Toy rapid assembling and adjusting method based on modular assembly

Abstract

The invention relates to the technical field of toy assembling, in particular to a toy rapid assembling and adjusting method based on modular assemblies. The method comprises the following steps: acquiring physical attribute data of a modular component through scanning equipment, and forming a modular component combined data set through three-dimensional reconstruction processing and feature extraction operation; performing feature matching and compatibility verification on the current assembly state data based on the data set, and determining the real-time assembly state of the modular component; a preset assembly path is planned, deviation of the real-time assembly state is predicted according to the path, and assembly deviation parameters are obtained; the deviation parameters are analyzed, and a multi-dimensional strategy containing structure alignment, connection strength and interaction response adjustment is obtained; and determining the multi-dimensional adjustment amount of the assembly process according to the strategy, and carrying out assembly compensation adjustment on the target modular component. The method can accurately obtain the component attributes, monitor the assembly state in real time, predict the deviation in advance and perform targeted adjustment, assist in improving the assembly efficiency and quality of the modular toy, and guarantee the realization of toy functions.

Description

Technical Field

[0001] This invention relates to the field of toy assembly technology, and in particular to a method for rapid assembly and adjustment of toys based on modular components. Background Technology

[0002] In the current toy manufacturing and consumer market, modular component toys have gradually become an important category due to their flexible combination and ability to meet diverse play needs. These toys are typically composed of multiple modular components with different structures and functions. Users can disassemble and reassemble the components according to their preferences to create toy products of different shapes, greatly enhancing the fun and interactivity of the toys. However, the actual assembly process of modular component toys faces many problems that urgently need to be solved. Traditional modular toy assembly relies heavily on manual experience for component matching and assembly. Because modular components from different batches may have slight dimensional variations, and the accuracy of human identification of component physical properties (such as shape, interface dimensions, and material hardness) is limited, component matching errors are prone to occur. Once a matching error occurs, not only does it require disassembly and reassembly, wasting considerable time, but forced assembly may also damage component interfaces, affecting the toy's lifespan. During assembly, there is a lack of effective monitoring and evaluation mechanisms for the real-time assembly status. When assembling manually, it's impossible to accurately determine whether the current assembly position is precise or whether the connections between components are secure; judgment relies solely on visual observation and touch. This method is highly subjective and makes it difficult to guarantee consistent assembly quality. This is especially true for modular toys with complex structures and numerous assembly steps. Even minor deviations in one step can accumulate, hindering subsequent assembly and potentially resulting in a toy that fails to perform its intended functions. When deviations occur during assembly, existing adjustment methods are often simplistic and delayed. Assembly is typically halted for troubleshooting only after significant problems arise (such as components failing to connect or becoming loose after assembly). Furthermore, adjustments are often made by completely disassembling and reassembling the entire assembly, lacking targeted, localized adjustment strategies. This approach is not only inefficient but may also exacerbate component wear due to repeated disassembly and reassembly. Additionally, it fails to predict the development trend of deviations, making it difficult to fundamentally prevent their recurrence. As modular toys become increasingly feature-rich, the demands on their interactive performance after assembly are also rising. For example, some modular toys contain electronic components, requiring precise circuit connections between components to ensure the proper functioning of interactive features such as lighting, sound effects, and movements. Traditional assembly methods cannot simultaneously guarantee structural assembly precision and electrical connection precision, often resulting in structurally sound assembly but malfunctioning electrical interactive functions, severely impacting the user experience. These problems not only hinder the improvement of assembly efficiency and quality for modular toys but also, to some extent, limit the further development and innovation of the modular toy category. Summary of the Invention

[0003] The main objective of this invention is to provide a method for rapid assembly and adjustment of toys based on modular components, aiming to solve the technical problems in the prior art.

[0004] This invention proposes a rapid assembly and adjustment method for toys based on modular components, including: The physical property data of modular components are collected by scanning equipment, and three-dimensional reconstruction and feature extraction are performed based on the physical property data to obtain a modular component combination dataset. Based on the modular component combination dataset, feature matching and compatibility verification are performed on the current assembly status data to determine the real-time assembly status of the modular components. A preset assembly path is determined, and the real-time assembly status of the modular components is predicted according to the preset assembly path to obtain assembly deviation parameters. The assembly deviation parameters are analyzed for trend judgment and strategy to obtain multi-dimensional adjustment strategies, including structural alignment adjustment strategies, connection strength adjustment strategies and interactive response adjustment strategies. The multi-dimensional adjustment strategy is used to analyze the real-time assembly status of the modular component and the assembly deviation parameters to determine the multi-dimensional adjustment amount in the assembly process, and to perform assembly compensation adjustment on the target modular component based on the multi-dimensional adjustment amount in the assembly process.

[0005] Preferably, the obtained modular component combination dataset includes: The physical attribute data is classified and extracted according to the component type to obtain geometric structure data and surface characteristic data; The geometric data is triangulated and surface reconstructed to generate an initial three-dimensional virtual assembly model; The surface characteristic data is mapped to the initial three-dimensional virtual assembly model for texture enhancement to obtain the target three-dimensional virtual assembly model; Based on the assembly complexity requirements, determine the partitioning method and partitioning granularity; The target 3D virtual assembly model is spatially partitioned according to the partitioning method and partitioning granularity to obtain the modular component combination dataset.

[0006] Preferably, determining the real-time assembly status of the modular components includes: Key features are extracted from the modular component combination dataset and the current assembly status data respectively to obtain the model feature point set and the real-time assembly feature point set; Multi-dimensional feature descriptions are performed sequentially on the model feature point set and the real-time assembled feature point set to obtain a model feature description subset and a real-time assembled feature description subset. The feature matching algorithm is used to perform matching calculations on the model feature point set and the real-time assembled feature point set according to the model feature description subset and the real-time assembled feature description subset to obtain the matching feature point set. Based on the set of matching feature points, state prediction processing and compatibility verification are performed to determine the real-time assembly status of modular components.

