0038] This invention also provides a file box location identification system based on RFID read / write timing and shelf position feedback, comprising:
[0039] The probability boundary calibration module is used to acquire file box tag reading event data, and based on the file box tag reading event data, to perform radio frequency effective reading boundary calibration task using the logistic probability boundary calibration mechanism, and output archive calibration data packet;
[0040] The position feedback filtering module is used to perform position feedback filtering tasks based on the archived calibration data package using an adaptive Kalman filter and a delay compensation mechanism, and output a dynamic identification parameter set.
[0041] The read / write timing calculation module is used to perform read / write timing calculation tasks based on the dynamic identification parameter set using kinematic extrapolation and window overlap pruning mechanism, and outputs a dynamic read window parameter set.
[0042] The tag storage module is used to perform tag storage tasks based on the dynamic reading window parameter set using a dynamic frame length anti-collision and a three-dimensional scheduling mechanism combining space, time slot and frequency points, and output tag storage reading records.
[0043] The in-situ determination module is used to perform the in-situ determination task of the file box based on the tag storage reading record, using a dual determination mechanism of recursive Bayes confidence and continuous missed reading, and output the in-situ identification result of the file box.
[0044] The present invention also provides a file box location identification device based on RFID read / write timing and shelf position feedback, including a memory, a processor, and a file box location identification program based on RFID read / write timing and shelf position feedback stored in the memory and executable on the processor; when the program is executed by the processor, the above-mentioned file box location identification method based on RFID read / write timing and shelf position feedback is implemented.
[0045] The present invention also provides a computer program product, the computer program product including a file box location identification program based on RFID read / write timing and shelf position feedback, the file box location identification program based on RFID read / write timing and shelf position feedback implements the above method when executed by a processor.
[0046] The beneficial effects of this invention are as follows:
[0047] By fitting the reading probability of each tag to the change of the frame position through the logistic probability boundary calibration mechanism, the boundary between the entry and exit of the reading area is obtained. This allows the effective reading area boundary of the radio frequency to be dynamically determined according to the actual response characteristics of the tag, avoiding the misalignment between the fixed empirical boundary and the actual entry and exit time of the tag. This enables the reading and writing window to be accurately triggered when the frame moves at different speeds, reducing the number of tag readings missed.
[0048] An adaptive Kalman filter and delay compensation mechanism is used to process the readings of the frame position sensor. The process noise and measurement noise are adjusted according to the motion mode, and abnormal information is suppressed by the fading factor. At the same time, the communication delay is corrected online, so that the filtered position, velocity and acceleration are closer to the actual motion state of the frame. This provides reliable dynamic parameters for subsequent read and write timing calculations and reduces the interference of position jitter and delay on window triggering.
[0049] The dynamic reading window is calculated by kinematic extrapolation and window overlap pruning mechanism. Combined with dynamic frame length anti-collision and spatial, time slot and frequency joint three-dimensional scheduling to perform tag inventory, the reading window length and start position are adjusted with the real-time movement status of the frame. At the same time, the tag response is dynamically scheduled according to position segmentation and collision rate to reduce time slot conflicts and collisions in high-density tag scenarios and improve the tag reading integrity rate of a single movement inventory. Attached Figure Description
[0050] Figure 1 This is a flowchart illustrating the first embodiment of a file box location identification method based on RFID read / write timing and shelf position feedback according to the present invention. Detailed Implementation
[0051] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0052] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0053] Example 1: As Figure 1 The diagram shown is a flowchart of the first embodiment of the file box location identification method based on RFID read / write timing and shelf position feedback of the present invention. The present invention proposes the first embodiment of the file box location identification method based on RFID read / write timing and shelf position feedback.
[0054] In the first embodiment, the file box location identification method based on RFID read / write timing and shelf position feedback includes:
[0055] Step S10: Obtain file box tag reading event data, and perform radio frequency effective reading boundary calibration task based on the file box tag reading event data using the logistic probability boundary calibration mechanism, and output archive calibration data packet;
[0056] As the electric mobile shelving unit moves at a preset calibrated speed through the effective reading area, the fixed RFID reader continuously collects reading events of the file box tags. Each event record is associated with a synchronized timestamp, tag identifier, number of reads, reading location, received signal strength indicator value, and shelving position sensor reading. Reading events with a received signal strength indicator value not lower than a preset effective threshold are marked as valid reading events. For each tag, the ratio of valid reads to total queries at the shelving location is calculated to obtain the reading probability at that location. Subsequently, a four-parameter logistic model is used to fit the curve of reading probability changing with shelving location. The four parameters represent the reading probability outside the reading area with low noise, the reading probability saturation at the center of the reading area, the midpoint of the reading probability, and the steepness of the reading probability change. By substituting the preset entry and exit probability thresholds into the entry and exit segment fitting curves, respectively, the entry and exit boundaries of each tag in the reading area can be determined. The archive calibration data package establishes a correspondence between file box identifiers, tag identifiers, and nominal locations, and associates and stores the entry into the reading area boundary, exit from the reading area boundary, reading area center saturation reading probability, reading probability change steepness, preset effective threshold for received signal strength indication, and calibration residual standard deviation. The calibration residual standard deviation reflects the random fluctuations during the reading probability curve fitting process, providing an input basis for delay correction in subsequent position feedback filtering and boundary matching in read / write timing calculation.
[0057] The archived calibration data packet converts the effective reading area of the fixed RFID reader into a tag-level boundary corresponding to the rack position, recording the changes in reading range caused by differences in installation location and RF field distribution among different tags in the data packet. Since boundary extraction uses a reading probability threshold as the criterion, the entry and exit positions of the reading area have clear repeatability. Subsequent steps, when calculating the trigger position and safety lead, can directly reference the entry and exit boundaries of the reading area and the standard deviation of the calibration residual from the archived calibration data packet, without relying on uniform reading area parameters. Step S20 can use the difference between the tag's first effective reading position and the entry boundary to correct communication delay online, and step S30 can calculate the trigger position and window length based on the tag's own boundary, thereby reducing timing deviations caused by inaccurate window trigger boundaries and providing a stable boundary basis for subsequent dynamic window scheduling.
