A method, system, product, and medium for producing a precast component

Abstract

A prefabricated component production method, system, product and medium, wherein the method comprises: obtaining geometric data and process data of a prefabricated component to be processed and spatial point cloud data of a target processing workshop; generating a production task based on the geometric data and the process data, and dividing the production task to obtain a plurality of subtasks; obtaining real-time state information of a first execution unit and a second execution unit; determining the task type of each subtask based on the production stage, matching an end effector, and associating and configuring the end effector with the first execution unit to obtain a target first execution unit; determining an execution combination of each subtask; determining a movement path of each target first execution unit; and controlling the target first execution unit to travel to a preset work position of the second execution unit in the execution combination according to the movement path, and execute each subtask according to work parameters. The application can improve the production efficiency of prefabricated components.

Description

Technical Field

[0001] This application relates to the field of component production technology, specifically to a method, system, product, and medium for producing prefabricated components. Background Technology

[0002] In the field of modern industrialized construction, precast component production has become an important means to improve construction efficiency and ensure project quality. The production process of precast components typically involves the coordinated operation of multiple procedures, including mold assembly, rebar laying, concrete pouring, curing, and demolding. These procedures need to be completed by different work units within the production workshop according to specific technological processes. With the increasing complexity of building structures and the diversification of component types, the requirements for the coordination, flexibility, and adaptability of work units in precast component production are becoming increasingly stringent.

[0003] In the existing technology, the control method for the production of precast components mainly adopts a combination of centralized scheduling and workstation instructions: the production management system plans the work process in advance according to the order requirements, generates a work task list corresponding to each workstation, and sends the instructions to the fixed workstation control system through fieldbus or wireless network; at the same time, it plans fixed travel paths and work time nodes for mobile work equipment, so that it can arrive at the designated workstation to perform tasks according to the predetermined schedule, and then move to the next workstation after completion.

[0004] However, due to the characteristics of precast component production tasks being multi-variety and small-batch, the production processes and quality requirements of different components vary significantly. In the existing technology, mobile operation equipment is usually only equipped with single-function execution tools. When production tasks are switched and process requirements change, the equipment is difficult to adapt to the operation requirements of the new task quickly. Manual intervention is often required to change tools or adjust equipment configuration, which leads to production process interruptions and increased waiting time, thus reducing the production efficiency of precast components. Summary of the Invention

[0005] This application provides a method, system, product, and medium for producing precast components, which can improve the production efficiency of precast components.

[0006] The first aspect of this application provides a method for producing precast components, the method comprising: Obtain geometric and process data of the prefabricated components to be processed, as well as spatial point cloud data of the target processing workshop. The process data includes production stages and quality requirements. Production tasks are generated based on the geometric data and process data, and the production tasks are divided into multiple sub-tasks, each of which corresponds to a work parameter. The real-time status information of the first execution unit and the second execution unit is obtained. The first execution unit is a mobile operation robot equipped with a replaceable end effector, and the second execution unit is a fixed work station. Based on the production stage, the task type of each sub-task is determined, and the end effector of the first execution unit that matches the sub-task is determined according to the task type. The end effector is then associated with the first execution unit to obtain the target first execution unit. The execution combination of each subtask is determined based on the job parameters and the real-time status information, and the execution combination includes at least one target first execution unit and one second execution unit; Based on the spatial point cloud data and the operation parameters, the movement path of each target first execution unit is determined; The target first execution unit is controlled to travel along the movement path to the preset work position of the second execution unit in the execution combination, and each of the sub-tasks is executed according to the work parameters.

[0007] Optionally, the production task is divided into multiple target tasks, and operation parameters for each sub-task are generated, specifically including: The production task is divided into multiple target tasks according to the production stage; Based on the geometric data, construct the intermediate geometric model of the prefabricated component to be processed at each of the production stages; Based on the production stage, environmental feature evolution data before and after the execution of each target task are determined. The environmental feature evolution data is used to characterize the point cloud feature changes of the fixed workstation due to the increase in components. The environmental feature evolution data is used as the operation parameter and associated with the corresponding target task to obtain each sub-task containing the operation parameter.

