A robot simulation walking method and system based on virtual-real coupling

Abstract

The application relates to the field of simulation data processing, and discloses a robot simulation walking method and system based on virtual-real coupling, which comprises the following steps: step 1, establishing a discrete time index sequence and obtaining a basic data set; step 2, generating a displacement paragraph label sequence according to position data; step 3, generating a left-right step event sequence and a step length level sequence; step 4, constructing a simulation walking semantic frame and forming a preliminary version of a simulation walking semantic sequence; step 5, performing consistency checking on the preliminary version of the simulation walking semantic sequence to generate a set of risk window sets; step 6, performing deterministic semantic editing repair on the semantic frame in the risk window set; and step 7, completing consistency checking within a preset round to output a final simulation walking semantic sequence and related marks. The application improves the stability, consistency and checkability of the robot simulation walking result in a virtual-real coupling scene.

Description

Technical Field

[0001] This invention belongs to the field of simulation data processing technology, specifically relating to a robot simulation walking method and system based on virtual-real coupling. Background Technology

[0002] With the widespread application of virtual simulation technology in robot design, demonstration, and interaction, how to accurately, stably, and cost-effectively present the visual effects of robot walking under real-world conditions has become a crucial issue in related technologies. Existing robot walking simulation solutions are typically based on continuous-time modeling or physical driving parameters, generating walking animations by directly simulating pose, joint states, or motion trajectories. However, in virtual-real coupled scenarios, the robot's real displacement data, virtual gait appearance, and appearance occlusion structures often originate from different sources, exhibiting differences in temporal granularity, data format, and semantic meaning. This can easily lead to problems during simulation, such as discontinuous gait, the display of steps even when stationary, and abrupt changes in occlusion relationships, affecting the overall consistency of the walking effect.

[0003] Furthermore, existing technologies for handling anomalies in simulated walking largely rely on manual parameter tuning or empirical rules, lacking a unified data structure and a verifiable processing flow. This makes it difficult to systematically identify and automatically correct visual inconsistencies that occur during walking. In cases of missing data, occlusion changes, or short-term state fluctuations, simulation results are prone to revealing flaws, which are difficult to locate and repair in subsequent processing. Summary of the Invention

[0004] This invention provides a robot simulation walking method and system based on virtual-real coupling, which solves the technical problems in related technologies, such as inconsistent time granularity of multi-source walking data and lack of unified rules for semantic binding, which leads to incoherent gait semantics, abrupt changes in visual occlusion relationships, and difficulty in verifying and repairing simulation results during the simulation walking process.

[0005] This invention provides a robot walking simulation method based on virtual-real coupling, comprising the following steps:

[0006] Step 1: Establish a discrete-time index sequence and obtain the basic data set corresponding to the discrete-time index sequence; wherein, the basic data set includes: location data sequence, gait template, and occlusion level sequence;

[0007] Step 2: Generate a displacement segment label sequence based on the adjacent index differences of the location data sequence;

[0008] Step 3: Generate left and right stepping event sequences based on the gait template, and generate stride level sequences based on the displacement segment label sequences;

[0009] Step 4: Convert the occlusion level sequence into a detection risk level sequence, bind the displacement segment label sequence, left and right step event sequence, stride level sequence, and detection risk level sequence with the same discrete time index, construct a simulated walking semantic frame and form an initial version of the simulated walking semantic sequence.

[0010] Step 5: Perform consistency verification on the initial simulated walking semantic sequence based on the set of hard constraint rules, and generate a violation marker sequence and a set of vulnerability risk windows;

[0011] Step 6: Based on the set of error risk windows and the sequence of violation markers, perform deterministic semantic editing and repair on the simulated walking semantic frames within the error risk windows to obtain the repaired simulated walking semantic sequence.

[0012] Step 7: Perform consistency verification on the repaired simulated walking semantic sequence, obtain the final simulated walking semantic sequence within a preset number of rounds, and output the final simulated walking semantic sequence, consistency verification pass marker, and window repair record.

[0013] This invention provides a robot simulation walking system based on virtual-real coupling, comprising:

[0014] The discrete index modeling module is used to establish discrete-time index sequences and obtain a basic data set that corresponds one-to-one with the discrete-time index sequences. The basic data set includes: location data sequences, gait templates, and occlusion level sequences.

[0015] The displacement segment determination module is used to generate a displacement segment label sequence based on the adjacent index difference of the position data sequence;

[0016] The gait semantic generation module is used to generate left and right stepping event sequences based on gait templates and to generate stride level sequences based on displacement segment label sequences.

[0017] The walking semantic construction module is used to convert the occlusion level sequence into the exposure risk level sequence, bind the displacement segment label sequence, left and right step event sequence, stride level sequence, and exposure risk level sequence according to the same discrete time index, construct the simulated walking semantic frame and form the initial version of the simulated walking semantic sequence.

[0018] The consistency verification and analysis module is used to perform consistency verification on the initial simulated walking semantic sequence based on the set of hard constraint rules, and generate a violation marker sequence and a set of vulnerability risk windows.

[0019] The deterministic semantic repair module is used to perform deterministic semantic editing and repair on the simulated walking semantic frames within the fault risk windows based on the set of fault risk windows and the sequence of violation markers, so as to obtain the repaired simulated walking semantic sequence.

[0020] The round convergence output module is used to perform consistency verification on the repaired simulated walking semantic sequence. It obtains the final simulated walking semantic sequence within a preset number of rounds and outputs the final simulated walking semantic sequence, consistency verification pass markers, and window repair records.

[0021] The beneficial effects of this invention are as follows: This invention unifies the binding of location data, gait templates, and occlusion information based on discrete-time index sequences, avoiding the misalignment problem of multi-source data in the time dimension, and ensuring a consistent index foundation for simulated walking-related data. By constructing simulated walking semantic frames and forming semantic sequences, continuous spatial displacement, gait appearance, and occlusion states are transformed into computable and verifiable discrete semantic data, improving the structuring level of the simulated walking process.

