Three-dimensional assembly path planning method and system for solar wing structure

Abstract

The invention relates to the technical field of geometric modeling of three-dimensional models, in particular to a three-dimensional assembly path planning method and system for a solar wing structure. The method comprises the following steps: acquiring three-dimensional assembly data of the solar wing, and constructing an assembly reachable atlas according to the three-dimensional assembly data of the solar wing to obtain assembly reachable atlas data; path evasion type behavior planning is carried out according to the assembly reachable atlas data, and assembly path data is obtained; performing optimization scheduling on the assembly path data to obtain assembly path optimization data; performing interference resolution reconstruction on the assembly path optimization data to obtain assembly path reconstruction data; and performing digital twinning scene generation according to the assembly path reconstruction data to obtain assembly path digital twinning data. According to the method, the whole-process digital mapping from action behavior modeling, assembly time sequence control to interactive linkage is realized, and the verifiability, the interactivity and the assessment capability before deployment of the assembly process are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of geometric modeling of three-dimensional models, and in particular to a three-dimensional assembly path planning method and system for a solar wing structure. Background Art

[0002] Digital twins, one of the core technologies of next-generation intelligent manufacturing, build dynamic, interactive models corresponding to physical entities in virtual space through multi-source sensing, 3D modeling, state mapping, and time-series synchronization. This allows for prediction, evaluation, and optimization of physical systems. They are widely used in the design verification, manufacturing scheduling, and operational monitoring of highly complex products, demonstrating significant engineering value in aerospace and equipment manufacturing. With the increasing complexity of aerospace equipment structures, especially in the design and manufacturing of deployable components like solar panels, the assembly process involves the coordinated operation of large, multi-degree-of-freedom, and high-precision components. Traditional assembly processes rely heavily on manual planning and empirical experience for path design and motion organization. This makes it difficult to accurately identify accessible assembly areas within complex component spaces and avoid potential interference risks, resulting in low assembly efficiency and high error rates. This also hinders the deployment of automated assembly and intelligent manufacturing. The question arises: how to integrate digital twins with intelligent manufacturing for solar panels. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes a three-dimensional assembly path planning method and system for a solar wing structure to solve at least one of the above technical problems.

[0004] This application provides a three-dimensional assembly path planning method for a solar wing structure, comprising the following steps:

[0005] Step S1: Acquire the three-dimensional assembly data of the solar wing, and construct an assembly reachable graph based on the three-dimensional assembly data of the solar wing to obtain assembly reachable graph data;

[0006] Step S2: Perform path avoidance behavior planning based on the assembly reachable graph data to obtain assembly path data;

[0007] Step S3: Optimizing and scheduling the assembly path data to obtain assembly path optimization data; performing interference elimination and reconstruction on the assembly path optimization data to obtain assembly path reconstruction data;

[0008] Step S4: Generate a digital twin scene based on the assembly path reconstruction data to obtain the assembly path digital twin data.

[0009] In the present invention, by introducing the construction of assembly reachable graphs and path avoidance behavior planning mechanisms, the system can fully identify the assembly accessibility and interference risk areas in the solar wing structure, realize the spatial constraint coordination of path planning and behavioral actions, and effectively avoid the assembly failure problems caused by rigid path strategies or delayed collision detection in traditional methods. On this basis, combined with path optimization and interference resolution and reconstruction processing, the path segments with interference are further reconstructed with action posture windows and path segment re-avoidance operations, which significantly improves the continuity and spatial adaptability of the assembly process. Through the digital twin scene generation module, the execution logic and spatial dynamics of the assembly path can be reproduced with high fidelity in a virtual environment, and visual preview and multi-dimensional verification of path feasibility, spatial constraint adaptability and behavioral stability can be achieved, forming an integrated "planning-reconstruction-verification" technical closed loop.

[0010] Preferably, step S1 is specifically:

[0011] Obtain solar wing 3D assembly data;

[0012] Perform component analysis on the solar wing 3D assembly data to obtain component analysis data;

[0013] Generate a spatial correlation diagram for component analysis data based on the solar wing three-dimensional assembly data to obtain spatial correlation diagram data;

[0014] According to the three-dimensional assembly data of the solar wing, the spatial association graph data is mapped to the assembly direction to obtain the assembly direction constraint graph data;

[0015] Performing attitude space mapping on the assembly direction constraint graph data according to the solar wing three-dimensional assembly data to obtain attitude constraint graph data;

[0016] The traversable space is solved according to the posture constraint graph data to obtain the assembly reachable graph data.

[0017] In the present invention, through component analysis and spatial association graph construction, the system can accurately capture the geometric connection relationship and spatial adjacency between the various components in the solar wing structure, providing a structured expression basis for path analysis. The introduction of the assembly direction mapping mechanism can effectively identify the assembly feasibility of each component in different spatial directions, avoiding the path blind spot problem caused by the lack of directional judgment in traditional methods. Construct a posture constraint graph, and clarify the posture freedom range of the assembly action through posture space mapping, so that the action behavior can be angle-restricted and the posture feasibility screening can be performed. Solving the assembly reachable map through posture constraint graph data not only realizes the quantitative modeling of spatial accessibility, but also provides the path planning algorithm with a high-dimensional feasible space input with directional and posture constraints.

[0018] Preferably, the traversable space solution is specifically:

[0019] Generate component posture domain-decomposition body according to posture constraint graph data to obtain component posture domain-decomposition body data;

[0020] Solve the posture drive path form according to the component posture solution domain data to obtain the posture drive path form data;

[0021] Perform spatial collision volume culling on the posture-driven path form data to obtain collision culling data;

[0022] The component path-posture-direction graph is reconstructed based on the collision elimination data to obtain the assembly reachable graph data.

[0023] In the present invention, the range of posture movements that can be achieved by each component in three-dimensional space can be expressed in volume form through the generation of component posture solution bodies, breaking through the limitation that traditional methods only deal with position constraints but cannot describe the feasibility of dynamic postures. Based on the posture solution body data to solve the posture-driven path window, the system can construct a dynamic path window, while maintaining the rationality of spatial movements, finely control the posture evolution process of the assembly path, and improve the responsiveness of the path strategy to the continuity of movements. On this basis, a spatial collision volume elimination mechanism is introduced to achieve dynamic filtering of potential interference areas and path window correction, effectively reducing the probability of conflict in the path planning process. Through the reconstruction of the component path-posture-direction graph, the system integrates the three types of constraint information into a unified graph structure to generate an assembly reachable graph with directional selectivity, posture continuity and path accessibility.

[0024] Preferably, step S2 is specifically:

[0025] Extract the reachable path based on the assembly reachable graph data to obtain the reachable path data;

[0026] Generate a path window according to the reachable path data to obtain path window data;

[0027] Performing path interference area detection on the path window data to obtain path interference area data;

[0028] Local behavior replacement is performed based on the path interference area data to obtain assembly path data.

[0029] In the present invention, during the reachable path extraction stage, the system can screen out effective paths that meet the constraints of posture, direction and spatial accessibility from the assembly reachable map, avoiding the problem of redundant paths or infeasible paths caused by the low level of path abstraction in traditional methods. Through path window generation, the local spatial segments in the path can be modeled in a windowed manner to achieve a continuous description of the dynamic characteristics of the path (such as posture turning, component rotation, etc.). Path interference area detection further accurately identifies the potential conflicting segments in these windows, and determines the interfering segments through criteria such as volume overlap, direction contradiction and posture conflict, thereby effectively avoiding the risk of spatial conflict. Through the local behavior replacement mechanism, the path segments with interference are dynamically replaced at the action level, rather than abandoning the entire path, which significantly improves the efficiency of path correction and path continuity.

[0030] Preferably, the optimized scheduling is specifically:

[0031] Constructing a component dependency graph for the assembly path data to obtain component dependency graph data;

[0032] Optimize the component dependency graph data for the shortest installation time and the minimum assembly action switching range to obtain first optimized sequence data and second optimized sequence data respectively;

[0033] The first optimized sequence data and the second optimized sequence data are used to mark the assembly path data with scheduling information to obtain assembly path optimized data.

[0034] The construction of the component dependency graph in the present invention is based on the structural connection logic, spatial nesting order and assembly sequence constraints between components. It can systematically characterize the dependency chain in the assembly task and avoid the path backtracking or action conflict problems caused by ignoring the mutual constraints between components in traditional assembly planning. For the shortest installation time optimization strategy, the system gives priority to generating the component scheduling sequence with the shortest overall installation time on the premise of ensuring that the component dependency constraints are not violated, effectively compressing the total assembly time; while the action switching amplitude minimum optimization strategy considers the physical action amplitudes such as posture switching and path switching between adjacent assembly tasks, and improves the continuity and stability of the assembly action by minimizing the action change range. The two optimization sequences are fused and marked into the original path data to form a scheduling optimization result with "optimal time efficiency + minimum action cost".

[0035] Preferably, the interference resolution and reconstruction is specifically:

[0036] Perform three-dimensional interference area detection on assembly path optimization data to obtain interference area data;

[0037] Conflict volume modeling is performed based on interference area data to obtain conflict volume model data;

[0038] Reconstruct the conflict segment action posture window according to the conflict volume model data to obtain action reconstruction data;

[0039] The assembly path optimization data is replanned according to the action reconstruction data to obtain the assembly path reconstruction data.

[0040] In the present invention, three-dimensional interference area detection is performed based on assembly path optimization data. The system can identify locations in complex component layouts where path segments and surrounding structures have volume overlap or posture contradictions, solving the problem of missed conflicts in traditional methods due to failure to consider dynamic posture changes. Combined with conflict volume modeling technology, a spatial conflict expression body with directionality, shape and variable boundary characteristics is constructed to achieve quantitative modeling and spatial topological constraint expression of the conflict area. Action posture window reconstruction based on the conflict model can not only perform local corrections on the action posture corresponding to the interference segment, but also generate alternative action sequences that can adapt to spatially restricted areas, ensuring the accessibility and continuity of the assembly process. The reconstructed action is integrated into the global path through the path segment replanning mechanism, and dynamic reconstruction is achieved while retaining the continuity of the original path.

