Mechanical arm track generation method and system for wooden hanger free-form surface machining

Abstract

The invention discloses a mechanical arm track generation method and system for wooden clothes hanger free-form surface machining, and relates to the technical field of mechanical arm track generation. Digital twin bodies of wooden clothes hangers are constructed, and the wooden clothes hangers are divided into a plurality of machining subareas; arranging process whistle point networks on the plurality of processing partitions; a coverage path of each subarea is generated based on the working space and the task rhythm of the double mechanical arms, the coverage paths are distributed to the mechanical arms, and a relative pose corridor constraint is introduced to generate a double-arm initial joint track; the initial joint tracks of the two arms act on the digital twin, and based on the feedback of the process whistle point network, the collision risk is predicted and a conflict report is generated; iterative optimization is conducted on the initial joint tracks of the two arms according to the conflict report, and the working track of the mechanical arm is generated; through off-line rehearsal and intelligent collaborative planning, traditional on-line passive collision avoidance is converted into off-line active prevention and control, and a double-arm collaborative track which is safe, free of collision, smooth and synchronous in motion and efficient is fundamentally generated.

Description

Technical Field

[0001] This invention relates to the field of robotic arm trajectory generation technology, specifically to a method and system for generating robotic arm trajectories for processing freeform surfaces of wooden coat hangers. Background Technology

[0002] In the polishing of workpieces with complex free-form surfaces, such as wooden coat hangers, traditional processing methods mainly face systemic challenges. Due to the large curvature changes, irregular geometric shapes, and local occlusion of these surfaces, traditional single-arm polishing processes suffer from low coverage efficiency and unreasonable trajectory planning, making it difficult to achieve efficient processing while ensuring consistent surface quality. To solve the efficiency bottleneck, the industry is gradually adopting a dual-arm collaborative operation mode. However, in practical applications, existing collaborative trajectory planning methods still have certain shortcomings. First, in the trajectory generation stage, they rely heavily on simple geometric partitioning and time synchronization, lacking deep integration of process semantics such as curvature features and processing stages, resulting in suboptimal trajectory allocation and easy generation of processing blind spots or repeated coverage. At the same time, in terms of collaborative safety, existing methods usually perform passive collision avoidance through real-time collision detection after trajectory generation, or apply simple spatial constraints in planning, which makes it difficult to proactively avoid interference between the two arms in the dynamic and continuous polishing process, resulting in high collision risk and slow response safety hazards. Furthermore, in terms of trajectory quality, existing methods often neglect the smoothness and precise synchronization of motion at partition boundaries and collaborative relay points, which can easily lead to sudden changes in speed and attitude, affecting the consistency and smoothness of the final polished surface. Therefore, it is of great significance to develop a method for generating the trajectory of a robotic arm for machining the freeform surface of a wooden clothes hanger. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for generating robotic arm trajectories for machining freeform surfaces of wooden clothes hangers, in order to solve the problems in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for generating the trajectory of a robotic arm for machining the free-form surface of a wooden clothes hanger, comprising: The surface point cloud data of the wooden clothes hanger is obtained by 3D scanning, a 3D mesh model of the wooden clothes hanger is constructed and process semantic tags are added to generate a digital twin of the wooden clothes hanger; Based on the geometric data of the digital twin and the semantic tags of the process, the wooden clothes hanger is divided into multiple processing zones; A process sentinel network is deployed across multiple processing zones on the surface of the digital twin. The process sentinel network includes process sentinels and virtual transmission channels. Based on the workspace and task cycle of the dual robotic arms, the coverage path of each processing zone is generated, the coverage path is assigned to the first robotic arm and the second robotic arm, and the relative pose corridor constraint is introduced to generate the initial joint trajectory of the dual arms. The initial joint trajectories of the two arms are applied to the digital twin, and based on the feedback from the process sentinel network, collision risks are predicted and conflict reports are generated. Based on the conflict report, the initial joint trajectory of the two arms is iteratively optimized until no conflict occurs in the collaborative pre-simulation, generating an optimized trajectory. The optimized trajectory is then subjected to collision detection and motion smoothness verification to generate the working trajectory of the robotic arm.

[0005] In a preferred embodiment, the steps of acquiring surface point cloud data of the wooden clothes hanger through 3D scanning, constructing a 3D mesh model of the wooden clothes hanger, adding process semantic tags, and generating a digital twin of the wooden clothes hanger are as follows: The surface of the wooden clothes hanger was scanned using a structured light scanner to obtain the surface point cloud data of the wooden clothes hanger; A 3D surface reconstruction was performed based on surface point cloud data to generate a 3D mesh model of the wooden clothes hanger. Process semantic tags are annotated on a 3D mesh model to generate a digital twin of the wooden clothes hanger; Process semantic tags include processing area type tags, curvature distribution tags, and surface normal tags.

[0006] In a preferred embodiment, the step of dividing the wooden clothes hanger into multiple processing zones based on the geometric data of the digital twin and the process semantic tags is as follows: Based on the curvature distribution tags in the process semantic tags, identify the high curvature areas and flat areas on the surface of the wooden clothes hanger; Based on the processing area type label in the process semantic label, distinguish the shoulder area, hook area and end area of ​​the wooden clothes hanger; By combining the geometric data of the digital twin with the surface normal labels, the contact data of the polishing tool is determined. The contact data includes contact parameters and coverage width parameters. Based on the distribution of high curvature and flat areas, the division of shoulder area, hook area and end area, and contact data, the surface of the wooden clothes hanger is divided into multiple processing zones; The processing zones include the flat shoulder area, the high curvature hook area, and the curved end area.

[0007] In a preferred embodiment, the step of deploying a process sentinel network across multiple processing zones on the surface of the digital twin, wherein the process sentinel network includes process sentinels and virtual transmission channels, is as follows: Process sentinels include status synchronization sentinels and collaborative monitoring sentinels; Among them, the status synchronization sentinel is used to mark the processing status of the robotic arm, and the collaborative monitoring sentinel is used to monitor safety risks; Status synchronization sentinels are set up at a preset density in the flat area on the shoulder of the wooden clothes hanger, and a virtual transmission channel is built between the status synchronization sentinels to transmit the status data of the robotic arm. Collaborative monitoring sentinels are deployed at a preset density in the high curvature area of ​​the hooks and the bending area at the ends of the wooden clothes hanger, and a virtual transmission channel is constructed between the collaborative monitoring sentinels to transmit collaborative data of the robotic arm. Configure a list of associated sentinels, a safe radius of action, and a safe threshold parameter for each collaborative monitoring sentinel to generate a process sentinel network.