[0022] This invention introduces a set of hard constraint rules to perform consistency checks on the semantic sequences of simulated walking, and locates abnormal intervals in the form of violation marker sequences and error risk windows, enabling automatic identification of semantic inconsistencies that occur during simulated walking. Simultaneously, by performing deterministic semantic editing and repair within a limited window, combined with a pre-defined round of consistency review mechanism, the semantic sequences of simulated walking achieve stable convergence output within a finite number of iterations, thus improving the overall stability, consistency, and verifiability of robot simulated walking results in virtual-real coupled scenarios. Attached Figure Description

[0023] Figure 1 This is a flowchart of a robot simulation walking method based on virtual-real coupling according to the present invention;

[0024] Figure 2 This is a schematic diagram of the overall framework of the present invention;

[0025] Figure 3 This is a top view schematic diagram of the robot body structure in an embodiment of the present invention;

[0026] Figure 4 This is a schematic diagram of the appearance of simulating the left leg stepping state in an embodiment of the present invention;

[0027] Figure 5 This is a schematic diagram of the appearance of simulating the right leg stepping state in an embodiment of the present invention;

[0028] Figure 6 This is a schematic diagram of the simulated static posture in an embodiment of the present invention;

[0029] In the picture: 1. Car chassis; 2. Motor-integrated wheel; 3. Front universal wheel; 4. Gear set anchoring frame; 5. Left leg rack; 6. Right leg rack; 7. Imitation left toe; 8. Imitation right toe; 9. Imitation left heel; 10. Imitation right heel; 11. Gear; 12. Frame with reserved motor mounting holes; 13. Two-way 360-degree servo motor; 14. Servo motor mounting screw; 15. Original servo motor rudder; 16. Waist support frame; 17. Simulated waist; 18. Floor-length skirt. Detailed Implementation

[0030] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0031] like Figures 1-6 As shown, a robot walking simulation method based on virtual-real coupling includes the following steps:

[0032] Step 1: Establish a discrete-time index sequence and obtain the basic data set corresponding to the discrete-time index sequence; wherein, the basic data set includes: location data sequence, gait template, and occlusion level sequence;

[0033] Step 2: Generate a displacement segment label sequence based on the adjacent index differences of the location data sequence;

[0034] Step 3: Generate left and right stepping event sequences based on the gait template, and generate stride level sequences based on the displacement segment label sequences;

[0035] Step 4: Convert the occlusion level sequence into a detection risk level sequence, bind the displacement segment label sequence, left and right step event sequence, stride level sequence, and detection risk level sequence with the same discrete time index, construct a simulated walking semantic frame and form an initial version of the simulated walking semantic sequence.

[0036] Step 5: Perform consistency verification on the initial simulated walking semantic sequence based on the set of hard constraint rules, and generate a violation marker sequence and a set of vulnerability risk windows;

[0037] Step 6: Based on the set of error risk windows and the sequence of violation markers, perform deterministic semantic editing and repair on the simulated walking semantic frames within the error risk windows to obtain the repaired simulated walking semantic sequence.

[0038] Step 7: Perform consistency verification on the repaired simulated walking semantic sequence, obtain the final simulated walking semantic sequence within a preset number of rounds, and output the final simulated walking semantic sequence, consistency verification pass marker, and window repair record.

[0039] In this invention, virtual-real coupling refers to the unified modeling and collaborative processing of robot displacement data from the real environment and simulation elements such as gait appearance and occlusion structure in the virtual environment at the data processing layer, without involving direct control of robot motion.

[0040] In one embodiment of the present invention, such as Figure 2 As shown, the robot simulation walking system includes a small chassis for generating spatial displacement, simulated leg devices for forming the gait appearance, and a floor-length skirt for creating an overall occlusion effect. To uniformly process information from the aforementioned different sources, this invention first establishes a discrete-time index sequence at the data layer, obtaining a basic data set corresponding one-to-one with the discrete-time index sequence, specifically including:

[0041] Step 11: Establish a discrete-time index sequence. Set the discrete-time index sequence as an ordered set of indices arranged in ascending order, and apply uniqueness and ordering constraints to the discrete-time index sequence. The uniqueness constraint ensures that each discrete-time index in the discrete-time index sequence appears only once in the entire index set, and the ordering constraint ensures that the discrete-time index sequence is strictly arranged in chronological order. Through these settings, the discrete-time index sequence can stably serve as a unified time identifier for various types of data during the simulation walkthrough.

[0042] Step 12: Obtain a basic data set including a position data sequence, a gait template, and an occlusion level sequence. The position data sequence is a set of position information records corresponding one-to-one with the discrete-time index sequence, used to characterize the robot's spatial position state at each discrete-time index. The gait template is a finite sequence composed of preset gait symbols, used to describe the symbolic gait structure of left and right steps and transition states during simulated walking. Different gait symbols in the gait template correspond to the appearance of left and right steps and transition states, respectively. The occlusion level sequence is a set of occlusion degree identifiers corresponding one-to-one with the discrete-time index sequence, used to characterize the robot's appearance occlusion state at each discrete-time index. The occlusion level sequence reflects the change in the degree of occlusion of the internal structure by the appearance occlusion structure under different walking states. It should be noted that in this invention, both the gait template and the occlusion level sequence participate in subsequent data processing in a symbolic or hierarchical form.

[0043] Step 13: After obtaining the basic data set, perform a consistency check on the basic data set. Specifically, verify the index coverage relationship and one-to-one correspondence between the location data sequence and the occlusion level sequence and the discrete-time index sequence, that is, confirm that each discrete-time index in the discrete-time index sequence corresponds to a location information record value in the location data sequence and an occlusion degree indicator in the occlusion level sequence. When it is detected that any discrete-time index in the discrete-time index sequence is missing a corresponding location information record value or occlusion degree indicator, write a corresponding null placeholder record at that discrete-time index, and simultaneously write a null placeholder indicator.

[0044] The null placeholder record is used to explicitly identify the presence of missing basic data at the discrete-time index. The null placeholder identifier is used to distinguish between missing data and valid data during subsequent data generation, semantic binding, and consistency verification. By explicitly placing missing data at the discrete-time index level, the basic data set remains complete and continuous in the index dimension, avoiding implicit inference or index breaks during subsequent simulation semantic frame construction and rule verification.

[0045] Through the above steps, this embodiment completes the establishment of discrete-time index sequences, the acquisition of basic data sets, and consistency verification and completion processing at the data processing level. This enables the location data sequences, gait templates, and occlusion level sequences to form a well-structured and indexed basic data set under a unified discrete-time index framework, providing a stable data foundation for the subsequent construction of simulated walking semantic frames, cross-domain consistency verification, and deterministic semantic editing and repair.

