A three-dimensional assembly path planning method and system for solar panel structures

Abstract

This invention relates to the field of geometric modeling technology for three-dimensional models, and particularly to a method and system for three-dimensional assembly path planning of a solar array structure. The method includes the following steps: acquiring three-dimensional assembly data of the solar array, and constructing an assembly reachability map based on the three-dimensional assembly data to obtain assembly reachability map data; performing path avoidance behavior planning based on the assembly reachability map data to obtain assembly path data; optimizing and scheduling the assembly path data to obtain optimized assembly path data; reconstructing the optimized assembly path data through interference resolution to obtain reconstructed assembly path data; and generating a digital twin scene based on the reconstructed assembly path data to obtain digital twin data of the assembly path. This invention achieves full-process digital mapping from action behavior modeling and assembly timing control to interactive linkage, significantly improving the verifiability, interactivity, and pre-deployment evaluation capabilities of the assembly process.

Description

Technical Field

[0001] This invention relates to the field of geometric modeling technology for three-dimensional models, and in particular to a three-dimensional assembly path planning method and system for a solar panel structure. Background Technology

[0002] Digital twins, as one of the core technologies of next-generation intelligent manufacturing, refer to the construction of a dynamic, interactive model corresponding to a physical entity in virtual space through multi-source sensing, 3D modeling, state mapping, and time synchronization, enabling the prediction, evaluation, and optimization of the physical system. They are widely used in the design verification, manufacturing scheduling, and operation monitoring of highly complex products, demonstrating significant engineering value, especially in aerospace and equipment manufacturing. With the increasing complexity of aerospace equipment structures, particularly in the design and manufacturing of space-deployable components such as solar arrays, the assembly process involves the collaborative operation of large-size, multi-degree-of-freedom, and high-precision components. Traditional assembly processes rely heavily on manual planning and experience for path design and action organization, making it difficult to accurately identify accessible assembly areas and avoid potential interference risks within complex component spaces. This results in low assembly efficiency, high error rates, and hinders the deployment of automated assembly and intelligent manufacturing. Therefore, combining digital twins with intelligent manufacturing of solar arrays has become a crucial issue. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes a three-dimensional assembly path planning method and system for solar array structures, thereby resolving at least one of the aforementioned technical issues.

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

[0005] Step S1: Obtain the three-dimensional assembly data of the solar array, and construct the assembly reachability map based on the three-dimensional assembly data of the solar array to obtain the assembly reachability map data;

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

[0007] Step S3: Optimize and schedule the assembly path data to obtain optimized assembly path data; reconstruct the assembly path data by resolving interference to obtain reconstructed assembly path data.

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

[0009] This invention introduces an assembly reachability graph construction and path-avoidance behavior planning mechanism, enabling the system to fully identify assembly accessibility and interference risk areas in the solar panel structure. This achieves spatial constraint coordination between path planning and behavioral actions, effectively avoiding assembly failures caused by rigid path strategies or lagging collision detection in traditional methods. Furthermore, by combining path optimization and interference resolution reconstruction, action posture windows are reconstructed and path segments are re-avoided for interference-prone path segments, significantly improving the continuity and spatial adaptability of the assembly process. Through a 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, enabling visualized pre-performance and multi-dimensional verification of path feasibility, spatial constraint adaptability, and behavioral stability, forming an integrated "planning-reconstruction-verification" technology closed loop.

[0010] Preferably, step S1 specifically includes:

[0011] Obtain 3D assembly data of the solar array;

[0012] Component analysis was performed on the three-dimensional assembly data of the solar array to obtain component analysis data;

[0013] Based on the three-dimensional assembly data of the solar array, a spatial correlation diagram is generated from the component analytical data to obtain spatial correlation diagram data;

[0014] Based on the three-dimensional assembly data of the solar array, the spatial correlation diagram data is mapped to the assembly direction to obtain the assembly direction constraint diagram data.

[0015] The attitude constraint diagram data is obtained by mapping the assembly orientation constraint diagram data to the three-dimensional assembly data of the solar array.

[0016] Based on the attitude constraint diagram data, the traversable space is solved to obtain the assembly reachability map data.

[0017] This invention utilizes component analysis and spatial association graph construction to accurately capture the geometric connections and spatial adjacencies between components in the solar array structure, providing a structured representation for path analysis. The introduction of an assembly direction mapping mechanism effectively identifies 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. An attitude constraint graph is constructed, clarifying the range of attitude degrees of freedom for assembly actions through attitude space mapping, thereby enabling angle restrictions and attitude feasibility screening for action behaviors. Solving the assembly reachability graph using attitude constraint graph data not only achieves quantitative modeling of spatial traversability but also provides a high-dimensional feasible space input with directional and attitude constraints for path planning algorithms.

[0018] Preferably, the solution for the passable space specifically involves:

[0019] The component attitude solution volume is generated based on the attitude constraint diagram data to obtain the component attitude solution volume data;

[0020] The attitude-driven path window is solved based on the component attitude solution volume data to obtain the attitude-driven path window data;

[0021] Spatial collision volume culling is performed on the attitude-driven path window data to obtain collision culling data;

[0022] Based on the collision rejection data, the component path-attitude-direction map is reconstructed to obtain the assembly reachability map data.

[0023] This invention, by generating component attitude solution volumes, expresses the achievable attitude range of each component in three-dimensional space in volumetric form, overcoming the limitation of traditional methods that only handle positional constraints and cannot describe the feasibility of dynamic attitudes. Based on the attitude solution volume data, the system solves for the attitude-driven path window, enabling the construction of a dynamic path window. While maintaining the rationality of spatial actions, it finely controls the attitude evolution process of the assembly path, improving the responsiveness of the path strategy to action continuity. Furthermore, a spatial collision volume culling mechanism is introduced to dynamically filter potential interference areas and correct the path window, effectively reducing the probability of conflicts during path planning. Through component path-attitude-direction graph reconstruction, the system integrates these three types of constraint information into a unified graph structure, generating an assembly reachability graph with direction selectivity, attitude continuity, and path accessibility.

[0024] Preferably, step S2 specifically includes:

[0025] Accessibility paths are extracted from the assembly accessibility map data to obtain accessibility path data;

[0026] Path windows are generated based on reachability data to obtain path window data.

[0027] Path interference region detection is performed on the path window data to obtain path interference region data;

[0028] The assembly path data is obtained by performing local behavior replacement based on the path interference region data.

[0029] In this invention, during the reachability path extraction stage, the system can filter out effective paths that meet attitude, orientation, and spatial accessibility constraints from the assembly reachability map, avoiding redundant or infeasible paths caused by low path abstraction levels in traditional methods. Through path window generation, local spatial segments within the path can be modeled in a window-like manner, enabling continuous description of path dynamic features (such as attitude transitions and component rotations). Path interference region detection further accurately identifies potential conflict segments within these windows, determining interference segments based on criteria such as volume overlap, directional contradictions, and attitude conflicts, thereby effectively avoiding spatial conflict risks. Through a local behavior replacement mechanism, path segments with interference are dynamically replaced at the action level, rather than the entire path being discarded, significantly improving path correction efficiency and path continuity.

[0030] Preferably, the optimized scheduling specifically includes:

[0031] Component dependency graphs are constructed from the assembly path data to obtain component dependency graph data;

[0032] The component dependency graph data is optimized for minimizing installation time and minimizing assembly action switching, resulting in the first and second optimized sequence data, respectively.

[0033] The assembly path data is marked with scheduling information using the first optimized sequence data and the second optimized sequence data to obtain the assembly path optimized data.

[0034] In this invention, the construction of the component dependency graph is based on the structural connection logic, spatial nesting order, and assembly sequence constraints between components. This systematically depicts the dependency chains in assembly tasks, avoiding path backtracking or action conflicts caused by neglecting inter-component constraints in traditional assembly planning. For the shortest installation time optimization strategy, the system prioritizes generating the component scheduling sequence with the shortest overall installation time, effectively compressing the total assembly time, while ensuring that component dependency constraints are not violated. The minimum action switching amplitude optimization strategy considers the physical action amplitudes such as attitude switching and path switching between adjacent assembly tasks, improving the continuity and stability of assembly actions by minimizing the range of action changes. The two optimized sequences are merged and marked into the original path data, forming a scheduling optimization result with "optimal time efficiency + minimum action cost".

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

[0036] Three-dimensional interference region detection is performed on the assembly path optimization data to obtain interference region data;

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

[0038] Based on the conflict volume model data, the action posture window of the conflict segment is reconstructed to obtain the action reconstruction data;

[0039] Based on the motion reconstruction data, the assembly path optimization data is replanned to obtain the assembly path reconstruction data.