[0041] Preferably, the conflict volume modeling is specifically as follows:

[0042] Sampling component motion trajectory according to interference area data to obtain component motion trajectory data;

[0043] Performing sweep envelope generation on component motion trajectory data to obtain sweep envelope data;

[0044] Performing Boolean difference operation of interference components according to the swept envelope data to obtain interference component data;

[0045] Segment and mark conflict volume segments according to interference component data to obtain conflict volume segment data;

[0046] A conflict volume aggregation graph is constructed according to the conflict volume fragment data to obtain conflict volume model data.

[0047] In the present invention, the motion trajectory of the components in the interference area is sampled at multiple moments to obtain the spatial movement path formed by the components in the dynamic assembly process, providing temporal support for spatial modeling. Subsequently, the swept envelope generation method is used to convert the motion trajectory of the components in the assembly process into a spatial envelope, effectively capturing the complete spatial range occupied by the components during the movement process, and making up for the problem that traditional static collision detection methods lack the expressive power of dynamic interference. Based on the envelope data, the system further performs Boolean difference operations on the interfering components, thereby accurately extracting the actual conflict areas where the volume overlap occurs, and improving the accuracy and resolution of interference identification. Then, by performing segmentation and semantic annotation on the interference component data, a set of conflict volume fragments with structural significance is constructed, and the spatial positioning and semantic association capabilities of the conflict parts are enhanced. A conflict volume aggregation graph is constructed to integrate the scattered volume fragments into a graph structure expression to form a conflict volume model with spatial topological relationships, conflict severity and modifiable properties.

[0048] Preferably, the conflict segment action posture window reconstruction is specifically as follows:

[0049] Perform degree of freedom domain extraction based on the collision volume model data to obtain degree of freedom domain data;

[0050] Conduct anti-constraint analysis of conflict volume based on degree of freedom domain data to obtain blocked area data;

[0051] Reconstruct the action posture form according to the blocked area data to obtain the action posture form data;

[0052] Extract the optimal posture domain of the action posture form data to obtain the optimal posture domain data;

[0053] The reconstructed action set is generated according to the optimal posture domain data to obtain the action reconstruction data.

[0054] In the present invention, by extracting the degree of freedom domain from the conflict volume model data, the system can identify the translation and rotation dimensions that the component can still execute under the current space-constrained conditions, clarify the feasible space boundaries of the action adjustment, and break through the limitations of the traditional path reconstruction method in that the degree of freedom is rough and the granularity is too large. Based on the anti-constraint analysis carried out on the degree of freedom domain, the system can infer the spatial blockade area of the assembly behavior from the conflict volume, and provide a precise constraint basis for the feasibility screening of the posture window. The action posture window reconstruction link expresses the feasible posture range in the form of a window body, which not only enhances the structured expression ability of the posture feasible solution, but also facilitates the generation and evaluation of multi-angle and multi-posture alternative strategies. On this basis, the optimal posture domain is extracted, and the action posture solution set with the most execution value can be screened through optimization functions (such as minimizing interference and maximizing posture continuity). The reconstructed action set finally generated can replace the original action sequence and be embedded in the path segment to achieve precise obstacle avoidance, continuous steering and posture connection, and ensure assembly smoothness. Unlike traditional methods that rely on a static discrete pose library for selection, this method implements an integrated reconstruction process from spatial constraint modeling, degree of freedom compression, to pose domain optimization. This significantly improves the system's adaptability, spatial coordination, and dynamic adjustment accuracy for assembly actions in complex interference environments. This method differs from traditional path interference modeling methods that rely solely on closest distance detection or simple bounding box analysis. It significantly improves the ability to accurately model dynamic spatial conflicts and the quality of path repair strategy generation, providing a highly reliable and accurate decision-making basis for intelligent assembly path reconstruction.

[0055] Preferably, step S4 is specifically:

[0056] Component action mapping is performed based on assembly path reconstruction data to obtain component action association data;

[0057] Construct an assembly state machine based on component action association data to obtain assembly state machine data;

[0058] Generating a virtual configuration space for the assembly state machine data to obtain virtual configuration space data;

[0059] Perform posture binding according to the virtual configuration space data to obtain posture binding data;

[0060] Perform assembly action timing synchronization on the posture binding data to obtain assembly synchronization data;

[0061] Interactive events are integrated based on assembly synchronization data to obtain assembly path digital twin data.

[0062] In the present invention, through component action mapping and assembly state machine construction, the system can convert the reconstruction path action of each assembly component into a controllable state transfer unit, realize the logical drive and state monitoring of assembly behavior, and break through the static presentation mode of traditional digital twins that only simulate geometric animation. The generation of virtual configuration space enables the system to have the ability to dynamically deduce and display the spatial combination state of multiple components and multiple stages, and supports high-fidelity restoration of the assembly sequence and spatial layout of multiple components. The posture binding link accurately binds the posture parameters in the action path with the node entities in the virtual configuration, so that the assembly action has real physical expression and action consistency in space. Through action timing synchronization, the system can achieve time-domain coordination of the entire assembly process, ensuring the consistency and controllability of the component assembly sequence, relative position and timing logic. The introduction of the interactive event integration mechanism enables the digital twin system to not only have a visual path demonstration function, but also support human-computer interaction, event triggering feedback and assembly abnormality simulation.

[0063] Preferably, the present application further provides a three-dimensional assembly path planning system for a solar wing structure, for executing the three-dimensional assembly path planning method for a solar wing structure as described above, the three-dimensional assembly path planning system for a solar wing structure comprising:

[0064] The 3D assembly perception and mapping module is used to obtain the 3D assembly data of the solar wing and construct an assembly reachable map based on the 3D assembly data of the solar wing to obtain the assembly reachable map data;

[0065] The path avoidance planning module is used to perform path avoidance behavior planning based on the assembly reachable graph data to obtain assembly path data;

[0066] The path optimization and interference reconstruction module is used to optimize the assembly path data to obtain the assembly path optimization data; perform interference elimination and reconstruction on the assembly path optimization data to obtain the assembly path reconstruction data;

[0067] The digital twin generation module is used to generate digital twin scenes based on the assembly path reconstruction data to obtain assembly path digital twin data.

[0068] The beneficial effects of the present invention are as follows: through component action mapping and assembly state machine construction, the system can convert the reconstruction path action of each assembly component into a controllable state transfer unit, realize the logical drive and state monitoring of assembly behavior, and break through the static presentation mode of traditional digital twins that only simulate geometric animation. The generation of virtual configuration space enables the system to have the ability to dynamically deduce and display the spatial combination state of multiple components and multiple stages, and supports high-fidelity restoration of the assembly sequence and spatial layout of multiple components. The posture binding link accurately binds the posture parameters in the action path with the node entities in the virtual configuration, so that the assembly action has real physical expression and action consistency in space. Through action timing synchronization, the system can achieve full-process time-domain coordination of the assembly process, ensuring the consistency and controllability of the component assembly sequence, relative position and timing logic. The introduction of the interactive event integration mechanism enables the digital twin system to not only have a visual path demonstration function, but also support human-computer interaction, event triggering feedback and assembly abnormality simulation. Different from traditional solutions that only provide path animation playback or CAD hierarchical display, this method realizes digital mapping of the entire process from action behavior modeling, assembly timing control to interactive linkage, significantly improving the verifiability, interactivity and pre-deployment evaluation capabilities of the assembly process, and providing efficient and reliable technical support for virtual assembly testing and intelligent optimization of complex component systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Other features, objects and advantages of the present application will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0070] Figure 1 A flowchart showing the steps of a three-dimensional assembly path planning method for a solar wing structure according to one embodiment is shown;

[0071] Figure 2 A flowchart showing the steps of a three-dimensional assembly perception mapping method according to an embodiment is shown;

[0072] Figure 3 A flowchart showing the steps of a path avoidance planning method according to an embodiment is shown;

[0073] Figure 4 A flowchart of the steps of a digital twin generation method according to one embodiment is shown. DETAILED DESCRIPTION

[0074] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative work are within the scope of protection of the present invention.

[0075] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0076] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0077] See also Figures 1 to 4 , the present application provides a three-dimensional assembly path planning method for a solar wing structure, comprising the following steps:

[0078] Step S1: Acquire the three-dimensional assembly data of the solar wing, and construct an assembly reachable graph based on the three-dimensional assembly data of the solar wing to obtain assembly reachable graph data;

[0079] In one embodiment, the 3D assembly data is obtained using methods including, but not limited to, standard 3D file formats (such as STEP or IGES) generated through CAD modeling, or 3D point cloud data acquired through methods such as laser scanning or structured light scanning. This 3D assembly data includes information such as the spatial layout of solar wing components, the structure of connectors between components, hole distribution, component installation orientation, distribution of occluded areas, and component geometric profile data. Based on this 3D assembly data, the system uses a geometric segmentation algorithm to perform component analysis and structural decomposition on the overall solar wing model. This geometric segmentation algorithm may include a surface segmentation method based on concave point recognition or a region clustering method based on boundary contour extraction. Through component analysis, the overall model is divided into several sub-components with physical assembly significance. Each component is assigned a unique identification number during the analysis process, and its corresponding assembly direction (such as insertion direction), degrees of freedom information (including movable dimensions and their constraint ranges), and connection relationships (such as connection logic implemented by bolts, slots, or plug-in structures) are extracted. After completing component analysis, the system further constructs an assembly reachability map. Each node in the graph represents an independent component or a group of structural units. The attribute information associated with each node includes the geometric center coordinates of the component or assembly, its bounding box, and the reference pose information required for its installation. Edges between nodes describe the assembly sequence or dependencies between assembly paths. Specifically, if component A must be assembled before component B, or if there is geometric interference with component B on component A's assembly path, a directed edge is created from A to B in the graph, indicating that A takes precedence over B. To determine whether a valid assembly path exists between components, the system employs a path reachability detection algorithm based on spatial Boolean operations. Specifically, by performing continuous Boolean intersection operations on the spatial volumes of components in different poses, it determines whether a component's assembly action will spatially conflict with already installed components during the path execution. In three-dimensional space, the path space is discretized using a voxel grid. Combined with the sequence of action pose changes, the spatial region of a feasible path that a component traverses from its initial position to its target installation location is constructed. If there is a path voxel sequence that does not intersect with obstacles at all intermediate posture points, the assembly path of the component is considered reachable under the current configuration. Through the above method, the system finally generates assembly reachability map data.