[0008] In a preferred embodiment, the steps of generating coverage paths for each processing zone based on the workspace and task cycle of the dual robotic arms, assigning the coverage paths to the first and second robotic arms, and introducing relative pose corridor constraints to generate the initial joint trajectories of the dual arms are as follows: Generate a sequence of processing path points covering the surface of each processing partition, which serves as the coverage path for that partition; Based on the workspace and joint range of motion of the first and second robotic arms, the accessibility of each robotic arm to different processing zones is evaluated. Based on the task cycle requirements, with the goal of balancing the processing load and minimizing the total processing time, the reachable processing zones are allocated to the first robotic arm and the second robotic arm. Assign a coverage path in the corresponding processing zone to each robotic arm, and generate the task path for that robotic arm by combining the contact data of the polishing tool; A relative pose corridor constraint is introduced on the task path of the two robotic arms. Based on the relative pose corridor constraint, the inverse kinematics solution is performed on the task path of the two robotic arms to generate the initial joint trajectory of the two arms. The relative pose corridor constraint is used to define the minimum distance that must be maintained between the ends of the two arms, the maximum allowable relative attitude angle, and the maximum relative velocity.

[0009] In a preferred embodiment, the step of applying the initial joint trajectories of the two arms to the digital twin, predicting collision risks and generating conflict reports based on feedback from the process sentinel network, is as follows: Load the initial joint trajectories of the first and second robotic arms in the simulation environment corresponding to the digital twin, and establish a unified simulation clock; Drive the simulation clock to run and calculate the positions of the ends of the two arms and each joint link on the digital twin in real time; When the robotic arm passes a state synchronization sentinel, the state synchronization sentinel is triggered. The triggered state synchronization sentinel broadcasts the robotic arm's state update message to other state synchronization sentinels in the process sentinel network through a virtual transmission channel. The status update message includes the trigger sentinel identifier, the trigger robot arm identifier, the trigger processing status type, and the precise simulation time; When the robotic arm passes by the collaborative monitoring outpost, the collaborative monitoring outpost is triggered. The triggered collaborative monitoring outpost sends collaborative data to all collaborative monitoring outposts in the associated outpost list through the virtual transmission channel. The collaborative data includes the identifier of the triggering sentinel, the identifier of the triggering robotic arm, the triggering time, and the estimated duration of occupation; After receiving collaborative data, the collaborative monitoring sentinels in the associated sentinel list predict the spatiotemporal proximity of the robotic arm and another robotic arm within their own monitoring area based on the robotic arm trajectory information. The predicted spatiotemporal proximity state is compared with its own configured safety threshold. If the spatiotemporal proximity state exceeds the safety threshold, a collision risk is determined and a conflict report is generated. The conflict report includes the estimated time of the conflict, the identifier of the robotic arm involved in the conflict, and the location of the collaborative monitoring outpost associated with the conflict.

[0010] In a preferred embodiment, the steps of iteratively optimizing the initial joint trajectories of the two arms based on the conflict report until no conflicts occur during the collaborative pre-simulation, generating an optimized trajectory, and performing multi-level collision detection and motion smoothness verification on the optimized trajectory to generate the robotic arm's working trajectory are as follows: Based on the conflict report, a trajectory optimization strategy is formulated and the initial joint trajectory is modified. Collaborative pre-simulation and conflict prediction are performed again. This process is iterated until no conflict occurs, and the optimized trajectory is obtained. The trajectory optimization strategies include using velocity scaling on the time axis to address conflicts caused by overlapping time windows, using a local path replanning strategy to address conflicts caused by excessively close spatial paths, and using a velocity curve correction strategy to address conflicts caused by relative attitude or velocity not meeting constraints. Collision verification and motion smoothness verification are performed on the optimized trajectory. The optimized trajectory that passes the verification is encoded into an executable file to generate the working trajectory of the robotic arm.

[0011] This invention also provides a robotic arm trajectory generation system for machining free-form surfaces of wooden coat hangers, comprising: The digital twin construction module acquires surface point cloud data of a wooden clothes hanger through 3D scanning, constructs a 3D mesh model of the wooden clothes hanger, adds process semantic tags, and generates a digital twin of the wooden clothes hanger. The zoning planning module, connected to the digital twin construction module, divides the wooden clothes hanger into multiple processing zones based on the geometric data and process semantic tags of the digital twin. The sentinel network deployment module is connected to the zoning planning module to deploy a process sentinel network on multiple processing zones on the surface of the digital twin. The process sentinel network includes process sentinels and virtual transmission channels. The trajectory initialization module is connected to the sentinel network deployment module. Based on the workspace and task cycle of the two robotic arms, it generates the coverage path of each processing zone, assigns the coverage path to the first robotic arm and the second robotic arm, and introduces relative pose corridor constraints to generate the initial joint trajectory of the two arms. The collaborative pre-simulation module, connected to the trajectory initialization module, applies the initial joint trajectories of the two arms to the digital twin. Based on the feedback from the process sentinel network, it predicts collision risks and generates conflict reports. The trajectory optimization generation module is connected to the collaborative pre-simulation module. It iteratively optimizes the initial joint trajectory of the two arms based on the conflict report until no conflict occurs in the collaborative pre-simulation, generates the optimized trajectory, and performs collision detection and motion smoothness verification on the optimized trajectory to generate the working trajectory of the robotic arm.