[0040] This invention utilizes assembly path optimization data for 3D interference region detection. The system can identify locations where path segments overlap in volume or exhibit posture conflicts with surrounding structures within complex component layouts, resolving the issue of missed conflicts in traditional methods due to neglect of dynamic posture changes. By combining conflict volume modeling technology, a spatial conflict representation with directional, shape, and variable boundary characteristics is constructed, enabling quantitative modeling and spatial topological constraint expression of conflict regions. Based on this conflict model, action posture window reconstruction not only allows for local correction of the action postures corresponding to interference segments but also generates alternative action sequences adaptable to spatially constrained areas, ensuring the accessibility and continuity of the assembly process. A path segment replanning mechanism integrates the reconstructed actions into the global path, achieving dynamic reconstruction while preserving the continuity of the original path.

[0041] Preferably, the conflict volume modeling specifically involves;

[0042] The component motion trajectory is sampled based on the interference region data to obtain the component motion trajectory data;

[0043] Sweep envelope data is generated by performing a sweep envelope operation on the component motion trajectory data.

[0044] Boolean difference operations are performed on the interferometric components based on the swept envelope data to obtain the interferometric component data;

[0045] The conflict volume fragments are segmented and labeled based on the interference component data to obtain the conflict volume fragment data;

[0046] The conflict volume aggregation graph is constructed based on the conflict volume fragment data to obtain the conflict volume model data.

[0047] This invention obtains the spatial movement path of components during dynamic assembly by sampling the motion trajectory of components in the interference region at multiple time points, providing temporal support for spatial modeling. Subsequently, a swept envelope generation method is used to transform the motion trajectory of the components during assembly into a spatial envelope, effectively capturing the complete spatial range occupied by the components during their movement and overcoming the lack of expressive power for dynamic interference in traditional static collision detection methods. Based on this envelope data, the system further performs Boolean difference operations on the interfering components to accurately extract the conflict areas where actual volume overlap occurs, improving the accuracy and resolution of interference recognition. Next, by segmenting and semantically labeling the interfering component data, a set of conflict volume segments with structural significance is constructed, enhancing the spatial localization and semantic association capabilities of the conflict sites. Finally, a conflict volume aggregation graph is constructed, integrating the scattered volume segments into a graph structure representation, forming a conflict volume model with spatial topological relationships, conflict severity, and correctable attributes.

[0048] Preferably, the reconstruction of the action pose window during the conflict segment specifically involves:

[0049] The domain of freedom is extracted from the conflict volume model data to obtain the domain of freedom data;

[0050] Based on the domain of freedom data, conflict volume inverse constraint analysis is performed to obtain the data of the blocked area;

[0051] Based on the data of the blocked area, the action posture window is reconstructed to obtain the action posture window data;

[0052] The optimal pose domain is extracted from the action pose window data to obtain the optimal pose domain data;

[0053] The action set is generated based on the optimal pose domain data to obtain action reconstruction data.

[0054] This invention extracts the domain of freedom from conflict volume model data, enabling the system to identify the translational and rotational dimensions that components can still perform under current spatial constraints. This clarifies the feasible spatial boundaries for action adjustments, overcoming the limitations of traditional path reconstruction methods that suffer from coarse-grained and overly complex handling of degrees of freedom. Based on the domain of freedom, the system performs inverse constraint analysis, deducing the spatially confined areas of assembly behavior from the conflict volume, providing precise constraints for the feasibility screening of the attitude window. The action attitude window reconstruction process expresses the feasible attitude range through a window, enhancing the structured representation of feasible attitude solutions and facilitating the generation and evaluation of multi-angle and multi-attitude alternative strategies. Based on this, the optimal attitude domain is extracted, and the most valuable action attitude solution set can be selected through optimization functions (such as interference minimization and attitude continuity maximization). The final reconstructed action set can replace the original action sequence embedded in the path segment, achieving precise obstacle avoidance, continuous turning, and attitude connection, ensuring smooth assembly. Unlike traditional methods that use a static discrete attitude library for selection, this method achieves an integrated reconstruction process from spatial constraint modeling and degree-of-freedom compression to attitude domain optimization, significantly improving the system's adaptability, spatial coordination, and dynamic adjustment accuracy in complex interference environments. This method also differs from traditional path interference modeling methods that rely solely on nearest-distance detection or simple bounding box analysis, significantly enhancing the accuracy of dynamic spatial conflict modeling and the quality of path repair strategy generation, providing highly reliable and accurate decision-making support for intelligent assembly path reconstruction.

[0055] Preferably, step S4 specifically includes:

[0056] Component actions are mapped based on the assembly path reconstruction data to obtain component action association data;

[0057] The assembly state machine is constructed based on the component action association data to obtain the assembly state machine data.

[0058] Virtual configuration space data is generated from the assembly state machine data.

[0059] Attitude binding is performed based on virtual configuration space data to obtain attitude binding data;

[0060] Synchronize the assembly action timing with the attitude binding data to obtain assembly synchronization data;

[0061] By integrating interactive events based on assembly synchronization data, digital twin data of the assembly path is obtained.

[0062] This invention, through component motion mapping and assembly state machine construction, enables the system to transform the reconfiguration path actions of each assembly component into controllable state transition units, achieving logical driving and state monitoring of assembly behavior, breaking through the static presentation method of traditional digital twins that only simulates geometric animation. The generation of the virtual configuration space allows the system to dynamically deduce and display the spatial combination states of multiple components and multiple stages, supporting high-fidelity restoration of the assembly sequence and spatial layout of multiple components. The attitude binding process precisely binds the attitude parameters in the motion path to the node entities in the virtual configuration, giving the assembly actions a realistic physical expression and consistent action in space. Through action timing synchronization, the system can achieve full-process temporal coordination of the assembly process, ensuring the consistency and controllability of component assembly sequence, relative position, and timing logic. The introduction of an 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-triggered feedback, and assembly anomaly simulation.

[0063] Preferably, this application also provides a three-dimensional assembly path planning system for a solar wing structure, used to execute the three-dimensional assembly path planning method for a solar wing structure as described above. The three-dimensional assembly path planning system for the solar wing structure includes:

[0064] The 3D assembly perception and mapping module is used to acquire the 3D assembly data of the solar array and construct the assembly reachability map based on the 3D assembly data of the solar array to obtain the assembly reachability map data.

[0065] The path avoidance planning module is used to plan path avoidance behaviors based on assembly reachability map data to obtain assembly path data.

[0066] The path optimization and interference reconstruction module is used to optimize and schedule assembly path data to obtain optimized assembly path data; and to perform interference resolution and reconstruction on the optimized assembly path data to obtain reconstructed assembly path data.

[0067] The digital twin generation module is used to generate digital twin scenarios based on the assembly path reconstruction data, thereby obtaining assembly path digital twin data.

[0068] The beneficial effects of this invention are as follows: By mapping component actions and constructing an assembly state machine, the system can transform the reconstructed path actions of each assembly component into controllable state transition units, realizing logical driving and state monitoring of assembly behavior, breaking through the static presentation method of traditional digital twins that only simulates geometric animation. The generation of virtual configuration space enables the system to dynamically deduce and display the spatial combination state of multiple components and multiple stages, supporting high-fidelity restoration of the assembly sequence and spatial layout of multiple components. The attitude binding process accurately binds the attitude parameters in the action path to the node entities in the virtual configuration, enabling the assembly actions to have a real physical expression and action consistency in space. Through action timing synchronization, the system can achieve full-process temporal coordination of the assembly process, ensuring the consistency and controllability of component assembly sequence, relative position, and timing logic. The introduction of an 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 trigger feedback, and assembly anomaly simulation. Unlike traditional solutions that only provide path animation playback or CAD-level display, this method realizes full-process digital mapping from action behavior modeling and assembly sequence 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. Attached Figure Description

[0069] Other features, objects, and advantages of this application will become more apparent from the following detailed description of the non-limiting embodiments, taken with reference to the accompanying drawings:

[0070] Figure 1 A flowchart illustrating the steps of a three-dimensional assembly path planning method for a solar panel structure according to an embodiment is shown.

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

[0072] Figure 3 A flowchart illustrating the steps of a path avoidance planning method according to one embodiment is shown.

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

[0074] The technical method of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0075] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

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

[0077] Please see Figures 1 to 4 This application provides a three-dimensional assembly path planning method for a solar panel structure, including the following steps:

[0078] Step S1: Obtain the three-dimensional assembly data of the solar array, and construct the assembly reachability map based on the three-dimensional assembly data of the solar array to obtain the assembly reachability map data;

[0079] In one embodiment, the acquisition of the three-dimensional assembly data includes, but is not limited to, standard three-dimensional file formats (such as STEP or IGES formats) generated through CAD modeling and design, or three-dimensional point cloud data acquired through laser scanning, structured light scanning, etc. This three-dimensional assembly data contains information such as the spatial arrangement of solar panel components, the connection structure between components, hole distribution, component installation direction, occlusion area distribution, and component geometric contour data. For the aforementioned three-dimensional assembly data, the system employs a geometric segmentation algorithm to perform component analysis and structural decomposition on the overall solar panel model. The 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 can be divided into several sub-components with assembly physical meaning. Each component is assigned a unique identifier during the analysis process, and its corresponding assembly direction (such as insertion direction), degrees of freedom information (including movable dimensions and their constraint range), and connection relationships (such as connection logic implemented through bolts, grooves, 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 associated attribute information for each node includes the geometric center coordinates, bounding box, and reference posture information required for its installation. Edges between nodes describe the dependencies between assembly sequences or assembly paths. Specifically, if the assembly operation of component A must be completed before the assembly of component B, or if there is geometric interference from component B occluding component A on its assembly path, a directed edge is established from A to B in the graph to represent the dependency logic that A is assembled before 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 posture states, it determines whether the assembly action of the component will cause spatial conflicts with already installed components during path execution. In three-dimensional space, the path space is discretized using a voxel grid, and combined with the sequence of action posture changes, a feasible path space region traversed by the component from its initial position to the target installation position is constructed. If there exists a path voxel sequence that does not intersect with any obstacle at any intermediate attitude point, then the assembly path of that component is considered reachable in the current configuration. Through this method, the system ultimately generates assembly reachability map data.