[0058] Traditional calibration typically uses a fixed empirical range for the reading area or a uniform reading distance threshold, ignoring the morphological differences in the reading probability curves of different tags. During the speed change or start-stop process of electric mobile shelving, the shelving position feedback itself is subject to jitter, and the actual boundary of the tag entering the reading area is prone to offset from the fixed empirical boundary. This causes the read / write window to be triggered prematurely or delayed, resulting in some tags not entering the reading area before the window starts, or the tag still being within the effective reading area after the window has ended, thus causing missed reads. Step S10 fits the reading probability curve of each tag using a four-parameter logistic model and analyzes the entry and exit boundaries with the same probability threshold, making the boundary data directly correspond to the shelving position. Even if the radio frequency field is non-uniform at different locations on the shelving, the differences in the reading range of different tags are recorded in the archived calibration data package. Subsequent window calculations can use the tag's own boundary for trigger control, reducing response window overlap and missed reads caused by inaccurate boundaries, and providing a reusable boundary basis for tag-level timing control during mobile inventory.
[0059] During calibration, the electric mobile shelving unit passes through the effective reading area of the fixed RFID reader at a preset calibration speed, while the shelving unit's position sensor synchronously records the shelving unit's position. When a file box tag located at the bottom layer moves from a lower position to a higher position on the shelving unit, the reading probability gradually increases from a low level to near saturation. A four-parameter model is fitted to obtain the midpoint position and steepness of the reading probability change. Using a preset reading probability threshold, the entry and exit boundaries of the reading area are analyzed, and the standard deviation of the tag's calibration residual is calculated. After archiving the calibration data package and saving this information, when subsequent shelving units pass through the same area, the system can directly use the tag's entry trigger boundary without re-estimating based on the reader's overall reading area, and without needing to re-determine the effective reading area range due to short-term jitter in the shelving unit's position feedback. This improves the certainty of the window boundary and the stability of the trigger position.
[0060] Step S20: Based on the archived calibration data package, an adaptive Kalman filter and delay compensation mechanism are used to perform a position feedback filtering task, and a dynamic identification parameter set is output;
[0061] The archive calibration data package stores the entry and exit boundaries of the reading area for each file box tag, while the frame position sensor continuously sends timing readings. These two types of data together constitute the input for this step. The filtering process first establishes a state vector composed of frame position, frame velocity, and frame acceleration. The frame acceleration is determined based on the sampling period of the position sensor and the difference in acceleration between adjacent moments, and is used to drive state prediction, obtaining the predicted state value and prediction covariance matrix. By comparing the filtered velocity, filtered acceleration, and rate of change of filtered acceleration at the previous moment with the preset motion state threshold, it can be determined whether the frame is currently in a stationary, uniform, accelerating, decelerating, or impact transition mode. The noise characteristics of the sensor readings differ significantly in different modes, and the process noise covariance and measurement noise variance need to be adaptively adjusted according to the motion mode and the difference in filtered acceleration, filtered velocity, and filtered acceleration between adjacent moments. The position information is obtained by subtracting the current reading of the frame position sensor from the predicted state value. When the information exceeds the anomaly boundary determined by the theoretical information standard deviation and the preset anomaly judgment coefficient, a fading factor is introduced to amplify the prediction covariance, thereby suppressing the impact of abnormal readings on the state update. The communication delay is corrected online using the difference between the entry reading area boundary in the archived calibration data packet and the filtered position when the tag is first effectively read. The corrected delay is used for kinematic compensation of the updated state vector. Finally, the compensated predicted position, predicted velocity, predicted acceleration, predicted jerk, position estimation covariance, velocity estimation covariance, filtered position error standard deviation, and online correction delay together form a dynamic identification parameter set, which includes the current motion state of the frame, as well as uncertainty information of the position estimation and delay estimation, for direct use in step S30 when calculating the dynamic reading window.
[0062] After the dynamic identification parameter set is formed, the frame position feedback is no longer just isolated sensor readings, but a motion state estimate with error statistics and delay correction. In step S30, when performing the read / write timing calculation task based on the dynamic identification parameter set using kinematic extrapolation and window overlap pruning mechanisms, the compensated predicted position, predicted velocity, predicted acceleration, and predicted jerk can be directly used for extrapolation to obtain the frame's position change trend over a future period, thereby determining the trigger position and duration of the dynamic reading window. The position estimation covariance and velocity estimation covariance reflect the reliability of the filtering results, and the standard deviation of the filtered position error provides a statistical basis for calculating the window trigger safety lead, ensuring that the window opening time avoids areas with large position estimation errors. Online delay correction aligns the timestamp of the predicted position with the tag RF reading event, avoiding window calculation deviations caused by communication delays. If the position sensor readings experience short-term fluctuations during frame startup or speed change, the filtered position in the dynamic identification parameter set has already suppressed high-frequency noise, and step S30 will not erroneously trigger the window due to individual abnormal readings. Position error information can also provide a reference for spatial segmentation and protection band settings in step S40, enabling the tag inventory task to avoid areas with uncertain positions in the spatial dimension, reducing missed reads and collisions. Therefore, the dynamic identification parameter set provides a reliable dynamic input for subsequent read / write timing calculations and tag inventory, avoiding window advances or lags caused by position feedback jitter or delay accumulation, thereby reducing the risk of missed and misjudged detection in file box in-situ identification.
[0063] Traditional position feedback processing methods typically use raw readings from the rack's position sensors directly or employ fixed-parameter filtering algorithms to smooth the position signal. During the start-up, shutdown, and speed changes of electric mobile shelving units, the rack's motion changes drastically. Sensor readings may experience short-term jumps due to mechanical vibration and communication delays. Fixed-parameter filtering cannot track the true position in a timely manner, leading to lag or fluctuations in the filtering results during impact phases. For example, when the rack suddenly accelerates, the raw position sensor readings may exhibit significant jitter within a short period. If directly used to trigger the read / write window, the window may open at the wrong location, causing missed or misread tags. Communication delays are usually taken as fixed values or ignored, resulting in inconsistencies between the position feedback timestamp and the tag reading record, further amplifying window calculation errors. This step, without altering the hardware structure, correlates the process noise covariance with the actual motion state of the rack through motion pattern identification. It maintains stable filtering in stationary and constant-speed modes, while increasing the process noise term in acceleration, deceleration, and impact transition modes. This allows the filter to track the true position in a timely manner even during rapid rack changes, while simultaneously suppressing high-frequency jitter. The fading mechanism constrains abrupt changes in abnormal locations, preventing single outliers from contaminating the prediction state. Online delay correction utilizes the external observation of the tag's first valid read position, continuously approximating the communication delay estimate to the true value, aligning the timestamp of the predicted position with the tag's radio frequency read event. After these processes, the position and velocity error information in the dynamic identification parameter set can directly provide a statistical basis for calculating the window trigger safety lead in subsequent step S30, and also provide a position error reference for spatial slicing and guard zone settings in step S40, making subsequent scheduling less prone to window calculation deviations caused by position feedback jitter or delay.