[0012] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention constructs a digital twin of a wooden clothes hanger and deploys a network of process sentinels, achieving proactive collaborative prediction and avoidance of collision risks during the offline planning stage, significantly improving the safety and efficiency of trajectory generation. Traditional methods typically perform collision detection and response after trajectory generation or during online execution, which carries the risk of computational lag and interruption. This solution uses the digital twin as a virtual test field and deploys a network of process sentinels with interactive functions on its key surface areas. During collaborative rehearsals in the planning stage, these sentinels can sense the passage of the robotic arm in real time and exchange status and risk information through virtual transmission channels. This allows for the prediction and assessment of potential spatiotemporal interference between the two arms in advance in the virtual environment, placing the high-load collision calculation and complex collaborative decision-making offline. The generated robotic arm working trajectory theoretically avoids all identified conflicts, thereby significantly reducing the real-time computational burden and emergency stop risk during online execution. 2. This invention achieves intelligent perception and adaptive collaborative control of complex processing processes through the division of labor and collaboration between state synchronization sentinels and collaborative monitoring sentinels in the process sentinel network. This effectively ensures the consistency of processing quality and improves system flexibility. In this invention, state synchronization sentinels deployed in flat areas form a state perception network that broadcasts the processing stage and progress information of each robotic arm in real time, forming a global collaborative context. Collaborative monitoring sentinels deployed in high-risk areas utilize this context information, combined with detailed motion data they monitor, to perform more accurate and process-stage-appropriate collision risk prediction and assessment. This design enables the system not only to prevent physical collisions but also to intelligently adjust safety assessment thresholds and collaborative strategies based on process semantics. When a conflict occurs, the system can flexibly select strategies such as time scaling, local replanning, or attitude adjustment for iterative optimization based on the type and context of the conflict report. This ensures absolute safety while maximizing the process rationality, motion smoothness, and collaborative synchronization of the processing path, ultimately generating a high-quality, safe-to-execute collaborative trajectory. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0014] Figure 1 This is a flowchart of the method of the present invention.

[0015] Figure 2 This is a system block diagram of the present invention.

[0016] Figure 3 This is a logic block diagram of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1 and Figure 3 As shown in this embodiment, a method for generating the trajectory of a robotic arm for machining a free-form surface of a wooden clothes hanger includes: S1. Obtain surface point cloud data of the wooden clothes hanger through 3D scanning, construct a 3D mesh model of the wooden clothes hanger and add process semantic tags to generate a digital twin of the wooden clothes hanger; S2. Based on the geometric data of the digital twin and the process semantic tags, the wooden clothes hanger is divided into multiple processing zones; S3. Deploy a process sentinel network in multiple processing zones on the surface of the digital twin. The process sentinel network includes process sentinels and virtual transmission channels. S4. Based on the workspace and task cycle of the dual robotic arms, generate the coverage path of each processing zone, assign the coverage path to the first robotic arm and the second robotic arm, and introduce relative pose corridor constraints to generate the initial joint trajectory of the dual arms. S5. Apply the initial joint trajectories of both arms to the digital twin, and predict collision risks and generate conflict reports based on feedback from the process sentinel network. S6. Iteratively optimize the initial joint trajectories of the two arms based on the conflict report until no conflict occurs in the collaborative pre-simulation, generate the optimized trajectory, and perform collision detection and motion smoothness verification on the optimized trajectory to generate the working trajectory of the robotic arm. As described in steps S1-S6 above, in the field of polishing workpieces with complex free-form surfaces such as wooden coat hangers, traditional processing methods mainly face systemic challenges. Due to the large curvature changes, irregular geometric shapes, and local occlusion of such surfaces, traditional single-arm polishing processes suffer from low coverage efficiency and unreasonable trajectory planning, making it difficult to achieve efficient processing while ensuring consistent surface quality. To solve the efficiency bottleneck, the industry is gradually adopting a dual-arm collaborative operation mode. However, in practical applications, existing collaborative trajectory planning methods still have certain shortcomings. First, in the trajectory generation stage, they rely heavily on simple geometric partitioning and time synchronization, lacking deep integration of process semantics such as curvature features and processing stages, resulting in suboptimal trajectory allocation and easy generation of processing blind spots or repeated coverage. At the same time, in terms of collaborative safety, existing methods usually perform passive collision avoidance through real-time collision detection after trajectory generation, or apply simple spatial constraints in planning, which makes it difficult to proactively avoid interference between the two arms in the dynamic and continuous polishing process, resulting in high collision risk and slow response safety hazards. Furthermore, in terms of trajectory quality, existing methods often neglect the smoothness and precise synchronization of motion at partition boundaries and collaborative relay points, which can easily lead to sudden changes in speed and attitude, affecting the consistency and smoothness of the final polished surface. This invention constructs a digital twin of a wooden clothes hanger and deploys a network of process sentinels, achieving proactive collaborative prediction and avoidance of collision risks during the offline planning stage, significantly improving the safety and efficiency of trajectory generation. Traditional methods typically perform collision detection and response after trajectory generation or during online execution, which carries the risk of computational lag and interruption. This solution uses the digital twin as a virtual test field and deploys a network of interactive process sentinels on its key surface areas. During collaborative rehearsals in the planning stage, these sentinels can sense the passage of the robotic arm in real time and exchange status and risk information through virtual transmission channels. This allows for the prediction and assessment of potential spatiotemporal interference between the two arms in advance in a virtual environment. By placing the heavy collision calculation and complex collaborative decision-making offline, the generated robotic arm working trajectory theoretically avoids all identified conflicts, thereby significantly reducing the real-time computational burden and emergency stop risk during online execution. By dividing the work and coordinating the state synchronization sentinels and collaborative monitoring sentinels in the process sentinel network, intelligent perception and adaptive collaborative control of complex processing processes are achieved, effectively ensuring the consistency of processing quality and improving system flexibility. In this invention, state synchronization sentinels deployed in flat areas form a state perception network, broadcasting the processing stage and progress information of each robotic arm in real time, forming a global collaborative context. Collaborative monitoring sentinels deployed in high-risk areas utilize this context information, combined with detailed motion data they monitor, to perform more accurate and process-stage-appropriate collision risk prediction and assessment. This design enables the system not only to prevent physical collisions but also to intelligently adjust safety assessment thresholds and collaborative strategies based on process semantics. When a conflict occurs, the system can flexibly select strategies such as time scaling, local replanning, or attitude adjustment for iterative optimization based on the type and context of the conflict report. This ensures absolute safety while maximizing the process rationality, motion smoothness, and collaborative synchronization of the processing path, ultimately generating a high-quality, safe-to-execute collaborative trajectory.