[0008] Optionally, determining the end effector of the first execution unit matching the subtask based on the task type specifically includes: Select a set of candidate end effectors from the preset tool library that meet the tool function parameters corresponding to the task type; Obtain the geometric envelope data of each candidate end effector in the candidate end effector set, and combine it with the sensor installation position of the first execution unit to construct the self-occlusion blind zone model of the candidate end effector in the mounted state; Based on the spatial point cloud data, the spatial distribution characteristics of obstacles within a preset range of each of the second execution units are identified; Calculate the perception conflict index between each self-occlusion blind spot model and the corresponding second execution unit for the spatial distribution characteristics of obstacles; The candidate end effector with the smallest perceived conflict index is determined as the end effector.

[0009] Optionally, the perceptual conflict index between each self-occlusion blind spot model and the corresponding obstacle spatial distribution features of the second execution unit is calculated, specifically including: The self-occlusion blind spot model is projected onto the target space point cloud data within the preset range of the second execution unit to obtain the blind spot coverage space; Identify target location feature points and physical obstacles located within the blind zone coverage space in the spatial distribution characteristics of the obstacles; Calculate the feature loss rate of the target location feature points within the blind zone coverage space, and the volume ratio of the physical obstacle within the blind zone coverage space; The perceptual conflict index is obtained by weighting and summing the feature loss rate and the volume ratio.

[0010] Optionally, the execution combination of each subtask is determined based on the job parameters and the real-time status information, specifically including: Based on the job parameters and the real-time status information, a set of candidate second execution units that meet the job conditions of the sub-task is selected. Construct the dynamic job envelope of the target first execution unit, and extract the neighborhood dynamic point cloud data of each of the candidate second execution units in the set of candidate second execution units within the target range from the spatial point cloud data; The dynamic operation envelope is placed into the neighborhood dynamic point cloud data of each of the candidate second execution units for virtual spatiotemporal interference verification; The alternative second execution unit that has passed the virtual spatiotemporal interference verification is combined with the target first execution unit to form the execution combination.

[0011] Optionally, the dynamic job envelope is placed into the neighborhood dynamic point cloud data of each of the candidate second execution units for virtual spatiotemporal interference verification, specifically including: Based on the operation parameters of the sub-task, the motion trajectory of each joint of the target first execution unit during the execution of the sub-task is analyzed, and each joint motion trajectory is mapped to a space occupancy set that changes over time. Dynamic feature points are extracted from the neighborhood dynamic point cloud data, and the predicted motion trajectory of the dynamic feature points within the execution cycle of the subtask is analyzed based on the displacement vector of the continuous frame point cloud of the neighborhood dynamic point cloud data. The spatial occupancy set is spatiotemporally overlapped with the predicted motion trajectory and the static obstacle point cloud in the neighborhood dynamic point cloud data, and the minimum safe distance is calculated under different preset time steps. If the minimum safety distance is greater than the preset safety threshold during the execution cycle of the subtask, then the virtual spatiotemporal interference verification is passed.

[0012] Optionally, based on the spatial point cloud data and the operation parameters, the movement path of each target first execution unit is determined, specifically including: Based on the spatial point cloud data, the target processing workshop is rasterized to generate a global navigation map containing obstacle information and passable areas. Determine the work stop point of the target first execution unit at the second execution unit based on the work parameters; Obtain the current position of the target first execution unit and map the geometric dimensions of the target first execution unit to a motion collision envelope; In the global navigation map, a collision-free trajectory from the current position to the work stop point is generated based on the mobile collision envelope, thus obtaining the mobile path.

[0013] In a second aspect, embodiments of this application provide a prefabricated component production system, the prefabricated component production system comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the prefabricated component production system to perform the method as described in the first aspect and any possible implementation thereof.

[0014] Thirdly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a prefabricated component production system, cause the prefabricated component production system to perform the method described in the first aspect and any possible implementation thereof.

[0015] Fourthly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a prefabricated component production system, cause the prefabricated component production system to perform the method described in the first aspect and any possible implementation thereof.