[0080] Step S2: Perform path avoidance behavior planning based on the assembly reachable graph data to obtain assembly path data;

[0081] In one embodiment, the system performs a path avoidance behavior planning process based on the constructed assembly reachable graph data, combined with the initial state of the component and the target assembly position, to generate assembly path data for each component. Specifically, the system uses the current position of each component to be assembled as the starting point for path planning, and the position corresponding to its target installation posture as the end point of the path. In the directed graph structure described by the assembly reachable graph, a heuristic path search algorithm is used to generate the path. The path search algorithm may include an A* algorithm, RRT (rapidly exploring random trees) or its improved variants. The A* algorithm can dynamically balance the search efficiency and path shortestness between the path cost and the heuristic function, while the RRT algorithm is suitable for quickly generating feasible paths in high-dimensional space, and is suitable for handling high-degree-of-freedom posture changes in the component assembly process. In order to be closer to the actual assembly scene, the system introduces a dynamic obstacle prediction model between components to simulate the movement or deformation of other components during the assembly process, thereby improving the robustness of the path planning results. To model avoidance behavior, the system constructs a voxel map of the assembly path based on three-dimensional space. This discretizes the assembly space into a number of voxel units and records the traversability status of each voxel unit within this space. Based on this, the geometric volume of the components involved in the assembly is spatially expanded, extending a certain volume beyond their geometric boundaries to simulate the operating space of the assembly tool (such as the robotic arm end effector or gripper). This allows the path planning process to account for the spatial impact of tool size. To prevent unintended interference when the assembly path approaches other assembled components, the system incorporates a safety buffer boundary rule into the path planning process. This rule ensures that the planned path maintains a preset spatial clearance from surrounding obstacles at all times by setting a minimum safe distance between components, thereby achieving obstacle avoidance control. The system also sets path constraint rules to ensure that the generated path meets the kinematic constraints of the actual assembly equipment. For example, during path changes, the system sets the maximum acceptable rotation and translation amplitudes of the components to prevent drastic posture changes of the robotic arm during execution, reducing execution risk. For assembly tasks that require high docking accuracy (such as precision plug-in components), the system also introduces a path smoothing mechanism to perform continuity processing on turning points in the path. Specifically, the Bezier curve interpolation method can be used to smooth the path curve, reduce angle jumps or mutations in assembly movements, and improve the stability and accuracy of path execution. In this way, the system generates assembly path data based on the assembly reachable map that meets spatial accessibility, safety obstacle avoidance, assembly equipment motion constraints, and movement smoothness.

[0082] Step S3: Optimizing and scheduling the assembly path data to obtain assembly path optimization data; performing interference elimination and reconstruction on the assembly path optimization data to obtain assembly path reconstruction data;

[0083] In one embodiment, the system performs an optimized scheduling process on the assembly path data initially generated. This process optimizes component sorting and assembly timing by introducing a component priority scheduling graph. The component priority scheduling graph is a directed graph structure constructed based on the structural hierarchical relationship of the components and the assembly dependency logic, also known as an assembly priority dependency graph, in which each node represents a component to be assembled, and the directed edges between the nodes represent the dependency relationship of the assembly order. In priority scheduling, the system prioritizes the installation order of the main load-bearing components (such as the skeleton, support base, etc.) according to the structural load-bearing relationship, and inserts and sorts non-critical components on this basis. In order to improve scheduling efficiency and resource utilization, the system adopts a greedy scheduling algorithm and combines the time window constraint strategy to schedule the overall assembly process in batches. Based on the component dependency graph, this scheduling method takes the "earliest executable" principle as the core scheduling logic, and arranges as many components as possible to enter the same time period assembly batch without violating the dependency constraints, thereby optimizing the total assembly time and path utilization density. Specifically, the scheduling objective function set by the system is: Where T i is the assembly time of the i-th component, L i is the path length of the i-th component, λ is the path length weight coefficient, a dimensionless constant (set to 1-5 for general industrial manipulators (6-axis), 0.1-1 for high-speed linear assembly platforms, 5-20 for flexible serial robots (collaborative), and 20-100 for ultra-high-precision micro-assembly equipment), i is the component index variable, and n is the total number of components involved in the scheduling. The objective function aims to balance the overall simplicity and execution efficiency of the assembly path while ensuring reasonable time scheduling. After completing the optimized scheduling, the system further performs spatial interference detection and resolution reconstruction on the optimized assembly path data. This process first implements rapid interference detection by constructing a three-dimensional collision volume model. Specifically, the system performs spatial overlap detection on the bounding volumes corresponding to the component poses in each frame of the assembly path to identify path segments with potential spatial interference risks. For detected interfering path segments, the system applies the motion pose window reconstruction method to perform local path replacement. This method generates a set of perturbation-type posture path windows near the interfering path segment, that is, introducing small perturbations (such as posture angle offset ±2°, position offset ±5mm) on the basis of the original path to form several candidate paths. Subsequently, the system evaluates the spatial collision score of each candidate path segment, evaluates its minimum spacing and collision probability with surrounding obstacle components during execution, and selects the posture combination with the lowest collision score as the replacement path. If no valid collision-free path can be generated in all perturbation path windows, the system automatically triggers the local fallback mechanism of the assembly sequence, rearranges the scheduling order of the components corresponding to the current path segment and its dependent components, and recalculates the assembly timing and path allocation until an alternative path that meets the obstacle avoidance constraints is generated.

[0084] Step S4: Generate a digital twin scene based on the assembly path reconstruction data to obtain the assembly path digital twin data.

[0085] In one embodiment, the system constructs a digital twin scene based on the obtained assembly path reconstruction data to realize three-dimensional visualization, timing simulation and interactive feedback functions of the assembly process. The assembly path reconstruction data includes the path segment information of the component, the complete assembly action sequence and the reconstruction action sequence adjusted due to interference conflicts. The system maps the above data structure into instruction nodes in the digital twin environment to construct a timing diagram of the assembly process. Specifically, the system identifies the action path segment corresponding to each component as one or more "action instruction nodes", each node is associated with a specific component ID, a corresponding posture change sequence, an action type (such as insertion, sliding, rotation) and its timing parameters (start time, duration, dependent actions, etc.), and all instruction nodes are organized into an assembly action timing diagram according to the assembly dependency relationship. This timing diagram can express behavioral logic such as multi-component parallel actions, sequential dependencies, and path mutual exclusion, and is the control structure that drives the execution of digital twin animation. In terms of visualization implementation, the system builds an assembly scene simulation platform based on a three-dimensional rendering engine (such as Unity3D or WebGL). The platform loads a complete solar wing structural component model and maps the above assembly action timing diagram into the dynamic transformation control flow of the corresponding component. The spatial position and posture of each component change synchronously with the progress of the time axis, forming a complete three-dimensional assembly demonstration process. The system supports a variety of visual control functions, including but not limited to timing control operations such as play, pause, replay, fast forward, and jump of assembly actions; as well as scene display options such as highlighting of assembly paths, transparency adjustment, and execution speed adjustment. The platform has set up a clickable feedback mechanism based on component granularity, which supports users to click on any component during operation to query its path segment, the status of executed actions, whether there is interference history, actual execution time and other data content. All feedback data is presented in real time in the form of pop-ups, sidebars or charts. The system integrates a path anomaly detection module to monitor whether the actual execution path deviates from the planned trajectory. Based on statistical deviation detection methods, this module analyzes the offset of the spatial path of each action sequence. If the offset of any execution path exceeds three times the standard deviation interval of its expected path (i.e., the offset is >3 times the standard deviation), the system automatically marks the path segment as abnormal and highlights it in the twin interface. At the same time, path tuning suggestions pop up, including recommended re-planning postures or path buffer adjustment plans to assist users in assembly accuracy optimization and path reconfiguration. In this way, the system realizes an integrated digital twin modeling process from path reconstruction data to three-dimensional dynamic scenes, timing control and intelligent feedback, and generates digital twin data for assembly paths.

[0086] Preferably, step S1 is specifically:

[0087] Step S11: Acquire solar wing three-dimensional assembly data;

[0088] In one embodiment, assembly model data in .step, .iges, .stl, or .obj formats is imported from mainstream 3D design software. The data includes the 3D geometry (mesh data) of each subcomponent, assembly relationships between components (such as constraints, screw connections, and insertion and removal directions), origin references, coordinate system definitions, and component hierarchical structures. Isolated meshes and non-assembly virtual parts are removed, and units and coordinate system orientations are unified.