[0019] In one embodiment, step S1, which involves acquiring surface point cloud data of a wooden hanger through 3D scanning, constructing a 3D mesh model of the wooden hanger, adding process semantic tags, and generating a digital twin of the wooden hanger, includes: S11. Scan the surface of the wooden clothes hanger using a structured light scanner to obtain the surface point cloud data of the wooden clothes hanger; S12. Based on the surface point cloud data, perform three-dimensional surface reconstruction to generate a three-dimensional mesh model of the wooden clothes hanger; S13. Label the process semantic tags on the 3D mesh model to generate a digital twin of the wooden clothes hanger; S14. Process semantic tags include processing area type tags, curvature distribution tags, and surface normal tags; As described in steps S11-S14 above, the wooden clothes hanger is scanned from multiple angles using a structured light scanner to obtain a high-precision point cloud data set of its surface. Then, the point cloud data is preprocessed by noise reduction filtering, multi-view registration, and point cloud fusion to form a complete and clean workpiece point cloud model. Next, the point cloud is converted into a continuous watertight triangular mesh surface based on Poisson reconstruction or greedy projection triangulation algorithms, thereby constructing a three-dimensional mesh model of the wooden clothes hanger. On this basis, according to process knowledge, the corresponding areas of the three-dimensional mesh model are manually labeled or automatically identified and labeled with processing area type labels based on curvature features, such as the main polishing area of ​​the shoulder and the fine processing area of ​​the hook. At the same time, curvature distribution labels are automatically generated by calculating the Gaussian curvature and average curvature of the mesh vertices and dividing the intervals. The surface normal label of each area is calculated using the vertex normal of the mesh triangular facet. Finally, the three-dimensional mesh model with processing area type labels, curvature distribution labels, and surface normal labels is output as a digital twin of the wooden clothes hanger that integrates geometric and process information.

[0020] In one embodiment, step S2, which divides the wooden clothes hanger into multiple processing zones based on the geometric data of the digital twin and the process semantic tags, includes: S21. Based on the curvature distribution label in the process semantic label, identify the high curvature area and flat area on the surface of the wooden clothes hanger; S22. Based on the processing area type label in the process semantic label, distinguish the shoulder area, hook area and end area of ​​the wooden clothes hanger; S23. Combining the geometric data of the digital twin with the surface normal label, determine the contact data of the polishing tool. The contact data includes contact parameters and coverage width parameters. S24. Based on the distribution of high curvature areas and flat areas, the division of shoulder areas, hook areas and end areas, and contact data, the surface of the wooden clothes hanger is divided into multiple processing zones. S25, wherein the processing zones include the flat shoulder area, the high curvature hook area, and the end bending area; As described in steps S21-S25 above, firstly, based on the curvature distribution label carried by the digital twin, the surface of the wooden hanger is divided into high-curvature regions and flat regions by setting a curvature threshold. High-curvature regions typically correspond to areas with drastic curvature changes, such as the inner side of the hook, while flat regions correspond to areas with gentler curvature, such as the shoulder. Then, based on the predefined semantic information in the processing area type label, the geometric boundaries of the shoulder region, hook region, and end region are clearly identified. Next, combining the triangular mesh geometric data of the digital twin with the surface normal labels of vertices or faces, the effective tilt angle and fit when the polishing tool contacts the curved surface are calculated. Contact parameters were determined, and the coverage width parameters without repetition or omission were calculated based on tool dimensions and process requirements. Finally, combining the distribution maps of high-curvature and flat areas, the division results of functional areas at the shoulder hook end, and the determined contact data, a region growth algorithm was used to automatically segment the wooden hanger surface into multiple processing zones. The steps were as follows: during algorithm initialization, representative triangular facets were selected as initial growth seed points within the shoulder, hook, and end areas based on the processing area type labels, ensuring that the seed points were distributed in typical positions within the high-curvature and flat areas; then, the region growth algorithm was defined. The merging criteria comprehensively consider the curvature continuity, surface normal angle, and spatial adjacency between candidate patches and the current growth region. Curvature continuity requires that the difference between the curvature distribution label value of the candidate patch and the average curvature value of the region is less than a preset threshold. Surface normal angle requires that the angle between the normal label of the candidate patch and the average normal of the region is less than the allowable angular deviation. Spatial adjacency requires that the candidate patch and the growth region are directly adjacent in the 3D mesh topology. The algorithm iterates from each seed point, continuously checking all its unassigned adjacent patches. If an adjacent patch simultaneously satisfies the above curvature continuity and normal consistency criteria... Based on the spatial adjacency merging criterion, the newly added facets are merged into the current growth region, and the region continues to expand outward with the newly added facets as the leading edge. This process is repeated until all facets have been visited and there are no more adjacent facets to be merged. After growth is completed, any small isolated regions that may be formed are merged into the adjacent main partitions according to their spatial location and geometric features. When the iteration ends, the algorithm outputs a series of continuous and uniform triangular facet sets. Each set corresponds to a processing partition. The partitions include the shoulder flat area suitable for high-speed and stable polishing, the hook high curvature area that requires precise trajectory control to conform to the curved surface, and the end curved area that needs to handle complex spatial orientation.

[0021] In one embodiment, step S3, which involves deploying a process sentinel network across multiple processing zones on the surface of the digital twin, the process sentinel network comprising process sentinels and virtual transmission channels, includes: S31. Process sentinels include status synchronization sentinels and collaborative monitoring sentinels; S32, where the state synchronization sentinel is used to mark the processing status of the robotic arm, and the collaborative monitoring sentinel is used to monitor safety risks; S33. Set up status synchronization sentinels at a preset density in the flat area on the shoulder of the wooden clothes hanger, and build a virtual transmission channel between the status synchronization sentinels to transmit the status data of the robotic arm. S34. Collaborative monitoring sentinels are set up at a preset density in the high curvature area of ​​the hook and the bending area at the end of the wooden clothes hanger, and a virtual transmission channel is constructed between the collaborative monitoring sentinels to transmit collaborative data of the robotic arm. S35. Configure the associated sentry list, safety radius of action, and safety threshold parameters for each collaborative monitoring sentry point to generate a process sentry point network; As described in steps S31-S35 above, two types of process sentinels are first defined: state synchronization sentinels are used to accurately mark the moment when the robotic arm enters or leaves a specific processing stage in the virtual simulation, while collaborative monitoring sentinels are used to monitor the potential interference and collision risks between the two robotic arms in real time. Then, based on the geometric characteristics and process importance of each processing zone, state synchronization sentinels are evenly distributed in the flat shoulder area at a preset density with equal spatial intervals. A virtual transmission channel is constructed between all state synchronization sentinels through a software-defined message queue. This channel can reliably transmit the processing status data of the robotic arm in real time during simulation, including the robotic arm number, the current processing stage identifier, and a precise timestamp. In the high-risk hook high-curvature area and end-bending area, based on the local curvature magnitude and space... Based on the reachability analysis results, collaborative monitoring outposts are adaptively deployed at a higher density, and independent virtual transmission channels are constructed between these outposts. These channels employ a lightweight communication protocol dedicated to transmitting real-time collaborative data of the robotic arm, such as end-effector position, velocity vector, attitude, and the expected time window for occupying the area. Finally, each collaborative monitoring outpost is configured with an associated outpost list containing unique identifiers of all other collaborative monitoring outposts logically related to it. A three-dimensional spherical safety radius and a set of dynamically adjustable safety threshold parameters are set for each outpost, including the minimum allowable Euclidean distance, maximum relative velocity, and maximum attitude deviation angle. This constructs a complete process outpost network with full coverage, real-time interaction, and risk quantification assessment capabilities.