[0064] After completing the calibration in step S10, the electric mobile shelving unit restarts and passes through the effective reading area of the fixed RFID reader. The shelving unit position sensor continuously outputs the shelving unit position reading, and the archive calibration data package stores the entry and exit boundaries of a certain file box tag within the reading area. Step S20 acquires the timing readings of the shelving unit position sensor, constructs a state vector consisting of the shelving unit position, velocity, and acceleration, and determines the shelving unit jerk based on the position sensor sampling period and the difference in acceleration between adjacent moments. The motion state is predicted using the shelving unit jerk, resulting in the predicted state value and the prediction covariance matrix. Based on the comparison between the filtered velocity, filtered acceleration, and rate of change of filtered acceleration at the previous moment and the corresponding preset motion state threshold, it is determined that the shelving unit is currently in acceleration mode. In acceleration mode, the process noise covariance increases accordingly, and the measurement noise variance is also adjusted based on the weighted increment of the absolute value of the shelving unit acceleration, the squared value of the shelving unit velocity, and the squared value of the difference in acceleration between adjacent moments relative to the measurement noise variance reference value. Subsequently, the position information is calculated. When the position information exceeds the anomaly boundary determined by the theoretical information standard deviation and the preset anomaly judgment coefficient, a fading factor is introduced. The predicted covariance matrix and state vector are updated using the process noise covariance, measurement noise variance, and fading factor. The communication delay is corrected online based on the deviation between the entry reading area boundary in the archived calibration data packet and the filtered position when the tag is first effectively read. Kinematic compensation is then performed on the updated state vector according to the communication delay. Finally, the compensated predicted position, predicted velocity, predicted acceleration, predicted jerk, position estimation covariance, velocity estimation covariance, filtered position error standard deviation, and online correction delay are combined to form a dynamic identification parameter set. This dynamic identification parameter set is passed to step S30 to calculate the trigger position and duration of the dynamic reading window, ensuring that the window opens accurately when the tag actually enters the effective reading area.
[0065] Step S30: Based on the dynamic identification parameter set, a kinematic extrapolation and window overlap pruning mechanism is used to perform read / write timing calculation tasks, and output the dynamic read window parameter set;
[0066] After the dynamic identification parameter set is entered into this step, the trigger position of the RFID reader / writer needs to be determined for each file box tag. This trigger position is jointly determined by the tag's entry into the reading area boundary, the trigger position advance, and the safety advance. The trigger position advance is obtained by kinematic extrapolation based on the advance start time (composed of RF on-time, frequency stabilization time, and PLL locking time) and the predicted velocity, predicted acceleration, and predicted jerk from the dynamic identification parameter set. The safety advance is formed by combining the position estimation covariance, velocity estimation covariance, online correction delay, and the standard deviation of the calibration residual in the archived calibration data packet; it is used to absorb the position uncertainty caused by filtering errors, communication delays, and calibration errors. The reading area length is determined by the difference between the tag's departure from the reading area boundary and its entry into the reading area boundary. The physical passage time is calculated based on the relationship between the current acceleration estimate and the preset minimum acceleration threshold, using uniform velocity or uniform acceleration kinematic relationships respectively, and then superimposed with the anti-collision storage time and the reader's termination protection time to obtain the initial reading window length. Subsequently, window overlap trimming is performed. The predicted end position of the current window is compared with the position of the next window trigger position minus the preset minimum protection gap. When the current window intrudes into the starting area of the next window, the maximum allowable window length is determined based on the distance between adjacent trigger positions, the minimum protection gap, the predicted average velocity, and the predicted average acceleration. The initial reading window length is then trimmed to not exceed this value. Windows that are shorter than the preset minimum storage time after trimming are marked as high-collision windows. Finally, the RFID read / write trigger position, final reading window length, predicted velocity, predicted acceleration, standard deviation of filter position error, online correction delay, standard deviation of calibration residual, nominal position, entry into the reading area boundary, exit from the reading area boundary, filter acceleration, and filter acceleration change rate are combined to form a dynamic reading window parameter set.
[0067] After the dynamic read window parameter set is formed, the read / write window of each tag is matched with the actual movement state of the frame and the tag's own boundary. Trigger position advance ensures the reader completes RF activation and frequency stabilization before the tag enters the effective read area, preventing the reader from being unprepared when the tag arrives. Safety advance prevents late window opening due to position estimation errors or communication delays, thus reducing missed reads of the tag header. The read area length and physical passage time are calculated based on the current acceleration estimate, allowing the window duration to change with the frame's acceleration or deceleration, avoiding premature window closure or excessive window occupancy caused by a fixed average speed assumption. The overlap pruning mechanism uses the minimum protective gap between adjacent window trigger positions to constrain the window length, reducing the probability of crosstalk from tags in overlapping areas being simultaneously stored in multiple windows. High-conflict windows are marked as priority processing objects for the subsequent three-dimensional scheduling in step S40, allowing storage resources to be tilted towards areas with insufficient window time or high tag density, improving the overall storage completion rate.
[0068] Traditional read / write timing calculations often use a fixed window length or estimate the passage time based on average speed, without considering the rack's acceleration and position feedback covariance. When the electric mobile shelving unit starts, stops, or changes speed, the fixed window will deviate from the actual time the tag passes through the reading area. When the rack decelerates through the reading area, the actual tag dwell time increases, and the fixed window may close prematurely, leading to missed readings at the end of the tag. When the rack accelerates, the actual tag dwell time decreases, and the fixed window may be too long, encroaching on the next tag's activation area, causing overlapping of adjacent windows and duplicate tag responses. This step uses kinematic extrapolation and error statistical weighting, utilizing predicted acceleration and position error to calculate lead time and window length. It also maintains protective gaps between adjacent windows through overlap trimming, ensuring that the window timing changes in tandem with the rack's position feedback, reducing response window conflicts and missed readings, and providing a more reasonable window boundary for subsequent inventory.