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

[0081] In one embodiment, the system, based on the constructed assembly reachability graph data and combining the initial state of the components with the target assembly position, executes a path-avoidance behavior planning process 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 of path planning and the position corresponding to its target installation posture as the ending point of the path. Within the directed graph structure described by the assembly reachability graph, a heuristic path search algorithm is used to generate paths. The path search algorithm may include the A* algorithm, RRT (Rapid Exploratory Random Tree), or improved variants thereof. The A* algorithm can dynamically balance search efficiency and path shortestness between path cost and heuristic function, while the RRT algorithm is suitable for quickly generating feasible paths in high-dimensional space and is well-suited for handling high-degree-of-freedom posture changes during component assembly. To further approximate actual assembly scenarios, the system introduces a dynamic obstacle prediction model between components to simulate the movement or deformation of other components during assembly, thereby improving the robustness of the path planning results. In terms of obstacle avoidance modeling, the system constructs a voxel map of the assembly path in 3D space, discretizing the assembly space into several voxel units and recording the accessibility status of each voxel unit. Based on this, the geometric volume of the components involved in the assembly is spatially expanded, i.e., a certain volume is added outside their geometric boundaries to simulate the operating space of assembly tools (such as robotic arm end effectors or gripping devices), ensuring that the path planning process considers the spatial impact of tool dimensions. To avoid accidental interference when the assembly path approaches other assembled components, the system introduces a safety buffer boundary rule in path planning. 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 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 amplitude of the components to prevent drastic attitude changes in the robotic arm during execution and reduce execution risks. For assembly tasks requiring high precision (such as precision-fitted components), the system also introduces a path smoothing mechanism to handle the continuity of turning points in the path. Specifically, Bézier curve interpolation can be used to smooth the path curve, reducing angle jumps or abrupt changes in assembly movements and improving the stability and accuracy of path execution. Through the above methods, the system generates assembly path data based on the assembly reachability map, which satisfies spatial accessibility, obstacle avoidance safety, assembly equipment motion constraints, and smoothness of movements.

[0082] Step S3: Optimize and schedule the assembly path data to obtain optimized assembly path data; reconstruct the assembly path data by resolving interference to obtain reconstructed assembly path data.

[0083] In one embodiment, the system optimizes the scheduling of the initially generated assembly path data. This process involves introducing a component priority scheduling graph to sort components and optimize assembly timing. The component priority scheduling graph is a directed graph structure constructed based on the structural hierarchy and assembly dependency logic of the components, also known as an assembly priority level dependency graph. Each node represents a component to be assembled, and the directed edges between nodes represent the dependency relationship of the assembly order. In priority scheduling, the system prioritizes the installation order of main load-bearing components (such as the skeleton and support base) according to the structural load-bearing relationship, and then sorts non-critical components accordingly. To improve scheduling efficiency and resource utilization, the system adopts a greedy scheduling algorithm combined with a time window constraint strategy to perform time-batch scheduling of the overall assembly process. Based on the component dependency graph, this scheduling method uses the "earliest executable" principle as the core scheduling logic, arranging as many components as possible into the same time period assembly batch without violating dependency constraints, thereby optimizing the total assembly time and path utilization density. Specifically, the system's scheduling objective function is: Where T i L represents the assembly time of the i-th component. i Let λ be the path length of the i-th component, λ be the path length weight coefficient, and λ be a dimensionless constant (set to 1-5 for general industrial robotic arms (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 be the component index variable, and n be the total number of components participating in the scheduling. The objective function aims to balance the overall simplicity and execution efficiency of the assembly path while ensuring reasonable time allocation. After optimizing the scheduling, the system further performs spatial interference detection and resolution reconstruction on the optimized assembly path data. This process first achieves rapid interference detection by constructing a three-dimensional collision volume model. Specifically, the system performs spatial overlap detection on the bounding volume corresponding to the component posture in each frame of the assembly path, thereby identifying path segments with potential spatial interference risks. For the detected interference path segments, the system applies the action posture window reconstruction method to perform local path replacement. This method generates a set of perturbation-based attitude path windows near the interference path segment. This involves introducing small perturbations (e.g., attitude angle offset ±2°, position offset ±5mm) into the original path to form several candidate paths. The system then evaluates each candidate path segment using a spatial collision score, assessing its minimum distance to surrounding obstacle components and the probability of collision during execution. The attitude combination with the lowest collision score is selected as the replacement path. If no effective collision-free path can be generated in any of the perturbation path windows, the system automatically triggers a local rollback mechanism for the assembly sequence. This rearranges the scheduling order of the components corresponding to the current path segment and their dependent components, and recalculates the assembly timing and path allocation until an alternative path that satisfies obstacle avoidance constraints is generated.

[0084] Step S4: Generate a digital twin scene based on the assembly path reconstruction data to obtain 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 3D visualization, timing simulation, and interactive feedback functions of the assembly process. The assembly path reconstruction data includes component path segment information, a complete assembly action sequence, and a reconstructed action sequence adjusted due to interference conflicts. The system maps the above data structure to instruction nodes in the digital twin environment to construct a timing graph 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.). All instruction nodes are organized into an assembly action timing graph according to assembly dependencies. This timing graph can express the behavioral logic of parallel actions of multiple components, sequential dependencies, and path mutual exclusion, and is the control structure driving the execution of the digital twin animation. In terms of visualization implementation, the system builds an assembly scene simulation platform based on a 3D rendering engine (such as Unity3D or WebGL). The platform loads a complete solar wing structure component model and maps the above assembly action timing graph to the dynamic transformation control flow of the corresponding component. The spatial position and orientation of each component change synchronously with the timeline, forming a complete 3D assembly demonstration process. The system supports various visual control functions, including but not limited to timing control operations such as playback, pause, rewind, fast forward, and jump of assembly actions; as well as scene display options such as highlighting the assembly path, adjusting transparency, and adjusting execution speed. The platform has a clickable feedback mechanism based on component granularity, allowing users to click on any component during operation to query its path segment, executed action status, whether there is interference history, actual execution time, and other data. 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. This module uses statistical deviation detection methods to analyze the spatial path of each action sequence. If the deviation of any execution path exceeds three times the standard deviation of its expected path (i.e., deviation > 3 times the standard deviation), the system automatically marks the path segment as an anomaly and highlights it on the digital twin interface. Simultaneously, path optimization suggestions appear, including suggested replanning of posture or path buffering adjustments, to assist users in optimizing assembly accuracy and reconfiguring paths. Through this approach, the system achieves an integrated digital twin modeling process from path reconstruction data to 3D dynamic scenes, timing control, and intelligent feedback, generating assembly path digital twin data.

[0086] Preferably, step S1 specifically includes:

[0087] Step S11: Obtain the three-dimensional assembly data of the solar array;

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

[0089] Step S12: Perform component analysis on the three-dimensional assembly data of the solar array to obtain component analysis data;

[0090] In one embodiment, the system divides components based on the hierarchical structure of the assembly tree or the topological features between components in the 3D assembly data. When the assembly model has an assembly tree structure, components and their hierarchical relationships can be identified level by level. When a clear assembly tree structure is lacking, spatial segmentation of components can be achieved through mesh connectivity analysis (e.g., identifying non-connected mesh regions), thereby identifying independent component entities. Extracted fields include component ID, geometric center (centroid), bounding box (AABB / OBB), principal axis direction, and connection interface location (e.g., hole positions, snap-fit). If the assembly model contains naming rules or additional metadata, such as a component named "WingLeft_Panel1", the system can automatically determine the semantic category of the component by combining the component's naming information with its geometric features. Component types include, but are not limited to, solar panels, battery modules, connecting brackets, drive mechanisms, etc. If naming information is missing, component clustering and classification are performed through geometric features (e.g., aspect ratio, surface features, hole distribution, etc.).