[0022] In one embodiment, step S4, which generates coverage paths for each processing zone based on the workspace and task cycle of the dual robotic arms, assigns the coverage paths to the first and second robotic arms, and introduces relative pose corridor constraints to generate the initial joint trajectories of the two arms, includes: S41. Generate a sequence of processing path points covering the surface of each processing partition, as the coverage path for that partition; S42. Based on the workspace and joint range of motion of the first and second robotic arms, evaluate the accessibility of each robotic arm to different processing zones; S43. Based on the task cycle requirements, with the goal of balancing the processing load and minimizing the total processing time, allocate the reachable processing zones to the first robotic arm and the second robotic arm. S44. Assign a coverage path in the corresponding processing zone to each robotic arm, and generate the task path for the robotic arm by combining the contact data of the polishing tool. S45. Introduce relative pose corridor constraints on the task path of the two robotic arms. Based on the relative pose corridor constraints, perform inverse kinematics solutions on the task path of the two robotic arms respectively to generate the initial joint trajectories of the two arms. S46. Among them, the relative pose corridor constraint is used to define the minimum distance that must be maintained between the ends of the two arms, the maximum allowable relative attitude angle, and the maximum relative velocity. As described in steps S41-S46 above, firstly, based on the triangular mesh geometry and surface normal labels of each processing partition, a dense and uniformly distributed sequence of processing path points is generated using an equidistant offset or contour path planning algorithm. This sequence ensures that the polishing tool can completely cover the partition surface with a specified coverage width and overlap rate, forming the coverage path of that partition. Next, using the kinematic models of the first and second robotic arms, the reachable workspace point cloud of the robotic arm ends in the global coordinate system is calculated. Combined with the joint angle constraints and link dimensions of each robotic arm, it is determined whether the centroid or key feature points of each processing partition fall within the reachable point set of the corresponding robotic arm, thereby evaluating the reach of each robotic arm to different processing partitions. The reachability results are then assessed. Based on the overall task cycle time requirements, an optimal allocation model is constructed with the goal of minimizing the maximum completion time. This model comprehensively considers the estimated processing time of each partition, load balancing among robotic arms, and process sequence constraints between partitions. It uses heuristic algorithms or integer programming to solve the problem, rationally allocating reachable processing partitions to the first and second robotic arms. The steps involve constructing an optimal allocation model with the goal of minimizing the maximum completion time. This model first calculates the total length of the covered path for each processing partition and combines it with a preset polishing feed rate to estimate the processing time of each partition as the model input. Simultaneously, the decision variable is defined as a binary variable to represent whether a certain processing partition is accessible. Assigning tasks to a specific robotic arm, the model's objective function is to minimize the maximum time taken for the first and second robotic arms to complete all assigned processing tasks, i.e., minimizing the maximum completion time. The constraints considered by the model include: an allocation integrity constraint that each processing partition must be assigned to only one robotic arm; a load balancing constraint that the sum of the processing times of all partitions assigned to each robotic arm must not exceed the theoretical time limit calculated based on cycle time requirements; and a sequence constraint between processing partitions defined according to the process logic. For example, fine polishing of the high-curvature area of ​​the hook must begin only after the rough polishing of the flat shoulder area is completed. Such sequence constraints are defined in the model using equations or inequalities relating the timing relationships between partitions. The following is an expression of the problem. Finally, for this mixed integer programming model, heuristic methods such as genetic algorithms are used for efficient approximate solution. The design of the genetic algorithm includes encoding the allocation scheme into chromosomes, evaluating the maximum completion time and load balancing degree of the allocation scheme using a fitness function, and iteratively evolving the population through selection, crossover, and mutation operators, finally outputting the processing partition allocation result that meets the cycle time and process requirements. After allocating processing partitions to each robotic arm, the coverage paths of all partitions it is responsible for are connected in process order, and the correct tool posture is calculated for each path point based on the contact parameters of the polishing tool, especially the preset tilt angle between the tool axis and the surface normal, thereby generating the continuous task path of the robotic arm.Based on this task path, a relative pose corridor constraint is introduced. This constraint explicitly specifies the minimum three-dimensional Euclidean distance that the end effectors of the two robotic arms must maintain at any time, the maximum allowable relative attitude angle deviation between their tool coordinate systems, and the maximum relative linear velocity limit between the end effectors. Finally, under the premise of strictly satisfying the above relative pose corridor constraint, numerical inverse kinematics solutions are performed on the discrete task path point sequences of the first and second robotic arms respectively. One or more sets of feasible joint angles are calculated for each path point, and the optimal solution is selected based on the principle of joint motion smoothness. Then, through time parameterization, the spatial path is converted into a time-stamped sequence of joint positions and velocities, ultimately generating a synchronized initial joint trajectory of the two arms that meets the basic cooperative safety requirements.