[0007] Preferably, determining the real-time assembly status of the modular components includes: The state prediction algorithm is initialized based on the initial assembly state of the target modular component and the error characteristics of the scanning device. Based on the historical assembly state data in the physical property data, the assembly process is fitted to construct a modular component dynamics model, and the estimated value of the assembly state sequence is predicted according to the modular component dynamics model. Based on the historical assembly state data and the estimated assembly state sequence, the state prediction algorithm is subjected to covariance prediction and iterative update to obtain an optimized state prediction algorithm. The optimized state prediction algorithm is used to estimate the state of the matching feature point set to determine the real-time assembly state of the modular component.

[0008] Preferably, the multi-dimensional adjustment strategy includes: The assembly deviation parameters are analyzed according to a time series to determine the trend of the assembly deviation. Based on the assembly deviation trend information, multi-dimensional adjustment strategies are analyzed to obtain structural alignment adjustment strategy, connection strength adjustment strategy, and interaction response adjustment strategy. Based on the aforementioned structural alignment adjustment strategy, connection strength adjustment strategy, and interaction response adjustment strategy, the multi-dimensional adjustment strategy is determined.

[0009] Preferably, the method for obtaining the assembly deviation parameters includes: Based on the optimized state prediction algorithm, the assembly process is predicted for the real-time assembly state of the modular component, and the assembly state prediction sequence values ​​at multiple preset time points are obtained. Based on the assembly state prediction sequence value, generate assembly prediction path information; Based on the preset assembly path, the assembly prediction path information is used to calculate the vector difference to obtain the assembly deviation parameter.

[0010] Preferably, determining the multidimensional adjustment amount in the assembly process includes: The structural alignment adjustment strategy is used to perform position compensation and interference detection analysis on the real-time assembly status of the modular components and the assembly deviation parameters to determine the structural alignment adjustment amount; Based on the connection strength adjustment strategy, stress distribution analysis and strength optimization compensation are performed on the real-time assembly status of the modular component and the assembly deviation parameters to determine the connection strength adjustment amount. The interactive response adjustment strategy is used to control and compensate the real-time assembly status of the modular component and the assembly deviation parameter to obtain the interactive response adjustment amount. Based on the structural alignment adjustment amount, the connection strength adjustment amount, and the interaction response adjustment amount, the multidimensional adjustment amount of the assembly process is determined.

[0011] Preferably, obtaining the interactive response adjustment amount includes: According to the interactive response adjustment strategy, the control compensator is initialized, and the parameter information of the control compensator includes linear control terms, integral control terms, and nonlinear control terms. Based on the control compensator, a closed-loop compensation calculation is performed on the real-time assembly status of the modular component and the assembly deviation parameter to obtain the interactive response adjustment amount.

[0012] Preferably, the method further includes: Real-time monitoring of user operation data, and updating the real-time assembly status of the modular components based on the user operation data; The multi-dimensional adjustment strategy is dynamically adjusted based on the real-time assembly status of the updated modular components. The adjusted multi-dimensional adjustment strategy will be applied to the real-time adjustment of the subsequent assembly process.

[0013] Preferably, the method further includes: Personalized assembly guidance is generated based on historical assembly records and user preference data; The assembly process is dynamically compensated for multidimensional adjustments based on real-time collected ambient light data and operating platform vibration data. The final output is an optimized modular component assembly configuration scheme.

[0014] The beneficial effects of this invention are as follows: This method utilizes scanning equipment to collect physical property data of modular components, and then performs 3D reconstruction and feature extraction based on this data to form a modular component assembly dataset. This process overcomes the limitations of traditional manual identification of component attributes, accurately capturing key information such as component shape, interface size, and material properties, thus constructing a comprehensive and accurate component data foundation. This dataset enables the digital representation of component physical attributes, avoiding misjudgments caused by subjective judgment or visual errors during manual identification. It provides precise data support for subsequent component matching and assembly operations, making component matching more accurate, reducing assembly errors caused by attribute identification mistakes, and lowering the possibility of component damage from the source. It also lays a reliable data foundation for subsequent assembly status monitoring and adjustment. In determining the real-time assembly status of modular components, this method performs feature matching and compatibility verification on the current assembly status data based on the component combination dataset. This operation enables real-time dynamic monitoring of the assembly process, allowing for timely detection of issues such as component mismatch and compatibility. Compared to traditional methods that rely on visual inspection and touch to judge the assembly status, this data-driven feature matching and compatibility verification is more objective and accurate. It allows for real-time monitoring of subtle changes during the assembly process, avoiding the accumulation of assembly deviations caused by delayed or inaccurate manual judgment. This ensures that the assembly process remains under control, maintaining stable assembly quality, reducing subsequent assembly difficulties caused by accumulated assembly deviations, and improving overall assembly efficiency. In terms of assembly path planning and deviation prediction, this method plans a preset assembly path and predicts deviations in the real-time assembly status according to the path to obtain assembly deviation parameters. This measure enables advance prediction of assembly deviations, breaking away from the passive adjustment approach adopted in traditional assembly methods where deviations only occur after they happen. By predicting potential deviations in advance, operators can prepare for contingencies during assembly, preventing deviations from escalating and reducing assembly downtime caused by deviations. Simultaneously, the planning of the preset assembly path provides clear guidance for assembly operations, making the assembly process more orderly and avoiding the inefficiencies caused by chaotic operation sequences in manual assembly, further improving the smoothness and efficiency of assembly. This method analyzes assembly deviation parameters to determine trends and formulate strategies, resulting in multi-dimensional adjustment strategies, including structural alignment adjustment, connection strength adjustment, and interactive response adjustment. This multi-dimensional approach overcomes the limitations of traditional, singular assembly deviation adjustment methods, enabling targeted adjustments based on the type and cause of the deviation. For example, when structural alignment deviations occur, precise positional correction can be performed using the structural alignment adjustment strategy; when component connection strength is insufficient, the connection method or force can be optimized using the connection strength adjustment strategy; and when deviations in components involving interactive functions affect interactive performance, the accuracy of electrical connections or mechanical interactions can be ensured using the interactive response adjustment strategy. This diversified approach not only improves the targeting and effectiveness of deviation adjustments, avoiding the inefficiency and component damage caused by traditional disassembly and reassembly, but also balances structural assembly accuracy with interactive performance requirements. This ensures that the assembled toy meets structural stability requirements while simultaneously fulfilling preset interactive functions, enhancing the user experience. This method determines multi-dimensional adjustment amounts during the assembly process by analyzing real-time assembly status and assembly deviation parameters, and then performs assembly compensation adjustments based on these adjustments. This process achieves precise compensation for assembly deviations, enabling fine-tuning of assembly operations according to specific deviations to ensure the final assembly result meets preset standards. Simultaneously, the entire adjustment process revolves around real-time data, exhibiting strong dynamic adaptability and capable of handling the impact of dimensional deviations in different batches of components, changes in the assembly environment, and other factors, ensuring the stability and consistency of assembly quality under various complex conditions. Through these technical means, this method effectively improves the assembly efficiency and quality of modular toys, reduces component loss, ensures the toy's functional realization, provides strong technical support for the further development and innovation of modular toys, and better meets users' diverse and high-quality demands for modular toys. Attached Figure Description