[0069] After completing the position feedback filtering in step S20, the electric mobile shelving unit stores the predicted velocity, predicted acceleration, predicted jerk, position estimation covariance, velocity estimation covariance, standard deviation of filtered position error, and online correction delay for a specific file box tag. The tag's entry into the reading area boundary is located 1300 mm from the shelving unit, and its exit from the reading area boundary is located 1395 mm from the shelving unit. Step S30 determines the advance start time based on the RF turn-on time, frequency stabilization time, and phase-locked loop locking time. The shelving unit movement within the advance start time is extrapolated using the predicted velocity, predicted acceleration, and predicted jerk to obtain the trigger position advance. Simultaneously, a safety advance is determined based on the position estimation covariance, velocity estimation covariance, online correction delay, and standard deviation of calibration residuals. Subtracting the trigger position advance and safety advance from the entry into the reading area boundary yields the RF identification read / write trigger position. The reading area length is 95 mm. The current acceleration estimate is lower than the preset minimum acceleration threshold. The physical passage time is calculated based on uniform motion, and then the anti-collision storage time and the reader's termination protection time are added to obtain the initial reading window length. Starting from the RFID read / write trigger position of the current reading window, kinematic extrapolation is performed using the predicted average velocity, predicted average acceleration, and the initial reading window length to obtain the predicted end position of the current window. This predicted end position is compared with the position of the next tag trigger position minus the preset minimum protection gap. It is found that the current window intrudes into the next window's activation area. Therefore, the maximum allowable window length is determined based on the distance between adjacent trigger positions, the minimum protection gap, the predicted average velocity, and the predicted average acceleration, and the initial reading window length is trimmed to this value. The final trimmed reading window length meets the minimum storage time requirement and is not marked as a high-collision window. The dynamic reading window parameter set is passed to step S40 to accurately open the storage window and complete tag storage when the tag actually enters the effective reading area.
[0070] Step S40: Based on the dynamic reading window parameter set, a dynamic frame length anti-collision and spatial, time slot and frequency joint three-dimensional scheduling mechanism is used to execute the tag inventory task and output the tag inventory reading record;
[0071] After the dynamic reading window parameter set enters this step, the inventory scheduling uses each reading window obtained in step S30 as the basic unit, rather than initiating simultaneously for the entire shelf or all file box tags. The expected number of tags within the window, prediction speed, standard deviation of filter position error, filter acceleration, rate of change of filter acceleration, nominal position, preset effective threshold of received signal strength indication, and preset saturation threshold of received signal strength indication together constitute the input constraints of the inventory scheduling. At the start of inventory, the initial Q value is determined based on the expected number of tags within the window and the preset target throughput. The Q value is the frame length parameter used to determine the number of time slots in a single frame. A single time slot is the time slot available for random selection by tags in this round of inventory. After each frame ends, the number of successfully read time slots, the number of collision time slots, and the number of idle time slots are counted. The number of collision time slots is used to estimate the number of remaining unidentified tags, and then the frame length of the next frame is determined by combining the remaining window time and the preset time slot time. The window remaining time limit prevents the frame length adjustment from increasing indefinitely, avoiding the inventory process from exceeding the read / write window. In the spatial dimension, tags within the window are sorted according to their nominal positions. The number of position segments is determined based on the preset maximum number of tags per segment, and the width of each position segment is determined based on the prediction speed, the average time for each position segment to complete one sub-inventory, and the preset minimum segment width. The nominal position of each tag is mapped to the corresponding position segment index. An adaptive guard band is determined near the segment boundary based on the standard deviation of the filtered position error, the product of the prediction speed and the online correction delay, the standard deviation of the calibration residual, and the tag density near the segment boundary. Tags within the guard band are added to adjacent position segments to prevent position estimation errors from directly causing missed readings of boundary tags. The corresponding tags are activated sequentially according to the position segment order for inventory. In the frequency and power dimensions, when the collision rate of a position segment exceeds a preset first collision rate threshold, the segment is split. When the collision rate exceeds a preset second collision rate threshold within a continuous inventory period, the operating frequency of the RFID reader is switched or the transmission power is adjusted within a preset transmission power range. Unidentified tags within a high-collision window undergo targeted inventory replenishment. Finally, the system records the single-round successful identification flag, reading position estimate, and single-round reading received signal strength indication mean for each tag. The system also associates the standard deviation of the filter position error, filter acceleration, filter acceleration change rate, nominal position, preset effective threshold of received signal strength indication, and preset saturation threshold of received signal strength indication with the records to form a tag inventory reading record.
[0072] After the tag inventory and reading record is formed, the following information is retained: whether each tag in the file box was successfully identified within the current reading window, the estimated position value at the time of identification, and the average value of the received signal strength indication. The record also includes the standard deviation of the filter position error, the filter acceleration, the rate of change of the filter acceleration, and the nominal position corresponding to the tag. Step S50 can no longer be judged solely based on the success or failure of a single reading, but can perform recursive Bayesian confidence updates based on multiple rounds of inventory records. The dual-judgment mechanism for continuous missed readings needs to distinguish between two situations: tags not being read due to temporarily leaving the effective reading area and tags not being read due to collisions or scheduling omissions. The estimated position value and the average value of the received signal strength indication in the tag inventory and reading record can provide a basis for this distinction. If a tag does not show a successful identification mark in multiple rounds of inventory, but the estimated reading position value is always near the effective reading area, it is more likely that the missed reading is caused by collisions or insufficient scheduling; if the estimated reading position value has moved far away from the effective reading area, it is more likely that the tag has left the reading range with the file box. Dynamic frame length adjustment and spatial segmentation reduce tag response conflicts within the window and improve the integrity of successfully identified markers in the tag inventory reading record. When performing confidence recursion in step S50, more positive samples and more reliable missed reading observations can be obtained, thereby avoiding misjudgments caused by directly determining that the file box is not in place due to a single round of missed reading.