[0046] In one embodiment of the present invention, the vehicle chassis provides a stable and continuous spatial displacement basis during simulated walking, constituting the spatial logical premise of simulated walking; the simulated leg device generates clear left and right stepping visual signals based on the spatial displacement, which is the core visual element for the recognition of simulated walking semantics; the floor-length skirt is used to comprehensively mask and integrate the local visual signals of the simulated leg device, thereby reducing the visual inconsistency caused by abrupt changes in the boundary between virtual and real, and thus reducing the probability of errors occurring during simulated walking. The above three components work together through functional complementarity and logical linkage to form a complete simulated walking visual semantic closed loop. The absence of any unit may lead to the interruption of the visual semantic chain, thereby affecting the overall consistency and credibility of the simulated walking effect.

[0047] In one embodiment of the present invention, generating a displacement segment label sequence based on the adjacent index difference of the location data sequence includes:

[0048] Step 21: Construct adjacent index pairs according to the discrete-time index sequence in ascending order. Each adjacent index pair consists of the current discrete-time index and its previous discrete-time index, and adjacent index pairs are constructed only for discrete-time indices other than the starting index. For each adjacent index pair, read the location information record value corresponding to that adjacent index pair from the location data sequence, thereby providing a paired data basis for subsequent location change calculations.

[0049] like Figure 3-4 As shown in the top view of the core component structure diagram, the overall layout of the robot body in planar space is illustrated. This diagram is used to help understand the spatial position meaning corresponding to the position information record values ​​in the position data sequence. Specifically, the robot body includes: a car chassis 1, an integrated motor wheel 2, a front universal wheel 3, a gear set anchoring frame 4, a left leg rack 5, a right leg rack 6, a simulated left toe 7, a simulated right toe 8, a simulated left heel 9, a simulated right heel 10, a gear 11, a frame with reserved motor mounting holes 12, a bidirectional 360-degree servo motor 13, servo motor mounting screws 14, an original servo motor servo disc 15, a waist support frame 16, a simulated waist 17, and a floor-length skirt 18.

[0050] In this embodiment, the position information record value in the position data sequence is used to describe the spatial position state of the robot body at the time corresponding to the discrete time index. The spatial position state can be understood as... Figure 2 The pose and position changes within the indicated planar range. Based on the above position data sequence, this invention only performs difference analysis on the position information record values ​​corresponding to adjacent discrete-time indices at the data processing level to determine whether spatial displacement has occurred, without involving any specific implementation of driving, control, or structural coordination.

[0051] Step 22: After constructing adjacent index pairs, calculate the adjacent index difference for each pair. Specifically, subtract the location information record value corresponding to the previous discrete-time index from the location information record value corresponding to the current discrete-time index in the adjacent index pair to obtain the adjacent index difference reflecting the spatial change between two adjacent discrete-time indices. Subsequently, arrange the adjacent index differences obtained from each adjacent index pair in the order of the discrete-time index sequence to form location change information, wherein the location change information corresponds only to discrete-time indices other than the starting index.

[0052] Step 23: After obtaining the position change information, determine the magnitude of the position change based on this information. Specifically, for each adjacent index difference, first square the components of the difference in each spatial dimension, then sum the squared results, and finally take the square root of the summation to obtain the magnitude of the position change at the corresponding discrete-time index. The magnitude of the position change is used to uniformly characterize the overall spatial change between two adjacent discrete-time indices. Specifically, the formula for calculating the magnitude of the position change is:

[0053] ;

[0054] in, This represents the magnitude of position change at the corresponding discrete-time index i, where i represents the discrete-time index, k represents the spatial dimension index, and n represents the number of spatial dimensions. Represents position vector The component in the k-th spatial dimension, Represents position vector The component in the k-th spatial dimension.

[0055] After determining the magnitude of position change, a displacement segment label sequence is generated. Specifically, when the magnitude of position change corresponding to a certain discrete-time index is zero, the displacement segment label corresponding to that discrete-time index is determined to be "stop"; when the magnitude of position change is not zero, the displacement segment label corresponding to that discrete-time index is determined to be "move". The displacement segment label sequence only includes two label states: stop and move. Simultaneously, the displacement segment label corresponding to the starting index in the discrete-time index sequence is determined to be "stop".

[0056] When the location information record corresponding to a discrete-time index in the location data sequence is a null placeholder, to avoid introducing implicit inference in subsequent processing, the displacement segment label corresponding to that discrete-time index is directly determined as a stop. This rule ensures that the displacement segment label sequence maintains determinism and processability even in the presence of missing data.

[0057] Through the above steps, this embodiment performs adjacent index difference analysis and semantic processing on the position data sequence at the data processing level, transforming continuous spatial position information into displacement segment label sequences that correspond one-to-one with discrete-time index sequences. These displacement segment label sequences abstractly describe the robot's displacement state at each discrete-time index using two discrete semantic states: stationary and moving. This maps the actual displacement process in physical space into computable and bindable discrete data objects, thus providing a definite data foundation for the subsequent fusion and verification of gait semantic data and occlusion semantic data within a unified discrete-time index framework.

[0058] In one embodiment of the present invention, such as Figure 4 and Figure 5 As shown, the left and right stepping states correspond to the appearance of left and right stepping during robot simulation walking, respectively. This invention generates left and right stepping event sequences based on gait templates and generates stride level sequences based on displacement segment label sequences, including:

[0059] Step 31: Read the gait template based on the discrete-time index sequence, and expand the gait template index by index in the discrete-time index sequence so that each discrete-time index in the discrete-time index sequence corresponds to a unique gait template gait symbol, thereby obtaining the initial left and right stepping event sequence. The gait template is used to describe the temporal arrangement relationship of left and right steps and transition states at the data layer.

[0060] Step 32: After obtaining the initial left and right stepping event sequence, the initial left and right stepping event sequence is corrected index by index based on the displacement segment label sequence. The displacement segment label sequence is used to identify whether the robot is in a moving or stopped state at the corresponding discrete time index, and it is derived from the segmented analysis of the position data sequence. When the discrete time index corresponding to the displacement segment label sequence is "stopped," the corresponding left and right stepping events in the initial left and right stepping event sequence are identified as transitions, indicating that no effective stepping action occurs at that time point; when the discrete time index corresponding to the displacement segment label sequence is "moving," the corresponding left and right stepping events in the initial left and right stepping event sequence remain unchanged. Through the above correction process, the left and right stepping event sequence can be kept consistent with the displacement state, avoiding stepping semantics in the stopped state, thereby improving the temporal consistency of the simulated walking semantics.