[0015] Figure 1 This is a schematic diagram illustrating the working principle of the toy rapid assembly and adjustment method based on modular components described in this invention. Figure 2 A flowchart for state prediction and real-time assembly state determination.

[0016] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0018] like Figure 1As shown, this application provides a rapid assembly and adjustment method for toys based on modular components, including: collecting physical property data of modular components through a scanning device; performing three-dimensional reconstruction and feature extraction based on the physical property data to obtain a modular component combination dataset; performing feature matching and compatibility verification on the current assembly state data based on the modular component combination dataset to determine the real-time assembly state of the modular components; planning and determining a preset assembly path; predicting the deviation of the real-time assembly state of the modular components according to the preset assembly path to obtain assembly deviation parameters; performing trend judgment and strategy analysis on the assembly deviation parameters to obtain multi-dimensional adjustment strategies, including structural alignment adjustment strategies, connection strength adjustment strategies, and interaction response adjustment strategies; using the multi-dimensional adjustment strategies to analyze the real-time assembly state and assembly deviation parameters of the modular components, determining the multi-dimensional adjustment amount in the assembly process, and performing assembly compensation adjustment on the target modular component based on the multi-dimensional adjustment amount in the assembly process.

[0019] In one embodiment, Example 1: The acquisition of physical property data is completed using a high-precision 3D scanning device. This device can acquire dense point cloud information of the modular component surface and the optical properties of the material. The point cloud data records the 3D coordinates of each sampling point on the component surface, forming a basic dataset describing the geometric shape of the object. The optical property data includes surface visual attributes such as color and reflectivity, providing information support for subsequent texture reconstruction. The component type is classified according to its geometric shape features and interface mechanical structure, such as distinguishing block components, shaft components, or connectors. The classification and extraction process adopts a feature vector-based machine learning algorithm. This algorithm identifies the contour features and structural patterns of different components through training, thereby automatically classifying the input raw point cloud data into predefined component types. The geometric structure data mainly includes the vertex coordinates, edge vectors, and surface information of the component. These data constitute the skeleton model of the component. The surface property data is extracted from the multispectral information acquired by scanning, including parameters such as color value, roughness coefficient, and material type.

[0020] The triangulation process employs an improved Delaunay triangulation algorithm, which transforms discrete point cloud data into a continuous set of triangular facets. First, the point cloud is spatially rasterized and partitioned. Then, the optimal connection method of points within each partition is calculated to ensure that the generated triangles are as close to equilateral as possible, avoiding sharp triangles. In the surface reconstruction stage, non-uniform rational B-spline (NURBS) technology is used to automatically place control points by calculating the curvature characteristics of the point cloud. The local support properties of B-splines are used to smoothly interpolate missing or noisy areas. The initial 3D virtual assembly model is thus generated. This model fully preserves the geometry of the components but has not yet added surface details.

[0021] The texture enhancement stage uses UV mapping technology to fuse surface characteristic data with the geometric model. First, the triangular mesh model is parametrically unfolded to map the three-dimensional surface to a two-dimensional coordinate space. Then, the texture image generated based on the optical characteristic data is matched with corresponding points. Color information is smoothly attached to the mesh surface through bilinear interpolation. Roughness data is used to generate normal maps. By simulating the effect of microscopic uneven surfaces on light, the visual realism is enhanced. The target three-dimensional virtual assembly model finally has high-fidelity appearance characteristics, which not only accurately reflects the geometric structure, but also presents a realistic material visual effect.

[0022] Assembly complexity assessment is based on multiple dimensions such as the number of components, the number of connections, and the diversity of interface types. The system automatically selects an appropriate partitioning method based on the assessment results. Uniform grid partitioning is used for simple structures, while an octree adaptive partitioning method is used for complex structures. The partitioning granularity is dynamically adjusted according to the assembly accuracy requirements. Fine-grained partitioning is used when the accuracy requirements are high, and coarse-grained partitioning is used when the accuracy requirements are low to improve processing efficiency. The spatial partitioning process decomposes the target 3D virtual assembly model into multiple logical units. Each unit contains local geometric data, texture data, and connection information with other units. The modular component combination dataset is finally stored in a hierarchical data structure, supporting fast spatial query and local update operations.

[0023] In one embodiment, Example 2: See Figure 2 Feature extraction from the modular component assembly dataset and the current assembly state data is performed using the Scale Invariant Feature Transform (SIFT) algorithm. This algorithm detects stable extrema in the image by constructing a Gaussian difference pyramid. The model feature point set is obtained from the pre-generated 2D projection map of the component's 3D model, with the projection angle dynamically adjusted according to the current assembly viewpoint. The real-time assembly feature point set is captured by a high-frame-rate industrial camera installed in the operation area. The camera acquires a video stream of the assembly site at a rate of 60 frames per second, from which keyframes are extracted for feature detection. Feature point screening sets contrast thresholds and edge response thresholds to exclude low-contrast regions and feature points with unstable edges, ensuring that the extracted feature points are rotation-invariant and scale-invariant.