[0089] Step S12: performing component analysis on the solar wing three-dimensional assembly data to obtain component analysis data;

[0090] In one embodiment, the system divides components based on the assembly tree hierarchical structure or topological features between components in the three-dimensional assembly data. When the assembly model has an assembly tree structure, the components and their subordinate relationships can be identified step by step according to the hierarchical structure; when there is no clear assembly tree structure, the spatial segmentation of the components can be achieved through mesh connectivity analysis (for example, determining non-connected mesh areas), thereby identifying independent component entities. The extracted fields include component ID, geometric center (centroid), size bounding box (AABB / OBB), main axis direction, and connection interface location (such as hole position, snap). If the assembly model contains naming rules or additional metadata (Metadata), for example, the component is named "WingLeft_Panel1", the system can automatically determine the semantic category of the component by combining the component naming information with its geometric features. Component types include but are not limited to solar wing panels, battery modules, connecting brackets, drive mechanisms, etc. If the naming information is missing, the components are clustered and classified based on geometric features (such as aspect ratio, surface features, hole group distribution, etc.).

[0091] Step S13: generating a spatial correlation graph for the component analysis data according to the solar wing three-dimensional assembly data to obtain spatial correlation graph data;

[0092] In one embodiment, each component is a node in the graph. If any of the following conditions are met between two components, an edge is added: the distance between the connecting surfaces is less than a threshold (e.g., <1.0 mm), there are overlapping holes, snaps, and connectors (detected by Boolean intersection), and the attributes of the edge include connection type, contact area, surface normal angle, and connection priority. KD-Tree is used for rapid screening of component proximity. The normal angle and inter-surface distance of the selected component pairs are judged to construct an edge set to obtain spatial association graph data.

[0093] Step S14: performing assembly direction mapping on the spatial association graph data according to the solar wing three-dimensional assembly data to obtain assembly direction constraint graph data;

[0094] In one embodiment, potential assembly directions (which are insertion directions) are extracted based on the normal vectors of the component connection surfaces. A set of main assembly directions, such as ±X, ±Y, and ±Z, is introduced to standardize all assembly directions. Each candidate direction vector will be projected into the main direction set and assigned to the closest main direction label. Determine whether there is any obstruction on the insertion path and whether it complies with the component connection structure design (such as a pin-hole structure that only allows one-way insertion). Each pair of connecting edges is labeled with its assembly direction vector. Construct a graph, and the edge attributes include direction vector, insertable interval, and spatial volume required for assembly.

[0095] Step S15: performing attitude space mapping on the assembly direction constraint graph data according to the solar wing three-dimensional assembly data to obtain attitude constraint graph data;

[0096] In one embodiment, the posture of the component includes two parts: position and direction, which are represented by quaternion or Euler angle + translation vector. For each assembly direction, a set of allowed component motion posture areas are constructed (for example: translation ±5mm, rotation ±10°). For each posture candidate, a path feasibility test is performed to determine whether it will collide (using an axis-aligned bounding box or a three-dimensional polyhedron enclosing structure to model the boundary of the component; discrete sampling is performed on the motion trajectory, and a collision detection algorithm is called at each sampling point to determine whether there is a geometric overlap with other components). The system constructs a posture constraint graph to describe the posture state space that the component can legally transform in a specific assembly direction. Let the graph structure be G pose =(V,E pose ); V represents the set of all legal posture states, and each node corresponds to a legal component posture configuration; E pose Indicates the connection edge from one posture state to another posture state under the continuous transformation path. Each edge e mn ∈E pose Indicates that from posture P m Smooth transition to posture P n A legal path transformation relationship that does not collide or violate assembly constraints in physical space.

[0097] Step S16: Solve the traversable space according to the posture constraint graph data to obtain assembly reachable graph data.

[0098] In one embodiment, the traversable space refers to the set of three-dimensional spaces that a component can occupy in the process of moving from the initial assembly position to the target assembly position without collision, occlusion or violation of structural constraints. This space reflects the degree of freedom and reachable range of the component in performing assembly actions, and constitutes the basis of the assembly channel. The motion path is voxelized to discretize the component motion path into multiple posture keyframes. Collision detection is performed on each frame, and if consecutive frames are legal, a feasible path is formed. All feasible path spaces are combined to form the assembly channel space body. Construct an assembly reachable graph, where each node is a component + posture and the edge is a traversable path segment.

[0099] Preferably, the traversable space solution is specifically:

[0100] Generate component posture domain-decomposition body according to posture constraint graph data to obtain component posture domain-decomposition body data;

[0101] In one embodiment, the posture constraint graph is a graph structure composed of multiple legal postures, denoted as G pose , containing the pose set P i , where each posture P i They are composed of three-dimensional position vectors (x, y, z) and corresponding attitude angle vectors (θx, θy, θz), which are used to describe the possible configuration of components during the assembly process. j , define its attitude dedomain Ω j It is a subset of the special Euclidean group SE(3), representing the component C j All feasible postures that meet the constraints in six-dimensional space. The solution domain must meet the following conditions at the same time, including that the position coordinates x, y, and z of the component in three-dimensional space must be limited to the boundary of its path planning constraints; the posture angles θx, θy, and θz must be within the preset rotation tolerance range; in order to avoid interference with adjacent components, the posture solution domain should avoid the influence range of adjacent components and reserve a minimum buffer distance, for example, not less than 2 mm. In the specific solution process, a six-dimensional rasterization sampling strategy is adopted, that is, the three-dimensional position and three-dimensional posture angle space are discretely divided, and each sampling point is checked one by one to see if it meets the above constraints at the same time, so as to screen out component C. j All feasible postures in space. For assembly scenarios involving surface insertion, such as inserting components along the tangent direction on a complex surface, the affine transformation method can be further used to fit the posture of the component. At this time, the posture dedomain body will construct a continuous posture dedomain surface along the tangent direction of the component insertion path, thereby ensuring the smoothness and accessibility of the action during the assembly process. The system outputs the posture dedomain body data corresponding to each component, recorded as D pose The data structure is identified by component C j is the index, and records its corresponding legal posture space Ωj .

[0102] Solve the posture drive path form according to the component posture solution domain data to obtain the posture drive path form data;

[0103] In one embodiment, for component C j , its posture dedomain body Ω j The legal configuration set of the component in the six-dimensional posture space SE(3) has been expressed. On this basis, its posture driving path form W is defined j is one or more continuous feasible posture path channels, which are composed of a set of time-ordered posture point sequences {P0, P1, ..., P n}, where each posture point P i Including position vectors (x, y, z) and attitude angles (θx, θy, θz), the changes between adjacent attitude points satisfy the continuity constraints and the physical motion model during the assembly process. To construct the above path form, a sampling-based trajectory planning method, such as the Rapid Random Tree Extended Connection Algorithm (RRT-Connect) or the Probabilistic Road Map Algorithm (PRM), is used to solve the attitude domain Ω corresponding to the component. j Search for one or more continuous feasible paths from the starting posture to the target posture in the P. The starting posture is set as the initial grasping posture before the assembly tool contacts the component, and the target posture is set as the position posture of the component after assembly. The above path sequence not only records a series of legal postures experienced by the component during the assembly process, but can also be converted into a path channel body in the posture space. The path channel body is composed of the swept volume between adjacent posture frames, that is, in P i To P i+1 The three-dimensional envelope formed by the continuous transformation of the component model in the posture space. The complete posture driving path form is obtained by combining the swept volumes of each segment. Thus, the feasible spatial corridor along which the components can be assembled is expressed geometrically. The output posture driven path form data W path Component identification C j For index, record its corresponding path form W j , forming a structured path channel volume dataset.

[0104] Perform spatial collision volume culling on the posture-driven path form data to obtain collision culling data;

[0105] In one embodiment, a collision detection acceleration strategy combining axis-aligned bounding box (AABB) and bounding volume hierarchy is used to perform preliminary screening on each pose point in the path form. Specifically, for each pose point P in the path sequence, i, construct the geometric envelope of the corresponding component in this posture, which includes the three-dimensional volume occupied by the posture and its directional range. Perform spatial interaction detection on the envelope with static scene elements such as assembled components and fixed obstacles to determine whether there is spatial overlap or interference. If a single posture point P i If a collision occurs midway along the path, the path is considered broken at that point and can be truncated at that location. If collisions are detected at multiple consecutive pose points, indicating that the path as a whole is infeasible, the path segment is discarded. If the collision point is close to the path endpoint and there is still room for pose adjustment, a local perturbation search algorithm is initiated to perform small-scale perturbations and resamples on the pose path in an attempt to restore path connectivity. Based on the collision detection and elimination results, the path segments in the path form are classified and labeled, including valid path form segments: continuous, collision-free areas of the path that can be directly used as assembly action channels; interrupted path segments: paths that are broken but potentially recoverable and can be used as references for local reconstruction and optimization; and high-risk path segments: areas that frequently collide or cannot be adjusted and should be marked for subsequent assembly sequence optimization or avoidance path design.

[0106] The component path-posture-direction graph is reconstructed based on the collision elimination data to obtain the assembly reachable graph data.

[0107] In one embodiment, the feasible path, posture sequence, and assembly direction information are integrated into a unified graph structure to form an assembly reachable graph, and a ternary graph G is constructed. reach =(V,E), where the node V is a triplet (C j ,P i ,D k ), representing components, postures, and assembly directions; edges E represent physical accessibility between nodes, which are then connected to form edges, representing posture transition path segments. The edge attribute structure includes path_cost: path cost (such as length, posture change amplitude, and time), interference_risk: path risk score, motion_type: action type such as insertion, rotation, and sliding, and safety_margin: safety clearance margin (unit: mm). Each valid posture path form The discrete frame points in the graph are used as nodes; the directed edges are constructed using the time series order; the assembly direction is written into the graph as the edge label attribute to achieve multi-directional support and path switching strategy. The complete assembly reachable graph data G is obtained. reach =(V,E).