[0061] Step 33: Generate a stride level sequence based on the position data sequence and the displacement segment label sequence. For discrete-time indices other than the starting index, the difference between adjacent indices is obtained by subtracting the position information record value corresponding to the previous discrete-time index from the corresponding position information record value in the current position data sequence. This difference is then converted into the position change amplitude to quantify the displacement magnitude between adjacent time points. When the discrete-time index corresponding to the displacement segment label sequence is stationary, the stride level is directly determined as no stride. When the discrete-time index corresponding to the displacement segment label sequence is moving, the stride level is determined as small stride, medium stride, or large stride based on the comparison between the current position change amplitude and the position change amplitude corresponding to the previous discrete-time index. For discrete-time indices that cannot be compared due to the lack of the position change amplitude of the previous discrete-time index, the stride level is determined as medium stride to ensure the continuity and completeness of the stride level sequence in the time dimension.

[0062] Through the above implementation methods, without introducing physical control parameters, the present invention constructs left and right stepping event sequences and stride level sequences based on discrete-time index sequences, gait templates, displacement segment label sequences, and position data sequences. This achieves a unified data representation of gait direction and stride magnitude during robot simulation walking, effectively reducing the complexity caused by the coupling of continuous-time modeling and physical parameters, and improving the consistency of the mapping between simulated walking semantics and real displacement data.

[0063] In one embodiment of the present invention, such as Figure 6 As shown, the robot in standby mode exhibits a simulated static posture, with its external occlusion structure creating varying degrees of occlusion on its internal structure. This invention converts the occlusion level sequence into a detection risk level sequence, and binds the displacement segment label sequence, left and right stepping event sequence, stride level sequence, and detection risk level sequence to the same discrete-time index, constructing a simulated walking semantic frame and forming an initial version of the simulated walking semantic sequence, including:

[0064] Step 41: Obtain the occlusion level sequence and verify whether there is a one-to-one correspondence between the occlusion level sequence and the discrete-time index sequence. The occlusion level sequence can be derived from the occlusion analysis results of the robot's local or overall occlusion in the virtual scene. If a corresponding occlusion level identifier is missing at any discrete-time index in the discrete-time index sequence, a null placeholder record is written at that discrete-time index position, along with a null placeholder identifier, to clearly indicate that the occlusion information at that time point is missing. In this way, the integrity of the occlusion level sequence in the time dimension can be maintained at the data structure level, avoiding index misalignment or semantic ambiguity in subsequent processing due to data missingness.

[0065] Step 42: Convert the occlusion level sequence into a flaw risk level sequence index by index according to a preset mapping rule. The flaw risk level describes the degree of risk of visual flaws or semantic inconsistencies occurring during robot simulation walking at the corresponding discrete-time index. The preset mapping rule is a one-to-one discrete mapping relationship between occlusion level identifiers and flaw risk levels, used to convert visual occlusion information into risk level data that can be directly used for subsequent semantic verification and repair. When the occlusion level identifier is a null placeholder record, write the null placeholder record of the flaw risk level at the corresponding discrete-time index, and simultaneously write the null placeholder identifier to maintain the consistency between the flaw risk level sequence and the occlusion level sequence in the index dimension.

[0066] Step 43: Read the values ​​of the displacement segment label sequence, left and right stepping event sequence, stride level sequence, and detection risk level sequence at the corresponding discrete-time indexes according to the discrete-time index sequence, and bind these values ​​with the corresponding discrete-time indexes to construct a simulated walking semantic frame. The simulated walking semantic frame is a data structure describing the robot's simulated walking state at a single discrete-time index, and its field set is limited to discrete-time index, displacement segment label, left and right stepping event, stride level, and detection risk level. After constructing all simulated walking semantic frames, arrange them in ascending order of the discrete-time index sequence to form an initial version of the simulated walking semantic sequence, ensuring that each discrete-time index in the discrete-time index sequence corresponds to a unique simulated walking semantic frame in the initial version of the simulated walking semantic sequence.

[0067] Through the above implementation methods, this invention unifies and structurally encapsulates multi-source data from real displacement data, gait semantic data, and virtual scene occlusion information within a discrete-time index space, forming a preliminary simulated walking semantic sequence with clear field definitions and time correspondences. By introducing a flaw detection risk level and co-organizing it with gait-related semantic fields, potential visual inconsistencies during simulated walking can be identified and quantified in advance at the data layer, improving the overall consistency and credibility of robot simulated walking results in virtual-real coupled scenarios.

[0068] In one embodiment of the present invention, a consistency check is performed on the initial simulated walking semantic sequence based on a set of hard constraint rules, generating a violation marker sequence and a set of vulnerability risk windows, including:

[0069] Step 51: Obtain the set of hard constraint rules and limit it to a set of rules used only for consistency checks on simulated walking semantic frames in the initial simulated walking semantic sequence. The set of hard constraint rules refers to a set of rules used to determine whether a simulated walking semantic frame meets preset consistency conditions. To ensure the determinism and verifiability of the rule determination process, the rule reference fields of each rule item in the hard constraint rule set are further limited to include only displacement segment labels, left and right step events, stride level, and slip-through risk level. Furthermore, the gait coherence rules are explicitly defined to only reference the left and right step events corresponding to the previous discrete-time index, thereby avoiding data dependencies across multiple indices or uncertain ranges.

[0070] Step 52: Traverse the initial simulation walk semantic sequence in ascending order of the discrete-time index sequence. For each discrete-time index in the discrete-time index sequence, perform a consistency check on the corresponding simulation walk semantic frame according to the set of hard constraint rules. When any rule is violated, the violation mark corresponding to that discrete-time index is determined to be a violation; when all rule items are not violated, the violation mark corresponding to that discrete-time index is determined to be non-violation. Through the above frame-by-frame verification method, a violation mark sequence corresponding one-to-one with the discrete-time index sequence is formed, where the value range of the violation mark sequence is limited to violation and non-violation, which is used to clearly identify the semantic compliance status at each discrete-time index.

