A method, device and medium for mobile robot positioning and shelf recognition

Abstract

This invention discloses a method, device, and medium for mobile robot alignment with shelving, relating to the field of robot perception and control technology. The method includes: an industrial control system receiving alignment tasks and cycle constraints, generating control time slots and safety interlock conditions; collecting perception data within the control time slots and performing time alignment and coordinate unification to obtain a scene evidence set, a task constraint set, and a motion state package; under the constraints of the scene evidence set, task constraint set, and motion state package, extracting anchor points of the shelving structure and generating a pool of reachable candidate poses; forming candidate confidence scores based on anchor point matching consistency and comparing them with a budget threshold; and outputting a budget status flag and a verification trigger flag. This invention achieves temporal consistency constraints for alignment recognition, command issuance, and effective windows by generating control time slots and safety interlock conditions under the industrial control system and implementing time alignment and coordinate unification of perception data, thus avoiding alignment fluctuations and sudden changes in action caused by temporal drift.

Description

Technical Field

[0001] This invention relates to the field of robot perception and control technology, and in particular to a method, device and medium for mobile robot positioning and shelf recognition. Background Technology

[0002] Mobile robots undertake tasks such as picking, handling, and replenishment in warehousing and discrete manufacturing logistics. Shelf positioning recognition, as a key prerequisite for entering workstations and completing pick-and-place operations, has been continuously developing along with the evolution of sensors from single ranging to multi-source fusion, environmental characterization from regular geometry to structured features, and scheduling control from single-machine planning to rhythmic collaboration relying on industrial control systems. With the increase in warehouse density and the acceleration of operation rhythm, positioning recognition has shifted from "being able to recognize" to "being able to stably recognize and verify within a limited time sequence." The focus is gradually shifting to reliably outputting the relative pose and positioning judgment criteria of the shelves that can be used for control execution under complex occlusion, reflective materials, similar structures, and dynamic disturbance conditions.

[0003] Existing methods have shortcomings. The alignment recognition process is not sufficiently coupled with the control cycle, trigger time slot, and command effective window of the industrial control system. Timing drift and delayed results are prone to occur between perception update, pose calculation, and command issuance, leading to abrupt changes in alignment actions and decreased stability. In addition, the recognition results lack quantifiable access and executability verification for the execution constraints of the industrial control system. They often rely on single scores or fixed thresholds, making it difficult to ensure both reliability and executability under constraints such as occlusion, similar structures, and safety boundaries. This can easily lead to misalignment, risk of exceeding limits, and conservative shutdown. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a mobile robot alignment shelf recognition method to solve the problems of unstable alignment timing and misalignment risk.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for mobile robot alignment with a shelf, comprising: an industrial control system receiving alignment tasks and cycle constraints, generating control time slots and safety interlock conditions, collecting sensing data within the control time slots and completing time alignment and coordinate unification to obtain a scene evidence set, a task constraint set, and a motion state data package; under the constraints of the scene evidence set, task constraint set, and motion state data package, extracting shelf structure anchor points and generating a reachable candidate pose pool, forming candidate confidence scores based on anchor point matching consistency and comparing them with a budget threshold, and outputting a budget status flag and a verification trigger flag; the verification trigger flag and the budget status flag are subject to safety interlock conditions. The locking conditions and motion state data constraints are used to perform micro-motion verification actions. The perception data is re-acquired and the scene evidence set is updated uniformly according to time and coordinates. The credible pose, tolerance window and alignment confirmation conditions are determined in the reachable candidate pose pool to form an alignment control table. The alignment control table, combined with the budgeted state flag, generates correction instructions under the motion state data and safety interlock conditions. The correction instructions are processed by limiting amplitude and variation before being issued, so that the pose error of the robot relative to the credible pose enters the tolerance window and the alignment confirmation condition is determined. If the alignment confirmation condition is met, the shelf alignment recognition result and operation status are output. Otherwise, a conservative alignment instruction is issued and a re-recognition flag is output.

[0008] In a preferred embodiment of the mobile robot positioning shelf recognition method of the present invention, the specific steps for generating the control time slot and safety interlock conditions are as follows:

[0009] The industrial control system receives the alignment task and cycle constraints and performs alignment task normalization and cycle constraint timing processing to obtain the target shelf type, target working area, allowed approach direction, alignment termination condition, control cycle and trigger time.

[0010] The start and end times of the execution time slot are calculated and bound to the time slot based on the control cycle and trigger time to form a control time slot;

[0011] Based on the target work area, the allowed approach direction, and the alignment termination conditions, interlock rules are generated and interlock conditions are solidified to form safe interlock conditions.

[0012] As a preferred embodiment of the mobile robot positioning shelf recognition method of the present invention, the specific steps for obtaining the scene evidence set, task constraint set, and motion state data package are as follows:

[0013] Collect sensing data within the control time slot and perform time alignment and coordinate unification to form a scene evidence set;

[0014] The target shelf type, target work area, allowed access direction, alignment termination condition and safety interlock condition are compiled into a task constraint set;

[0015] The pose and velocity information within the control time slot is extracted to form a motion state data packet.

[0016] In a preferred embodiment of the mobile robot positioning shelf recognition method of the present invention, the specific steps for generating the reachable candidate pose pool are as follows:

[0017] Based on the scene evidence set, task constraint set and motion state data package, evidence screening and noise suppression processing are performed, and line segment fitting and contour extraction are performed on the screened evidence to obtain a set of structural features.

[0018] Based on the target shelf type, the structural feature set is subjected to structural consistency screening and clustering merging to extract the shelf structure anchor point set, forming the shelf structure anchor point set;

[0019] Anchor point pairing and topology combination are performed on the set of anchor points of the shelf structure to obtain the original candidate pose pool;

[0020] Based on the target working area, allowed approach direction, and pose and velocity boundaries in the motion state data packet, reachability verification and non-executable item elimination are performed to obtain the reachable candidate pose pool.

[0021] As a preferred embodiment of the mobile robot positioning shelf recognition method of the present invention, the specific steps for outputting the budget status flag and the verification trigger flag are as follows:

[0022] Based on the set of anchor points of the shelf structure and the pool of reachable candidate poses, perform anchor point matching consistency calculation and residual statistics to form candidate confidence scores;

[0023] The candidate confidence level is compared with the budget threshold to generate budget status markers and verification trigger markers.