[0024] Multi-dimensional feature description employs directional FAST and rotated BRIEF (ORB) descriptor generation techniques, creating a 256-bit binary description vector for each feature point. The model feature description subset is pre-calculated and stored in the database offline, while the real-time assembled feature description subset is calculated immediately after feature point extraction. The descriptor generation process includes two key operations: first, the principal orientation of the feature points is determined using the gray-scale centroid method, ensuring rotation invariance; then, a circular neighborhood with a radius of 20 pixels is constructed, and 256 pairs of pixels are selected in the rotated-corrected coordinate system for brightness comparison, generating a binary string. Both the model feature description subset and the real-time assembled feature description subset are stored using an optimized Hamming tree data structure, supporting fast feature matching retrieval. The feature matching stage employs the Approximate Nearest Neighbor Fast Library (FLANN) matcher. This matcher constructs a multi-layer hash table to spatially partition the descriptors. During the matching calculation process, a matching distance ratio threshold is set, retaining only matching pairs where the nearest and second nearest neighbor distance ratio is less than 0.7, effectively eliminating erroneous matches. The output of the matching feature point set is a list of coordinate correspondences. Each matching pair contains the coordinates of the model feature points and the coordinates of the real-time image feature points. At the same time, the Hamming distance between the descriptors is recorded as the matching confidence.

[0025] In the initialization phase of state prediction processing, Kalman filter parameters are configured. The initial assembly state of the target modular component is acquired through a laser displacement sensor and an inertial measurement unit (IMU), including position coordinates and Euler angle information. Error characteristic information of the scanning equipment is obtained from the equipment calibration report, including a systematic bias of 0.05 mm and a random error model conforming to a normal distribution. The state vector is defined as a six-degree-of-freedom pose and its first derivative, and the state transition matrix is ​​constructed based on the Newtonian kinematics model. The assembly process fitting uses cubic spline interpolation to process historical assembly state data. The physical property database stores motion trajectory samples from past successful assemblies. The modular component dynamics model is constructed considering mass distribution and the friction coefficient of connectors, and the differential relationship between component pose and force is established through the Lagrange equation. The assembly state sequence estimate predicts the component pose and velocity for the next five time steps (100 milliseconds each), and Gaussian process regression is incorporated into the prediction process to correct model errors. The covariance prediction algorithm calculates the uncertainty ellipse of the state estimate. The process noise covariance matrix is ​​set according to the equipment vibration characteristics. During the iterative update phase, anomaly measurements are detected using a novelty sequence. A measurement rejection mechanism is triggered when the measurement residual exceeds three times the standard deviation. The optimized state prediction algorithm dynamically adjusts the weights of the predicted and measured values ​​using the Kalman gain matrix. The final state estimate output includes the optimal pose estimate and its covariance matrix. The compatibility verification phase checks the geometric constraints of the connection interface, including diameter tolerance verification for shaft-hole fits and parallelism verification for planar contacts. When the predicted state results in an interface gap of less than 0.1 mm, it is considered an interference risk, triggering the state correction process. The real-time assembly state output of the modular components is a six-DOF pose parameter with confidence intervals, while also marking the locations of potentially conflicting interfaces.

[0026] The entire implementation process adopts a pipelined architecture, with feature extraction, description, matching, and state estimation executed in parallel, and a double-buffering mechanism ensuring real-time performance. The data processing unit is equipped with a dedicated hardware accelerator; SIFT feature extraction is implemented using FPGA, with a processing time of less than 10 milliseconds per frame, and Kalman filtering operations are parallelized using GPUs, increasing the iteration update speed by 8 times. The system includes an exception handling mechanism that automatically switches to a backup visual recognition algorithm when the feature matching failure rate exceeds 30% for three consecutive frames.

[0027] Taking the assembly of a large space shuttle model as an example, when the user is assembling the connection between the right wing and the fuselage, a binocular vision system installed above the operating area begins to acquire real-time assembly images. This system uses a 20-megapixel industrial camera to capture the assembly process at 30 frames per second, and the image data is transmitted to the processing unit via gigabit Ethernet. Simultaneously, a pre-generated 3D model of the wing assembly is retrieved from the database. This model contains 32,768 triangular facets and 256 feature points. The feature extraction process is performed synchronously in the image processing unit, matching SIFT feature points extracted from the real-time video stream with the model's feature points. In the wing assembly scenario, the system successfully matched 173 pairs of feature points, mainly concentrated in the wing connection tenon and fuselage interface areas. Each feature point carries a 256-dimensional descriptor vector. Hamming distance calculations between descriptors show that 83% of the matching point pairs have a distance below the threshold of 0.15, indicating high matching quality. The state prediction algorithm begins initialization; the Kalman filter's state vector contains the six-degree-of-freedom pose information of the wing assembly. The initial state was acquired using a laser tracker, measuring the wing's current pose relative to the fuselage as follows: X-axis offset 2.3mm, Y-axis rotation angle 1.7 degrees. Error parameters from the scanning equipment were loaded into a filter, including camera calibration error of 0.05mm and laser ranging error of 0.02mm. The assembly process fitting module retrieved similar assembly records from the historical database, finding the average trajectory data of the past 37 wing assemblies. This data was input into the dynamic model, which considered the wing's mass distribution (2.4kg) and the friction coefficient of the connecting mechanism (0.15). The system predicted the assembly state sequence within the next 500 milliseconds, showing that, based on the current trend, the wing would continue to offset 0.3mm along the X-axis. The compatibility verification step checked the geometric constraints of the connection interface, finding that the fit clearance between the wing tenon and the fuselage slot was decreasing. When the prediction showed the clearance would be less than 0.1mm, the system determined there was a risk of interference. The verification algorithm also checked the mating status of the electrical connectors, finding a 0.2mm misalignment at pin 12, which could potentially affect subsequent circuit conduction. Real-time status updates are achieved through multi-sensor fusion. Inertial Measurement Unit (IMU) data shows the wing is rotating at an angular velocity of 0.5 degrees per second. Force sensors detect a user-applied grip force of 6.2 N, slightly higher than the standard value of 5 N. This data is integrated into the status estimate, with the Kalman gain matrix dynamically adjusting the weight ratio of predicted and measured values. The system performs covariance update calculations to assess the uncertainty of the status estimate. The process noise covariance is set based on the table vibration characteristics, while the measurement noise covariance is based on sensor accuracy parameters. During iterative updates, when a decrease in visual feature matching confidence is detected, the system automatically increases the weighting coefficient of the IMU data.The final output of the real-time assembly status includes the precise orientation of the wing relative to the fuselage: X-axis offset 2.28±0.03mm, Y-axis rotation 1.72±0.05 degrees, and identifies potential risk points in the electrical connectors. This status data is refreshed at a frequency of 20Hz, providing input for subsequent deviation prediction and adjustment strategies. An anomaly handling mechanism continuously monitors the entire process; when five consecutive frames of image matching fail, the system automatically switches to backup positioning mode, using pre-attached QR code markers for orientation calculation. The time delay of all processing steps is strictly controlled within 50 milliseconds to ensure that real-time requirements are met.