[0071] Step 53 involves traversing the violation marker sequence, identifying the discrete-time indices of the violation markers, and merging consecutive violation discrete-time indices into a detection risk window. The detection risk window describes the range of indices containing semantic inconsistencies or potential visual detection risks within a continuous time interval. For each detection risk window, its starting discrete-time index is determined as the minimum discrete-time index within the window, and its ending discrete-time index is determined as the maximum discrete-time index within the window. Subsequently, all detection risk windows are arranged in ascending order of their starting discrete-time indices to obtain a set of detection risk windows, thereby clearly identifying high-risk intervals requiring further processing in the time dimension.

[0072] Through the above implementation methods, this invention introduces a set of hard constraint rules on the basis of the initial simulated walking semantic sequence, performs frame-by-frame consistency verification on the simulated walking semantics, and outputs the verification results in the form of violation marker sequences and a set of error risk windows, thereby realizing structured identification and interval-based localization of simulated walking semantic anomalies. By transforming the semantic inconsistency problem into a computable and mergeable risk window form, it not only reduces the manual inspection cost of simulated walking results in virtual-real coupled scenarios, but also provides a clear and limited processing range for subsequent deterministic semantic editing and repair, thereby effectively improving the stability and consistency of robot simulated walking methods in complex virtual environments.

[0073] In one embodiment of the present invention, the generation of the risk window set for revealing flaws in step 5 further includes:

[0074] Step 61: Generate an initial set of violation risk windows based on the violation marker sequence. The violation marker sequence is a marker sequence that corresponds one-to-one with the discrete-time index sequence, and its value is used to indicate whether the simulated walking semantic frame at the corresponding discrete-time index violates the set of hard constraint rules. In this embodiment, the discrete-time indices marked as violations are identified, and consecutive violation discrete-time indices are merged into a violation risk window to obtain the initial set of violation risk windows. The initial set of violation risk windows is used to describe the basic interval where semantic inconsistency risks occur consecutively in the time dimension.

[0075] Step 62: After generating the initial set of vulnerability risk windows, iterate through them to determine the number of discrete-time indices between any two adjacent vulnerability risk windows. The number of discrete-time indices refers to the number of discrete-time indices located between the end discrete-time index of the previous vulnerability risk window and the start discrete-time index of the next vulnerability risk window, and not belonging to any vulnerability risk window. In this embodiment, it is further determined whether the number of discrete-time indices is one, and whether the violation mark corresponding to this unique discrete-time index in the violation mark sequence is non-violation. This determination is used to identify adjacent vulnerability risk windows that are separated only because they briefly meet the rules at a single time point.

[0076] Step 63: For two adjacent violation risk windows that meet the above judgment conditions, the discrete time index of the interval is merged into the previous violation risk window, and the end discrete time index of the previous violation risk window is updated to the end discrete time index of the next violation risk window. At the same time, the next violation risk window is deleted, thereby completing the window merging process. In this way, the window breakage phenomenon caused by the non-violation mark of a single frame can be eliminated at the data structure level, and the final set of violation risk windows is obtained.

[0077] Through the above implementation methods, this invention introduces a secondary merging mechanism for the set of violation risk windows based on the violation marker sequence. This allows the violation risk windows to not only reflect strictly continuous violation intervals but also cover potentially continuous risk segments masked by transient rule satisfaction. By enhancing the continuity representation of the violation risk window set in the time dimension, the robustness of risk interval identification during virtual-real coupled robot simulation walking is effectively improved. This provides a more reasonable and stable processing range for subsequent deterministic semantic editing and repair, thereby further improving the consistency and credibility of simulation walking results in complex virtual environments.

[0078] In one embodiment of the present invention, based on a set of procedural risk windows and a sequence of violation markers, deterministic semantic editing and repair are performed on the simulated walking semantic frames within the procedural risk windows to obtain a repaired simulated walking semantic sequence, including:

[0079] Step 71: Read the set of detection risk windows and the sequence of violation markers, and confirm their consistency. Specifically, confirm that all discrete-time indices covered by the set of detection risk windows correspond to violation markers in the violation marker sequence. Then, define the editing scope of semantic editing using the start and end discrete-time indices of each detection risk window in the set of detection risk windows. Simulation walking semantic frames within this editing scope in the initial version of the simulation walking semantic sequence are identified as editable objects, while simulation walking semantic frames outside the editing scope are identified as non-editable objects. This method strictly limits semantic editing and repair to the identified high-risk range, avoiding unnecessary modifications to data in risk-free ranges.

[0080] Step 72: After determining the editing scope, further define the editing fields and rules for deterministic semantic editing and repair. Deterministic semantic editing and repair refers to performing a unique editing operation on a specified semantic field under fixed rules and fixed priority constraints to ensure the reproducibility of the repair process. In this embodiment, the editing fields are limited to left and right stepping events and stride levels; displacement segment labels and detection risk levels are not included in the editing fields. Simultaneously, a fixed-priority set of editing rules is set, with the priority order being: no-detection repair, stop-walk consistency repair, and gait continuity repair. Specifically, no-detection repair is used to handle simulated walking semantic frames with a high detection risk level; stop-walk consistency repair is used to handle simulated walking semantic frames with a stop displacement segment label; and gait continuity repair is used to handle simulated walking semantic frames where the left and right stepping events are the same as those corresponding to the previous discrete-time index. By fixing the editing fields and rule priorities, at most one type of rule is effective at the same discrete-time index, avoiding ambiguity caused by multiple rules overlapping.

[0081] Step 73: Traverse the initial version of the simulation walking semantic sequence in ascending order of the discrete-time index sequence. During the traversal, when the simulation walking semantic frame corresponding to the current discrete-time index is determined to be the object to be edited, the triggering conditions of the editing rules are judged sequentially according to the fixed priority mentioned above. After the high-priority rule is triggered and the editing is completed, the low-priority rule will not be executed on the simulation walking semantic frame. When the simulation walking semantic frame corresponding to the current discrete-time index is not the object to be edited, all fields of the simulation walking semantic frame are kept unchanged. After the traversal is completed, all simulation walking semantic frames are rearranged according to the order of the discrete-time index sequence to obtain the repaired simulation walking semantic sequence.