[0024] In a preferred embodiment of the mobile robot alignment shelf recognition method of the present invention, the specific steps for forming the alignment control table are as follows:

[0025] The check trigger flag and budget status flag are combined with safety interlock conditions and motion status data to perform an executability determination, and a micro-motion check action sequence is generated when the executability determination is allowed;

[0026] The industrial control system executes a micro-motion verification action sequence and re-collects sensing data. It updates the scene evidence set in a unified manner according to time alignment and coordinates. Based on the updated scene evidence set and the reachable candidate pose pool, it performs anchor point matching consistency calculation to update the candidate confidence and obtain the candidate confidence update result.

[0027] Based on the candidate confidence update results, a reliable pose is determined in the reachable candidate pose pool, and a tolerance window and alignment confirmation conditions are generated based on the reliable pose.

[0028] The reliable pose, tolerance window, and alignment confirmation conditions are compiled into an alignment control table.

[0029] As a preferred embodiment of the mobile robot alignment shelf recognition method of the present invention, the specific steps for outputting the shelf alignment recognition result and the operating status are as follows:

[0030] The pose error of the robot relative to the reliable pose is calculated based on the reliable pose in the alignment control table and the current position and current orientation in the motion state packet.

[0031] The positional error is calculated and processed to generate the initial value of the correction command. The amplitude and variation are limited in combination with the motion state data packet. At the same time, the interlock verification is performed in combination with the safety interlock conditions to obtain the correction command and the issueable mark.

[0032] When marked as allowed, a correction command can be issued and the pose error can be put into the tolerance window. Based on the alignment confirmation conditions, the shelf alignment recognition result and operation status can be determined and output.

[0033] When the flag is set to prohibited or the alignment confirmation condition is not met, a conservative alignment instruction is issued and a re-identification flag is output.

[0034] As a preferred embodiment of the mobile robot alignment shelf recognition method of the present invention, the budget threshold refers to a candidate confidence threshold fixed corresponding to the target shelf type. The budget threshold is determined by alignment verification as the lowest candidate confidence corresponding to the verification result that satisfies the alignment termination condition and the safety interlock condition remains permissible.

[0035] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein the computer program, when executed by the processor, implements any step of the mobile robot alignment shelf recognition method as described in the first aspect of the present invention.

[0036] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the mobile robot alignment shelf recognition method as described in the first aspect of the present invention.

[0037] The beneficial effects of this invention are as follows: By generating control time slots and safety interlock conditions under the industrial control system and implementing time alignment and coordinate unification of the sensing data, the timing consistency constraints of alignment recognition, command issuance, and effective windows are realized, avoiding alignment fluctuations and sudden changes in action caused by timing drift, and improving the stability and predictability of the work cycle; by constructing a reachable candidate pose pool and triggering micro-motion verification and alignment confirmation condition determination by comparing candidate confidence and budget threshold, the access verifiability and executability of the recognition results are guaranteed, reducing the risk of misalignment and out-of-bounds, reducing repeated trial and error and downtime waiting, and improving the alignment success rate and consistency. Attached Figure Description

[0038] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart of a method for mobile robots to identify and align shelves.

[0040] Figure 2 A flowchart for generating a pool of reachable candidate poses.

[0041] Figure 3 A flowchart for creating the alignment control table.

[0042] Figure 4 A flowchart for generating shelf alignment recognition results, operation status, and re-identification markers.

[0043] Figure 5 A comparison chart of command out-of-bounds data under communication jitter conditions.

[0044] Figure 6 A data comparison chart is constructed for alignment results under different occlusion ratios.

[0045] Figure 7 This is a comparison chart of alignment error timing and local differences. Detailed Implementation

[0046] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0047] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0048] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0049] Reference Figures 1-7 As one embodiment of the present invention, this embodiment provides a method for mobile robot positioning and shelf recognition, including the following steps:

[0050] S1. The industrial control system receives the alignment task and cycle constraints, generates control time slots and safety interlock conditions, collects sensing data within the control time slots and completes time alignment and coordinate unification, and obtains the scene evidence set, task constraint set and motion state quantity package.

[0051] S1.1 The industrial control system receives the alignment task and cycle constraint, performs alignment task normalization processing and cycle constraint timing processing, and obtains the target shelf type, target working area, allowed approach direction, alignment termination condition, control cycle and trigger time.

[0052] It should be noted that after receiving the alignment task and cycle constraints, the industrial control system sequentially completes frame header positioning, length verification, and check code verification according to the alignment task message format to extract the effective payload. It then parses the shelf type field, work area field, approach direction field, and termination condition field according to the fixed field positions. It performs legal value and enumeration value verification on the parsed fields. If the approach direction field or termination condition field is missing, it is filled in according to the default code rules in the message format. Finally, it outputs the target shelf type, target work area, allowed approach direction, and alignment termination condition.

[0053] The payload is extracted and fixed fields are parsed according to the clock constraint message format to obtain the clock cycle field and trigger time field. The format and timing consistency check is then performed. After the check is passed, the clock cycle field is converted into the control cycle, and the trigger time field is aligned to the trigger time.

[0054] It should also be noted that the fixed field position refers to the fact that in the "alignment task message format and clock constraint message format", the byte offset and byte length of each field in the payload are fixed. The industrial control system directly extracts the corresponding byte segment according to the field offset and converts it into the field value, without the need for variable delimiter lookup or dynamic parsing. The default code rule refers to the correspondence between the default code and the completion semantic agreed upon for optional fields in the message format. That is, when the approach direction field or termination condition field is missing, or the field value is equal to the default code specified by the message format, the industrial control system interprets the default code as the default value and completes the completion according to the agreement, so that the output still obtains the definite value of "allowed approach direction and alignment termination condition".

[0055] S1.2. Based on the control cycle and trigger time, perform time slot start and end calculation and time slot binding processing to form a control time slot. Based on the target work area, allowed approach direction and alignment termination conditions, perform interlock rule generation and interlock condition solidification to form a safety interlock condition.

[0056] It should be noted that the industrial control system uses the trigger time as the start point of the control time slot and the control cycle as the time slot length. The start point and the time slot length are summed to obtain the end point of the control time slot. The start point, the end point, and the trigger time are associated and written into the control time slot identifier to complete the time slot binding process, thereby forming the control time slot. The control time slot identifier is generated by performing time formatting on the trigger time to obtain the time code, performing cycle number calculation on the control cycle to obtain the cycle number code, and concatenating the time code and the cycle number code.

[0057] The industrial control system uses the target operating area, the allowed approach direction, and the alignment termination condition to generate safety interlock conditions. The target operating area is used to determine the prohibited boundary and the allowed passage boundary and form the area interlock condition. The allowed approach direction is used to determine the approach direction constraint and form the direction interlock condition. The alignment termination condition is used to determine the termination trigger decision item and form the termination interlock condition. The area interlock condition, direction interlock condition, and termination interlock condition are summarized to form the interlock rule, and the interlock rule is bound to the control time slot and solidified as the safety interlock condition.