[0023] In one embodiment, step S5, which applies the initial joint trajectories of both arms to the digital twin and predicts collision risks and generates conflict reports based on feedback from the process sentinel network, includes: S51. Load the initial joint trajectories of the first and second robotic arms in the simulation environment corresponding to the digital twin, and establish a unified simulation clock. S52, drive the simulation clock to run and calculate the position of the ends of the two arms and each joint link on the digital twin in real time; S53. When the robotic arm passes through the state synchronization sentinel, the state synchronization sentinel is triggered. The triggered state synchronization sentinel broadcasts the robotic arm's state update message to other state synchronization sentinels in the process sentinel network through the virtual transmission channel. S54. The status update message includes the trigger sentinel identifier, the trigger robotic arm identifier, the type of processing status triggered, and the precise simulation time. S55. When the robotic arm passes through the collaborative monitoring sentry point, the collaborative monitoring sentry point is triggered. The triggered collaborative monitoring sentry point sends collaborative data to all collaborative monitoring sentry points in the associated sentry point list through the virtual transmission channel. S56. The collaborative data includes the identifier of the triggering sentinel, the identifier of the triggering robotic arm, the triggering time, and the estimated occupation duration; S57. After receiving collaborative data, the collaborative monitoring sentinels in the associated sentinel list predict the spatiotemporal proximity of the robotic arm and another robotic arm within their own monitoring area based on the robotic arm trajectory information. S58. Compare the predicted spatiotemporal proximity state with the configured safety threshold. If the spatiotemporal proximity state exceeds the safety threshold, it is determined that there is a collision risk and a conflict report is generated. S59. The conflict report includes the estimated time of occurrence of the conflict, the identification of the robotic arm involved in the conflict, and the location of the collaborative monitoring sentinel associated with the conflict. As described in steps S51-S59 above, firstly, the initial joint trajectory data of the first and second robotic arms are loaded into the simulation engine corresponding to the digital twin, and a unified simulation clock is initialized as a virtual time reference. Then, the simulation clock is driven to iterate at fixed steps. Within each simulation step, the precise position and orientation of all joints, links, and end effectors of both arms in the three-dimensional space of the digital twin are calculated in real time using forward kinematics. When the simulation detects that the end effector of a robotic arm enters the physical proximity range of any state synchronization sentinel, the state synchronization sentinel is triggered. The triggering event immediately drives the sentinel to broadcast a state update message to all other state synchronization sentinels in the process sentinel network through its pre-built virtual transmission channel. This message encapsulates the unique identifier of the triggering sentinel, the number of the triggering robotic arm, the processing state type it represents, and the simulation time accurate to milliseconds. Simultaneously, when the end effector of a robotic arm enters the preset safe operating radius of any collaborative monitoring sentinel, the collaborative monitoring sentinel is triggered. The triggering event drives it to broadcast a state update message to its associated sentinel list through its dedicated virtual transmission channel. All defined associated collaborative monitoring outposts send collaborative data, which includes the triggering outpost identifier, the triggering robotic arm identifier, the triggering time, and the estimated duration of occupation of the monitoring area based on the current robotic arm speed. Upon receiving such collaborative data, any collaborative monitoring outpost in the associated outpost list immediately queries the trajectory information of the other robotic arm stored in the digital twin. It then uses linear interpolation or motion extrapolation algorithms to predict the spatiotemporal proximity of the two robotic arm trajectories within its own monitoring area in a future time window, including the minimum predicted distance and the precise time when the closest possible state may occur. Subsequently, the collaborative monitoring outpost compares the calculated spatiotemporal proximity parameters, such as the minimum predicted distance, relative speed, and attitude angle, with its own configured safety threshold parameters. If the minimum predicted distance is lower than the safety distance threshold or the relative speed exceeds the safety speed threshold, a potential collision risk is identified, and a structured conflict report is automatically generated. This report details the estimated time of the conflict, the identifiers of the first and second robotic arms involved, and the three-dimensional spatial coordinates of the collaborative monitoring outpost that serves as the starting point of the risk source.

[0024] In one embodiment, step S6, which iteratively optimizes the initial joint trajectories of the two arms based on conflict reports until no conflicts occur during collaborative pre-simulation, generates an optimized trajectory, and performs multi-level collision detection and motion smoothness verification on the optimized trajectory to generate the robotic arm's working trajectory, includes: S61. Based on the conflict report, formulate a trajectory optimization strategy and modify the initial joint trajectory. Perform collaborative pre-simulation and conflict prediction again. Iterate this process until no conflict occurs to obtain the optimized trajectory. S62. The trajectory optimization strategy includes using velocity scaling on the time axis to address conflicts caused by overlapping time windows, using a local path replanning strategy to address conflicts caused by excessively close spatial paths, and using a velocity curve correction strategy to address conflicts caused by relative attitude or velocity not meeting constraints. S63. Perform collision verification and motion smoothness verification on the optimized trajectory, encode the verified optimized trajectory into an executable file, and generate the working trajectory of the robotic arm. As described in steps S61-S63 above, firstly, based on the conflict type, estimated occurrence time, and specific robotic arm and collaborative monitoring sentinel locations recorded in the conflict report, a corresponding trajectory optimization strategy is intelligently formulated, and the initial joint trajectories of the two arms are modified accordingly. Specific strategies include: when the conflict report analysis shows overlapping time windows (i.e., both robotic arms are expected to enter the same high-risk area within the same time period), the system uses a speed scaling strategy on the time axis to proportionally adjust the movement speed of the relevant robotic arms at trajectory points before and after the conflict period to change their arrival time and thus achieve peak shifting; when the conflict report analysis shows that the spatial path is too short (i.e., the predicted minimum distance is lower than the safe distance), the system optimizes the trajectory by adjusting the movement speed of the relevant robotic arms at trajectory points before and after the conflict period to change their arrival time. At full threshold, the system employs a local path replanning strategy to define a local spatial window centered on the conflict point in the digital twin. Within this window, a smooth local path segment avoiding the interference region is regenerated based on random sampling or the potential field method, ensuring smooth connection with the original trajectory segments before and after it in terms of position and velocity. When the conflict report analysis shows that the relative attitude angle exceeds the limit or the relative velocity is too large, the system uses a velocity curve correction strategy to refit and smooth the joint velocity or end effector linear velocity curves of the relevant robotic arm based on polynomials or spline functions to reduce the peak velocity or adjust the motion phase. After modifying the joint trajectory using the selected strategy, the system automatically re-... Initiating the entire collaborative pre-simulation process involves re-executing a new round of conflict prediction. This iterative optimization loop continues until no new conflict reports are generated during a complete collaborative pre-simulation. The joint trajectories obtained at this point are marked as optimized trajectories. Subsequently, rigorous collision verification is performed on the optimized trajectories. First, a rapid global interference check is conducted using the bounding box models of each link of the robotic arm, the digital twin, and a simplified model of another robotic arm to eliminate obvious collision risks. Then, any potentially dangerous sections marked in the rapid check are switched to high-precision mode, using a complete triangular mesh model of the robotic arm and the workpiece to perform precise distance calculations to ensure no geometric penetration. Simultaneously, motion is performed. Smoothness verification ensures the absence of abrupt changes by calculating the continuity of the first and second derivatives of velocity and acceleration in the joint space of the optimized trajectory. It also verifies the smoothness of tool posture changes at key process connection points in the operating space, especially at the boundaries of partitions. Furthermore, it verifies whether the actual occurrence time difference of key collaborative events marked by state synchronization sentinels in the dual-arm trajectory is less than the preset synchronization error threshold. Finally, the optimized trajectory that has passed all collision and smoothness verifications, along with its precise timestamps, joint positions, velocities, and acceleration information, is encoded into an executable file in a standard format such as binary or a specific text format. The executable file is then used as the working trajectory of the robotic arm.