[0091] Step S13: Generate a spatial correlation diagram from the component analysis data based on the three-dimensional assembly data of the solar array to obtain spatial correlation diagram data;

[0092] In one embodiment, each component is a node in the graph. An edge is added if two components meet any of the following conditions: the distance between their connecting surfaces is less than a threshold (e.g., <1.0 mm); there are overlapping holes, snap-fit ​​devices, or connectors (detected via Boolean intersection); and the edge attributes include connection type, contact area, face normal angle, and connection priority. KD-Tree is used for rapid component proximity filtering. Edge sets are constructed by judging the normal angle and inter-face distance of the filtered component pairs, resulting in spatial association graph data.

[0093] Step S14: Map the assembly direction of the spatial correlation diagram data based on the three-dimensional assembly data of the solar array to obtain the assembly direction constraint diagram data;

[0094] In one embodiment, potential assembly directions (for insertion) are extracted based on the normal vectors of the component connection surfaces. A set of master assembly directions, such as ±X, ±Y, and ±Z, is introduced to standardize all assembly directions. Each candidate direction vector is projected into the master direction set and assigned to the nearest master direction label. It is determined whether there are obstructions on the insertion path and whether it conforms to the component connection structure design (e.g., pin-hole structures only allow unidirectional insertion). For each pair of connection edges, its assembly direction vector is labeled. A graph is constructed, with edge attributes including direction vector, insertable interval, and required assembly space volume.

[0095] Step S15: Map the assembly orientation constraint diagram data to attitude space based on the solar array 3D assembly data to obtain attitude constraint diagram data;

[0096] In one embodiment, the orientation of the component includes both position and orientation, represented by quaternions or Euler angles + translation vectors. For each assembly direction, a set of allowed component motion orientation regions is constructed (e.g., translation ±5mm, rotation ±10°). For each orientation candidate, path feasibility is checked to determine whether a collision will occur (by modeling the component boundary using an axis-aligned bounding box or a 3D polyhedron bounding structure; the motion trajectory is discretely sampled, and a collision detection algorithm is called at each sampling point to determine whether there is geometric overlap with other components). The system constructs an orientation constraint graph to describe the orientation state space of the component under a specific assembly direction, denoted as G. pose =(V,E) pose V represents the set of all legal attitude states, and each node corresponds to a legal component attitude configuration; E pose This represents a connecting edge that allows access from one attitude state to another along a continuously changing path. Each edge e mn ∈E pose Indicates from attitude P m Smooth transition to attitude P n A valid path transformation relationship, in which the path does not collide or violate assembly constraints in physical space.

[0097] Step S16: Solve the traversable space based on the attitude constraint graph data to obtain the assembly reachability graph data.

[0098] In one embodiment, the traversable space refers to the set of three-dimensional spaces that a component can occupy during its movement from the initial assembly position to the target assembly position without collision, occlusion, or violation of structural constraints. This space reflects the degrees of freedom and reachability of the component's assembly actions, forming the basis of the assembly channel. Motion path voxelization is used to discretize the component's motion path into multiple attitude keyframes. Collision detection is performed on each frame; if consecutive frames are valid, a feasible path is formed. All feasible path spaces are combined to form the assembly channel space volume. An assembly reachability graph is constructed, where each node represents a component and its attitude, and edges represent traversable path segments.

[0099] Preferably, the solution for the passable space specifically involves:

[0100] The component attitude solution volume is generated based on the attitude constraint diagram data to obtain the component attitude solution volume data;

[0101] In one embodiment, the attitude constraint graph is a graph structure composed of multiple legal attitudes, denoted as G. pose , containing the pose set P i Each pose P i Each component is composed of a three-dimensional position vector (x, y, z) and a corresponding attitude angle vector (θx, θy, θz), used to describe the possible configurations of the component during assembly. For each component C to be assembled... j Define its attitude solution volume Ω j Let C be a subset of its special Euclidean group SE(3), representing the component C. j The solution volume must simultaneously satisfy the following conditions: the position coordinates (x, y, z) of the component in three-dimensional space must be confined within its path planning constraint boundaries; the attitude angles (θx, θy, θz) must be within a preset allowable rotation range; and to avoid interference with adjacent components, the attitude solution volume should avoid the influence range of adjacent components, reserving a minimum buffer distance, for example, not less than 2 mm. In the specific solution process, a six-dimensional rasterized sampling strategy is adopted, that is, the three-dimensional position and three-dimensional attitude angle spaces are discretized, and each sampling point is checked to see if it simultaneously satisfies the above constraints, thereby selecting component C. j All feasible poses in space. For assembly scenarios involving surface insertion, such as inserting components along the tangential direction on complex surfaces, affine transformation methods can be used to further fit the component's pose. In this case, the pose solution volume will construct a continuous pose solution surface along the tangential direction of the component insertion path, thus ensuring the smoothness and accessibility of actions during assembly. The system outputs the pose solution volume data corresponding to each component, denoted as D. pose This data structure uses component identifiers C. j Use it as an index to record the corresponding legal pose space Ωj .

[0102] The attitude-driven path window is solved based on the component attitude solution volume data to obtain the attitude-driven path window data;

[0103] In one embodiment, for component C j Its attitude solution volume Ω j The set of legal configurations for this component in the six-dimensional attitude space SE(3) has been defined. Based on this, its attitude driving path window W is defined. j A channel is one or more consecutive feasible attitude paths, wherein the channel consists of a time-ordered sequence of attitude points {P0, P1, ..., P...}. n} constitutes, where each attitude point P i This includes position vectors (x, y, z) and attitude angles (θx, θy, θz), where changes between adjacent attitude points satisfy continuity constraints and the physical motion model during assembly. To construct the aforementioned path window, a sampling-based trajectory planning method is employed, such as the Fast Random Tree Extended Connect (RRT-Connect) algorithm or the Probabilistic Path Graph (PRM) algorithm, within the attitude solution volume Ω corresponding to the component. j The system searches for one or more consecutive feasible paths from the initial pose to the target pose. The initial pose is set as the initial gripping pose before the assembly tool contacts the component, and the target pose is set as the position pose of the component after assembly. This path sequence not only records a series of legal poses experienced by the component during assembly but can also be converted into a path channel volume in the pose space. This path channel volume consists of the sweep volume between adjacent pose frames, i.e., in P... i To P i+1 The three-dimensional envelope is formed by the continuous transformation of component models in attitude space. The swept volumes of each segment are combined to obtain the complete attitude-driven path window. This geometrically represents a feasible spatial corridor along which components can be assembled. The output is the attitude-driven path window data W. path Component Identifier C j For indexing, record the corresponding path window W j This forms a structured path channel data set.

[0104] Spatial collision volume culling is performed on the attitude-driven path window data to obtain collision culling data;

[0105] In one embodiment, a collision detection acceleration strategy combining axis-aligned bounding boxes (AABB) and bounding volume hierarchy is employed to perform preliminary screening of each pose point in the path window. Specifically, for each pose point P in the path sequence... iA geometric envelope is constructed for the corresponding component in that posture. This envelope includes the three-dimensional volume occupied by the posture and its directional range. Spatial interaction detection is then performed between this envelope and static scene elements such as assembled components and fixed obstacles to determine if there is spatial overlap or interference. If a single posture point P... i If a collision occurs in the middle of the path, the path is considered broken at that point, and the path can be truncated at that location. If multiple consecutive attitude points are detected to be colliding, it indicates that the entire path is infeasible, and the path segment is discarded. If the collision point is close to the end of the path and the attitude still has room for adjustment, a local perturbation search algorithm is activated to perform small-scale perturbations and resampling on the attitude path to attempt to repair path connectivity. Based on the above collision detection and elimination results, the path segments in the path window are classified and labeled, including: valid path window segments: continuous areas in the path that have not experienced collisions, which can be directly used as assembly action channels; interrupted path segments: paths that have broken but have the possibility of recovery, which can be used as a reference for local reconstruction and optimization; high-risk path segments: areas that frequently collide or cannot be adjusted, which should be marked for subsequent assembly sequence optimization or avoidance path design.

[0106] Based on the collision rejection data, the component path-attitude-direction map is reconstructed to obtain the assembly reachability map data.

[0107] In one embodiment, feasible paths, attitude sequences, and assembly directions are integrated into a unified graph structure to form an assembly reachability graph, thus constructing a ternary graph G. reach = (V, E), where node V is a triplet (C) for each node. j ,P i D k The component, attitude, and assembly direction are represented by an edge E. If there is physical traversability between nodes, they are connected to form an edge, representing an attitude transition path segment. The edge attribute structure includes path_cost: path cost (e.g., length, attitude change amplitude, time), interference_risk: path risk score, motion_type: motion type such as insertion, rotation, and sliding, and safety_margin: safety margin (unit: mm). Each valid attitude path window... Discrete frame points in the graph are used as nodes; directed edges are constructed using time-series order; assembly direction is written as an edge label attribute into the graph to achieve multi-directional support and path switching strategy. This yields the complete assembly reachability graph data G. reach = (V, E).