[0058] S1.3. Collect sensing data within the control time slot and perform time alignment and coordinate unification to form a scene evidence set. Compile the target shelf type, target work area, allowed approach direction, alignment termination condition and safety interlock condition to form a task constraint set. Extract the pose and velocity information within the control time slot to form a motion state data package.

[0059] It should be noted that sensing data is acquired within the control time slot, and pose and velocity information within the control time slot are acquired simultaneously. The sensing data is then uniformly timestamped to form a sampling sequence, with the start of the control time slot as the reference time. Each timestamp in the sampling sequence is converted into a time offset relative to the reference time, and then resampling and alignment are performed. Missing sampling points are compensated by interpolation to obtain time-aligned sensing data. The expression for generating the time offset is as follows:

[0060] ;

[0061] in, Indicates the first sampled sequence The time offset of each timestamp; Indicates the first sampled sequence A timestamp; This indicates the reference time, i.e., the starting point of the control time slot.

[0062] The time-aligned perception data undergoes coordinate unification processing. This process uses the mobile robot's body coordinate system or its odometry coordinate system as a unified reference coordinate system. Based on the coordinate transformation relationship obtained from sensor installation and calibration, the time-aligned perception data is transformed into the unified reference coordinate system. The pose within the control time slot is then synchronously applied to the unified reference coordinate system to complete spatial alignment, resulting in coordinate-unified perception data. Using the control time slot identifier as an archiving index, the coordinate-unified perception data within the control time slot is organized in timestamp order. Each coordinate-unified perception data entry is then bound to a corresponding timestamp and control time slot identifier, forming a scene evidence set composed of all coordinate-unified perception data within the control time slot, along with timestamps and control time slot identifiers.

[0063] Constraint assembly is performed on the target shelf type, target work area, allowed access direction, alignment termination condition and safety interlock condition. The constraint assembly includes collecting each fixed field in the order of target shelf type, target work area, allowed access direction, alignment termination condition and safety interlock condition and establishing the field correspondence to form a task constraint set. The pose and velocity information within the control time slot are sorted by timestamp and clipped to the control time slot range to obtain the pose sequence and velocity sequence, and the pose sequence and velocity sequence are collected to form a motion state quantity packet.

[0064] S2. Under the constraints of scene evidence set, task constraint set and motion state quantity package, extract the shelf structure anchor points and generate a reachable candidate pose pool. Based on the anchor point matching consistency, form candidate confidence and compare it with the budget threshold. Output budget state flag and verification trigger flag.

[0065] S2.1. Based on the scene evidence set, task constraint set and motion state quantity package, perform evidence screening and noise suppression processing, and perform line segment fitting and contour extraction on the screened evidence to obtain a set of structural features. Perform structural consistency screening and cluster merging on the set of structural features according to the target shelf type to extract the set of shelf structure anchor points, forming a set of shelf structure anchor points.

[0066] It should be noted that, based on the target operation area in the task constraint set, the scene evidence set is spatially pruned to obtain an evidence subset, and the evidence subset is filtered based on the pose sequence and velocity sequence in the motion state data package. The evidence filtering includes removing perception data with missing timestamps, duplicate timestamps, data exceeding the control time slot range corresponding to the scene evidence set, and data corresponding to velocity anomalies. The velocity sequence includes linear velocity values ​​and angular velocity values. A velocity anomaly is defined as the linear velocity value at any time exceeding the maximum permissible linear velocity in the task constraint set, or the angular velocity value at any time exceeding the maximum permissible angular velocity in the task constraint set. The corresponding time is then determined to be a velocity anomaly.

[0067] The maximum permissible linear velocity and maximum permissible angular velocity in the task constraint set are predetermined by the industrial control system based on the target shelf type, target work area, mobile robot chassis motion capability parameters, and safety interlock conditions, and written into the task constraint set. Specifically, within the corresponding target work area, the mobile robot can perform straight-line driving tests and turning tests respectively, and record the maximum stable linear velocity and maximum stable angular velocity corresponding to the condition that the safety interlock conditions are allowed and the emergency stop judgment condition, no-entry boundary judgment condition, and minimum safe distance judgment condition are not triggered. The linear velocity value that is not greater than the maximum stable linear velocity is fixed as the maximum permissible linear velocity, and the angular velocity value that is not greater than the maximum stable angular velocity is fixed as the maximum permissible angular velocity.

[0068] After evidence screening, noise suppression processing is performed on the evidence subset. Noise suppression processing includes median smoothing and outlier removal of the perceived data to obtain clean evidence. Line segment fitting and contour extraction are performed on the clean evidence. Line segment fitting includes segmenting and aggregating the point set in the clean evidence and performing least squares fitting to obtain a set of line segments. Contour extraction includes edge detection and contour tracking of the clean evidence to obtain a set of contours. The set of line segments and the set of contours together constitute a set of structural features.

[0069] Based on the target shelving type in the task constraint set, structural consistency screening is performed on the set of structural features. Structural consistency screening includes checking the parallelism and spacing consistency of candidate uprights, and checking the horizontality and relative height consistency of candidate beams, and eliminating structural features that do not meet the structural constraints of the target shelving type. After structural consistency screening, the remaining structural features are clustered and merged. Clustering and merging includes aggregating duplicate endpoints and intersections according to spatial proximity and merging similar structural features to obtain unique anchor point positions, forming a set of shelving structural anchor points.

[0070] S2.2. Perform anchor point pairing and topology combination on the set of anchor points of the rack structure to obtain the original candidate pose pool. Based on the target working area, the allowed approach direction and the pose and velocity boundaries in the motion state package, perform reachability verification and remove unexecutable items to obtain the reachable candidate pose pool.