[0073] Traditional tag inventory management often uses fixed frame lengths or a uniform full-window activation method. When tags are densely packed, the number of collision slots increases, and some tags may not be completed before the window ends, leading to missed reads. When there is no spatial fragmentation, tags in different locations respond simultaneously, significantly increasing the probability of collisions; fixed frequencies and power make it impossible to schedule for high-collision local areas. This step introduces dynamic frame length, location fragmentation, adaptive guard bands, and high-collision scheduling based on dynamic reading window parameters. Dynamic frame length ensures that the number of slots within the window matches the number of remaining unidentified tags, avoiding excessive collisions when tags are dense or wasted slots when tags are sparse. Location fragmentation reduces the number of tags responding simultaneously, and the adaptive guard band provides redundant scheduling for tags near the fragmentation boundary, reducing the risk of boundary missed reads caused by location errors. Collision splitting, frequency switching, and power adjustment further reduce the collision rate in areas with concentrated collisions, and targeted inventory management in high-collision windows allows tags that have not completed their inventory management within the window to have another chance to be inventoried. These processes reduce tag response collisions within the window and improve the tag inventory completion rate per unit window without adding additional hardware, providing more sufficient multi-round reading records for subsequent on-the-spot determination. Especially when the labels of multi-layer and multi-column file boxes are densely distributed, location segmentation can limit the simultaneous response of labels to a small range, reduce the collision rate, and make the suppression of missed readings during the mobile inventory process not only by extending the window time, but also by the coordination of space and time slots.
[0074] After completing the read / write timing calculation in step S30, the electric mobile shelving unit stores the start and end positions of the read window for a specific file box tag, predicted velocity, predicted acceleration, predicted jerk, standard deviation of filter position error, filter acceleration, rate of change of filter acceleration, nominal position, preset effective threshold for received signal strength indication, and preset saturation threshold for received signal strength indication. The read window length for this tag is trimmed to meet the minimum storage time requirement and is not marked as a high-collision window. At the start of step S40, the expected number of tags within the window is provided by the dynamic read window parameter set. The initial Q value is determined based on the expected number of tags and the preset target throughput, and this Q value determines the number of time slots in the initial frame. After the first frame is stored, the number of successfully read time slots, collision time slots, and idle time slots are counted. The number of collision time slots is used to estimate the remaining unidentified tags, and the frame length of the next frame is determined by combining the remaining window time and the preset time slot time. After the tags within the window are sorted according to their nominal positions, they are divided into several position segments. Tags near the segment boundaries are added to adjacent segments according to the adaptive guard band. Then, the corresponding tags are activated sequentially according to the position segment order for inventory. During the inventory process, if the collision rate of a certain position segment exceeds a preset first collision rate threshold, the segment is split into two sub-segments and the inventory continues. If the collision rate still exceeds a preset second collision rate threshold within a continuous inventory cycle, the RFID reader switches its operating frequency and adjusts its transmission power within a preset transmission power range. Unidentified tags in the high-collision window are re-acquired. When the window ends, the tag is successfully identified. The single-round successful identification flag, the estimated reading position value, and the average value of the single-round received signal strength indication are recorded. The standard deviation of the filtered position error, the filtered acceleration, the rate of change of the filtered acceleration, the nominal position, the preset effective threshold of the received signal strength indication, and the preset saturation threshold of the received signal strength indication are associated with the record to form a tag inventory reading record, which is then passed to step S50.
[0075] Step S50: Based on the tag inventory reading record, the file box presence determination task is performed using a dual determination mechanism of recursive Bayes confidence and continuous missed reading, and the file box presence identification result is output.
[0076] After the tag inventory reading records reach this step, the in-situ status of each tag is determined individually. Within one inventory cycle, if the number of times a tag is effectively read reaches the preset minimum number of effective reads, the single-cycle successful identification flag is set to one; otherwise, it is set to zero. The position consistency weight is jointly determined by the deviation between the estimated read position and the nominal position, and the standard deviation of the filtered position error. A smaller deviation and a more stable standard deviation result in a higher weight. The read quality weight is determined based on the position of the average received signal strength indication in a single cycle between the preset effective threshold and the preset saturation threshold. If it is below the effective threshold, the weight is zero; if it is above the saturation threshold, the weight is one; and if it is in between, the weight is determined by the relative position. The adaptive learning rate is affected by the filtering acceleration and the rate of change of the filtering acceleration in the previous cycle. The more obvious the frame acceleration or impact transition, the lower the learning rate. The position confidence is recursively fused with the position confidence in the previous cycle, the single-cycle successful identification flag in this cycle, the position consistency weight, and the read quality weight according to the adaptive learning rate to form a continuous estimate of the tag's in-situ status. The consecutive missed read count is reset to zero when the current round of recognition is successful, and is accumulated based on the previous round when the current round of recognition fails. Finally, the location confidence score is compared with the preset location confidence threshold and the preset location confidence threshold, and the consecutive missed read count is compared with the preset maximum number of consecutive missed reads. The results of the in-place recognition of the file box are output as follows: in-shelf, in-shelf and triggered targeted review, off-shelf, or pending and triggered review and inventory.
[0077] The file box presence identification result serves as the final conclusion of the entire identification process, providing the file management system with information to update the status of the shelving files. Fluctuations in a single reading round do not directly change the file box status; only when multiple readings consistently point in the same direction will the location confidence exceed the judgment threshold. Location consistency weights and reading quality weights contribute significantly to high-confidence readings, reducing the impact of readings with location deviations or weak received signal strength, preventing low-quality readings from rapidly lowering the location confidence. The adaptive learning rate decreases during shelving acceleration or impact transitions, resulting in slower convergence of location confidence and avoiding state jumps caused by instantaneous interference during dynamic inventory checks. Continuous missed reading counts serve as an independent constraint, distinguishing between temporary and continuous missed readings. Only when both low confidence and continuous missed reading conditions are met is the file box considered off-shelf, avoiding misjudgments due to single-window collisions or signal fluctuations. For cases where the confidence level is between the upper and lower thresholds, the system maintains a pending state and triggers a review inventory, avoiding direct status conclusions based on insufficient information and improving the adaptability of the presence determination to inventory fluctuations.
[0078] Traditional inventory checks often rely on a single reading of a tag to determine whether an item is on or off the shelf. However, in dynamic inventory checks, window collisions, positional errors, and fluctuations in the radio frequency (RF) environment can cause missed readings in a single round, easily misclassifying an item that is actually on the shelf as off, or being overly sensitive to short-term reading fluctuations. This new step employs a dual approach: confidence level progression and continuous missed readings. The status is not immediately changed upon a missed reading in a single round; off-shelf status is only determined when the confidence level remains consistently below the lower threshold and the maximum number of consecutive missed readings is reached. For cases where the confidence level is between the upper and lower thresholds, a review inventory is triggered. This preserves the cumulative effect of multiple reading rounds while preventing the status from remaining in an uncertain region for extended periods through continuous missed readings, reducing misjudgments and missed judgments. Compared to simply relying on the success of a single reading, this mechanism is more suitable for item on-shelf identification under conditions of variable-speed shelf movement, dense tag distribution, and limited RF windows, resulting in more stable output results.