[0082] Through the above implementation methods, this invention employs a deterministic semantic editing and repair mechanism to perform targeted corrections on the semantics of simulated walking within the editing scope defined by the set of error detection risk windows. This ensures that the simulated walking results can simultaneously meet the requirements of gait consistency, displacement consistency, and error detection risk control in a virtual-real coupled scenario. By limiting semantic editing to reproducible and verifiable deterministic operations, not only is the consistency and stability of simulated walking semantics improved in complex virtual environments, but a reliable data foundation is also provided for subsequent consistency verification and round convergence output.

[0083] In one embodiment of the present invention, a consistency check is performed on the repaired simulated walking semantic sequence, and the final simulated walking semantic sequence is obtained within a preset number of rounds. The final simulated walking semantic sequence, a consistency check pass marker, and a window repair record are output, including:

[0084] Step 81: Read the repaired simulated walking semantic sequence and reuse the set of hard constraint rules used in the aforementioned consistency verification step to perform consistency verification frame by frame on the repaired simulated walking semantic sequence. The repaired simulated walking semantic sequence is the semantic sequence obtained after performing deterministic semantic editing and repair within the flaw risk window, and its structure is consistent with the initial version of the simulated walking semantic sequence. By performing rule judgment on the simulated walking semantic frame corresponding to each discrete time index in the discrete time index sequence, a violation mark sequence corresponding one-to-one with the discrete time index sequence is generated, and the consistency verification result of the current repaired simulated walking semantic sequence is determined based on whether there are violation marks in the violation mark sequence. By reusing the same set of hard constraint rules, the consistency verification criteria before and after repair can be kept consistent.

[0085] Step 82: After completing one consistency check, a preset number of rounds is set and the round count is recorded. The preset number of rounds limits the maximum number of repetitions of consistency checks and semantic editing repair to avoid infinite loop processing in complex virtual-real coupled scenarios. When the consistency check result shows that there are no violation markers, the current repaired simulated walking semantic sequence is directly determined as the final simulated walking semantic sequence. When the consistency check result shows that there are still violation markers and the round count has not reached the preset number of rounds, the current repaired simulated walking semantic sequence is used as a new input object, and deterministic semantic editing repair is re-executed, and consistency checks are performed again after the repair is completed. When the round count reaches the preset number of rounds, regardless of whether there are still violation markers, the repaired simulated walking semantic sequence when the round count reaches the preset number of rounds is determined as the final simulated walking semantic sequence. Through this round control mechanism, the semantic repair and check process converges within a finite number of rounds.

[0086] Step 83: After obtaining the final simulated walking semantic sequence, output the final simulated walking semantic sequence and generate a consistency verification pass flag based on the last consistency verification result. This flag indicates whether the final simulated walking semantic sequence satisfies the set of hard constraint rules. Simultaneously, summarize the set of flaw detection risk windows generated during each round of consistency verification within a preset number of rounds to form a window repair record. The window repair record describes the flaw detection risk intervals identified and processed throughout the entire repair and verification process.

[0087] Through the above implementation methods, after completing the semantic repair of the simulated walking, this invention introduces a consistency verification mechanism controlled by preset rounds, enabling the simulated walking semantic sequence to reach a stable state and output results within a finite number of attempts in a virtual-real coupled scenario. By simultaneously outputting the final simulated walking semantic sequence, consistency verification pass markers, and window repair records, not only is the reliability and verifiability of the simulated walking semantic results guaranteed, but the engineering applicability and system stability of the robot simulated walking method under complex virtual environment and real data coupling conditions are also enhanced.

[0088] This invention provides a robot simulation walking system based on virtual-real coupling, comprising:

[0089] The discrete index modeling module is used to establish discrete-time index sequences and obtain a basic data set that corresponds one-to-one with the discrete-time index sequences. The basic data set includes: location data sequences, gait templates, and occlusion level sequences.

[0090] The displacement segment determination module is used to generate a displacement segment label sequence based on the adjacent index difference of the position data sequence;

[0091] The gait semantic generation module is used to generate left and right stepping event sequences based on gait templates and to generate stride level sequences based on displacement segment label sequences.

[0092] The walking semantic construction module is used to convert the occlusion level sequence into the exposure risk level sequence, bind the displacement segment label sequence, left and right step event sequence, stride level sequence, and exposure risk level sequence according to the same discrete time index, construct the simulated walking semantic frame and form the initial version of the simulated walking semantic sequence.

[0093] The consistency verification and analysis module is used to perform consistency verification on the initial simulated walking semantic sequence based on the set of hard constraint rules, and generate a violation marker sequence and a set of vulnerability risk windows.

[0094] The deterministic semantic repair module is used to perform deterministic semantic editing and repair on the simulated walking semantic frames within the fault risk windows based on the set of fault risk windows and the sequence of violation markers, so as to obtain the repaired simulated walking semantic sequence.

[0095] The round convergence output module is used to perform consistency verification on the repaired simulated walking semantic sequence. It obtains the final simulated walking semantic sequence within a preset number of rounds and outputs the final simulated walking semantic sequence, consistency verification pass markers, and window repair records.

[0096] It should be noted that the interval and threshold sizes are set for ease of comparison. The size of the threshold depends on the amount of sample data and the base number set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless calculations, and the formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0097] The embodiments of the present invention have been described above, but the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of the present embodiments, all of which are within the protection scope of the present embodiments.