[0071] It should be noted that, based on the structural constraints corresponding to the target shelf type, anchor point pairing and topological combination are performed on the set of anchor points of the shelf structure. Anchor point pairing and topological combination include candidate pairing of the set of anchor points according to spatial proximity, combination of candidate pairings according to the relative positional relationship corresponding to the target shelf type, and structural constraint verification of each combination to eliminate combinations with unreasonable spacing or conflicting relative orientations, resulting in a set of valid anchor point combinations. Unreasonable spacing refers to anchor point spacing not conforming to the structural dimension range given by the corresponding specification parameters of the target shelf type in the alignment task. Conflicts in relative orientation refer to the left-right or front-back relationship of the anchor point combination being inconsistent with the relative positional relationship corresponding to the target shelf type. Candidate pose solving is then performed on each of the valid anchor point combinations to form an original candidate pose pool. Candidate pose solving includes determining the candidate position using the geometric center of the valid anchor point combination set, and determining the candidate orientation using the main direction of the valid anchor point combination set combined with the allowed approach direction, ensuring that each candidate pose in the original candidate pose pool includes both position and orientation. The expressions for determining the candidate position and candidate orientation are as follows:

[0072] ;

[0073] ;

[0074] in, Indicates the candidate position; This represents the anchor index in the set of valid anchor combinations, with a value ranging from 1 to... ; This indicates the number of anchor points in the set of valid anchor point combinations; Indicates the first The location of each shelf structure anchor point; Indicates the candidate orientation; This represents the set of allowed directions of approach, typically containing one or more discrete angle values; Indicates the permitted approach direction; Indicates the main direction of the set of valid anchor point combinations; This represents the absolute value of the angle difference.

[0075] Based on the target operating area, a regional feasibility check is performed on the original candidate pose pool to eliminate candidate poses located outside the boundary of the target operating area. A directional feasibility check is also performed on the original candidate pose pool based on the allowed approach direction to eliminate candidate poses whose orientation does not match the allowed approach direction. Reachability checks and inoperable item elimination are performed based on the pose and velocity boundaries in the motion state packet. The reachability check includes calculating the displacement from the current position to the candidate position and the rotation angle from the current orientation to the candidate orientation using the current position and current orientation in the motion state packet, expressed as follows:

[0076] ;

[0077] , ;

[0078] in, Indicates the amount of displacement; Indicates the robot's current position; Indicates the angle measurement; Indicates the robot's current facing angle; This represents the arctangent function in the four quadrants, used to normalize the angle difference to the standard range.

[0079] The displacement and rotation angle are converted into the required linear velocity and angular velocity values ​​according to the control cycle, and the expression is as follows:

[0080] ;

[0081] in, Indicates the required linear velocity value; Indicates the control cycle; This indicates the required angular velocity value.

[0082] The required linear velocity and angular velocity values ​​are compared with the velocity boundaries in the motion state quantity package. If any candidate pose has a required velocity that exceeds the velocity boundary, it is determined to be an unexecutable item and removed from the original candidate pose pool. After the removal is completed, the reachable candidate pose pool is obtained.

[0083] S2.3. Based on the set of anchor points of the shelf structure and the reachable candidate pose pool, perform anchor point matching consistency calculation and residual statistics to form candidate confidence and compare it with the budget threshold to generate budget status flags and verification trigger flags.

[0084] It should be noted that for each reachable candidate pose in the reachable candidate pose pool, anchor point matching consistency calculation and residual statistical processing are performed sequentially. Anchor point matching consistency calculation includes: performing spatial transformation on the set of shelf structure anchor points based on the reachable candidate poses to obtain a predicted anchor point position set; specifically, rotating the coordinates of the shelf structure anchor points according to the candidate orientation to obtain rotated coordinates, translating the rotated coordinates according to the candidate position to obtain transformed coordinates, and using the transformed coordinates of each shelf structure anchor point as the corresponding predicted anchor point position in the predicted anchor point position set, thus forming the predicted anchor point position set from the shelf structure anchor point set; and sequentially performing anchor point type consistency checks and spatial proximity matching on the predicted anchor point position set and the shelf structure anchor point set to form a set of matched anchor point pairs; specifically, matching anchor point pairs according to the shelf structure anchor point type... The set of predicted anchor points is grouped according to the type of the shelf structure anchor point, and matching is performed only within the same shelf structure anchor point type group. Spatial nearest neighbor matching is performed within each shelf structure anchor point type group. The matching process includes calculating the spatial distance between the predicted anchor point position and other shelf structure anchor points of the same type for each predicted anchor point position, selecting the shelf structure anchor point with the smallest spatial distance as the matching object and generating a pair of matching anchor points, marking the selected shelf structure anchor points as occupied to avoid duplicate matching, and traversing all predicted anchor point positions to obtain the set of matching anchor point pairs. Predicted anchor points that fail to find a matching object are counted in the number of unmatched anchor points. Occupancy marking means removing matched shelf structure anchor points from the candidate set of the same type, so that a shelf structure anchor point only participates in spatial nearest neighbor matching once.

[0085] The residual statistical processing includes calculating the positional and orientation deviations for each pair of matched anchor points and summarizing the results. The orientation deviation is calculated only for rack structure anchor points that include orientation attributes; for rack structure anchor points that do not include orientation attributes, only the positional deviation is calculated, and the orientation deviation is recorded as zero. The residual statistical results include at least the number of matched anchor points, the number of unmatched anchor points, the positional deviation statistic, and the orientation deviation statistic. The matching coverage is obtained as the ratio of the number of matched anchor points to the total number of rack structure anchor points, expressed as follows:

[0086] ;

[0087] in, Indicates the matching coverage; Indicates the number of matched anchor points; This represents the total number of anchor points in the shelving structure.

[0088] The expression for the positional deviation statistic is:

[0089] ;

[0090] in, This represents the positional deviation statistic; Indicates the first The first prediction anchor point and the second Positional deviation between observation anchor points.

[0091] The expression for the orientation deviation statistic is,

[0092] ;

[0093] in, This represents the orientation deviation statistic; Indicates the first The first prediction anchor point and the second Orientation deviation between observation anchor points.

[0094] Using matching coverage as candidate confidence levels, the positional deviation statistics and orientation deviation statistics are compared with the upper limit of the allowable deviation corresponding to the alignment termination condition in the task constraint set to obtain a residual consistency judgment. The candidate confidence levels are then compared with the budget threshold in the task constraint set, and a budget status flag and a verification trigger flag are generated based on the residual consistency judgment. When the candidate confidence level reaches the budget threshold and the residual consistency judgment passes, a budget status flag of "budget satisfied" is generated. When the candidate confidence level does not reach the budget threshold, or the residual consistency judgment fails, a budget status flag of "budget insufficient" is generated, and a verification trigger flag of "triggered" is generated. The expression for the budget status flag is...

[0095] ;

[0096] in, Indicates the budget status flag; Indicates the confidence level of the candidate; Indicates the budget threshold.