[0108] Preferably, step S2 is specifically:

[0109] Step S21: extracting reachable paths based on the assembly reachable graph data to obtain reachable path data;

[0110] In one embodiment, a set of all feasible paths from the starting component posture to the target posture is extracted from the assembly reachable graph. The system uses the initial assembly posture of the component as the starting node in the graph and the posture after its expected installation is completed as the end node. Based on the above-mentioned start and end nodes, the system uses a path search algorithm to extract all path sequences that meet the feasibility conditions in the graph structure. The search algorithms that can be used include breadth-first search, depth-first search, Dijkstra shortest path algorithm, and A* heuristic path search algorithm. Different algorithms can respectively achieve shortest path search or path coverage enhancement according to the design of the path cost function or heuristic function. During the path extraction process, for each path, the system records its corresponding node sequence, that is, the set of posture states that the component experiences during the path process, which can be formally expressed as {v1, v2, ..., v n}, where each node represents the specific posture configuration of the component at a certain moment. At the same time, the system calculates the path cost for each path, which may include dimensions such as the total path length, the total amount of posture angle change, and the number of assembly direction switches. The system sets path legality constraints, and specific constraints include but are not limited to the following rules: (1) The total assembly time corresponding to the path shall not exceed the preset threshold; (2) The posture changes corresponding to any two adjacent nodes in the path shall not have discontinuous "derailment jump" phenomena, that is, the translation and rotation differences between any two frames should be within the allowable change range, ensuring that the path has interpolation and dynamic continuity during actual execution. The system outputs a set of candidate path sequences corresponding to each component as path extraction results. Each set of results represents a set of selectable paths for the component from the initial state to the assembly position, collectively referred to as reachable path data.

[0111] Step S22: Generate a path window based on the reachable path data to obtain path window data;

[0112] In one embodiment, after completing the extraction of the reachable path, the system performs a structured division of the time and posture information of the path data to achieve path interference detection and local behavior reconstruction, and generates path window data with space-time consistency. The so-called path window refers to a time series data block consisting of the continuous posture and assembly direction state sequence of the component within a specific time period, which is used to describe the local action evolution process during the component assembly process. The system converts the posture sequence corresponding to each reachable path into a path window set. The posture sequence can be expressed as the spatial posture information of the component at each key frame during the assembly process. The posture includes the position vector and the posture angle (such as Euler angle or quaternion form), which is defined in the three-dimensional rigid body motion group and is recorded as SE (3) space. The path window is recorded as Its form can be expressed as a path window consisting of a set of posture-time pairs, that is, each path window consists of several (P t,t) state, where P t represents the spatial pose of the component at time point t. During the path window generation process, the discrete pose sequence of each path is interpolated. This interpolation employs linear interpolation or curve interpolation (such as cubic spline, B-spline, or Bezier interpolation). The system uses a sliding window mechanism to divide the path sequence into multiple local segment windows. For example, with 10 frames as a window, each window contains the continuous pose information of the component at 10 adjacent time points.

[0113] Step S23: performing path interference area detection on the path window data to obtain path interference area data;

[0114] In one embodiment, the system dynamically simulates the spatial volume of the component corresponding to each posture time point in the path window. The simulation environment considered in the interference detection process includes (1) the assembled component model, that is, other components that have been assembled before the current window time period; (2) the tooling fixture boundary model, including the fixture restriction area and structural positioning elements; (3) the dynamic safety boundary model, such as the working range volume of the robot execution arm, the safety buffer shell and other virtual collision areas. For each path window, the system classifies and labels the interference area type according to the following judgment rules. If more than 30% of the posture frames in the window have geometric overlap (i.e., collision) with objects in the simulation environment, the window is marked as a high-risk interference window. In principle, this type of path segment needs to be partially replaced or avoided. If a brief collision occurs only in individual time frames within the path window (such as only one or two frames) and the spatial interference volume is small, it is marked as a recoverable interference point, indicating that the interference phenomenon can be adjusted through a perturbation correction strategy without completely replacing the path. If the motion trajectories of two or more path windows in the posture space are close to each other, resulting in the minimum distance between a window and other path segments being less than the set safety threshold (for example, a linear spacing less than 5 mm, or a posture angle less than 5 degrees), the window is marked as a path compression interference segment. Although such paths do not touch each other, there is an extremely high risk of collision.

[0115] Step S24: Perform local behavior replacement according to the path interference area data to obtain assembly path data.

[0116] In one embodiment, a limited range of sampling is performed around the spatial position and posture corresponding to the interference segment, with the perturbation amplitude set to ±5 mm in the position dimension and ±5 degrees in the posture angle dimension, to form a local legal posture space. Based on the sampling results, a set of candidate motion trajectories is constructed. For each candidate path, the posture change cost, including metrics such as translation length and total angular change, is calculated. The path with the lowest total cost is selected as the replacement path segment, ensuring that the replacement path achieves interference avoidance without sacrificing execution stability. The replacement path must meet the condition of good connectivity with the subsequent segments of the original path, that is, its ending posture must be able to be continuously connected to the starting posture of the subsequent path segment through interpolation. Once the replacement path segment is determined, the system embeds it into the original path sequence, replacing the original interfering path segment.

[0117] In one embodiment, if the local disturbance path cannot avoid interference, the system will call the standard action template in the assembly action behavior library to replace the assembly strategy corresponding to the interference path segment. The action behavior library contains several predefined assembly methods, such as direct insertion (axial insert); offset slide-in; S-curve twist-in, etc. In a specific implementation, the system selects multiple applicable standard action path generators based on the spatial characteristics of the interference segment and the component type, and regenerates the path segment. The system evaluates the path smoothness, motion cost, directional consistency and interference safety of all generated candidate action paths, selects the action type with the highest degree of compliance as the replacement path, and splices it as part of the final path of the component.

[0118] Preferably, the optimized scheduling is specifically:

[0119] Constructing a component dependency graph for the assembly path data to obtain component dependency graph data;

[0120] In one embodiment, based on the generated component assembly path data, a systematic analysis of the sequential dependencies between components during the assembly process is performed to construct a component dependency graph during the assembly process. The dependency graph expresses the assembly priority relationships between components in terms of time sequence, spatial occlusion, and structural load-bearing in the form of a directed graph. The input data includes a component assembly path dataset, wherein each component corresponds to one or more complete assembly paths, which contain information about the target assembly position, posture change direction, and time of action execution. Based on the path data, the following three types of assembly dependency rules are defined: a spatial occlusion dependency rule, wherein if the assembly path of component A needs to pass through the occupied space of component B during execution, and there is a spatial overlap or occlusion relationship between the two, component A must be installed first to avoid obstruction of B's assembly path. A connection structure dependency rule, wherein if component A is the load-bearing structure or connector of component B (for example, B is a subcomponent or fixture installed on A), A should be installed first, followed by B, to meet structural integrity and connection strength requirements. Tool path dependency rule: if the assembly operation of component A (such as screw tightening, sliding insertion) requires the help of an external tool, and the operation path of the tool will cross the installation area of component B, then A should be assembled first, and then B should be assembled to avoid tool operation restrictions. In the process of building the dependency graph, each component is a node in the graph structure to form a node set V; any two components C i with C j If any of the above dependency rules are met, and C i Must take precedence over C j Installation, then in the graph structure from node C i Point to C j Establish directed edges to form edge set E dep .

[0121] Optimize the component dependency graph data for the shortest installation time and the minimum assembly action switching range to obtain first optimized sequence data and second optimized sequence data respectively;

[0122] In one embodiment, the scheduling process is based on a set of components C1, C2, ..., C n As input, construct a dependency graph (G dep ) Scheduling sequence of topological structure Among them, each component C j The corresponding assembly time t j It consists of the following parts: the execution time of the component assembly path; the equipment switching time required for component grasping, moving and positioning; and the idle waiting time of the workstation (if there is a resource conflict). totalIt is defined as the maximum time of each component installation completion time, that is, the critical path completion time. It can be expressed as total assembly time = critical path length = the maximum of the latest completion time of all components. dep Perform topological sorting to generate all legal assembly sequence candidate sets that satisfy dependency constraints; use the critical path method to calculate the earliest start time and the latest acceptable start time for each component node in the graph; set minimization of total assembly time (minimize T_total) as the optimization goal; based on the scheduling graph structure and path weight information, use dynamic programming methods or heuristic search algorithms based on cost functions (such as A*, greedy backtracking method) to search for the optimal scheduling path. The first optimized sequence data generated is recorded as S1 opt , which represents an execution path with the shortest overall assembly time under the premise of satisfying the assembly dependency constraints of all components.

[0123] By quantifying the variation of path posture, assembly direction and tool type between adjacent component assembly actions, a switching cost matrix is constructed. i with C j The action sequence between them is used to calculate the rotation amplitude difference, movement direction change angle, and fixture / tool switching frequency in the posture space (SE(3)); the total action switching cost is constructed min is minimization, k is the action frame index, n is the total length of the posture sequence, is the action switching cost between adjacent postures, indicating that k Switch to i k+1 The required cost, the unit can be a dimensionless score or time / energy consumption index, M ij =α·||x j ―x i ||+β·∠(θ i ,θ j )+γ·D dir (i,j), M ij is the switching cost from posture i to j, α is the translation weight factor, ranging from 0.2 to 0.5, ||x j ―x i || is the spatial translation distance, that is, the position vector distance between posture i and posture j, in units of mm or m, β is the posture angle weight factor, ranging from 0.3 to 0.6, ∠(θ i ,θ j ) is the attitude angle difference, the rotation angle change between attitudes i and j (such as Euler angle difference, quaternion angle difference), the unit is degree (°) or radian (rad), γ is the action direction weight factor, the value range is 0.1~0.3, D dir(i, j) is the action direction difference data, which represents the action type (such as translation → rotation) or the structural cost of the direction jump, which can be defined by the type matching distance matrix. For example, the weighted score of mismatch is 1 and that of match is 0. i is the action posture frame index and j is the adjacent action posture frame index. Among all the legal sequences that meet the topological constraints of the dependency graph, select The smallest sequence, denoted as S2 opt .