[0025] Example 2, please refer to Figure 2 As shown in this embodiment, a robotic arm trajectory generation system for machining freeform surfaces of wooden coat hangers includes: The digital twin construction module acquires surface point cloud data of a wooden clothes hanger through 3D scanning, constructs a 3D mesh model of the wooden clothes hanger, adds process semantic tags, and generates a digital twin of the wooden clothes hanger. The zoning planning module, connected to the digital twin construction module, divides the wooden clothes hanger into multiple processing zones based on the geometric data and process semantic tags of the digital twin. The sentinel network deployment module is connected to the zoning planning module to deploy a process sentinel network on multiple processing zones on the surface of the digital twin. The process sentinel network includes process sentinels and virtual transmission channels. The trajectory initialization module is connected to the sentinel network deployment module. Based on the workspace and task cycle of the two robotic arms, it generates the coverage path of each processing zone, assigns the coverage path to the first robotic arm and the second robotic arm, and introduces relative pose corridor constraints to generate the initial joint trajectory of the two arms. The collaborative pre-simulation module, connected to the trajectory initialization module, applies the initial joint trajectories of the two arms to the digital twin. Based on the feedback from the process sentinel network, it predicts collision risks and generates conflict reports. The trajectory optimization generation module is connected to the collaborative pre-simulation module. It iteratively optimizes the initial joint trajectory of the two arms based on the conflict report until no conflict occurs in the collaborative pre-simulation, generates the optimized trajectory, and performs collision detection and motion smoothness verification on the optimized trajectory to generate the working trajectory of the robotic arm.

[0026] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for generating the trajectory of a robotic arm for machining the freeform surface of a wooden clothes hanger, characterized in that, The surface point cloud data of the wooden clothes hanger is obtained by 3D scanning, a 3D mesh model of the wooden clothes hanger is constructed and process semantic tags are added to generate a digital twin of the wooden clothes hanger; Based on the geometric data of the digital twin and the semantic tags of the process, the wooden clothes hanger is divided into multiple processing zones; A process sentinel network is deployed across multiple processing zones on the surface of the digital twin. The process sentinel network includes process sentinels and virtual transmission channels. Based on the workspace and task cycle of the dual robotic arms, the coverage path of each processing zone is generated, the coverage path is assigned to the first robotic arm and the second robotic arm, and the relative pose corridor constraint is introduced to generate the initial joint trajectory of the dual arms. The initial joint trajectories of the two arms are applied to the digital twin, and based on the feedback from the process sentinel network, collision risks are predicted and conflict reports are generated. Based on the conflict report, the initial joint trajectory of the two arms is iteratively optimized until no conflict occurs in the collaborative pre-simulation, generating an optimized trajectory. The optimized trajectory is then subjected to collision detection and motion smoothness verification to generate the working trajectory of the robotic arm.

2. The method for generating the trajectory of a robotic arm for machining the free-form surface of a wooden clothes hanger according to claim 1, characterized in that, The steps for obtaining surface point cloud data of the wooden clothes hanger through 3D scanning, constructing a 3D mesh model of the wooden clothes hanger, adding process semantic tags, and generating a digital twin of the wooden clothes hanger are as follows: The surface of the wooden clothes hanger was scanned using a structured light scanner to obtain the surface point cloud data of the wooden clothes hanger; A 3D surface reconstruction was performed based on surface point cloud data to generate a 3D mesh model of the wooden clothes hanger. Process semantic tags are annotated on a 3D mesh model to generate a digital twin of the wooden clothes hanger; Process semantic tags include processing area type tags, curvature distribution tags, and surface normal tags.

3. The method for generating the trajectory of a robotic arm for machining a free-form surface of a wooden clothes hanger according to claim 1, characterized in that, The step of dividing the wooden clothes hanger into multiple processing zones based on the geometric data and process semantic tags of the digital twin is as follows: Based on the curvature distribution tags in the process semantic tags, identify the high curvature areas and flat areas on the surface of the wooden clothes hanger; Based on the processing area type label in the process semantic label, distinguish the shoulder area, hook area and end area of ​​the wooden clothes hanger; By combining the geometric data of the digital twin with the surface normal labels, the contact data of the polishing tool is determined. The contact data includes contact parameters and coverage width parameters. Based on the distribution of high curvature and flat areas, the division of shoulder area, hook area and end area, and contact data, the surface of the wooden clothes hanger is divided into multiple processing zones; The processing zones include the flat shoulder area, the high curvature hook area, and the curved end area.

4. The method for generating the trajectory of a robotic arm for machining a free-form surface of a wooden clothes hanger according to claim 1, characterized in that, The step of deploying a process sentinel network across multiple processing zones on the surface of the digital twin, wherein the process sentinel network includes process sentinels and virtual transmission channels, is as follows: Process sentinels include status synchronization sentinels and collaborative monitoring sentinels; Among them, the status synchronization sentinel is used to mark the processing status of the robotic arm, and the collaborative monitoring sentinel is used to monitor safety risks; Status synchronization sentinels are set up at a preset density in the flat area on the shoulder of the wooden clothes hanger, and a virtual transmission channel is built between the status synchronization sentinels to transmit the status data of the robotic arm. Collaborative monitoring sentinels are deployed at a preset density in the high curvature area of ​​the hooks and the bending area at the ends of the wooden clothes hanger, and a virtual transmission channel is constructed between the collaborative monitoring sentinels to transmit collaborative data of the robotic arm. Configure a list of associated sentinels, a safe radius of action, and a safe threshold parameter for each collaborative monitoring sentinel to generate a process sentinel network.