[0108] Preferably, step S2 specifically includes:

[0109] Step S21: Extract reachable paths based on the assembly reachability map data to obtain reachable path data;

[0110] In one embodiment, the system extracts a complete set of feasible paths from the initial component posture to the target posture from the assembly reachability graph. The system uses the initial assembly posture of the component as the starting node in the graph and its expected posture after installation as the ending node. Based on these starting and ending nodes, the system uses a path search algorithm to extract all path sequences that satisfy the feasibility conditions within the graph structure. Possible search algorithms include breadth-first search, depth-first search, Dijkstra's shortest path algorithm, and A* heuristic path search algorithm. Different algorithms can achieve shortest path finding or path coverage enhancement based on the design of the path cost function or heuristic function. During path extraction, for each path, the system records its corresponding node sequence, i.e., the set of posture states experienced by the component during the path process, which can be formally represented as {v1, v2, ..., v...}. n Each node represents the specific attitude configuration of the component at a certain moment. Simultaneously, the system calculates the path cost for each path, which may include dimensions such as the total path length, the total amplitude of attitude angle changes, and the number of assembly direction switching times. The system sets path validity constraints, including but not limited to the following rules: (1) the total assembly time corresponding to the path must not exceed a preset threshold; (2) the attitude changes corresponding to any two adjacent nodes in the path must not exhibit discontinuous "derailment jumps," meaning the translation and rotation differences between any two frames should be within the allowable range, ensuring the path has interpolability 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 the set of paths selectable from the initial state to the assembly position for that component, 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 extracting reachable paths, the system performs a structured partitioning of the path data in terms of time and attitude information to achieve path interference detection and local behavior reconstruction, generating path window data with spatial-temporal consistency. A path window is a temporal data block consisting of a sequence of continuous attitudes and assembly orientation states of a component within a specific time period, used to describe the local motion evolution process during component assembly. The system converts the attitude sequence corresponding to each reachable path into a set of path windows. This attitude sequence can be represented as the spatial pose information of the component in each keyframe during assembly. The pose includes position vectors and attitude angles (such as Euler angles or quaternions), defined in a three-dimensional rigid body motion group, denoted as SE(3) space. The path window is denoted as... Its form can be described as a path window consisting of a set of pose-time pairs, that is, each path window consists of several pairs of the form (P tThe state composition of ,t), where P t This represents the spatial pose of the component at time point t. During the generation of the path window, interpolation is performed on the discrete pose sequence of each path. The interpolation operation uses linear interpolation methods or curve interpolation methods (such as cubic spline interpolation, B-spline interpolation, or Bezier interpolation). The system uses a sliding window mechanism to divide the path sequence into multiple local segment windows. Taking 10 frames as an example, each window contains the continuous pose information of the component at 10 adjacent time points.

[0113] Step S23: Detect path interference regions in the path window data to obtain path interference region 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 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, such path segments need to be partially replaced or avoided. If only a brief collision occurs in a few time frames (such as only one or two frames) in the path window, and the spatial interference volume is small, it is marked as a recoverable interference point, indicating that the interference phenomenon can be adjusted by the perturbation correction strategy without completely replacing the path. If the motion trajectories of two or more path windows in the attitude space are close to each other, causing the minimum distance between a certain window and other path segments to be less than a set safety threshold (e.g., less than 5 mm of linear spacing, or attitude angle less than 5 degrees), then the window is marked as a path compression interference segment. Such paths, although not touched, have an extremely high risk of collision.

[0115] Step S24: Perform local behavior replacement based on the path interference region data to obtain assembly path data.

[0116] In one embodiment, a limited range of sampling is performed around the spatial position and attitude corresponding to the interference segment. The perturbation amplitude is set to ±5 mm in the position dimension and ±5 degrees in the attitude angle dimension to form a local legal attitude space. Based on the sampling results, a set of candidate action trajectories is constructed. The attitude change cost of each candidate path is calculated, including indicators such as translation length and total angle change. The path with the minimum 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 connection with the subsequent segments of the original path, that is, its terminating attitude must be able to be continuously connected with the starting attitude 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 interference 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 axial insert, offset slide-in, and S-curvetwist-in. In specific implementation, the system selects multiple applicable standard action path generators based on the spatial characteristics of the interference segment and the component type to regenerate the path segment. The system evaluates all generated candidate action paths for path smoothness, motion cost, directional consistency, and interference safety, selecting the action type with the highest degree of compliance as the replacement path and splicing it as part of the final component path.

[0118] Preferably, the optimized scheduling specifically includes:

[0119] Component dependency graphs are constructed from 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. This dependency graph expresses the assembly priority relationships between components in terms of temporal order, spatial occlusion, and structural load-bearing capacity in the form of a directed graph. The input data includes a component assembly path dataset, where each component corresponds to one or more complete assembly paths. Each path includes the target assembly position, orientation change direction, and execution time information. Based on this path data, three types of assembly dependency rules are defined, including spatial occlusion dependency rules: if the assembly path of component A needs to traverse the space occupied by component B during execution, and there is spatial overlap or occlusion between the two, then component A must be installed first to avoid obstructing the assembly path of B. Connection structure dependency rules: if component A is a load-bearing structure or connector of component B (e.g., B is a sub-component or fastener installed on A), then A should be installed first, followed by B, to meet structural integrity and connection strength requirements. The tool path dependency rule states that if the assembly operation of component A (e.g., screw tightening, inserting a tool) requires the use of an external tool, and the tool's operation path crosses the installation area of ​​component B, then A should be assembled first, followed by B, to avoid restricting tool operation. During dependency graph construction, each component is treated as a node in the graph structure, forming a node set V; any two components C... i With C j Between them, if any of the above dependency rules are satisfied, and C i It must take priority over C j Installation, then from node C in the graph structure i Pointing to C j Construct directed edges to form the edge set E. dep .

[0121] The component dependency graph data is optimized for minimizing installation time and minimizing assembly action switching, resulting in the first and second optimized sequence data, respectively.

[0122] In one embodiment, the scheduling process uses a component set C1, C2, ..., C n As input, construct a dependency graph (G) that satisfies the condition. dep Scheduling sequence of topology Among them, each component C j The corresponding assembly time t j It consists of the following components: execution time of the component assembly path; equipment changeover time required for component grabbing, moving, and positioning; and workstation idle waiting time (if resource conflicts exist). Total assembly time T totalDefined as the maximum completion time of all components, i.e., the critical path completion time. Formally, it can be expressed as: Total assembly time = Critical path length = Maximum of the latest completion times of all components. (See component dependency graph G.) dep Perform topological sorting to generate a candidate set of all legal assembly sequences that satisfy dependency constraints; calculate the earliest start time and latest acceptable start time for each component node in the graph using the critical path method; set minimizing the total assembly time (T_total) as the optimization objective; and search for the optimal scheduling path using dynamic programming or a cost function-based heuristic search algorithm (such as A* or greedy backtracking) based on the scheduling graph structure and path weight information. The generated first optimized sequence data is denoted as S1. opt This represents an execution path with the shortest total assembly time, provided that the assembly sequence of all components is satisfied.

[0123] A switching cost matrix is ​​constructed by quantifying the changes in path posture, assembly direction, and tool type of assembly actions of adjacent components. For any component C... i With C j The sequence of actions between them is used to calculate the difference in rotation amplitude, the angle of change in movement direction, and the frequency of fixture / tool ​​switching in the attitude space (SE(3)); the total action switching cost is then constructed. min represents minimization, k is the action frame index, and n is the total length of the pose sequence. Let $\mathbf{i}$ be the cost of action switching between adjacent poses, representing the cost of switching from pose $i$. k Switch to i k+1 The required cost, in units of dimensionless scores or time / energy consumption indicators, in megalitermes (M). ij =α·||x j ―x i ||+β·∠(θ i ,θ j )+γ·D dir (i,j), M ij Let be the action switching cost from pose i to j, and α be the translation weight factor, ranging from 0.2 to 0.5. j ―x i || represents the spatial translation distance, i.e., the position vector distance between attitude i and attitude j, in mm or m; β is the attitude angle weighting factor, ranging from 0.3 to 0.6; ∠(θ) i ,θ j ) represents the attitude angle difference, the rotation angle change between attitudes i and j (such as Euler angle difference, quaternion angle difference), in degrees (°) or radians (rad), γ is the motion direction weighting factor, with a value range of 0.1 to 0.3, D dir(i,j) represents the action direction difference data, indicating the structural overhead of action type (e.g., translation → rotation) or direction jump. This can be defined using a type matching distance matrix, where a mismatch is weighted at 1 and a match at 0. i is the action pose frame index, and j is the index of the adjacent action pose frame. Among all valid sequences satisfying the dependency graph topological constraints, the sequence that makes the action pose frame index the most suitable is selected. The smallest sequence is denoted as S2. opt .

[0124] The assembly path data is marked with scheduling information using the first optimized sequence data and the second optimized sequence data to obtain the assembly path optimized data.