[0079] After the electric mobile shelving completes the tag inventory in step S40, the inventory reading record of a certain file box tag saves the single-round successful identification flag, reading position estimate, single-round reading received signal strength indication mean, filter position error standard deviation, filter acceleration, filter acceleration change rate, nominal position, preset received signal strength indication effective threshold, and preset received signal strength indication saturation threshold. This tag was successfully identified in this inventory round, the single-round successful identification flag is set to one, the reading position estimate deviates little from the nominal position, the position consistency weight is high, the single-round reading received signal strength indication mean is between the effective threshold and the saturation threshold, and the reading quality weight is determined according to the relative position. The previous round's filter acceleration and filter acceleration change rate were small, and the adaptive learning rate remained at a high level. The position confidence level increased progressively from the previous round's value, but remained below the preset position confidence threshold but above the preset distance position confidence threshold, and the continuous missed reading count was reset to zero. The system outputs "pending" and triggers a review inventory, and the file management system schedules the next inventory round to continue observing this tag. If the tag is successfully read consecutively in subsequent rounds, and the location confidence exceeds the preset location confidence threshold, the system outputs "on shelf". If another tag fails to be read for multiple consecutive rounds, and the consecutive missed read count reaches the preset maximum number of consecutive missed reads, the location confidence will decrease recursively due to each round of failure until it falls below the preset location confidence threshold. In this case, the system outputs "off shelf" and triggers a review.
[0080] Example 2: Furthermore, the present invention provides a file box location identification system based on RFID read / write timing and shelf position feedback, employing a file box location identification method based on RFID read / write timing and shelf position feedback from the above embodiments. This system can solve the technical problem of file box location identification based on RFID read / write timing and shelf position feedback. The beneficial effects of the file box location identification system based on RFID read / write timing and shelf position feedback provided by the present invention are the same as those of the file box location identification method based on RFID read / write timing and shelf position feedback provided in the above embodiments. Other technical features of the file box location identification system based on RFID read / write timing and shelf position feedback are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.
[0081] In this embodiment, the system includes a probability boundary calibration module, a position feedback filtering module, a read / write timing calculation module, a tag storage module, and an in-situ determination module, with each module transmitting processing results sequentially.
[0082] The probability boundary calibration module acquires file box tag reading event data, and performs radio frequency effective reading boundary calibration task based on the file box tag reading event data using the logistic probability boundary calibration mechanism, and outputs archive calibration data packet.
[0083] The position feedback filtering module performs the position feedback filtering task based on the archived calibration data package, using an adaptive Kalman filter and a delay compensation mechanism, and outputs a dynamic identification parameter set.
[0084] Based on the dynamically identified parameter set, the read / write timing calculation module performs the read / write timing calculation task using kinematic extrapolation and window overlap pruning mechanism, and outputs the dynamic read window parameter set.
[0085] The tag storage module, based on the dynamic reading window parameter set, uses a dynamic frame length anti-collision and a three-dimensional scheduling mechanism combining space, time slot and frequency points to execute tag storage tasks and output tag storage reading records.
[0086] The in-situ determination module reads records from the tag inventory and performs the in-situ determination task of the file box using a dual determination mechanism of recursive Bayesian confidence and continuous missed reading, and outputs the in-situ identification result of the file box.
[0087] Example 3: This invention provides a file box location identification device based on RFID read / write timing and shelf position feedback. The device includes at least one processor and a memory communicatively connected to the processor. The memory stores instructions executable by the processor, which are then executed to enable the processor to perform the file box location identification method based on RFID read / write timing and shelf position feedback described in Example 1. The file box location identification device based on RFID read / write timing and shelf position feedback in this invention can include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. This file box location identification device based on RFID read / write timing and shelf position feedback is merely an example and should not limit the functionality or scope of the invention. A file box location identification device based on RFID read / write timing and shelf position feedback may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) that can perform various appropriate actions and processes according to a program stored in read-only memory or a program loaded from a storage device into random access memory. The random access memory also stores various programs and data required for the operation of the file box location identification device based on RFID read / write timing and shelf position feedback. The processing unit, read-only memory, and random access memory are interconnected via a bus. An I / O interface is also connected to the bus. Typically, the following systems can be connected to the I / O interface: input devices including touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices including liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices including magnetic tapes, hard disks, etc.; and communication devices. The communication device allows the file box location identification device based on RFID read / write timing and shelf position feedback to exchange data with other devices wirelessly or via wired communication. While a file box location identification device based on RFID read / write timing and shelf position feedback with various systems has been described, it should be understood that implementation of all the described systems is not required. More or fewer systems may be implemented alternatively.
[0088] Example 4: This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method for identifying the location of a file box based on RFID read / write timing and shelf position feedback. The computer program product provided by this invention can solve the technical problem of identifying the location of a file box based on RFID read / write timing and shelf position feedback. Compared with the prior art, the beneficial effects of the computer program product provided by this invention are the same as those of the above-described method for identifying the location of a file box based on RFID read / write timing and shelf position feedback, and will not be repeated here.
[0089] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a read-only memory. When the computer program is executed by a processing device, it performs the functions defined in the methods of the embodiments disclosed in this invention.
[0090] It should be understood that the various parts disclosed in this invention can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.
[0091] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the present invention and its equivalents, the present invention also intends to include these modifications and variations.
Claims
1. A method for identifying the location of file boxes based on RFID read / write timing and shelf position feedback, characterized in that, The method includes: Step S10: Obtain file box tag reading event data, and perform radio frequency effective reading boundary calibration task based on the file box tag reading event data using the logistic probability boundary calibration mechanism, and output archive calibration data packet; Step S20: Based on the archived calibration data package, an adaptive Kalman filter and delay compensation mechanism are used to perform a position feedback filtering task, and a dynamic identification parameter set is output; Step S30: Based on the dynamic identification parameter set, a kinematic extrapolation and window overlap pruning mechanism is used to perform read / write timing calculation tasks, and output the dynamic read window parameter set; Step S40: Based on the dynamic reading window parameter set, a dynamic frame length anti-collision and spatial, time slot and frequency joint three-dimensional scheduling mechanism is used to execute the tag inventory task and output the tag inventory reading record; Step S50: Based on the tag inventory reading record, the file box presence determination task is performed using a dual determination mechanism of recursive Bayes confidence and continuous missed reading, and the file box presence identification result is output.