[0097] The budget threshold is a candidate confidence threshold that is fixed according to the target shelf type. The budget threshold setting process is as follows: For each target shelf type, control time slot generation and safety interlock condition generation are repeatedly performed in the standard operating area, sensing data is collected and time alignment and coordinate unification are completed to form a scene evidence set, shelf structure anchor point set is extracted and reachable candidate pose pool is generated, anchor point matching consistency calculation and residual statistics are performed to form candidate confidence and complete residual consistency judgment. When the residual consistency judgment is passed and the safety interlock condition remains permissible, the corresponding candidate confidence is recorded to form a verification record. The standard operating area refers to the passable area in the target operating area where the safety interlock condition remains permissible and does not contain prohibited boundaries. The lowest candidate confidence that can still maintain the residual consistency judgment in multiple verification records is selected and fixed as the budget threshold, so that the industrial control system can directly call the budget threshold to generate budget status flags and verification trigger flags when performing alignment tasks.

[0098] It should also be noted that existing technologies typically output alignment poses directly through perceptual feature matching or fitting, lacking a unified usability determination. They are susceptible to mismatches due to occlusion and similar structures, resulting in large fluctuations and poor controllability in alignment results. This solution, through the processing of reachable candidate pose pools, candidate confidence, budget threshold comparison, budget state markers, and verification trigger markers, provides clear admission criteria for the recognition results. It can eliminate candidates that do not meet the admission criteria in advance, reduce the spread of mismatches, improve the stability and consistency of alignment recognition, reduce the probability of alignment failure and repeated trial and error, reduce unnecessary movement and waiting time, and improve the predictability of the operation cycle.

[0099] S3. The verification trigger mark and budget status mark are constrained by the safety interlock conditions and motion state data to perform micro-motion verification actions, re-collect the perception data and update the scene evidence set in a unified manner according to time and coordinates, determine the credible pose, tolerance window and alignment confirmation conditions in the reachable candidate pose pool, and form an alignment control table.

[0100] S3.1 Combine the verification trigger flag and the budget status flag with the safety interlock condition and motion state quantity packet to perform an executability determination, and generate a micro-motion verification action sequence when the executability determination is allowed.

[0101] It should be noted that the executability determination is performed by combining the verification trigger flag and the budget status flag with the safety interlock conditions and motion state data package. The executability determination performs an equivalence check on the verification trigger flag and the budget status flag. If the verification trigger flag is triggered and the budget status flag indicates insufficient budget, the interlock determination is initiated; otherwise, the executability determination is prohibited. The interlock determination is performed in a fixed order: emergency stop determination, speed determination, area determination, and safety distance determination. The emergency stop determination involves reading the status of the emergency stop condition in the safety interlock conditions and confirming that it is in an allowed state. The speed determination involves comparing the linear velocity and angular velocity values ​​in the motion state data package with the maximum allowable linear velocity and maximum allowable angular velocity in the safety interlock conditions, respectively. If the angular velocity is allowed to be compared and confirmed to be within the allowed range, the area determination is to compare the current position in the motion state data packet with the boundary of the target work area and confirm that the current position is within the target work area. The safe distance determination is to form a safe distance check line segment in front of the current position with the current position in the motion state data packet as the center and in combination with the allowed approach direction. The safe distance check line segment is then compared with the nearest obstacle position corresponding to the scene evidence set to calculate the nearest distance and confirm that it is not less than the minimum safe distance. The nearest obstacle position is the position corresponding to the unified perception data point with the smallest Euclidean distance from the current position selected from the scene evidence set. If the emergency stop determination, speed determination, area determination and safe distance determination are all passed, the executability determination is allowed.

[0102] The minimum safe distance is used as a distance reference for determining the minimum safe distance. The process is as follows: select a representative passable position within the target work area, and let the mobile robot approach the fixed obstacle along the allowed approach direction under the maximum allowable linear velocity and maximum allowable angular velocity allowed by the safety interlock conditions. Trigger the emergency stop condition to stop the mobile robot. Repeat this process multiple times and record the closest distance between the mobile robot and the obstacle when the stop is completed. Take the maximum value of the recorded closest distance and fix it as the minimum safe distance. The representative passable position refers to a number of positions within the target work area that meet the safety interlock conditions, allow passage, and have continuous passable space along the allowed approach direction.

[0103] Preferably, the exemplary range of the minimum safe distance is 0.05m to 1.00m. When the minimum safe distance is too small, it is difficult to reserve enough buffer space for sensor detection deviation, actuator response delay and motion fluctuation, which may lead to insufficient safe distance when the mobile robot approaches the target shelf, which is not conducive to the stable completion of the alignment operation. When the minimum safe distance is too large, the mobile robot will be restricted from approaching too early, resulting in an increase in the residual deviation between the alignment position and the target position, while compressing the passable space and affecting the operation efficiency.

[0104] When the executability is deemed permissible, a micro-motion verification action sequence is generated. The micro-motion verification action sequence includes a micro-yawing action, a micro-displacement action, and a centering action in a fixed order. The direction of the micro-yawing action and the micro-displacement action is defined with reference to the current orientation in the motion state packet. The amplitude and duration of the micro-yawing action and the micro-displacement action are calculated from the velocity boundary in the motion state packet and are constrained by safety interlock conditions to ensure that the action execution process does not trigger the prohibited boundary judgment condition and the minimum safe distance judgment condition.

[0105] S3.2 The industrial control system executes the micro-motion verification action sequence and re-collects the sensing data. It updates the scene evidence set in a unified manner according to time alignment and coordinates. Based on the updated scene evidence set and the reachable candidate pose pool, it performs anchor point matching consistency calculation to update the candidate confidence and obtains the candidate confidence update result.

[0106] It should be noted that after the executability is determined to be permissible, the micro-motion verification action sequence is executed and the execution receipt of the micro-motion verification action sequence is obtained. The consistency of the execution receipt is checked and the execution of the micro-motion verification action sequence is confirmed to be completed. Within the control time slot, the sensing data is re-collected and unified timestamp marking, time alignment processing and coordinate unification processing are performed in sequence to update the scene evidence set. After updating the scene evidence set, for each reachable candidate pose in the reachable candidate pose pool, the set of anchor points of the shelf structure is spatially transformed according to the reachable candidate pose to obtain the set of predicted anchor point positions. Anchor point type consistency check, spatial proximity matching and residual statistics are performed in sequence between the set of predicted anchor point positions and the corresponding anchor point positions in the updated scene evidence set. Candidate confidence forming processing is performed to recalculate the candidate confidence, and the recalculated candidate confidence is aggregated to form the candidate confidence update result.

[0107] S3.3. Based on the candidate confidence update results, determine the credible pose in the reachable candidate pose pool, and generate a tolerance window and alignment confirmation conditions based on the credible pose. Compile the credible pose, tolerance window and alignment confirmation conditions into an alignment control table.