[0125] In one embodiment, the scheduling order result is written back to the assembly path to generate optimized assembly path data with scheduling priority, time tag, and action score. For each component C j The corresponding path sequence will be appended with the following scheduling metadata fields, including the scheduling order. j This indicates the execution order of the component in the overall assembly sequence, which can be started from the first optimized sequence S1. opt Or the second optimized sequence S2 opt Obtained by mapping; estimated start time This indicates the earliest point in time when the component can begin assembly, obtained through critical path analysis; estimated completion time. This represents the theoretical time point at which the component completes installation, calculated based on its path duration and scheduling sequence; Action Score This indicates that the quality score of the assembly actions involved in the path is derived from a pre-built assembly action switching score matrix (denoted as M).

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

[0127] Three-dimensional interference region detection is performed on the assembly path optimization data to obtain interference region data;

[0128] In one embodiment, path segments that would cause conflicts during actual spatial operation are detected in the optimized path data, and the precise range of the interference region is output. Continuum Collision Detection (CCD) is used to detect whether components penetrate the environment within the path segment; a component buffer (e.g., a 2-5mm safety gap) is introduced to provide early warning of interference; the path attitude sequence is projected onto a spatial voxel grid, and a voxel map of component spatial occupancy is overlaid to detect overlapping areas. The detected interference regions are classified according to their severity and impact type as follows: Type A: Continuous penetration of the path segment → Reconstruction is required; Type B: Intermittent contact only at the boundary → Perturbation repair can be attempted; Type C: Pre-conflict with future components → Scheduling order adjustment is recommended.

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

[0130] In one embodiment, the component attitude envelope corresponding to the interference path segment is extracted and subjected to Boolean intersection with the interference object to obtain the conflict cross body; the conflict region is encapsulated using Alpha Shape or Concave Hull to generate a non-convex body model; if multiple attitude envelopes in continuous attitude segments have high overlap, they can be merged into a time-dynamic volume body. The model parameter definitions include conflict body center: Centroid, maximum bounding box: AABB, conflict influence range: influence attitude range ΔP, and hazard level: a severity index based on volume × coverage × contact frequency.

[0131] Based on the conflict volume model data, the action posture window of the conflict segment is reconstructed to obtain the action reconstruction data;

[0132] In one embodiment, a local path perturbation search (based on RRT or KinodynamicRRT) is performed on the conflict volume region to generate a new action sequence in the attitude space without penetrating the obstacle. A flexible attitude sliding window mechanism is introduced to allow for attempts at actions with small amplitude deformations (e.g., ±5mm, ±5°). If the local search fails, an alternative action template is invoked (e.g., changing from insertion to side-slide entry). The reconstruction constraint is set such that the change amplitude between any adjacent attitude configurations must not exceed a set continuity threshold ΔP. i,i+1 <δ pose , where ΔP i,i+1 The attitude change amplitude, i.e., the difference in attitude of the component between two consecutive action frames i and i+1, is δ. pose The main direction of movement of the newly generated path should be consistent with the original path, and there should be no drastic reversal of direction or sudden stop; the total installation time should not be increased by more than 15%.

[0133] Based on the motion reconstruction data, the assembly path optimization data is replanned to obtain the assembly path reconstruction data.

[0134] In one embodiment, replacement The conflict segment [i:j] is new_pose_sequence[i:j]; simultaneously, the path segment time label, action type label, and scheduling successor dependency are updated (if there is a time delay, cascade adjustments are made). To prevent discontinuous, unnatural, or kinematically constraint-violation abrupt changes at path replacement points, the connection verification between the beginning and end of the newly accessed path segment and the states before and after the original path is required. Spline interpolation (such as B-splines, Bezier) is used to smoothly connect boundary frames, or the attitude difference is determined by calculating the angle between direction vectors to determine whether it is less than the preset tolerance. When the direction difference is large but does not exceed the upper limit, transitional attitude frames are allowed to be inserted for buffering to ensure a natural transition of the overall path. Path segment simplification and cleaning are performed, such as merging consecutive frames with attitude change amplitudes less than a set threshold (such as position difference <1mm, angle difference <1°); identifying local abnormal direction abrupt changes in the path and replacing them with locally interpolated paths. Based on the above, the assembly path reconstruction data is obtained.

[0135] Preferably, the conflict volume modeling specifically involves;

[0136] The component motion trajectory is sampled based on the interference region data to obtain the component motion trajectory data;

[0137] In one embodiment, the system extracts the frame segment information where the component interferes in the path based on the output interference region data. Each interference record contains a component identifier and the corresponding start and end frame index range, which can be represented as component C. j The corresponding interference path segment interval is [start] idx end idx ], where start idx The start frame of the interval, end idx The system extracts the component's attitude sequence data within this interval, which is the set of discrete attitude points of the component from the start frame to the end frame. Each attitude point contains the component's position and orientation information in three-dimensional space, defined in the three-dimensional rigid body motion group SE(3). The position data is described by a ternary vector (x, y, z) in Cartesian coordinates; the attitude data can be expressed in Euler angles (pitch, yaw, roll) or quaternions (q0, q1, q2, q3). The system performs a dual-domain interpolation operation on the above original discrete attitude sequence, performing interpolation sampling 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 (e.g., inserting 5 intermediate points between each two frames). The output component motion trajectory data is a set of high-precision sampled trajectory point sequences of the component within the interference path segment, covering the continuous change process of spatial position and rotation angle.

[0138] Sweep envelope data is generated by performing a sweep envelope operation on the component motion trajectory data.

[0139] In one embodiment, the system samples each attitude point P in the component's motion trajectory. i Perform frame-by-frame processing. For each frame, the pose P... i The system applies a corresponding rigid body transformation T to the original geometric model M of the component. i This involves transforming a component from its static geometric state in its local coordinate system to its spatial arrangement in the global coordinate system, resulting in 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 x, y, z directions; rotating the model's attitude along three axes, which can be described using Euler angles or quaternions; and embedding the local component model into the global assembly scene. After transforming all sampled attitude points, the system obtains a set of geometric instances {M1, M2, ..., M...} of the component in a series of discrete attitudes along the path segment. n The system performs a spatial volume fusion operation on the above model set to construct the overall swept volume corresponding to the path segment of the component, denoted as the swept envelope V. swept This volume represents the spatial extent that a component may reach during all its pose changes within the path segment. The swept volume can be constructed in one of two ways: 1. By mapping the transformed component model to a 3D voxel mesh, and using Boolean union rules to accumulate voxel occupancy states, a 3D raster data structure representing the spatial occupancy probability or occupancy boundary is generated. 2. A Boolean union operation is performed on the mesh model generated in each frame, fusing the model contours under all poses to form a high-fidelity swept envelope with continuous boundary surfaces.

[0140] Boolean difference operations are performed on the interferometric components based on the swept envelope data to obtain the interferometric component data;

[0141] In one embodiment, for component C j sweep volume V swept , and other component sets {M} in the scene k} k≠j Perform Boolean intersection operation one by one; if the intersection result Then component k is determined to be an interference component, and the interference volume is in the form of a difference set. Use Boolean meshing libraries (such as OpenCSG, CGAL, and Volumetric MeshDifference); it is recommended to use AABB first to quickly eliminate non-overlapping candidates to improve efficiency. The output interference component data includes structured information such as component identifiers that spatially overlap with the current component sweep path, the volume model of the corresponding intersection group, the spatial location information of the intersection region, and the size of the intersection volume.

[0142] The conflict volume fragments are segmented and labeled based on the interference component data to obtain the conflict volume fragment data;

[0143] In one embodiment, when using a voxel grid to represent the interference volume, the 3D voxel model can be sliced ​​along the time axis based on the time dimension of the attitude change. The system sets a fixed time step (e.g., every Δt seconds or every Δframe) and projects the voxel set onto the time axis, dividing it into several segment volumes. Each segment represents the conflict region of a component within a certain time interval. When the interference volume is stored as a polygonal mesh, the system can append the corresponding attitude frame number to the attribute information of the mesh vertices or faces. Based on these frame numbers, the system can divide the entire conflict volume mesh into multiple sub-mesh blocks, each sub-block corresponding to a set of adjacent attitude states, forming a geometric segment with temporal semantics. After segmentation, the system appends a structured annotation field to each conflict volume segment unit, specifically including the attitude frame range (frame_range), which records the frame number range in the original motion trajectory corresponding to the conflict segment, such as frames 12 to 18, for associating path segments. The collision center (collision_center) refers to the geometric centroid coordinates of the collision fragment volume, which can be obtained by calculating the fragment's 3D centroid or bounding box center, and is used to locate the collision's location in space. The collision severity score (severity_score) measures the severity of the collision fragment within the overall sweep volume, defined as the percentage of the current fragment volume relative to the original sweep volume. This metric reflects the spatial impact weight of the collision in this path segment, facilitating priority-based processing. The recommended resolution strategy (recommended_resolution) is generated by the system based on the collision fragment's structural characteristics, collision type, and occurrence area, combined with a pre-defined motion strategy library, through parameter mapping and rule inference. Optional strategies include, but are not limited to, offset slide-in (suitable for boundary interference); re-orientation (suitable for angled collisions); and schedule rollback (suitable for occlusion collisions caused by component sequence).