[0028] In one embodiment, Example 3: the time series analysis of the assembly deviation parameters is processed using an Autoregressive Integral Moving Average (ARIMA) model. This model can effectively capture the trend and seasonal changes in the time series. The assembly deviation parameters are stored in vector form, containing six components: positional deviation and angular deviation. Each component is sampled at 100-millisecond intervals to form a time series. The trend is determined by calculating the first-order forward difference of the deviation series to obtain the instantaneous rate of change. Simultaneously, the second-order difference is calculated using the sliding window method, with the window size set to 5 sampling points to capture the acceleration information of the rate of change. The assembly deviation trend information is ultimately represented as a change vector with directional attributes, containing two dimensions: magnitude rate of change and direction rate of change.

[0029] The structural alignment adjustment strategy generates pose correction commands based on the deviation trend. When a continuous increase in positional deviation is detected, the strategy calculates a reverse compensation vector. The magnitude of the compensation vector is proportional to the rate of change of deviation, and its direction is opposite to the deviation trend. The connection strength adjustment strategy analyzes the impact of deviation on the stress of mechanical connectors. It calculates the stress distribution cloud map of the connection area through finite element analysis. When the deviation causes the stress concentration factor to exceed 60% of the material's yield strength, strength optimization compensation is triggered. The interactive response adjustment strategy uses an adaptive control algorithm to dynamically adjust the proportional gain and derivative gain of the control system according to the rate of change of deviation, ensuring that the system response speed is synchronized with the deviation change. The multi-dimensional adjustment strategy integrates the outputs of the three strategies through a weighted fusion algorithm. The weight coefficients are dynamically adjusted according to the current assembly stage. The initial assembly stage focuses on structural alignment, while the final assembly stage focuses on connection strength. The assembly process prediction uses a state-space model and performs state deduction based on the modular component dynamic equations. The state vector contains 12 components, including position, velocity, acceleration, and angular velocity. The preset time points are set to five prediction points evenly distributed within the next 500 milliseconds, with a time interval of 100 milliseconds. The assembly state prediction sequence values ​​are calculated using numerical integration, and the fourth-order Runge-Kutta method is used to ensure calculation accuracy. The assembly prediction path information is connected into a smooth curve by cubic spline interpolation, and the path curvature constraint ensures that the generated path conforms to the motion characteristics of the mechanical structure.

[0030] The pre-defined assembly path extracts the ideal motion trajectory from the computer-aided design (CAD) model. This trajectory consists of a series of critical path points, each represented by a quaternion of position and orientation. Vector difference calculation employs a modified Euclidean distance metric, considering the coupling effect of position and orientation deviations. The calculation formula is as follows: in: This represents the overall deviation. and These represent the actual position and the desired position vectors, respectively. and Let these represent the actual attitude and the desired attitude quaternions, respectively. and These are the weighting coefficients. This represents a quaternion multiplication operation. The assembly deviation parameter output is a six-dimensional vector containing three translational deviation components and three rotational deviation components.

[0031] The deviation trend analysis module incorporates an anomaly detection mechanism. When the deviation change rate exceeds a threshold for three consecutive sampling periods, it is identified as an abnormal trend, triggering an emergency adjustment mode. The multi-strategy coordinator employs fuzzy logic control, dynamically adjusting the priority of the three strategies based on the deviation magnitude and change rate to ensure a smooth transition during the adjustment process. The predicted path generation module is equipped with a path smoothing filter, using Kalman filtering to smooth the predicted path and reduce the impact of measurement noise on path prediction. Real-time data processing utilizes a circular buffer structure to store deviation data for the most recent 30 seconds, supporting sliding window analysis and historical data backtracking. All algorithm modules implement multi-threaded parallel computation, with feature extraction, trend analysis, and strategy generation executed synchronously in different threads, exchanging data through lock-free queues. The system includes a runtime status monitor to detect the computational load and data processing latency of each module in real time. When the latency exceeds 50 milliseconds, the calculation accuracy is automatically reduced to ensure real-time performance. The deviation parameter calibration module periodically executes a sensor calibration program, measuring the systematic error between the actual pose and visual measurements using a laser tracker, and updating the camera calibration parameters and hand-eye conversion matrix. The environmental adaptability adjustment module monitors changes in workplace temperature and humidity, and dynamically adjusts deviation tolerance thresholds based on the material's coefficient of thermal expansion. All adjustment strategy parameters are visually managed through a graphical interface, allowing engineers to adjust control parameters and observe the effects online.