[0124] The first optimized sequence data and the second optimized sequence data are used to mark the assembly path data with scheduling information to obtain assembly path optimized data.

[0125] In one embodiment, the scheduling sequence result is written back to the assembly path to generate optimized assembly path data with scheduling priority, time label, and action score. j The corresponding path sequence will be appended with the following scheduling metadata fields including the scheduling order j , which represents the execution order of the component in the overall assembly sequence, can be obtained from the first optimization sequence S1 opt Or the second optimized sequence S2 opt Mapping obtained; estimated start time Indicates the earliest time point at which the component can start execution in the assembly task, obtained by combining the critical path method analysis; estimated completion time Indicates the theoretical time point when the component is installed, which is calculated based on its path time and scheduling sequence; action score It represents the quality score of the assembly actions involved in the path, which comes from the pre-constructed assembly action switching score matrix (denoted as M).

[0126] Preferably, the interference resolution and reconstruction is specifically:

[0127] Perform three-dimensional interference area detection on assembly path optimization data to obtain interference area data;

[0128] In one embodiment, the optimized path data is used to detect the path segments that may cause conflicts during actual spatial operation, and the precise interference area range is output. Continuum collision detection (CCD) is used to detect whether the component penetrates the environment in the path segment; a component buffer zone (such as a 2-5mm safety gap) is introduced to warn of premature interference; the path posture sequence is projected into a spatial voxel grid, and the component space occupancy voxel map is superimposed to detect overlapping areas. The detected interference areas are classified according to their severity and impact type as follows: Type A: Continuous penetration of the path segment → Reconstruction is required; Type B: Only intermittent contact at the boundary → Perturbation repair can be attempted; Type C: Pre-collision with future components → Scheduling order adjustment is recommended.

[0129] Conflict volume modeling is performed based on interference area data to obtain conflict volume model data;

[0130] In one embodiment, the component pose envelope corresponding to the interfering path segment is extracted and Boolean intersection is performed with the interfering object to obtain the conflicting intersection volume. Alpha Shape or Concave Hull is used to encapsulate the conflicting region to generate a non-convex volume model. If multiple pose envelopes within a continuous pose segment have a high degree of overlap, they can be fused into a temporally dynamic volume. Model parameters include the collision volume center (Centroid), the maximum circumscribed bounding box (AABB), the collision impact range (affected pose range ΔP), and the hazard level (a severity indicator based on volume, coverage, and contact frequency).

[0131] Reconstruct the conflict segment action posture window according to the conflict volume model data to obtain action reconstruction data;

[0132] In one embodiment, a local path perturbation search (based on RRT or KinodynamicRRT) is performed on the conflict volume area to generate a new action sequence in the posture space that does not penetrate the obstacle. A flexible posture sliding window mechanism is introduced to allow small deformation (such as ±5mm, ±5°) action attempts. If the local search fails, an alternative action template is called (such as changing from inserting to sliding in). The reconstruction constraint is set as the change amplitude between any adjacent posture configurations shall not exceed the set continuity threshold ΔP i,i+1 <δ pose , where ΔP i,i+1 is the amplitude of posture change, i.e., the posture difference between two consecutive action frames i and i+1, δ pose The main movement direction of the newly generated path should be consistent with the original path as a whole, and no violent direction reversal, sudden stop, etc. should occur; the total installation time should not be increased by more than 15%.

[0133] The assembly path optimization data is replanned according to the action reconstruction data to obtain the assembly path reconstruction data.

[0134] In one embodiment, replacing The conflicting segment [i:j] in the pose sequence is replaced with new_pose_sequence[i:j]; the path segment time label, action type label, and scheduling dependency are also updated (cascaded adjustments are performed if there are timing delays). To prevent discontinuities, unnatural behavior, or sudden changes that violate kinematic constraints at the path replacement, the connection between the beginning and end of the newly added path segment and the previous and subsequent states of the original path is verified. Spline interpolation (such as B-spline or Bezier) is used to smoothly connect the boundary frames, or the angle between the direction vectors is calculated to determine whether the pose difference is within a preset tolerance. When the direction difference is large but does not exceed the upper limit, transition pose frames are inserted for buffering to ensure a smooth transition of the overall path. Path segment simplification and cleaning are performed, such as merging consecutive frames with pose changes less than a set threshold (such as position difference <1mm, angle difference <1°). Local abnormal directional mutation points in the path are identified and replaced with local interpolated paths. Based on the above, the assembly path reconstruction data is obtained.

[0135] Preferably, the conflict volume modeling is specifically as follows:

[0136] Sampling component motion trajectory according to interference area data to obtain component motion trajectory data;

[0137] In one embodiment, the system extracts the frame segment information of the component that interferes in the path based on the output interference area data. Each interference record contains the component identifier and the corresponding start and end frame index interval, which can be formally represented as component C j The corresponding interference path segment interval is [start idx ,end idx ], where start idx is the start frame of the interval, end idx is the end frame of the interval. The system extracts the posture sequence data of the component in the interval, that is, the set of discrete posture points of the component from the start frame to the end frame. Each posture point contains the position and orientation information of the component in three-dimensional space, which is defined in the three-dimensional rigid body motion group SE (3). The position data is described by a three-element vector (x, y, z) in the Cartesian coordinate system; the posture data can be expressed in the form of Euler angles (pitch, yaw, roll) or quaternions (q0, q1, q2, q3). The system performs a dual-domain interpolation operation on the above original discrete posture sequence, and interpolates and samples on the time axis and spatial trajectory respectively. Multiple intermediate sampling points are inserted between each two adjacent frames, and the number of samples can be set according to the accuracy requirements (for example, 5 intermediate points are inserted between every two frames). The output component motion trajectory data is a set of high-precision sampling trajectory point sequences of the component in the interference path segment, covering the continuous change process of spatial position and rotation angle.

[0138] Performing sweep envelope generation on component motion trajectory data to obtain sweep envelope data;

[0139] In one embodiment, the system samples each posture point P in the component motion trajectory. i Process frame by frame. For each frame pose P i , the system applies the corresponding rigid body transformation T to the original geometric model M of the component i , that is, transforming the component from its static geometric state in the local coordinate system to the spatial arrangement state of the current posture in the global coordinate system, and obtaining the transformed component model M i =T i (M). The above rigid body transformation T i This includes displacing the model along the three-dimensional directions x, y, and z; rotating the model's posture along three axes, which can be described by Euler angles or quaternions; and embedding the local component model into the global assembly scene. After completing the transformation of all sampling posture points, the system obtains a set of geometric instances {M1, M2, ..., M n The system performs a spatial volume fusion operation on the above model set to construct the overall swept volume corresponding to the component path segment, which is recorded as the swept envelope V swept . This volume represents the spatial range that the component may reach during all changes in posture within the path segment. The construction of the swept volume can be done in one of the following two ways: 1. By mapping the transformed model of the component to a three-dimensional voxel grid, the voxel occupancy state is accumulated using the Boolean union rule to generate a three-dimensional grid data structure representing the spatial occupancy probability or occupation boundary. 2. Perform a Boolean union operation on the mesh model generated for each frame, fuse the model contours under all postures, and form a high-fidelity swept envelope with a continuous boundary surface.

[0140] Performing Boolean difference operation of interference components according to the swept envelope data to obtain interference component data;

[0141] In one embodiment, the component C j The swept volume V swept , and other components in the scene {M k} k≠j Perform Boolean intersection operations one by one; if the intersection result Then the component k is judged to be an interference component, and the interference volume is in the form of a difference set Use a Boolean mesh operation library (such as OpenCSG, CGAL, or Volumetric MeshDifference). It's recommended to first use AABB to quickly eliminate non-overlapping candidates for improved efficiency. The output interference component data includes structured information such as the identifier of the component that spatially overlaps with the current component's sweep path, the volume model of the corresponding intersection, the spatial location of the intersection area, and the size of the intersection volume.

[0142] Segment and mark conflict volume segments according to interference component data to obtain conflict volume segment data;

[0143] In one embodiment, when the interference volume is expressed in voxel grid form, the three-dimensional voxel model can be sliced along the time axis according to the time dimension of the posture change. The system sets a fixed time step (e.g., every Δt seconds or every Δ frame) and projects the voxel set on the time axis into several segment volumes, each segment representing the collision area of the component within a certain time interval. When the interference volume is stored in the form of a polygonal mesh, the system can attach the corresponding posture frame number to the attribute information of the mesh vertex or patch. Based on these frame numbers, the system can divide the entire collision volume mesh into multiple sub-grid blocks, each sub-block corresponding to a set of adjacent posture states, forming geometric segments with temporal semantics. After the segmentation is completed, the system adds a structured annotation field to each collision volume segment unit, specifically including the posture frame range (frame_range), which records the frame number range in the original motion trajectory corresponding to the collision segment, such as frames 12 to 18, for associating path segments. The collision center point position (collision_center) refers to the geometric center of gravity coordinates of the collision fragment volume, which can be obtained by calculating the three-dimensional center of mass of the fragment or the center of the bounding box, and is used to locate the location of the collision in space. The conflict intensity score (severity_score) is used to measure the severity of the collision fragment in the overall swept volume, and is defined as the percentage of the current fragment volume to the original swept volume. This indicator reflects the spatial impact weight of the conflict in this path segment, which is convenient for priority sorting. Recommended correction strategy (recommended_resolution): The system performs parameter mapping and rule reasoning based on the structural characteristics, collision type and occurrence area of the collision fragment, combined with the preset action strategy library, to generate candidate action correction templates. Optional strategies include but are not limited to offset slide-in: applicable to boundary interference; re-orientation: applicable to angle conflicts; and schedule rollback: applicable to occlusion conflicts caused by component sequence.