[0108] It should be noted that, based on the candidate confidence update results, each reachable candidate pose in the reachable candidate pose pool is associated with its corresponding candidate confidence update result, and a candidate ranking process is performed. The candidate ranking process sorts the candidate confidence update results from high to low. If the candidate confidence update results are the same, the order is determined by the priority rules of higher matching coverage, smaller position deviation statistics, and smaller orientation deviation statistics. The reachable candidate pose ranked first is determined as the reliable pose. After the reliable pose is determined, the allowable deviation corresponding to the alignment termination condition in the task constraint set is used. The upper limit is used to generate a tolerance window, which converts the upper limit of allowable deviation into allowable ranges of position deviation and orientation deviation, and combines them to form a tolerance window. Based on the tolerance window, budget status flag, and safety interlock conditions, alignment confirmation conditions are generated. The alignment confirmation conditions include the judgment requirement that the robot's pose error relative to the reliable pose enters the tolerance window, the judgment requirement that the budget status flag is satisfied, and the judgment requirement that the safety interlock conditions remain within the allowable range. The alignment control table is formed by aggregating the fixed fields of the reliable pose, tolerance window, and alignment confirmation conditions one by one and establishing the field correspondence.

[0109] The budget status marker adopts the budget comparison judgment result corresponding to the candidate confidence update result. The judgment of the pose error entering the tolerance window is to compare the position deviation and orientation deviation with the allowable range of position deviation and orientation deviation respectively and confirm that they are both satisfied.

[0110] Figure 6The differences in the alignment results (successful alignment / misalignment / out-of-bounds risk) between the baseline scheme and the present invention scheme were compared as the occlusion ratio increased from low to high. For each occlusion level, two parallel bars were used to represent the two schemes, with the proportions of the three types of results presented in a stacked manner within each bar. It was observed that as occlusion worsened, the proportions of misalignment and out-of-bounds risk increased faster in the baseline scheme, while the success rate decreased more significantly. In contrast, the present invention scheme maintained a high success rate and reduced misalignment and out-of-bounds risk even in high-occlusion scenarios. This demonstrates that by constructing an achievable candidate pose pool and triggering micro-motion verification and alignment confirmation conditions by comparing candidate confidence with budget thresholds, the recognition output possesses "verifiable and executable access guarantees." This allows for reduced repeated trial and error and downtime while improving alignment consistency even with observation degradation and increased interference.

[0111] It should also be noted that existing technologies mostly drive the alignment action directly after a one-time recognition, lacking executability determination and active verification. When encountering occlusion, reflection interference, or timing fluctuations, they are prone to misjudgment and cause alignment instability. This solution uses verification trigger markers and budget status markers to drive executability determination and micro-motion verification. Then, based on the updated candidate confidence level, it determines the credible pose and forms an alignment control table. This ensures that the alignment process has clear safety and executability prerequisites before entering the action. The recognition results have verifiable credible evidence and unified confirmation standards, reducing the risk of misalignment and jitter, reducing repeated trial and error and downtime, improving the alignment success rate and cycle stability, and making subsequent control inputs more stable and traceable.

[0112] S4. The alignment control table, combined with the budget status flag, generates a correction command under the motion state quantity package and safety interlock conditions. After performing amplitude and variation limit processing on the correction command, it is issued to make the pose error of the robot relative to the reliable pose enter the tolerance window and complete the alignment confirmation condition judgment. If the alignment confirmation condition is met, the shelf alignment recognition result and operation status are output; otherwise, a conservative alignment command is issued and a re-identification flag is output.

[0113] S4.1 Calculate the pose error of the robot relative to the reliable pose based on the reliable pose in the alignment control table and the current position and current orientation in the motion state packet.

[0114] It should be noted that the current position and current orientation are obtained by taking the latest pose record with the latest timestamp from the motion state data packet. The position deviation is obtained by performing coordinate difference between the current position and the position of the credible pose. The orientation deviation is obtained by performing angle difference between the current orientation and the orientation of the credible pose. The orientation deviation is then normalized to avoid abrupt changes caused by crossing angle boundaries. The position deviation and orientation deviation are combined to form the pose error of the robot relative to the credible pose, where the pose error consists of position deviation component and orientation deviation component.

[0115] S4.2 Calculate and process the alignment correction command for the pose error to generate the initial value of the correction command, and perform amplitude and variation limiting processing in combination with the motion state data packet. At the same time, perform interlock verification processing in combination with the safety interlock conditions to obtain the correction command and the mark that can be issued.

[0116] It should be noted that the alignment correction command calculation and processing based on the pose error is used to generate the initial value of the correction command. The alignment correction command calculation and processing includes decomposing the position deviation component in the pose error into a forward deviation component along the allowable approach direction and an alignment deviation component orthogonal to the allowable approach direction, and taking the orientation deviation component as the orientation alignment deviation component. The forward deviation component, alignment deviation component, and orientation alignment deviation component are mapped to forward correction component, alignment correction component, and orientation correction component, respectively. Specifically, the forward deviation component, alignment deviation component, and orientation alignment deviation component are converted into velocity command components used to reduce the corresponding deviations. The direction is determined by the deviation sign to be the direction of reducing the deviation. The amplitude is converted into the displacement change rate of the forward deviation component and alignment deviation component per control cycle based on the control cycle, and the rotation change rate of the orientation alignment deviation component per control cycle. When the corresponding deviation has fallen within the allowable range of the tolerance window, the corresponding correction component is set to zero, and the forward correction component, alignment correction component, and orientation correction component are obtained and combined as the initial value of the correction command.

[0117] After the initial value of the correction command is formed, amplitude and variation limiting processing is performed in conjunction with the same motion state quantity packet. This process includes comparing the forward correction component, alignment correction component, and orientation correction component of the initial value of the correction command with the velocity boundaries in the motion state quantity packet to perform amplitude clipping, and comparing them with the rate of change boundaries in the motion state quantity packet to perform rate of change clipping, thus obtaining the amplitude-limited and variation-limited correction command. The rate of change boundaries are calculated from the maximum permissible linear velocity / maximum permissible angular velocity in the control cycle and velocity boundaries. Specifically, the ratios of the maximum permissible linear velocity and the maximum permissible angular velocity to the control cycle are calculated to obtain the linear velocity change. The boundary of the rate of change of angular velocity and the boundary of the rate of change of angular velocity; after the correction command after the limit of amplitude and change is formed, the interlock verification process is performed in combination with the same safety interlock condition. The interlock verification process includes the emergency stop judgment condition, the no-entry boundary judgment condition, the minimum safe distance judgment condition, and the maximum permissible speed judgment condition. The status is read and compared in sequence to confirm that they are all in the permissible state. When the interlock verification process passes, the output is marked as permissible and the output correction command is the correction command after the limit of amplitude and change. When the interlock verification process fails, the output is marked as prohibited and the correction command is kept as the correction command after the limit of amplitude and change.