2. The method for identifying the location of file boxes based on RFID read / write timing and shelf position feedback as described in claim 1, characterized in that, Step S10, which involves acquiring file box tag reading event data, performing radio frequency effective read boundary calibration tasks based on the file box tag reading event data using a logistic probability boundary calibration mechanism, and outputting an archive calibration data packet, specifically includes: Step S101: Control the electric mobile shelving to pass through the effective reading area of the fixed RFID reader at a preset calibration speed, and simultaneously record the tag identification, number of reads, reading position, received signal strength indication value and the corresponding shelf position sensor reading at the time of each tag, and mark the reading event with the received signal strength indication value not lower than the preset effective threshold as a valid reading event, forming the file box tag reading event data; Step S102: For each tag, calculate the ratio of valid reads to total queries according to the shelf position to obtain the read probability of the corresponding position, and use a four-parameter logistic model to fit the curve of read probability changing with shelf position to obtain the reading probability outside the reading area, the reading probability saturation at the center of the reading area, the midpoint of the reading probability, and the steepness of the reading probability change. Step S103: Substitute the preset entry probability threshold and the preset exit probability threshold into the entry segment fitting curve and the exit segment fitting curve, respectively, to determine the entry and exit boundaries of each tag; establish the correspondence between the file box identifier, the tag identifier, and the nominal position, and associate the entry boundary, the exit boundary, the reading center saturation probability, the reading probability change steepness, the preset effective threshold for received signal strength indication, and the standard deviation of the calibration residual to form the archive calibration data packet.
3. The method for identifying the location of file boxes based on RFID read / write timing and shelf position feedback as described in claim 1, characterized in that, Step S20, which involves performing a position feedback filtering task based on the archived calibration data packet using an adaptive Kalman filter and delay compensation mechanism, and outputting a dynamic identification parameter set, specifically includes: Step S201: Obtain the time-series readings of the frame position sensor, construct a state vector consisting of frame position, frame velocity and frame acceleration, determine the frame jerk based on the sampling period of the position sensor and the difference in frame acceleration between adjacent times, use the frame jerk to predict the motion state of the state vector, and obtain the state prediction value and prediction covariance matrix. Step S202: Obtain the filtered velocity, filtered acceleration, and rate of change of filtered acceleration at the previous moment, and determine the stationary mode, uniform speed mode, acceleration mode, deceleration mode, or impact transition mode based on the comparison results between the filtered velocity, filtered acceleration, and rate of change of filtered acceleration at the previous moment and the corresponding preset motion state threshold; adaptively determine the process noise covariance and measurement noise variance based on the determined motion mode and the difference between the filtered acceleration, the filtered velocity, and the filtered acceleration at adjacent moments. Step S203: Determine the position information based on the predicted state value and the current reading of the frame position sensor. When the position information exceeds the anomaly boundary determined by the theoretical information standard deviation and the preset anomaly judgment coefficient, introduce a fading factor. Update the predicted covariance matrix and the state vector using the process noise covariance, the measurement noise variance, and the fading factor. Correct the communication delay online based on the deviation between the entry reading area boundary in the archived calibration data packet and the filtered position when the tag is first effectively read. Perform kinematic compensation on the updated state vector according to the communication delay. Combine the compensated predicted position, predicted velocity, predicted acceleration, predicted jerk, position estimation covariance, velocity estimation covariance, filtered position error standard deviation, and online correction delay to form the dynamic identification parameter set.
4. The method for identifying the location of file boxes based on RFID read / write timing and shelf position feedback as described in claim 3, characterized in that, In step S202, when adaptively determining the process noise covariance based on the motion mode, the filtered acceleration, the filtered velocity, and the difference between the filtered accelerations at adjacent time points, the process noise covariance is: in, For the first Process noise covariance at time step This is the process noise reference matrix. , and These are the process noise matrices corresponding to the impact transition mode, acceleration mode, and deceleration mode, respectively. , and In the shock transition mode Acceleration Mode and deceleration mode The default mode weights enabled below. The rate of change of filtered acceleration. For filtered acceleration, The filter time number is used; the measurement noise variance is determined by the weighted increment of the absolute value of the frame acceleration, the square value of the frame velocity, and the square value of the difference between the frame accelerations at adjacent times relative to the measurement noise variance reference value.
5. The method for identifying the location of file boxes based on RFID read / write timing and shelf position feedback as described in claim 3, characterized in that, In step S30, the step of performing read / write timing calculation based on the dynamic identification parameter set using kinematic extrapolation and window overlap pruning mechanism, and outputting the dynamic reading window parameter set, specifically includes: Step S301: Determine the advance start time based on the RFID reader's RF turn-on time, frequency stabilization time, and phase-locked loop lock time. Extrapolate the frame movement within the advance start time using the predicted velocity, predicted acceleration, and predicted jerk from the dynamic identification parameter set to obtain the trigger position advance amount. Determine the safety advance amount based on the position estimation covariance, velocity estimation covariance, online correction delay, and the standard deviation of the calibration residual in the archived calibration data packet. Finally, determine the RFID read / write trigger position based on the boundary of each tag's entry into the reading area, the trigger position advance amount, and the safety advance amount. Step S302: Determine the reading area length based on the entry and exit boundaries of each tag; when the current acceleration estimate is less than a preset minimum acceleration threshold, determine the physical passage time based on the ratio of the reading area length to the predicted velocity when entering the reading area; when the current acceleration estimate is not less than the preset minimum acceleration threshold, determine the physical passage time based on the kinematic relationship between the reading area length, the predicted velocity, and the current acceleration estimate; add the physical passage time, the anti-collision storage time, and the reader termination protection time to obtain the initial reading window length. Step S303: Determine the prediction end position based on the RFID read / write trigger position of the current reading window, the predicted average speed, the predicted average acceleration, and the initial reading window length. Compare the prediction end position with the position obtained by subtracting the preset minimum protection gap from the RFID read / write trigger position of the next reading window. When the current reading window intrudes into the starting area of the next reading window, determine the maximum allowable window length based on the distance between adjacent RFID read / write trigger positions, the preset minimum protection gap, the predicted average speed, and the predicted average acceleration. Determine the smaller value between the initial reading window length and the maximum allowable window length as the final reading window length. Mark the final reading window shorter than the preset minimum inventory time as a high-conflict window. Combine the RFID read / write trigger position, the final reading window length, the predicted speed, the predicted acceleration, the standard deviation of the filter position error, the online correction delay, the standard deviation of the calibration residual, the nominal position, the entry boundary of the reading area, the exit boundary of the reading area, the filter acceleration, and the rate of change of the filter acceleration to form the dynamic reading window parameter set.