[0144] A conflict volume aggregation graph is constructed according to the conflict volume fragment data to obtain conflict volume model data.

[0145] In one embodiment, the graph structure is constructed as node V = {volume fragment}, where each node represents a conflicting volume fragment. Edges E are constructed according to the following rules: For example, sequential edges are established between adjacent time segments; if two fragments are associated with the same interfering component, a shared interference volume edge is established; if two fragments have similar conflict severity and consistent gesture switching directions, a potential action switching edge is established. Each graph node is assigned the following attribute fields: the time frame interval corresponding to the fragment in the original path (frame_range), the severity score of the conflict for the fragment (severity), the geometric center coordinates of the conflicting fragment in 3D space (center_point), and the recommended action correction strategy for the fragment (recommended_resolution).

[0146] Preferably, the conflict segment action posture window reconstruction is specifically as follows:

[0147] Perform degree of freedom domain extraction based on the collision volume model data to obtain degree of freedom domain data;

[0148] In one embodiment, the motion space of the component is defined as a six-dimensional posture space SE(3), and its degrees of freedom are divided into the following two parts: linear translation freedom, which is the translation along the three spatial axes (x, y, z); angular rotation freedom, which is the rotation angle around the three coordinate axes (θ x ,θ y ,θ z ). In the area where the conflict volume exists, the system uses a dimension-by-dimension analysis method to evaluate the feasibility of movement in each degree of freedom direction. i (such as the x-direction), the system sets a certain range of disturbance steps in this direction to simulate the movement of the component in this direction; if any disturbance of the component in a certain degree of freedom direction will cause spatial interference with surrounding components or obstacles, then the dimension is considered unusable in the current environment and will be eliminated; if there is still a continuous interval in a certain degree of freedom direction, and its disturbance will not trigger spatial conflict, then the dimension is considered a legal action direction and its passable interval is recorded [a i ,b i ]. The system extracts the remaining legal action dimension set for each conflict segment component fragment, which is defined as the degree of freedom domain DoF j This set expresses all the legal transformation dimensions that the component can achieve in the conflict space and their feasible range, namely DoF j = {dimension i | Δ i ∈[a i ,b i ], no interference}.

[0149] Conduct anti-constraint analysis of conflict volume based on degree of freedom domain data to obtain blocked area data;

[0150] In one embodiment, the blockade area (i.e., the spatial range that the component cannot enter) is inferred based on the distribution of the degree of freedom domain and the conflict volume, providing guidance on the prohibited area for obstacle avoidance reconstruction. In the extracted degree of freedom domain space (including three-dimensional translation and three-dimensional rotation dimensions), a discrete posture sampling grid is constructed to exhaustively traverse all possible posture states; for each posture sampling point, the spatial occupied volume generated by the component in this posture is simulated, and the spatial intersection calculation is performed with the conflict volume model: if there is spatial overlap, that is, the component contacts or interferes with other components or obstacles in this posture state, it is determined to be an infeasible posture, and its corresponding area is added to the blockade area set; if the component occupied volume corresponding to the posture point does not overlap with any conflict area, it is determined to be a free posture. The above-mentioned spatial interference judgment result is mapped to the degree of freedom dimension to extract the sub-interval that should be prohibited in the degree of freedom domain. The system outputs a blockade area data set, which is represented as a set of multi-dimensional posture subspace blocks that need to be excluded in the degree of freedom space.

[0151] Reconstruct the action posture form according to the blocked area data to obtain the action posture form data;

[0152] In one embodiment, the action posture form refers to a sequence of legal postures that can be executed continuously during the assembly or movement of a component. i It is not in the blocked area; the posture changes are continuous and interpolated; it can be connected to the original path. The component's action posture window is reconstructed through a search algorithm in the remaining degree of freedom domain. Use heuristic search (such as RRT-Connect, BFS) to find the posture path in the legal free domain; each window segment has a maximum length (such as 20 frames) to form a slidable action segment. If the path changes drastically (the distance between adjacent posture frames exceeds the set threshold, the difference in posture angles between adjacent frames exceeds the set angle threshold, or the position change rate (or angular velocity) is unstable in multiple consecutive frames), use Bezier curves or spline interpolation for smoothing; add a flexible interpolation strategy to control posture jitter.

[0153] Extract the optimal posture domain of the action posture form data to obtain the optimal posture domain data;

[0154] In one embodiment, the evaluation index settings include smoothness_score: posture change continuity score (such as curvature change), energy_score: motion amplitude score (rotation and translation composite cost), collision_margin: minimum distance from the blocked area, execution_time: execution time or timestamp span. A scoring function is established for all candidate posture points: is the posture score data, α is the continuity coefficient of posture change, and the recommended range is [0.3, 0.5]. i is the smoothness_score, β is the action cost weight coefficient, and the recommended range is [0.2, 0.4]. i is energy_score, γ is the minimum distance weight coefficient from the blocked area, and the recommended range is [0.2, 0.4]. i For collision_margin, extract the high-scoring interval to form the optimal posture form subdomain where Ω best is the optimal posture subdomain, This is the original action posture form.

[0155] The reconstructed action set is generated according to the optimal posture domain data to obtain the action reconstruction data.

[0156] In one embodiment, the optimal posture domain Ω best The high-dimensional posture point sequence (6-dimensional SE(3) space) in the image is converted into an equivalent kinematic description unit to form a simplified motion vector, which contains content such as position, direction and angle changes. The system matches the converted motion segment with the predefined motion library (motion type template library). If the match is successful, a standard motion label is added to the motion segment, such as "rotation insertion" or "slide insertion". The matching criteria may include posture transformation type, interpolation direction consistency, posture smoothness, etc. For each identified motion type segment, its corresponding control parameters are extracted, including path direction, motion duration, speed curve and control accuracy requirements, etc., to form an executable parameter configuration unit. The motion type, parameters and posture sequence fragments are organized into a structured action set.

[0157] Preferably, step S4 is specifically:

[0158] Step S41: performing component action mapping according to the assembly path reconstruction data to obtain component action association data;

[0159] In one embodiment, the assembly path of each component (including posture sequence, action type, and timing) is mapped to an executable action unit in the digital twin. The assembly path reconstruction data is recorded as P final =C j :[P1,...,P n ], where P k (Value ranges from 1 to n) Contains action segments, reconstruction markers, timing, etc. The system divides each component path into several action frame units ActionUnit jEach action frame unit represents an abstract representation of a complete action process of a component during assembly. Each unit contains the following attributes, including component ID (component), component number C to which the action belongs j ; Action type (action_type), such as "slide_in", "rotate", "insert", etc., is used for action classification; posture frame sequence (frame_sequence), used to describe the posture evolution interval of the action segment, in the form of {P i ,...,P j}; Duration, which indicates the execution time required for the action segment (e.g., 1.8 seconds); Scheduling dependencies, which indicate that this action must be executed after certain actions of other components are completed, such as "C_k must finish". Component action associated data A map =C j :ActionUnit j , where ActionUnit j Represents component C j The corresponding action frame unit set, each action unit contains its action attributes and execution constraints.

[0160] Step S42: constructing an assembly state machine according to the component action association data to obtain assembly state machine data;

[0161] In one embodiment, the assembly process of each component is modeled as a state machine, and the state transition conditions, action trigger conditions, and termination conditions are clearly defined. States are defined as Idle (waiting for assembly), Engaging (entering action), Assembling (executing assembly), and Completed (installation completed). State transition triggering methods include time-driven, based on the start and end time periods of the action; event-driven, depending on the completion of the component action; and space-driven, with the current position / posture reaching the assembly position. The state machine graph structure includes nodes: representing states; edges: trigger conditions, with weights (such as time consumption, delay); and the output assembly state machine data is represented as. FSM = {C j :{states,transitions}}, each C j The state machine structure records its complete state sequence, trigger mechanism and timing logic as the input of the dynamic control model for assembly scheduling and visualization execution. j:Idle→(t_ready)→Engaging→(path_halfway)→Assembling→(done)→Completed, where t_ready is the current system time reaching the action start time, path_halfway is when the component has moved more than halfway or met a certain progress threshold, and done is when the assembly path segment is completed and the dependencies are met.

[0162] Step S43: generating a virtual configuration space for the assembly state machine data to obtain virtual configuration space data;

[0163] In one embodiment, the virtual configuration space refers to the three-dimensional spatial configuration set of all components in the system at any state moment or state interval; each state node corresponds to a component posture / position set (such as initial state, in progress, intermediate position, and completed position); the multi-state combination of multiple components constitutes the entire state layer of the system configuration space, supporting state-driven visual layout playback and simulation. The configuration space is generated by extracting the corresponding posture frame fragment for each state node in the assembly state machine. For example, the start and end frame intervals are extracted from the state transition path to form a spatial representation sequence of the component in this state; the posture combinations of all components in the current state are spatially superimposed, and a dynamic configuration layer is constructed hierarchically according to the topological order (or scheduling order) of the state machine; for the state evolution path of each component, a three-dimensional transformation instruction stack is established to dynamically control its spatial trajectory from the initial state to the target state; and the virtual configuration space data V is obtained. space ={State i :{C j :pose ij}}, where State i For assembly state (such as Engaging, Assembling, etc.), C j is the component number, pose ij For component C j In state S i The spatial posture below.