[0032] Taking the assembly of a large wind turbine generator model as an example, when the blades are docked with the hub, the system begins to perform time-series analysis of the deviation parameters. A laser displacement sensor installed on the hub docking surface collects the pose data of the blade flange at a frequency of 100Hz, forming a six-dimensional time series including XYZ axis offsets and rotation angles. This data is input into an autoregressive integral moving average model, configured with parameters p=2, d=1, and q=2, which effectively captures the periodic vibrations and trend offsets of the blades during the docking process. The trend judgment module detects a continuous positive offset of the blades in the Y-axis direction, with an average rate of change of 0.12 mm / s over the past 5 sampling periods. Simultaneously, calculations using a sliding window reveal that the Z-axis rotational angular velocity is accelerating; the window size is 8 sampling points, and the second-order difference value shows an angular acceleration of 0.05 deg / s². 2Assembly deviation trend information is quantified into a vector group with directional attributes, where the position change vector is [0.08, 0.12, -0.03] mm / s and the angle change vector is [0.02, 0.01, 0.05] deg / s. The structural alignment adjustment strategy generates compensation commands based on the detected trends. For the case of continuous Y-axis offset, a reverse compensation vector of [-0.10, -0.15, 0.02] mm / s is calculated. The connection strength adjustment strategy initiates finite element analysis to simulate the stress distribution on the bolt connection surface under the current deviation conditions. It is found that stress concentration occurs at the position of bolt number 3, with the equivalent stress value reaching 55% of the material's yield strength. The interactive response adjustment strategy adjusts the gain parameters of the hydraulic control system accordingly, increasing the proportional gain by 15% and decreasing the differential gain by 8%. The multi-dimensional adjustment strategy integrates these outputs through a weighted fusion algorithm, assigning a weight of 0.6 to the structural alignment strategy and a weight of 0.3 to the connection strength strategy during the blade docking stage. The assembly process prediction module performs state deduction based on the blade dynamics model, with the state vector containing 12 degrees of freedom parameters. The system predicts the pose sequence at five time points within the next 300 milliseconds, with a time interval of 60 milliseconds. The prediction results show that, based on the current trend, the blade will interfere with the hub by 0.25 mm after 180 milliseconds. The assembly prediction path generates a smooth curve through cubic spline interpolation, and the path curvature constraint ensures that the predicted trajectory conforms to the flexible deformation characteristics of the blade. The preset assembly path extracts the ideal docking trajectory from the CAD model. This path consists of 21 key points, each containing millimeter-accurate coordinates and attitude data with 0.01-degree accuracy. Vector difference calculation uses an improved spatial distance metric method, while also considering the inertial effects caused by the blade weight distribution. The calculated comprehensive deviation at the current moment is 0.38 mm, with radial deviation being the main component. The deviation trend analysis module sets up a multi-level early warning mechanism. When the deviation change rate exceeds 0.15 mm / s for three consecutive sampling periods, a yellow warning state is triggered. The multi-strategy coordinator uses fuzzy logic control to dynamically adjust the strategy priority based on the magnitude and rate of change of the deviation. When the deviation exceeds the 0.4 mm threshold, the system automatically increases the weight of the structure alignment strategy to 0.8. Real-time data processing employs a dual-buffer structure. The main buffer stores deviation data from the most recent 200 sampling points, while the auxiliary buffer retains historical feature patterns. All algorithm modules run in parallel on multi-core processors, with feature extraction, trend analysis, and strategy generation each utilizing independent computation threads. The system monitoring module tracks the computational latency of each thread in real time, ensuring that the overall processing latency does not exceed 40 milliseconds. The environmental adaptation module monitors changes in temperature and humidity in the assembly environment and adjusts deviation tolerances based on the thermal expansion characteristics of the composite material. When the ambient temperature increases by 5 degrees Celsius, the system automatically relaxes the radial deviation threshold by 0.05 mm. All adjustment parameters are transmitted to the actuators via industrial Ethernet, employing a deterministic Ethernet protocol to ensure transmission timing accuracy.

[0033] In one embodiment, Example 4: the determination of the structural alignment adjustment amount is completed collaboratively through position compensation calculation and interference detection. Specifically, the system obtains the pose matrix and position offset in the assembly deviation parameters of the modular components in real-time assembly status. The position compensation calculation adopts the inverse kinematics solution method. Taking the assembly of a car model gearbox by a six-degree-of-freedom industrial robotic arm as an example, when a deviation of 0.25mm between the axis of the driving gear and the axis of the driven gear is detected, the compensation algorithm calculates that the translation vector to be adjusted is [0.22, -0.07, 0.13]mm and the rotation vector is [0.5°, -0.3°, 0.1°]. The interference detection analysis adopts the hierarchical bounding box (AABB) algorithm to establish the axial bounding box tree of each component. In the gear assembly scenario, the potential collision between the gear teeth and the adjacent bearing seats is detected. When the bounding box overlap exceeds the safety threshold of 0.1mm, an interference warning is generated. The final output of the structural alignment adjustment amount is an instruction set containing the compensation vector and the interference avoidance path.

[0034] The calculation of the connection strength adjustment is based on stress distribution analysis and a strength optimization compensation strategy. In the gearbox assembly case, the system first constructs a finite element analysis model, discretizing the gear shaft assembly into 18,753 tetrahedral elements, and applies a preload caused by assembly deviations. Stress distribution analysis shows local stress concentration in the bearing housing contact area, with a maximum equivalent stress value of 143 MPa. The strength optimization compensation module retrieves the yield strength parameter of 45 steel from the material property database. When the stress concentration factor exceeds 0.6 times the yield limit, the compensation mechanism is triggered, calculating that the preload of the connecting bolts needs to be reduced by 15% to optimize the stress distribution. The connection strength adjustment output is a torque adjustment command; specific parameters are shown in the table below. Table 1: Gearbox Assembly Connection Strength Adjustment Parameters The generation of the interactive response adjustment depends on the closed-loop calculation of the control compensator. In the gearbox assembly scenario, the PID structural compensator is initialized with parameters set as proportional coefficient Kp=0.85, integral coefficient Ki=0.12, and derivative coefficient Kd=0.05. The compensation closed-loop calculation runs at a frequency of 100Hz, and each iteration collects real-time pose deviation as input. During gear meshing, when the axial deviation is detected to be continuously increasing, the nonlinear control term activates the anti-saturation algorithm to dynamically limit the accumulation rate of the integral term. In specific execution, when the deviation exceeds the 0.2mm threshold, the derivative term weight is automatically increased by 30% to enhance system damping. The interactive response adjustment is finally converted into a pulse width modulation (PWM) signal to drive the servo motor, with the adjustment being a 200Hz pulse sequence with a 65% duty cycle. The synthesis of multi-dimensional adjustment in the assembly process adopts a hierarchical fusion strategy, with structural alignment adjustment, connection strength adjustment, and interactive response adjustment entering three parallel processing channels. In the gearbox assembly example, the structural channel outputs the pose correction matrix, the strength channel outputs the bolt torque curve, and the response channel outputs the motor control signal. The fusion processor assigns weight coefficients according to the assembly stage: the initial positioning stage is weighted at [0.6, 0.2, 0.2], the precision leveling stage at [0.3, 0.5, 0.2], and the final locking stage at [0.1, 0.7, 0.2]. The multi-dimensional adjustment output is a structured data packet containing a timestamp, adjustment type identifier, and binary parameter stream. The implementation system is equipped with a dual-redundancy verification mechanism; when the required positional compensation for structural alignment adjustment exceeds the safety limit of 5mm, a manual confirmation process is automatically triggered. Each modification to the strength adjustment is logged, including the stress distribution diagram before adjustment and the simulation verification results after adjustment. A control variable change rate limiter is set during the interactive response adjustment generation process to prevent severe jitter in the actuator. All adjustment parameters are transmitted to the execution unit via industrial Ethernet, using the Time-Sensitive Networking (TSN) standard to ensure real-time performance.