[0118] Figure 7The figure compares the alignment error evolution over time of the baseline scheme and the scheme of the present invention under the same perturbation conditions. The upper figure shows the full-time domain error curves and highlights the typical range with more obvious fluctuations in red. The lower figure magnifies the typical range and marks several characteristic peaks with arrows. At the moment when the difference between the two curves is the greatest, the difference is marked with dashed lines and double-headed arrows. From the global curves, it can be seen that the error peak of the baseline scheme is higher and the fluctuations are more frequent, while the overall error of the scheme of the present invention is in a lower and smoother range. The magnified view further shows that the peak amplitude and abrupt change of the scheme of the present invention are suppressed near the characteristic peaks. The error difference between the two schemes is most prominent at the point of maximum difference. This indicates that by controlling the timing consistency management under the time slot constraint and linking the candidate admission and micro-motion verification mechanism, the alignment error is more likely to converge in the key fluctuation range and the risk of action abrupt change is reduced, thus demonstrating a more stable alignment effect and a more predictable operation rhythm.

[0119] S4.3 When the flag is set to allow, a correction command is issued and the pose error is put into the tolerance window. Based on the alignment confirmation conditions, the shelf alignment recognition result and operation status are determined and output. When the flag is set to prohibit or the alignment confirmation conditions are not met, a conservative alignment command is issued and a re-identification flag is output.

[0120] It should be noted that the "allowable" flag is used as the path selection condition. When the "allowable" flag is set, a correction command is issued, and a confirmation acknowledgment is executed within the control cycle to update the current position and orientation in the motion state data packet. Based on the updated current position and orientation, the pose error of the robot relative to the reliable pose is recalculated and compared with the tolerance window to confirm that both the position deviation component and the orientation deviation component fall within the corresponding allowable range of the tolerance window. This forms the determination result of whether the pose error has entered the tolerance window, and the determination result of whether the pose error has entered the tolerance window is marked as a requirement that the budget is met, along with the budget state. The judgment requirements for maintaining the safety interlock conditions are combined and executed to determine the alignment confirmation conditions. When the alignment confirmation conditions are met, the shelf alignment recognition result and operation status are output. The shelf alignment recognition result is taken from the alignment control table and includes the target shelf identifier and reliable pose. The operation status is used to indicate that the alignment confirmation is completed. When the flag can be issued as prohibited or the alignment confirmation conditions are not met, a conservative alignment instruction is issued. The conservative alignment instruction includes deceleration and alignment correction suppression to avoid amplification of pose error, and a re-identification flag is output. The re-identification flag is used to trigger the subsequent recalculation of candidate confidence.

[0121] Alignment correction suppression is a slowed-down output that sets the alignment correction component and the orientation correction component to zero and retains only the forward correction component.

[0122] Figure 5The differences in the distribution of instruction slips between the baseline scheme and the scheme of this invention under different levels of communication jitter are shown. The horizontal axis is grouped by jitter level, and box plots are given for both schemes. The dashed baseline represents the reference position of "no slip". It can be seen that as the communication jitter increases, the slip distribution of the baseline scheme becomes significantly wider and the upper edge is more likely to cross the reference line, indicating that the alignment recognition result and the instruction issuance / effectiveness window are more prone to timing deviation. In contrast, the slip distribution of the scheme of this invention is more concentrated under each jitter level and the amplitude of crossing the reference line is smaller. This shows that by generating control time slots in the industrial control system and implementing time alignment and coordinate unification of the sensing data, a more stable timing consistency constraint is formed for alignment recognition, instruction issuance and effectiveness window, thereby reducing the risk of timing drift caused by jitter and improving the predictability of the cycle time.

[0123] Figure 5 Figure 6 Figure 7 The baseline scheme refers to the conventional alignment shelf identification and control process without the improvements of this invention, under the same alignment task, communication jitter, and occlusion interference conditions: the industrial control system only performs sensing acquisition and identification calculation according to a fixed control cycle, without generating control time slots and safety interlock conditions to implement rigid timing consistency constraints on "alignment identification - command issuance - effective window", and without aligning the execution time of sensing data and unifying coordinates to suppress time offset and coordinate drift of multi-source data; on the identification output side, the best pose obtained from a single identification is directly used as the alignment result for issuance and execution, without building a reachable candidate pose pool for reachability screening, and without adopting the admission mechanism of "candidate confidence and budget threshold comparison triggering micro-motion verification and alignment confirmation condition judgment", thus making it more prone to command effective out-of-bounds, misalignment, or out-of-bounds risks when communication jitter, occlusion, or similar structure interference are enhanced, and causing repeated trial and error and downtime waiting, which is used as a performance comparison baseline.

[0124] Figure 5 Figure 6 Figure 7 The present invention refers to the complete process executed according to the steps in the specification: After receiving the alignment task and cycle constraints, the industrial control system generates control time slots and safety interlock conditions. Within the control time slots, time alignment and coordinate unification are performed on the multi-source sensing data to ensure that the alignment recognition results and command issuance and effective windows meet the timing consistency constraints. On the recognition output side, a reachable candidate pose pool is constructed and candidate confidence is calculated for the candidate poses. The candidate confidence is compared with the budget threshold to generate a trigger flag. When triggered, micro-motion verification is performed and the alignment confirmation conditions are combined for judgment. Only alignment results that meet the access verification and are executable are output, thereby reducing the risk of misalignment and out-of-bounds errors, reducing repeated trial and error and downtime waiting, and improving the alignment success rate and consistency.

[0125] It should also be noted that existing technologies typically convert the identified pose directly into motion commands, lacking amplitude and variation limits and interlocking gating, as well as verifiable alignment confirmation criteria, which easily leads to sudden command changes, misconceptions, and alignment jitter. This solution, through controlled generation and issuance of correction commands, makes the motion output smoother and only takes effect when safety conditions permit. It provides clear alignment confirmation standards with tolerance windows and confirmation conditions, thereby reducing the risk of collisions and boundary violations, reducing misalignment and repeated trial and error, improving alignment convergence stability and success rate, improving the consistency and predictability of the work cycle, and promptly switching to conservative alignment and re-identification when conditions are not met to maintain process continuity and safety.