Claims

1. A robot walking simulation method based on virtual-real coupling, characterized in that, Includes the following steps: Step 1: Establish a discrete-time index sequence and obtain the basic data set corresponding to the discrete-time index sequence; wherein, the basic data set includes: location data sequence, gait template, and occlusion level sequence; Step 2: Generate a displacement segment label sequence based on the adjacent index differences of the location data sequence; Step 3: Generate left and right stepping event sequences based on the gait template, and generate stride level sequences based on the displacement segment label sequences; Step 4: Convert the occlusion level sequence into a detection risk level sequence, bind the displacement segment label sequence, left and right step event sequence, stride level sequence, and detection risk level sequence with the same discrete time index, construct a simulated walking semantic frame and form an initial version of the simulated walking semantic sequence. Step 5: Perform consistency verification on the initial simulated walking semantic sequence based on the set of hard constraint rules, and generate a violation marker sequence and a set of vulnerability risk windows; Step 6: Based on the set of detection risk windows and the sequence of violation markers, perform deterministic semantic editing and repair on the simulated walking semantic frames within the detection risk windows to obtain the repaired simulated walking semantic sequence, including: Step 71: Read the set of error risk windows and the sequence of violation markers, and confirm that all violation markers corresponding to the discrete time indices covered by the set of error risk windows are violations. Use the starting discrete time index and the ending discrete time index of each error risk window in the set of error risk windows to limit the editing range. Determine the simulation walking semantic frames in the initial version of the simulation walking semantic sequence that are within the editing range as the editable objects, and determine the simulation walking semantic frames that are outside the editing range as non-editable objects. Step 72: The editing fields for deterministic semantic editing and repair are limited to left and right stepping events and stride level. A set of editing rules with fixed priorities is set. The fixed priorities are, in order, non-detection repair, stop-walk consistency repair, and gait continuity repair. Among them, non-detection repair is performed on simulated walking semantic frames with a high detection risk level, stop-walk consistency repair is performed on simulated walking semantic frames with a displacement segment label of "stop", and gait continuity repair is performed on simulated walking semantic frames where the left and right stepping events are the same as the left and right stepping events corresponding to the previous discrete time index. Step 73: Traverse the initial version of the simulation walking semantic sequence in ascending order of the discrete time index sequence. Perform deterministic semantic editing and repair at the discrete time index corresponding to the edited object according to a fixed priority. After the high priority rule is triggered and the editing is completed, the low priority rule will not be executed. Keep the simulation walking semantic frame unchanged at the discrete time index corresponding to the non-edited object. Arrange all the traversed simulation walking semantic frames in the order of the discrete time index sequence to obtain the repaired simulation walking semantic sequence. Step 7: Perform consistency verification on the repaired simulated walking semantic sequence, obtain the final simulated walking semantic sequence within a preset number of rounds, and output the final simulated walking semantic sequence, consistency verification pass marker, and window repair record.

2. The robot simulation walking method based on virtual-real coupling according to claim 1, characterized in that, Establish discrete-time index sequences and obtain the basic data sets that correspond one-to-one with the discrete-time index sequences, including: Step 11: Establish a discrete-time index sequence, set the discrete-time index sequence as an ordered set of indices arranged in ascending order, and apply uniqueness constraints and ordering constraints to the discrete-time index sequence. Step 12: Obtain a basic data set including location data sequence, gait template and occlusion level sequence. The location data sequence is a set of location information record values ​​that correspond one-to-one with the discrete time index sequence. The gait template is a finite sequence composed of preset gait symbols. The occlusion level sequence is a set of occlusion degree identifiers that correspond one-to-one with the discrete time index sequence. Step 13: Perform a consistency check on the basic data set, check the index coverage relationship and one-to-one correspondence between the location data sequence and the occlusion level sequence and the discrete time index sequence, and when any discrete time index in the discrete time index sequence is missing a location information record value or occlusion degree identifier, write a corresponding null value placeholder record in that discrete time index and write a null value placeholder identifier in that discrete time index.

3. The robot simulation walking method based on virtual-real coupling according to claim 1, characterized in that, A displacement segment label sequence is generated based on the adjacent index differences of the location data sequence, including: Step 21: Construct adjacent index pairs according to the discrete time index sequence in ascending order of index. Each adjacent index pair consists of the current discrete time index and the previous discrete time index. Adjacent index pairs are constructed only for discrete time indices other than the starting index. Read the location information record value corresponding to the adjacent index pair in the location data sequence. Step 22: For each adjacent index pair, subtract the location information record value corresponding to the previous discrete time index from the location information record value corresponding to the current discrete time index of the adjacent index pair to obtain the adjacent index difference. Arrange the adjacent index differences in the discrete time index sequence as location change information, and the location change information only corresponds to the discrete time index other than the starting index. Step 23: Determine the position change amplitude based on the position change information. Square each dimension component of the difference between adjacent indices and sum them. Then take the square root of the summation result to obtain the position change amplitude. When the position change amplitude is zero, the corresponding displacement segment label is determined as "stop". When the position change amplitude is not zero, the corresponding displacement segment label is determined as "move". The displacement segment label sequence only includes "stop" and "move". The displacement segment label corresponding to the starting index of the discrete time index sequence is determined as "stop". When the position information record value of the position data sequence is a null value, the corresponding displacement segment label is determined as "stop".

4. The robot simulation walking method based on virtual-real coupling according to claim 1, characterized in that, Generate left and right stepping event sequences based on gait templates, and generate stride level sequences based on displacement segment label sequences, including: Step 31: Read the gait template according to the discrete time index sequence, and expand the gait template index by index according to the discrete time index sequence so that each discrete time index in the discrete time index sequence corresponds to a unique gait template gait symbol, and obtain the initial left and right stepping event sequence. Step 32: Correct the initial left and right step event sequence index by index according to the displacement segment label sequence. When the discrete time index corresponding to the displacement segment label sequence is "stop", determine the left and right step event corresponding to the initial left and right step event sequence as "transition". When the discrete time index corresponding to the displacement segment label sequence is "move", keep the left and right step event corresponding to the initial left and right step event sequence unchanged, and obtain the left and right step event sequence. Step 33: Generate position change amplitude based on the position data sequence and stride level sequence based on the displacement segment label sequence. Specifically, for discrete time indices other than the starting index, subtract the position information record value corresponding to the previous position data sequence from the current position information record value to obtain the adjacent index difference, and convert the adjacent index difference into position change amplitude. When the discrete time index corresponding to the displacement segment label sequence is stationary, the stride level is determined to be no stride. When the discrete time index corresponding to the displacement segment label sequence is moving, the stride level is determined to be small stride, medium stride, or large stride based on the comparison between the position change amplitude and the position change amplitude of the previous discrete time index. For discrete time indices that lack the position change amplitude of the previous discrete time index, the stride level is determined to be medium stride.