[0164] Step S44: performing posture binding according to the virtual configuration space data to obtain posture binding data;

[0165] In one embodiment, posture binding is to convert the spatial configuration (including position and posture) of a component in a certain assembly state into a set of posture control instructions that can be recognized and executed in the rendering system to drive three-dimensional animation or visual state switching. Each binding unit contains the following attribute fields, including target object ID; binding type (absolute / relative); binding reference system (such as tooling reference point); posture parameter sequence (x, y, z, yaw, pitch, roll), of which the first three are spatial translation vectors and the last three are rotation postures expressed by Euler angles. Convert each component state into rendering layer transform binding (WebGL / Unity, etc.); use LERP or SLERP (linear / spherical interpolation) to bind posture changes during the action. Get posture binding data B pose =C j :[state1:pose1,...,state n :pose n ], where B pose Bind data for posture, C j is the component number, state i Pose is a state of a component in the assembly state machine. i is the binding posture parameter in this state, where i takes the value of 1…n.

[0166] Step S45: synchronizing the assembly action timing of the posture binding data to obtain assembly synchronization data;

[0167] In one embodiment, the action timeline of all components in the assembly is unified to realize the visual synchronous execution flow of parallel or sequential assembly of multiple components. A Gantt-type action scheduling diagram is constructed, where the X-axis is time, the Y-axis is component ID, and each line segment represents an action segment. The action time is extracted according to the timestamp transition[t] in the state machine; key frame synchronization points are set (such as "3 components need to be inserted synchronously at the 8th second"); logical tags are introduced to specify that the execution of an action depends on the completion of the preceding action, such as wait_for(C_k.completed) (indicating that the component C j The current action needs to wait for C k Completed); Set the time trigger mark start_at to specify that a certain action must be started at a specified absolute time, such as start_at(t=12.0); Set the synchronization group sync_group to define the component group set that needs to be executed collaboratively, sync_group={C_j,C_m} (indicates that two components need to complete concurrent assembly within a certain synchronization window). Get assembly synchronization data Where: t k Indicates a synchronization time point (or the starting point of a time period); C j A component that performs an action at that point in time; Represents component C j The corresponding action unit labels include slide_in, align, insert, etc., and the same applies to other parameters.

[0168] Step S46: Integrate interactive events based on assembly synchronization data to obtain assembly path digital twin data.

[0169] In one embodiment, the interactive events include user-initiated triggering behaviors, system automatic response events, and simulation operations for intervention simulation, which are mainly divided into the following three categories: 1. User-triggered events: click on the component to display the current assembly status; replay the complete assembly path of the component (such as from the starting position to the completion position); zoom in to view a local area or state feature. 2. System-triggered events: when the system detects a collision or path penetration during the posture change, it automatically triggers an alarm; when the assembly action is not executed in the scheduled order (such as early execution due to unmet dependencies), a sequence exception prompt is triggered; and real-time feedback on the current component execution delay. 3. Simulation intervention events (used for assembly plan verification and decision optimization): allow simulation of operations such as early insertion and delayed execution to observe the system's behavioral response under constraint conflicts or resource competition; record intervention results and provide path recovery options. The system defines listenable interactive event interfaces for each component action segment or posture node, including but not limited to onEnter: triggered when a component enters a certain state or posture segment; onExit: triggered when a component leaves a certain state segment; onClick: triggered when a user clicks on the component model; onCollision: triggered when an interference event is detected; and onViolation: triggered when a sequence, time, or logical constraint is violated. Each interactive event defines a standardized event script structure, including the event type, target component, response action, and parameter data. Event logic can be mapped one-to-one with state machine nodes, assembly sequence diagram nodes, or bound posture nodes, forming a bidirectional behavior-state linkage model, i.e., the assembly path digital twin data.

[0170] Preferably, the present application further provides a three-dimensional assembly path planning system for a solar wing structure, for executing the three-dimensional assembly path planning method for a solar wing structure as described above, the three-dimensional assembly path planning system for a solar wing structure comprising:

[0171] The 3D assembly perception and mapping module is used to obtain the 3D assembly data of the solar wing and construct an assembly reachable map based on the 3D assembly data of the solar wing to obtain the assembly reachable map data;

[0172] The path avoidance planning module is used to perform path avoidance behavior planning based on the assembly reachable graph data to obtain assembly path data;

[0173] The path optimization and interference reconstruction module is used to optimize the assembly path data to obtain the assembly path optimization data; perform interference elimination and reconstruction on the assembly path optimization data to obtain the assembly path reconstruction data;

[0174] The digital twin generation module is used to generate digital twin scenes based on the assembly path reconstruction data to obtain assembly path digital twin data.

[0175] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, the scope of the present invention is limited by the attached application documents rather than the above description, and it is intended that all changes that fall within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0176] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A three-dimensional assembly path planning method for a solar wing structure, characterized in that: The following steps are involved: Step S1: Acquire the three-dimensional assembly data of the solar wing, and construct an assembly reachable graph based on the three-dimensional assembly data of the solar wing to obtain assembly reachable graph data; Step S2: Perform path avoidance behavior planning based on the assembly reachable graph data to obtain assembly path data; Step S3: Optimizing and scheduling the assembly path data to obtain assembly path optimization data; Perform interference elimination and reconstruction on the assembly path optimization data to obtain assembly path reconstruction data; Step S4: Generate a digital twin scene based on the assembly path reconstruction data to obtain the assembly path digital twin data.

2. The method according to claim 1, characterized in that Step S1 is specifically as follows: Obtain solar wing 3D assembly data; Perform component analysis on the solar wing 3D assembly data to obtain component analysis data; Generate a spatial correlation diagram for component analysis data based on the solar wing three-dimensional assembly data to obtain spatial correlation diagram data; According to the three-dimensional assembly data of the solar wing, the spatial association graph data is mapped to the assembly direction to obtain the assembly direction constraint graph data; Performing attitude space mapping on the assembly direction constraint graph data according to the solar wing three-dimensional assembly data to obtain attitude constraint graph data; The traversable space is solved according to the posture constraint graph data to obtain the assembly reachable graph data.

3. The method according to claim 2, characterized in that The specific solution of the traversable space is: Generate component posture domain-decomposition body according to posture constraint graph data to obtain component posture domain-decomposition body data; Solve the posture drive path form according to the component posture solution domain data to obtain the posture drive path form data; Perform spatial collision volume culling on the posture-driven path form data to obtain collision culling data; The component path-posture-direction graph is reconstructed based on the collision elimination data to obtain the assembly reachable graph data.

4. The method according to claim 1, wherein Step S2 is specifically as follows: Extract the reachable path based on the assembly reachable graph data to obtain the reachable path data; Generate a path window according to the reachable path data to obtain path window data; Performing path interference area detection on the path window data to obtain path interference area data; Local behavior replacement is performed based on the path interference area data to obtain assembly path data.

5. The method according to claim 1, wherein The optimized scheduling is as follows: Constructing a component dependency graph for the assembly path data to obtain component dependency graph data; Optimize the component dependency graph data for the shortest installation time and the minimum assembly action switching range to obtain first optimized sequence data and second optimized sequence data respectively; The first optimized sequence data and the second optimized sequence data are used to mark the assembly path data with scheduling information to obtain assembly path optimized data.

6. The method according to claim 1, characterized in that The interference elimination and reconstruction are specifically as follows: Perform three-dimensional interference area detection on assembly path optimization data to obtain interference area data; Conflict volume modeling is performed based on interference area data to obtain conflict volume model data; Reconstruct the conflict segment action posture window according to the conflict volume model data to obtain action reconstruction data; The assembly path optimization data is replanned according to the action reconstruction data to obtain the assembly path reconstruction data.

7. The method according to claim 6, characterized in that The conflict volume modeling is specifically as follows: Sampling component motion trajectory according to interference area data to obtain component motion trajectory data; Performing sweep envelope generation on component motion trajectory data to obtain sweep envelope data; Performing Boolean difference operation of interference components according to the swept envelope data to obtain interference component data; Segment and mark conflict volume segments based on interference component data to obtain conflict volume segment data; A conflict volume aggregation graph is constructed according to the conflict volume fragment data to obtain conflict volume model data.

8. The method according to claim 6, characterized in that The conflict segment action posture window reconstruction is specifically as follows: Perform degree of freedom domain extraction based on the collision volume model data to obtain degree of freedom domain data; Conduct anti-constraint analysis of conflict volume based on degree of freedom domain data to obtain blocked area data; Reconstruct the action posture form according to the blocked area data to obtain the action posture form data; Extract the optimal posture domain of the action posture form data to obtain the optimal posture domain data; The reconstructed action set is generated according to the optimal posture domain data to obtain the action reconstruction data.

9. The method according to claim 1, characterized in that Step S4 is specifically as follows: Component action mapping is performed based on assembly path reconstruction data to obtain component action association data; Construct an assembly state machine based on component action association data to obtain assembly state machine data; Generating a virtual configuration space for the assembly state machine data to obtain virtual configuration space data; Perform posture binding according to the virtual configuration space data to obtain posture binding data; Perform assembly action timing synchronization on the posture binding data to obtain assembly synchronization data; Interactive events are integrated based on assembly synchronization data to obtain assembly path digital twin data.

10. A three-dimensional assembly path planning system for a solar wing structure, characterized in that: For executing the three-dimensional assembly path planning method of the solar wing structure according to claim 1, the three-dimensional assembly path planning system of the solar wing structure comprises: The 3D assembly perception and mapping module is used to obtain the 3D assembly data of the solar wing and construct an assembly reachable map based on the 3D assembly data of the solar wing to obtain the assembly reachable map data; The path avoidance planning module is used to perform path avoidance behavior planning based on the assembly reachable graph data to obtain assembly path data; Path optimization and interference reconstruction module, used to optimize and schedule assembly path data to obtain assembly path optimization data; Perform interference elimination and reconstruction on the assembly path optimization data to obtain assembly path reconstruction data; The digital twin generation module is used to generate digital twin scenes based on the assembly path reconstruction data to obtain assembly path digital twin data.

Images