[0144] The conflict volume aggregation graph is constructed based on the conflict volume fragment data to obtain the conflict volume model data.

[0145] In one embodiment, the graph structure is constructed as nodes V = {volume fragment}, each node representing a conflict volume fragment; edges E are constructed according to the following rules: sequential edges are established between adjacent time periods; if two fragments are associated with the same interferometer, a shared interferometer edge is established; if two fragments have similar conflict severity and the attitude switching direction is consistent, a potential action switching edge is established. Each graph node is appended with the following attribute fields, including the time frame interval (frame_range) corresponding to the fragment in the original path, the severity score of the conflict of the fragment, the geometric center coordinates (center_point) of the conflict fragment in 3D space, and the recommended action correction strategy (recommended_resolution) for the fragment.

[0146] Preferably, the reconstruction of the action pose window during the conflict segment specifically involves:

[0147] The domain of freedom is extracted from the conflict volume model data to obtain the domain of freedom data;

[0148] In one embodiment, the motion space of the component is defined as a six-dimensional attitude space SE(3), whose degrees of freedom are divided into the following two parts: linear translational degrees of freedom, translations (x, y, z) along the three spatial axes; and angular rotational degrees of freedom, rotations (θ) around the three coordinate axes. x ,θ y ,θ z Within regions with conflict volumes, the system employs a dimensional analysis approach to evaluate the feasibility of motion in each degree of freedom direction. For each degree of freedom dimension Δ... i (e.g., the x-direction), the system sets a certain range of perturbation step size in this direction to simulate the movement of the component in this direction; if any perturbation of the component in a certain degree of freedom direction will cause spatial interference with surrounding components or obstacles, then this dimension is considered unusable in the current environment and will be discarded; if a continuous interval still exists in a certain degree of freedom direction, and its perturbation will not trigger spatial conflict, then this dimension is considered a legal action direction and its passable interval is recorded [a i ,b i The system extracts the set of remaining legal action dimensions for each conflict segment component, defined as the domain of freedom (DoF). j This set represents all legal transformation dimensions that a component can achieve within the conflict space and their feasible range, i.e., DoF. j ={dimension i | Δ i ∈[a i ,b i [, no interference]

[0149] Based on the domain of freedom data, conflict volume inverse constraint analysis is performed to obtain the data of the blocked area;

[0150] In one embodiment, the blocked area (i.e., the spatial range in which components cannot enter) is deduced based on the degree-of-freedom domain and conflict volume distribution, providing guidance for obstacle avoidance reconstruction. In the extracted degree-of-freedom domain space (including three-dimensional translation and three-dimensional rotation dimensions), a discrete attitude sampling grid is constructed, and all possible attitude states are exhaustively traversed. For each attitude sampling point, the spatial volume occupied by the component in that attitude is simulated, and spatial intersection calculation is performed with the conflict volume model. If spatial overlap exists, i.e., the component contacts or interferes with other components or obstacles in that attitude state, it is determined to be an infeasible attitude, and its corresponding region is added to the blocked area set. If the component's occupied volume corresponding to the attitude point does not overlap with any conflict area, it is determined to be a free pose. The above spatial interference determination results are mapped to the degree-of-freedom dimension to extract the sub-intervals that should be prohibited in the degree-of-freedom domain. The system outputs a blocked area data set, represented as a set of multi-dimensional attitude subspace blocks that need to be excluded in the degree-of-freedom space.

[0151] Based on the data of the blocked area, the action posture window is reconstructed to obtain the action posture window data;

[0152] In one embodiment, the action posture window refers to a legal sequence of continuously executable postures of a component during assembly or movement. Each posture P i The requirements are: not located in a blocked area; attitude changes are continuous and interpolable; and the path can be connected to the original path. The component's motion attitude window is used to reconstruct the path within the remaining degrees of freedom using a search algorithm. Heuristic search (such as RRT-Connect, BFS) is used to find attitude paths in the legal free domain; each window segment has a maximum length (e.g., 20 frames) to form a sliding motion segment. If the path changes drastically (the distance between adjacent attitude frames exceeds a set threshold, the attitude angle difference between adjacent frames exceeds a set angle threshold, or the rate of position change (or angular velocity) changes unstablely across multiple consecutive frames), Bézier curves or spline interpolation are used for smoothing; a flexible interpolation strategy is added to control attitude jitter.

[0153] The optimal pose domain is extracted from the action pose window data to obtain the optimal pose domain data;

[0154] In one embodiment, the evaluation metrics include smoothness_score: a score for the continuity of attitude changes (e.g., curvature change), energy_score: a score for the amplitude of motion (combined cost of rotation and translation), collision_margin: the minimum distance from the blockade area, and execution_time: the execution time or timestamp span. A scoring function is established for all candidate attitude points: For attitude rating data, α is the attitude change continuity coefficient, with a recommended range of [0.3, 0.5], and S... i Let E be the smoothness score, β be the action cost weight coefficient, and its recommended range is [0.2, 0.4]. i Let M be the energy score, γ be the weight coefficient for the minimum distance from the blocked area, and its recommended range is [0.2, 0.4]. i For collision_margin, extract the high-scoring interval to form the optimal pose window subdomain. Where Ω best This is the optimal pose subdomain. This is the original action / pose window.

[0155] The action set is generated based on the optimal pose domain data to obtain action reconstruction data.

[0156] In one embodiment, the optimal attitude domain Ω best The high-dimensional attitude point sequence (6-dimensional SE(3) space) is converted into an equivalent kinematic description unit to form a simplified motion vector, which includes changes in position, direction, and angle. The system matches the converted motion segment with a 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". Matching criteria may include attitude transformation type, interpolation direction consistency, attitude stability, etc. For each identified motion type segment, its corresponding control parameters are extracted, including path direction, motion duration, velocity curve, and control accuracy requirements, to form an executable parameter configuration unit. The motion type, parameters, and attitude sequence segments are organized into a structured motion set.

[0157] Preferably, step S4 specifically includes:

[0158] Step S41: Map component actions based on assembly path reconstruction data to obtain component action association data;

[0159] In one embodiment, the assembly path (including posture sequence, action type, and timing) of each component is mapped to an executable action unit in the digital twin. The reconstructed assembly path data is denoted as P. final =C j :[P1,...,P n ], where P k (Values ​​range from 1 to n) Includes action segments, refactoring markers, timing, etc. The system divides each component path into several ActionUnits. jEach action frame unit represents an abstract representation of a complete action process of a component during assembly. Each unit contains the following attributes: component ID, component number C to which the action belongs. j Action type (action_type), such as "slide_in", "rotate", "insert", etc., used for action classification; pose frame sequence (frame_sequence), used to describe the pose evolution range of the action segment, in the form {P i ,...,P j}; Duration, indicating the execution time required for this action segment (e.g., 1.8 seconds); Dependencies, indicating that this action can only be executed after certain actions of other components have been completed, such as "C_k must finish". Component Action Association Data A map =C j ActionUnit j ActionUnit j Indicates component C j The corresponding set of action frame units, each of which contains its action attributes and execution constraints.

[0160] Step S42: Construct the assembly state machine based on the component action association data to obtain the assembly state machine data;

[0161] In one embodiment, the assembly process of each component is modeled as a state machine, clearly defining the state transition conditions, action triggering conditions, and termination conditions. States are defined as Idle (to be assembled), Engaging (entering an action), Assembling (performing assembly), and Completed (installation complete). State transition triggering methods include time-driven (based on the start and end time of the action), event-driven (dependent on the completion of component actions), and space-driven (the current position / attitude reaches the assembly position). The state machine diagram structure includes nodes: representing states; edges: triggering conditions with weights (e.g., time consumption, delay); 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, triggering mechanism, and timing logic, serving as input to the dynamic control model for assembly scheduling and visual execution. For example, C... 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 progressed more than halfway or met a certain progress threshold, and done is when the assembly path segment is completed and the dependencies are satisfied.

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

[0163] In one embodiment, the virtual configuration space refers to the set of three-dimensional spatial configurations of all components in the system at any given time or state interval; each state node corresponds to a set of component attitudes / positions (e.g., initial state, in execution, intermediate position, completed position); the multi-state combination of multiple components constitutes the entire state layer of the system configuration space, supporting state-driven visualization, layout playback, and simulation. The configuration space is generated by extracting the corresponding attitude frame fragments 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 that state; the attitude combinations of all components in the current state are spatially superimposed, and a dynamic configuration layer is constructed layer by layer 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 advancement from the initial state to the target state; thus, the virtual configuration space data V is obtained. space ={State i :{C j :pose ij}}, where State i For assembly states (such as Engaging, Assembling, etc.), C j Number the component, pose ij For component C j In state S i The spatial attitude below.