5. The method for generating the trajectory of a robotic arm for machining a free-form surface of a wooden clothes hanger according to claim 1, characterized in that, The steps for generating coverage paths for each processing zone based on the workspace and task cycle of the dual robotic arms, assigning the coverage paths to the first and second robotic arms, and introducing relative pose corridor constraints to generate the initial joint trajectories of the two arms are as follows: Generate a sequence of processing path points covering the surface of each processing partition, which serves as the coverage path for that partition; Based on the workspace and joint range of motion of the first and second robotic arms, the accessibility of each robotic arm to different processing zones is evaluated. Based on the task cycle requirements, with the goal of balancing the processing load and minimizing the total processing time, the reachable processing zones are allocated to the first robotic arm and the second robotic arm. Assign a coverage path in the corresponding processing zone to each robotic arm, and generate the task path for that robotic arm by combining the contact data of the polishing tool; A relative pose corridor constraint is introduced on the task path of the two robotic arms. Based on the relative pose corridor constraint, the inverse kinematics solution is performed on the task path of the two robotic arms to generate the initial joint trajectory of the two arms. The relative pose corridor constraint is used to define the minimum distance that must be maintained between the ends of the two arms, the maximum allowable relative attitude angle, and the maximum relative velocity.

6. The method for generating the trajectory of a robotic arm for machining a free-form surface of a wooden clothes hanger according to claim 1, characterized in that, The steps of applying the initial joint trajectories of both arms to the digital twin, predicting collision risks and generating conflict reports based on feedback from the process sentinel network are as follows: Load the initial joint trajectories of the first and second robotic arms in the simulation environment corresponding to the digital twin, and establish a unified simulation clock; Drive the simulation clock to run and calculate the positions of the ends of the two arms and each joint link on the digital twin in real time; When the robotic arm passes a state synchronization sentinel, the state synchronization sentinel is triggered. The triggered state synchronization sentinel broadcasts the robotic arm's state update message to other state synchronization sentinels in the process sentinel network through a virtual transmission channel. The status update message includes the trigger sentinel identifier, the trigger robot arm identifier, the trigger processing status type, and the precise simulation time; When the robotic arm passes by the collaborative monitoring outpost, the collaborative monitoring outpost is triggered. The triggered collaborative monitoring outpost sends collaborative data to all collaborative monitoring outposts in the associated outpost list through the virtual transmission channel. The collaborative data includes the identifier of the triggering sentinel, the identifier of the triggering robotic arm, the triggering time, and the estimated duration of occupation; After receiving collaborative data, the collaborative monitoring sentinels in the associated sentinel list predict the spatiotemporal proximity of the robotic arm and another robotic arm within their own monitoring area based on the robotic arm trajectory information. The predicted spatiotemporal proximity state is compared with its own configured safety threshold. If the spatiotemporal proximity state exceeds the safety threshold, a collision risk is determined and a conflict report is generated. The conflict report includes the estimated time of the conflict, the identifier of the robotic arm involved in the conflict, and the location of the collaborative monitoring outpost associated with the conflict.

7. The method for generating the trajectory of a robotic arm for machining a free-form surface of a wooden clothes hanger according to claim 1, characterized in that, The steps for iteratively optimizing the initial joint trajectories of the two arms based on the conflict report until no conflicts occur during the collaborative pre-simulation, generating an optimized trajectory, and performing multi-level collision detection and motion smoothness verification on the optimized trajectory to generate the robotic arm's working trajectory are as follows: Based on the conflict report, a trajectory optimization strategy is formulated and the initial joint trajectory is modified. Collaborative pre-simulation and conflict prediction are performed again. This process is iterated until no conflict occurs, and the optimized trajectory is obtained. The trajectory optimization strategies include using velocity scaling on the time axis to address conflicts caused by overlapping time windows, using a local path replanning strategy to address conflicts caused by excessively close spatial paths, and using a velocity curve correction strategy to address conflicts caused by relative attitude or velocity not meeting constraints. Collision verification and motion smoothness verification are performed on the optimized trajectory. The optimized trajectory that passes the verification is encoded into an executable file to generate the working trajectory of the robotic arm.

8. A robotic arm trajectory generation system for machining freeform surfaces of wooden clothes hangers, used to implement the robotic arm trajectory generation method for machining freeform surfaces of wooden clothes hangers as described in any one of claims 1-7, characterized in that, The digital twin construction module acquires surface point cloud data of a wooden clothes hanger through 3D scanning, constructs a 3D mesh model of the wooden clothes hanger, adds process semantic tags, and generates a digital twin of the wooden clothes hanger. The zoning planning module, connected to the digital twin construction module, divides the wooden clothes hanger into multiple processing zones based on the geometric data and process semantic tags of the digital twin. The sentinel network deployment module is connected to the zoning planning module to deploy a process sentinel network on multiple processing zones on the surface of the digital twin. The process sentinel network includes process sentinels and virtual transmission channels. The trajectory initialization module is connected to the sentinel network deployment module. Based on the workspace and task cycle of the two robotic arms, it generates the coverage path of each processing zone, assigns the coverage path to the first robotic arm and the second robotic arm, and introduces relative pose corridor constraints to generate the initial joint trajectory of the two arms. The collaborative pre-simulation module, connected to the trajectory initialization module, applies the initial joint trajectories of the two arms to the digital twin. Based on the feedback from the process sentinel network, it predicts collision risks and generates conflict reports. The trajectory optimization generation module is connected to the collaborative pre-simulation module. It iteratively optimizes the initial joint trajectory of the two arms based on the conflict report until no conflict occurs in the collaborative pre-simulation, generates the optimized trajectory, and performs collision detection and motion smoothness verification on the optimized trajectory to generate the working trajectory of the robotic arm.

Images