[0126] This embodiment also provides a computer device applicable to the mobile robot alignment shelf recognition method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the mobile robot alignment shelf recognition method proposed in the above embodiment.

[0127] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0128] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the method for identifying a mobile robot positioning shelf as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0129] In summary, this invention achieves consistency constraints on the timing of alignment recognition, command issuance, and effective windows by generating control time slots and safety interlock conditions under the industrial control system and implementing time alignment and coordinate unification of the sensed data. This avoids alignment fluctuations and sudden changes in action caused by timing drift, improving the stability and predictability of the work cycle. Furthermore, by constructing a pool of reachable candidate poses and triggering micro-motion verification and alignment confirmation condition determination by comparing candidate confidence and budget thresholds, the invention ensures the verifiability and executability of the recognition results, reduces the risk of misalignment and out-of-bounds errors, minimizes repeated trial and error and downtime, and improves the alignment success rate and consistency.

[0130] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for mobile robot positioning and shelf recognition, characterized in that, include: The industrial control system receives the alignment task and cycle constraints, generates control time slots and safety interlock conditions, collects sensing data within the control time slots and completes time alignment and coordinate unification, and obtains scene evidence set, task constraint set and motion state quantity package. Under the constraints of scene evidence set, task constraint set and motion state quantity package, the anchor points of the shelf structure are extracted and a pool of reachable candidate poses is generated. Based on the consistency of anchor point matching, candidate confidence is formed and compared with the budget threshold. Budget state flag and verification trigger flag are output. The verification trigger flag and budget status flag are constrained by safety interlock conditions and motion state data packets to perform micro-motion verification actions. The perception data is re-acquired and the scene evidence set is updated uniformly according to time alignment and coordinates. The reliable pose, tolerance window and alignment confirmation conditions are determined in the reachable candidate pose pool to form an alignment control table. The specific steps are as follows: The check trigger flag and budget status flag are combined with safety interlock conditions and motion status data to perform an executability determination, and a micro-motion check action sequence is generated when the executability determination is allowed; The industrial control system executes a micro-motion verification action sequence and re-collects sensing data. It updates the scene evidence set in a unified manner according to time alignment and coordinates. Based on the updated scene evidence set and the reachable candidate pose pool, it performs anchor point matching consistency calculation to update the candidate confidence and obtain the candidate confidence update result. Based on the candidate confidence update results, a reliable pose is determined in the reachable candidate pose pool, and a tolerance window and alignment confirmation conditions are generated based on the reliable pose. The reliable pose, tolerance window, and alignment confirmation conditions are compiled into an alignment control table. The alignment control table, combined with the budget status flag, generates correction instructions under the motion state quantity package and safety interlock conditions. After performing amplitude and variation limit processing on the correction instructions, they are issued to make the robot's pose error relative to the reliable pose enter the tolerance window and complete the alignment confirmation condition judgment. If the alignment confirmation condition is met, the shelf alignment recognition result and operation status are output; otherwise, a conservative alignment instruction is issued and a re-identification flag is output.

2. The mobile robot positioning shelf recognition method as described in claim 1, characterized in that, The specific steps for generating the control time slot and security interlock conditions are as follows: The industrial control system receives the alignment task and cycle constraints and performs alignment task normalization and cycle constraint timing processing to obtain the target shelf type, target working area, allowed approach direction, alignment termination condition, control cycle and trigger time. The start and end times of the execution time slot are calculated and bound to the time slot based on the control cycle and trigger time to form a control time slot; Based on the target work area, the allowed approach direction, and the alignment termination conditions, interlock rules are generated and interlock conditions are solidified to form safe interlock conditions.

3. The mobile robot positioning shelf recognition method as described in claim 2, characterized in that, The specific steps for obtaining the scene evidence set, task constraint set, and motion state data packet are as follows: Collect sensing data within the control time slot and perform time alignment and coordinate unification to form a scene evidence set; The target shelf type, target work area, allowed access direction, alignment termination condition and safety interlock condition are compiled into a task constraint set; The pose and velocity information within the control time slot is extracted to form a motion state data packet.

4. The mobile robot positioning shelf recognition method as described in claim 1, characterized in that, The specific steps for generating the reachable candidate pose pool are as follows: Based on the scene evidence set, task constraint set and motion state data package, evidence screening and noise suppression processing are performed, and line segment fitting and contour extraction are performed on the screened evidence to obtain a set of structural features. Based on the target shelf type, the structural feature set is subjected to structural consistency screening and clustering merging to extract the shelf structure anchor point set, forming the shelf structure anchor point set; Anchor point pairing and topology combination are performed on the set of anchor points of the shelf structure to obtain the original candidate pose pool; Based on the target working area, allowed approach direction, and pose and velocity boundaries in the motion state data packet, reachability verification and non-executable item elimination are performed to obtain the reachable candidate pose pool.

5. The mobile robot positioning shelf recognition method as described in claim 4, characterized in that, The specific steps for setting the output budget status flag and the verification trigger flag are as follows: Based on the set of anchor points of the shelf structure and the pool of reachable candidate poses, perform anchor point matching consistency calculation and residual statistics to form candidate confidence scores; The candidate confidence level is compared with the budget threshold to generate budget status markers and verification trigger markers.

6. The mobile robot positioning shelf recognition method as described in claim 1, characterized in that, The specific steps for outputting the shelf alignment recognition result and operation status are as follows: The pose error of the robot relative to the reliable pose is calculated based on the reliable pose in the alignment control table and the current position and current orientation in the motion state packet. The positional error is calculated and processed to generate the initial value of the correction command. The amplitude and variation are limited in combination with the motion state data packet. At the same time, the interlock verification is performed in combination with the safety interlock conditions to obtain the correction command and the issueable mark. When marked as allowed, a correction command can be issued and the pose error can be put into the tolerance window. Based on the alignment confirmation conditions, the shelf alignment recognition result and operation status can be determined and output. When the flag is set to prohibited or the alignment confirmation condition is not met, a conservative alignment instruction is issued and a re-identification flag is output.

7. The mobile robot positioning shelf recognition method as described in claim 1 or 5, characterized in that, The budget threshold refers to the candidate confidence threshold that is fixed for the target shelf type. The budget threshold is determined by alignment verification as the lowest candidate confidence level corresponding to the verification result that meets the alignment termination condition and the safety interlock condition remains permissible.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the mobile robot positioning shelf recognition method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the mobile robot positioning shelf recognition method according to any one of claims 1 to 7.

Images