6. The method for identifying the location of file boxes based on RFID read / write timing and shelf position feedback as described in claim 5, characterized in that, Step S40, which involves executing the tag inventory task based on the dynamic reading window parameter set using a dynamic frame length anti-collision and a joint three-dimensional scheduling mechanism of space, time slot, and frequency, and outputting the tag inventory reading record, specifically includes: Step S401: Determine the initial Q value based on the expected number of tags in the reading window, the preset target throughput, and the preset Q value range. The Q value is a frame length parameter used to determine the number of time slots in a single frame. Perform tag inventory in units of frames within the reading window. After each frame is completed, count the number of successfully read time slots, the number of collision time slots, and the number of idle time slots. Estimate the remaining number of unidentified tags based on the number of collision time slots. Determine the Q value of the next frame based on the remaining number of unidentified tags, the preset target throughput, the remaining time of the reading window, and the preset time slot time. Step S402: Sort the tags in the reading window according to the nominal position, determine the number of position segments according to the preset maximum number of tags per segment, and determine the width of the position segments according to the prediction speed, the average time for each position segment to complete one sub-inventory, and the preset minimum segment width. Map the nominal position of each tag to the corresponding position segment index. Determine the adaptive protection band according to the standard deviation of the filtered position error, the product of the prediction speed and the online correction delay, the standard deviation of the calibration residual, and the tag density near the segment boundary. Add the tags whose nominal positions are located within the adaptive protection band of the segment boundary to the adjacent position segments. Step S403: Activate the corresponding tags sequentially according to the location segmentation order for inventory; split the location segment when the collision rate of the location segment exceeds the preset first collision rate threshold; switch the RFID reader's operating frequency or adjust the transmission power within the preset transmission power range when the collision rate exceeds the preset second collision rate threshold within the continuous inventory period; and perform targeted inventory replenishment for unidentified tags in the high-collision window; record the single-round successful identification flag, reading position estimate, and single-round reading received signal strength indication mean of each tag; and associate the filtered position error standard deviation, the filtered acceleration, the filtered acceleration change rate, the nominal position, the preset received signal strength indication effective threshold, and the preset received signal strength indication saturation threshold with the records to form the tag inventory reading record.
7. The method for identifying the location of file boxes based on RFID read / write timing and shelf position feedback as described in claim 6, characterized in that, Step S50, which involves performing the file box presence determination task based on the tag inventory read records using a dual determination mechanism of recursive Bayesian confidence and continuous missed reads, and outputting the file box presence identification result, specifically includes: Step S501: For each tag, determine the single-round successful identification flag based on the comparison between the number of valid reads in a single round and the preset minimum number of valid reads; determine the position consistency weight based on the magnitude of the deviation between the estimated reading position and the corresponding nominal position relative to the standard deviation of the filter position error; and determine the reading quality weight based on the relative position of the average value of the received signal strength indication in a single round between the preset effective threshold and the preset saturation threshold of the received signal strength indication. Step S502: Determine the adaptive learning rate based on the weighted normalized result of the filtering acceleration and the rate of change of filtering acceleration from the previous round relative to the preset base learning rate, and recursively determine the label. In the In-position reliability of the wheel: in, For tags In the The reliability of the wheel in position. For tags In the The reliability of the wheel in position. For the first The adaptive learning rate of the wheel, For tags In the The single-wheel successful identification mark of the wheel, For tags In the Wheel position consistency weight, For tags In the The read quality weight of the wheel, For label serial numbers, This is the inventory cycle number; Step S503: Update the continuous missed reading count according to the single-round successful recognition flag of the current round. When the current round of recognition is successful, the continuous missed reading count is cleared to zero. When the current round of recognition fails, the continuous missed reading count of the previous round is incremented by one. The in-position confidence is compared with the preset in-position confidence threshold and the preset out-of-position confidence threshold, and the continuous missed reading count is compared with the preset maximum number of continuous missed readings. Based on the comparison results, the in-position recognition results of the file box are output as follows: in-shelf, in-shelf and triggered targeted review, out-of-shelf, or pending and triggered review and inventory.
8. A file box location identification system based on RFID read / write timing and shelf position feedback, applied to the file box location identification method based on RFID read / write timing and shelf position feedback as described in any one of claims 1 to 7, characterized in that, The system includes: The probability boundary calibration module is used to acquire file box tag reading event data, and based on the file box tag reading event data, to perform radio frequency effective reading boundary calibration task using the logistic probability boundary calibration mechanism, and output archive calibration data packet; The position feedback filtering module is used to perform position feedback filtering tasks based on the archived calibration data package using an adaptive Kalman filter and a delay compensation mechanism, and output a dynamic identification parameter set. The read / write timing calculation module is used to perform read / write timing calculation tasks based on the dynamic identification parameter set using kinematic extrapolation and window overlap pruning mechanism, and outputs a dynamic read window parameter set. The tag storage module is used to perform tag storage tasks based on the dynamic reading window parameter set using a dynamic frame length anti-collision and a three-dimensional scheduling mechanism combining space, time slot and frequency points, and output tag storage reading records. The in-situ determination module is used to perform the in-situ determination task of the file box based on the tag storage reading record, using a dual determination mechanism of recursive Bayes confidence and continuous missed reading, and output the in-situ identification result of the file box.
9. A file box location identification device based on RFID read / write timing and shelf position feedback, characterized in that, The file box location identification device based on RFID read / write timing and shelf position feedback includes: a memory, a processor, and a file box location identification program based on RFID read / write timing and shelf position feedback stored in the memory and executable on the processor. When the file box location identification program based on RFID read / write timing and shelf position feedback is executed by the processor, it implements the file box location identification method based on RFID read / write timing and shelf position feedback as described in any one of claims 1 to 7.
10. A computer program product, characterized in that, The computer program product includes a file box location identification program based on RFID read / write timing and shelf position feedback. When the file box location identification program based on RFID read / write timing and shelf position feedback is executed by the processor, it implements a file box location identification method based on RFID read / write timing and shelf position feedback as described in any one of claims 1 to 7.