5. The robot simulation walking method based on virtual-real coupling according to claim 1, characterized in that, The occlusion level sequence is converted into a detection risk level sequence. The displacement segment label sequence, left and right stepping event sequence, stride level sequence, and detection risk level sequence are bound together by the same discrete-time index to construct a simulated walking semantic frame and form an initial version of the simulated walking semantic sequence, including: Step 41: Obtain the occlusion level sequence and verify the one-to-one correspondence between the occlusion level sequence and the discrete time index sequence. If any discrete time index in the discrete time index sequence is missing an occlusion degree identifier, write a null value placeholder record to that discrete time index and write a null value placeholder identifier to that discrete time index. Step 42: Convert the occlusion level sequence into a detection risk level sequence index by index according to the preset mapping rule. The preset mapping rule is a one-to-one discrete mapping between the occlusion degree identifier and the detection risk level. When the occlusion degree identifier is a null placeholder record, write the null placeholder record of the detection risk level into the corresponding discrete time index and write the null placeholder identifier into the discrete time index. Step 43: Read the corresponding values ​​of the displacement segment label sequence, left and right step event sequence, stride level sequence, and slip-through risk level sequence index by index according to the discrete time index sequence, and bind the corresponding values ​​with the corresponding discrete time index to construct a simulated walking semantic frame. The field set of the simulated walking semantic frame is discrete time index, displacement segment label, left and right step event, stride level, and slip-through risk level. Arrange all simulated walking semantic frames in ascending order of the discrete time index sequence to form the initial version of the simulated walking semantic sequence, so that each discrete time index in the discrete time index sequence corresponds to a unique simulated walking semantic frame in the initial version of the simulated walking semantic sequence.

6. The robot simulation walking method based on virtual-real coupling according to claim 1, characterized in that, Based on the set of hard constraint rules, a consistency check is performed on the initial simulation walking semantic sequence, generating a violation marker sequence and a set of vulnerability risk windows, including: Step 51: Obtain the set of hard constraint rules and limit the set of hard constraint rules to the set of rules for performing consistency checks on the simulated walking semantic frames in the initial simulated walking semantic sequence. The rule reference fields of the hard constraint rule set are limited to displacement segment label, left and right step events, stride level, and slip-through risk level. The gait coherence rules are limited to referencing only the left and right step events corresponding to the previous discrete time index. Step 52: Traverse the initial simulation walk semantic sequence in ascending order of the discrete-time index sequence. Perform consistency check on the simulation walk semantic frame corresponding to each discrete-time index according to the set of hard constraint rules. If any rule item is violated, the violation mark corresponding to the discrete-time index is determined to be a violation. If all rule items are not violated, the violation mark corresponding to the discrete-time index is determined to be non-violation. This forms a violation mark sequence that corresponds one-to-one with the discrete-time index sequence. The violation mark sequence includes violations and non-violations. Step 53: Traverse the violation marker sequence, identify the violation marker as the discrete time index of the violation, and merge consecutive violation discrete time indices into a violation risk window. Determine the starting discrete time index of each violation risk window as the minimum discrete time index within the violation risk window and the ending discrete time index as the maximum discrete time index within the violation risk window. Arrange all violation risk windows in ascending order of their starting discrete time indices to obtain a violation risk window set.

7. The robot simulation walking method based on virtual-real coupling according to claim 6, characterized in that, Step 5, generating the set of risk windows for revealing errors, also includes: Step 61: Generate an initial set of exposure risk windows based on the violation mark sequence. The initial set of exposure risk windows is obtained by merging consecutive discrete time indices of violations into exposure risk windows. Step 62: Traverse the initial set of exposure risk windows, determine the number of discrete time indexes between any two adjacent exposure risk windows, and determine that the number of discrete time indexes is one and the corresponding violation mark is not a violation. Step 63: For two adjacent defect risk windows that satisfy the judgment in step 62, merge the interval discrete time index into the previous defect risk window, update the end discrete time index of the previous defect risk window to the end discrete time index of the next defect risk window, delete the next defect risk window, and obtain the defect risk window set.

8. The robot simulation walking method based on virtual-real coupling according to claim 1, characterized in that, Perform a consistency check on the repaired simulated walking semantic sequence, obtain the final simulated walking semantic sequence within a preset number of rounds, and output the final simulated walking semantic sequence, consistency check pass markers, and window repair records, including: Step 81: Read the repaired simulated walking semantic sequence, and reuse the set of hard constraint rules to perform consistency verification on the repaired simulated walking semantic sequence frame by frame, generate a violation mark sequence that corresponds one-to-one with the discrete time index sequence, and determine the consistency verification result based on the violation mark sequence; Step 82: Set a preset number of rounds and record the round count. When the consistency check result shows that there is no violation mark, determine the repaired simulated walking semantic sequence as the final simulated walking semantic sequence. When the consistency check result shows that there is a violation mark and the round count has not reached the preset number of rounds, re-execute the deterministic semantic editing and repair and execute the consistency check again. When the round count reaches the preset number of rounds, determine the repaired simulated walking semantic sequence when the round count reaches the preset number of rounds as the final simulated walking semantic sequence. Step 83: Output the final simulated walking semantic sequence, generate a consistency check pass mark based on the last consistency check result, and summarize the set of error risk windows generated by each round of consistency check within the preset rounds to form a window repair record.

9. A robot simulation walking system based on virtual-real coupling, characterized in that, The robot simulation walking method based on virtual-real coupling as described in any one of claims 1-8 includes: The discrete index modeling module is used to establish discrete-time index sequences and obtain a basic data set that corresponds one-to-one with the discrete-time index sequences. The basic data set includes: location data sequences, gait templates, and occlusion level sequences. The displacement segment determination module is used to generate a displacement segment label sequence based on the adjacent index difference of the position data sequence; The gait semantic generation module is used to generate left and right stepping event sequences based on gait templates and to generate stride level sequences based on displacement segment label sequences. The walking semantic construction module is used to convert the occlusion level sequence into the exposure risk level sequence, bind the displacement segment label sequence, left and right step event sequence, stride level sequence, and exposure risk level sequence according to the same discrete time index, construct the simulated walking semantic frame and form the initial version of the simulated walking semantic sequence. The consistency verification and analysis module is used to perform consistency verification on the initial simulated walking semantic sequence based on the set of hard constraint rules, and generate a violation marker sequence and a set of vulnerability risk windows. The deterministic semantic repair module is used to perform deterministic semantic editing and repair on the simulated walking semantic frames within the fault risk windows based on the set of fault risk windows and the sequence of violation markers, so as to obtain the repaired simulated walking semantic sequence. The round convergence output module is used to perform consistency verification on the repaired simulated walking semantic sequence. It obtains the final simulated walking semantic sequence within a preset number of rounds and outputs the final simulated walking semantic sequence, consistency verification pass markers, and window repair records.

Images