[0035] The anomaly handling module monitors conflict events during the adjustment process. In the gearbox assembly case, a conflict was detected between the downward displacement required for structural alignment adjustment and the increased bolt preload required for connection strength optimization. The conflict resolution strategy prioritizes ensuring structural alignment requirements while marking the strength adjustment amount for recalculation. The system's operating status is visualized on a dashboard, displaying real-time heatmaps of adjustment application points and adjustment forces on the 3D component model. Historical adjustment data is stored in a time-series database, supporting retrieval and analysis by component number or time range.

[0036] In one embodiment, Example 5: Real-time monitoring of user operation data is achieved through a multimodal sensing system. Taking the assembly of a large dinosaur skeleton model as an example, a piezoelectric thin-film sensor array mounted on the assembly workbench collects hand-held force distribution data at a sampling rate of 1000Hz, while a top-mounted binocular vision system tracks the user's finger movement trajectory at a frame rate of 30fps. When the user assembles the 23rd vertebral component, the system detects an abnormal increase in pressure applied by the left thumb to 8.7N (the standard range is 5-6N), and the assembly speed drops to 1.5 components per minute (the baseline speed is 2 components / minute). These data trigger a real-time assembly status update process, recalculating the contact angle and preload parameters of the vertebral connector. The dynamic adjustment of the multi-dimensional adjustment strategy occurs within 200 milliseconds after the status update. The system analyzes that the connection characteristics of the current vertebral component belong to a precision snap-fit ​​structure, automatically increasing the priority weight of the structural alignment adjustment strategy to 0.7, while decreasing the weight of the connection strength strategy to 0.2. The adjusted strategy was immediately applied to the subsequent assembly process of the 24th rib assembly. The position control accuracy of the robotic arm end effector was improved from the default ±0.2mm to ±0.1mm, and the torque control sensitivity was correspondingly reduced by 15% to accommodate the user's cautious operating style.

[0037] The personalized assembly guidance is generated based on in-depth analysis of the user's historical database. This user has completed seven dinosaur skeleton assemblies, and the database records show a preference for an assembly sequence from tail to head, and a habit of pausing to check complex joints. The system combines these characteristics to generate a customized 3D guided view. During the assembly of the scapula (number 25), a magnifying glass mode automatically pops up, highlighting the alignment marks. Voice prompts are adjusted to twice per step (the standard is three times) to reduce interference for experienced users. The environmental adaptability compensation module simultaneously processes multiple environmental parameters. When the light sensor detects that the illumination in the operating area drops to 150 lux (the standard requires 300 lux), the visual enhancement algorithm automatically increases the gamma value in image processing to 1.8, increasing the contrast of shadow areas. The vibration sensor detects a 0.15g, 2Hz low-frequency vibration on the worktable (originating from nearby equipment). The motion control algorithm immediately injects a vibration compensation signal, canceling the interference with an inverted waveform. During the assembly of the skull component (number 26), environmental compensation maintains stable positioning accuracy, keeping the displacement error caused by vibration within 0.05mm.

[0038] The optimized configuration scheme output adopts a hierarchical data structure. The final dinosaur skeleton assembly scheme contains three core layers: the basic structure layer defines the spatial topology of 87 components, including the three-dimensional coordinates and direction vectors of each connection point; the dynamic parameter layer records 214 parameters adjusted during assembly, such as the torque compensation value of 0.3 N·m for the 12th caudal vertebra and the angle fine-tuning of 1.2 degrees for the 5th cervical vertebra; and the environmental configuration layer stores the illumination compensation curve and vibration filtering parameter set. The scheme file is transmitted to the cloud knowledge base via an encrypted link, and a simplified AR guidance map is generated and projected onto the user interface. A smart rollback mechanism is implemented in the system settings. When the user adjusts the same joint component three times consecutively (such as the 15th knee joint), the current state is automatically saved as a temporary configuration point, allowing the user to go back to the three most recent operation nodes. The quality verification module performs an automatic scan after assembling every five components, comparing the deviation between the actual assembly state and the virtual model. During the dinosaur pelvis assembly stage, an axial offset of 0.3 mm was detected, triggering a fine-tuning cycle to recalibrate the connection angles. The user feedback channel is integrated into the haptic interface of the control handle, confirming operation commands through different vibration modes: short pulses indicate positioning completion, and long vibrations indicate proper connection. During the final assembly stage, when installing the tooth component, the system detects insufficient force applied by the user, triggering a three-level haptic warning mode (three strong vibrations at 0.5-second intervals), while automatically reducing the connection strength threshold of that component by 30%. All interaction events are recorded in the operation log, including timestamps, event types, and system response actions, forming a closed-loop optimization data chain.