[0164] Step S44: Perform attitude binding based on the virtual configuration space data to obtain attitude binding data;

[0165] In one embodiment, pose binding transforms the spatial configuration (including position and pose) of a component in a certain assembly state into a set of pose control instructions that can be recognized and executed in the rendering system to drive 3D animation or visual state switching. Each binding unit contains the following attribute fields: target object ID; binding type (absolute / relative); binding reference frame (e.g., tooling reference point); pose parameter sequence (x, y, z, yaw, pitch, roll), where the first three are spatial translation vectors and the last three are rotational poses represented by Euler angles. Each component state is converted into a rendering layer transform binding (WebGL / Unity, etc.); during the action, pose changes are bound using LERP or SLERP (linear / spherical interpolation). This yields pose binding data B. pose =C j :[state1:pose1,...,state n :pose n ], where B pose To bind data to the pose, C j For component numbering, state i pose represents a component in a certain state of the assembly state machine. i The bound attitude parameters in this state are , where i takes the value 1…n.

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

[0167] In one embodiment, a unified action timeline for all components in the assembly is implemented to achieve a visualized synchronous execution flow for parallel or sequential assembly of multiple components. A Gantt-type action scheduling graph is constructed, where the X-axis represents time and the Y-axis represents component ID, with each line segment representing an action segment. Action times are extracted based on the timestamp `transition[t]` in the state machine; keyframe synchronization points are set (e.g., "3 components need to be inserted synchronously at the 8th second"); logical flags are introduced to specify that action execution depends on the completion of a preceding action, such as `wait_for(C_k.completed)` (representing component C). j The current action needs to wait for C k Complete); Set the time trigger flag `start_at` to specify that a certain action must start at a specified absolute time, such as `start_at(t=12.0)`; Set the synchronization group `sync_group` to define the set of component groups that need to be executed collaboratively, `sync_group={C_j,C_m}` (meaning that two components need to be assembled concurrently within a certain synchronization window). Obtain the assembly synchronization data. Where: t k Indicates a specific point in time (or the start of a time period); C j The component that performs the action at that point in time; Indicates component C j The corresponding action unit labels include slide_in, align, insert, etc., and so on for other parameters.

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

[0169] In one embodiment, the interactive events include user-triggered behaviors, system-automatic response events, and simulated operations for intervention simulations. These are mainly categorized into three types: 1. User-triggered events: Clicking on a component displays its current assembly status; replaying the complete assembly path of the component (e.g., from the starting position to the finished position); zooming in to view a specific local area or state feature. 2. System-triggered events: When the system detects a collision or path penetration during attitude change, it automatically triggers an alarm; when assembly actions are not executed in the scheduled order (e.g., premature execution without fulfilling dependencies), it triggers an abnormal order prompt; it provides real-time feedback on the current component's execution delay. 3. Simulation intervention events (used for assembly scheme verification and decision optimization): Allowing simulations of premature insertion, delayed execution, etc., to observe the system's behavioral response under constraint conflicts or resource competition; recording intervention results and providing 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 the user clicks on the component model); onCollision (triggered when an interference event is detected); and onViolation (triggered when sequence, timing, or logical constraints are violated). Each interactive event can define a standardized event script structure, including event type, target component, response action, and parameter data. Event logic can correspond one-to-one with state machine nodes, assembly sequence diagram nodes, or bound posture nodes, forming a behavior-state bidirectional linkage model, i.e., assembly path digital twin data.

[0170] Preferably, this application also provides a three-dimensional assembly path planning system for a solar wing structure, used to execute the three-dimensional assembly path planning method for a solar wing structure as described above. The three-dimensional assembly path planning system for the solar wing structure includes:

[0171] The 3D assembly perception and mapping module is used to acquire the 3D assembly data of the solar array and construct the assembly reachability map based on the 3D assembly data of the solar array to obtain the assembly reachability map data.

[0172] The path avoidance planning module is used to plan path avoidance behaviors based on assembly reachability map data to obtain assembly path data.

[0173] The path optimization and interference reconstruction module is used to optimize and schedule assembly path data to obtain optimized assembly path data; and to perform interference resolution and reconstruction on the optimized assembly path data to obtain reconstructed assembly path data.

[0174] The digital twin generation module is used to generate digital twin scenarios based on the assembly path reconstruction data, thereby obtaining assembly path digital twin data.

[0175] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended application documents rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the application documents be incorporated into the invention.

[0176] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the 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 invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A three-dimensional assembly path planning method for a solar panel structure, characterized in that, Includes the following steps: Step S1: Obtain the three-dimensional assembly data of the solar array; Component analysis was performed on the three-dimensional assembly data of the solar array to obtain component analysis data; Based on the 3D assembly data of the solar array, a spatial correlation diagram is generated from the component analytical data to obtain spatial correlation diagram data; based on the 3D assembly data of the solar array, an assembly direction mapping is performed on the spatial correlation diagram data to obtain assembly direction constraint diagram data; based on the 3D assembly data of the solar array, an attitude space mapping is performed on the assembly direction constraint diagram data to obtain attitude constraint diagram data; based on the attitude constraint diagram data, the traversable space is solved to obtain assembly reachability map data. Step S2: Extract reachable paths from the assembly reachable map data to obtain reachable path data; generate path windows from the reachable path data to obtain path window data; detect path interference regions from the path window data to obtain path interference region data; perform local behavior replacement from the path interference region data to obtain assembly path data. Step S3: Construct a component dependency graph from the assembly path data to obtain the component dependency graph data; The component dependency graph data is optimized for minimizing installation time and minimizing assembly action switching, resulting in first and second optimized sequence data, respectively. The first and second optimized sequence data are then used to mark the assembly path data with scheduling information to obtain optimized assembly path data. The assembly path optimization data is reconstructed by interference elimination to obtain the reconstructed assembly path data. Step S4: Generate a digital twin scene based on the assembly path reconstruction data to obtain assembly path digital twin data.

2. The method according to claim 1, characterized in that, The solution for the passable space is specifically as follows: The component attitude solution volume is generated based on the attitude constraint diagram data to obtain the component attitude solution volume data; The attitude-driven path window is solved based on the component attitude solution volume data to obtain the attitude-driven path window data; Spatial collision volume culling is performed on the attitude-driven path window data to obtain collision culling data; Based on the collision rejection data, the component path-attitude-direction map is reconstructed to obtain the assembly reachability map data.

3. The method according to claim 1, characterized in that, Specifically, the interference resolution and reconstruction are as follows: Three-dimensional interference region detection is performed on the assembly path optimization data to obtain interference region data; Conflict volume modeling is performed based on the interference region data to obtain conflict volume model data; Based on the conflict volume model data, the action posture window of the conflict segment is reconstructed to obtain the action reconstruction data; Based on the motion reconstruction data, the assembly path optimization data is replanned to obtain the assembly path reconstruction data.

4. The method according to claim 3, characterized in that, The specific details of conflict volume modeling are as follows: The component motion trajectory is sampled based on the interference region data to obtain the component motion trajectory data; Sweep envelope data is generated by performing a sweep envelope operation on the component motion trajectory data. Boolean difference operations are performed on the interferometric components based on the swept envelope data to obtain the interferometric component data; The conflict volume fragments are segmented and labeled based on the interference component data to obtain the conflict volume fragment data; The conflict volume aggregation graph is constructed based on the conflict volume fragment data to obtain the conflict volume model data.

5. The method according to claim 3, characterized in that, The specific reconstruction of the action pose window during the conflict segment is as follows: The domain of freedom is extracted from the conflict volume model data to obtain the domain of freedom data; Based on the domain of freedom data, conflict volume inverse constraint analysis is performed to obtain the data of the blocked area; Based on the data of the blocked area, the action posture window is reconstructed to obtain the action posture window data; The optimal pose domain is extracted from the action pose window data to obtain the optimal pose domain data; The action set is generated based on the optimal pose domain data to obtain action reconstruction data.

6. The method according to claim 1, characterized in that, Step S4 is as follows: Component actions are mapped based on the assembly path reconstruction data to obtain component action association data; The assembly state machine is constructed based on the component action association data to obtain the assembly state machine data. Virtual configuration space data is generated from the assembly state machine data. Attitude binding is performed based on virtual configuration space data to obtain attitude binding data; Synchronize the assembly action timing with the attitude binding data to obtain assembly synchronization data; By integrating interactive events based on assembly synchronization data, digital twin data of the assembly path is obtained.

7. A three-dimensional assembly path planning system for a solar panel structure, characterized in that, For executing the three-dimensional assembly path planning method for the solar wing structure as described in claim 1, the three-dimensional assembly path planning system for the solar wing structure includes: The 3D assembly perception and mapping module is used to acquire the 3D assembly data of the solar array and construct the assembly reachability map based on the 3D assembly data of the solar array to obtain the assembly reachability map data. The path avoidance planning module is used to plan path avoidance behaviors based on assembly reachability map data to obtain assembly path data. The path optimization and interference reconstruction module is used to optimize and schedule assembly path data to obtain optimized assembly path data; and to perform interference resolution and reconstruction on the optimized assembly path data to obtain reconstructed assembly path data. The digital twin generation module is used to generate digital twin scenarios based on the assembly path reconstruction data, thereby obtaining assembly path digital twin data.

Images