[0039] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for rapid assembly and adjustment of toys based on modular components, characterized in that, The method includes: The physical property data of modular components are collected by scanning equipment, and three-dimensional reconstruction and feature extraction are performed based on the physical property data to obtain a modular component combination dataset. Based on the modular component combination dataset, feature matching and compatibility verification are performed on the current assembly status data to determine the real-time assembly status of the modular components. A preset assembly path is determined, and the real-time assembly status of the modular components is predicted according to the preset assembly path to obtain assembly deviation parameters. The assembly deviation parameters are analyzed for trend judgment and strategy to obtain multi-dimensional adjustment strategies, including structural alignment adjustment strategies, connection strength adjustment strategies and interactive response adjustment strategies. The multi-dimensional adjustment strategy is used to analyze the real-time assembly status of the modular component and the assembly deviation parameters to determine the multi-dimensional adjustment amount in the assembly process, and to perform assembly compensation adjustment on the target modular component based on the multi-dimensional adjustment amount in the assembly process.

2. The rapid assembly and adjustment method for toys based on modular components as described in claim 1, characterized in that, The obtained modular component combination dataset includes: The physical attribute data is classified and extracted according to the component type to obtain geometric structure data and surface characteristic data; The geometric data is triangulated and surface reconstructed to generate an initial three-dimensional virtual assembly model; The surface characteristic data is mapped to the initial three-dimensional virtual assembly model for texture enhancement to obtain the target three-dimensional virtual assembly model; Based on the assembly complexity requirements, determine the partitioning method and partitioning granularity; The target 3D virtual assembly model is spatially partitioned according to the partitioning method and partitioning granularity to obtain the modular component combination dataset.

3. The rapid assembly and adjustment method for toys based on modular components as described in claim 1, characterized in that, Determining the real-time assembly status of modular components includes: Key features are extracted from the modular component combination dataset and the current assembly status data respectively to obtain the model feature point set and the real-time assembly feature point set; Multi-dimensional feature descriptions are performed sequentially on the model feature point set and the real-time assembled feature point set to obtain a model feature description subset and a real-time assembled feature description subset. The feature matching algorithm is used to perform matching calculations on the model feature point set and the real-time assembled feature point set according to the model feature description subset and the real-time assembled feature description subset to obtain the matching feature point set. Based on the set of matching feature points, state prediction processing and compatibility verification are performed to determine the real-time assembly status of modular components.

4. The rapid assembly and adjustment method for toys based on modular components as described in claim 3, characterized in that, Determining the real-time assembly status of modular components includes: The state prediction algorithm is initialized based on the initial assembly state of the target modular component and the error characteristics of the scanning device. Based on the historical assembly state data in the physical property data, the assembly process is fitted to construct a modular component dynamics model, and the estimated value of the assembly state sequence is predicted according to the modular component dynamics model. Based on the historical assembly state data and the estimated assembly state sequence, the state prediction algorithm is subjected to covariance prediction and iterative update to obtain an optimized state prediction algorithm. The optimized state prediction algorithm is used to estimate the state of the matching feature point set to determine the real-time assembly state of the modular component.

5. The rapid assembly and adjustment method for toys based on modular components as described in claim 4, characterized in that, The obtained multi-dimensional adjustment strategy includes: The assembly deviation parameters are analyzed according to a time series to determine the trend of the assembly deviation. Based on the assembly deviation trend information, multi-dimensional adjustment strategies are analyzed to obtain structural alignment adjustment strategy, connection strength adjustment strategy, and interaction response adjustment strategy. Based on the aforementioned structural alignment adjustment strategy, connection strength adjustment strategy, and interaction response adjustment strategy, the multi-dimensional adjustment strategy is determined.

6. The rapid assembly and adjustment method for toys based on modular components as described in claim 4, characterized in that, The obtained assembly deviation parameters include: Based on the optimized state prediction algorithm, the assembly process is predicted for the real-time assembly state of the modular component, and the assembly state prediction sequence values ​​at multiple preset time points are obtained. Based on the assembly state prediction sequence value, generate assembly prediction path information; Based on the preset assembly path, the assembly prediction path information is used to calculate the vector difference to obtain the assembly deviation parameter.

7. The rapid assembly and adjustment method for toys based on modular components as described in claim 1, characterized in that, The determination of the multidimensional adjustment amount in the assembly process includes: The structural alignment adjustment strategy is used to perform position compensation and interference detection analysis on the real-time assembly status of the modular components and the assembly deviation parameters to determine the structural alignment adjustment amount; Based on the connection strength adjustment strategy, stress distribution analysis and strength optimization compensation are performed on the real-time assembly status of the modular component and the assembly deviation parameters to determine the connection strength adjustment amount. The interactive response adjustment strategy is used to control and compensate the real-time assembly status of the modular component and the assembly deviation parameter to obtain the interactive response adjustment amount. Based on the structural alignment adjustment amount, the connection strength adjustment amount, and the interaction response adjustment amount, the multidimensional adjustment amount of the assembly process is determined.

8. The rapid assembly and adjustment method for toys based on modular components as described in claim 7, characterized in that, The obtained interactive response adjustment amount includes: According to the interactive response adjustment strategy, the control compensator is initialized, and the parameter information of the control compensator includes linear control terms, integral control terms, and nonlinear control terms. Based on the control compensator, a closed-loop compensation calculation is performed on the real-time assembly status of the modular component and the assembly deviation parameter to obtain the interactive response adjustment amount.

9. The rapid assembly and adjustment method for toys based on modular components as described in claim 1, characterized in that, The method further includes: Real-time monitoring of user operation data, and updating the real-time assembly status of the modular components based on the user operation data; The multi-dimensional adjustment strategy is dynamically adjusted based on the real-time assembly status of the updated modular components. The adjusted multi-dimensional adjustment strategy will be applied to the real-time adjustment of the subsequent assembly process.

10. The rapid assembly and adjustment method for toys based on modular components as described in claim 9, characterized in that, The method further includes: Personalized assembly guidance is generated based on historical assembly records and user preference data; The assembly process is dynamically compensated for multidimensional adjustments based on real-time collected ambient light data and operating platform vibration data. The final output is an optimized modular component assembly configuration scheme.

Images