[0016] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: 1. By acquiring the geometric and process data of the prefabricated components to be processed, the operating parameters for each sub-task can be accurately generated according to the production stage and quality requirements. Based on this, by combining the real-time status information of the first execution unit (a mobile robot equipped with a replaceable end effector) and the second execution unit (a fixed workstation), the corresponding end effector is dynamically matched and configured according to the sub-task type determined by the production stage. This allows the mobile robot to quickly adapt to the operational needs of different processes, avoiding production interruptions caused by manual tool changes. Furthermore, based on the operating parameters and real-time status information, this invention determines the execution combination of each sub-task (including at least one target first execution unit and one second execution unit), and plans the movement path of the target first execution unit by combining spatial point cloud data and operating parameters, controlling it to travel along the path to the preset working position of the second execution unit in the execution combination to execute the sub-task. Through the above mechanism, this invention achieves dynamic collaboration and flexible adaptation between the mobile robot and the fixed workstation, enabling the same mobile robot to undertake different process tasks through rapid replacement of the end effector, thus improving the production efficiency of prefabricated components.

[0017] 2. When dividing the production task into multiple sub-tasks and generating operation parameters, a dynamic association mechanism based on intermediate geometric models and environmental feature evolution data is further introduced: After dividing the production task into multiple sub-tasks according to the production stage, an intermediate geometric model of the prefabricated component to be processed is constructed based on geometric data for each production stage. This allows for the determination of environmental feature evolution data before and after the execution of each sub-task. This data is used to characterize the point cloud feature changes at fixed workstations due to component increments and is then associated with the corresponding sub-tasks as operation parameters. Through this mechanism, the mobile robot can anticipate environmental changes such as the shift of positioning feature points and changes in obstacle distribution caused by component morphology evolution. This enables proactive adaptation in path planning and perception strategies, effectively avoiding the failure of preset operation parameters and the robot's inability to accurately locate or perform operations due to environmental evolution. This improves the environmental adaptability of operation parameters and the reliability of task execution during multi-process continuous production.

[0018] 3. When matching end effectors according to task type, a perception conflict index screening mechanism based on self-occlusion blind zone model and obstacle spatial distribution characteristics is further introduced: First, a set of candidate end effectors that meet the tool function parameters is selected from the preset tool library. Then, a self-occlusion blind zone model of each candidate end effector under the mounted state is constructed in combination with the sensor installation position of the first execution unit. At the same time, the spatial distribution characteristics of obstacles around the second execution unit are identified based on spatial point cloud data. By calculating the perception conflict index between the self-occlusion blind zone model and the obstacle spatial distribution characteristics, that is, by weighted summing the target positioning feature point loss rate and the physical obstacle volume ratio in the blind zone coverage space, the candidate end effector with the smallest perception conflict index is determined as the actual mounted end effector. This mechanism enables proactive assessment of the blind spot impact on workstation environment perception after the end effector is mounted during the switching of production tasks for multiple types of components. It selects the end effector with the least perception occlusion and the best field of view for mounting, effectively avoiding the problem of the robot being unable to identify positioning feature points or avoid obstacles after arriving at the workstation due to improper selection of end effectors. It significantly improves the accessibility, positioning accuracy and operational safety in complex workshop environments, and provides reliable perception guarantee for flexible collaborative production of multiple processes. Attached Figure Description

[0019] Figure 1 This is a schematic flowchart of a method for producing prefabricated components according to an embodiment of this application; Figure 2 This is a schematic diagram of the process of matching the end effector in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a prefabricated component production system provided in an embodiment of this application.

[0020] Explanation of reference numerals in the attached drawings: 301, Central Processing Unit; 302, Read-Only Memory; 303, Random Access Memory; 304, Bus; 305, Input / Output Interface; 306, Input Section; 307, Output Section; 308, Storage Section; 309, Communication Section; 310, Driver; 311, Removable Media. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0022] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0023] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0024] Figure 1 This is a schematic flowchart of a method for producing prefabricated components according to an embodiment of this application.

[0025] S101. Obtain the geometric data and process data of the prefabricated component to be processed, as well as the spatial point cloud data of the target processing workshop. The process data includes the production stage and quality requirements. The system acquires the geometric data of the precast components to be processed through a Building Information Modeling (BIM) interface. This geometric data primarily originates from the digital model at the design end and includes the three-dimensional outline dimensions of the precast components, the topology of the internal steel reinforcement mesh, and the precise coordinates of the embedded connectors. This geometric data provides an absolute spatial reference for subsequent processing actions. Simultaneously, the system retrieves associated process data, which is a structured description of the production logic. This data includes different production stages for the precast components, from mold cleaning and steel reinforcement laying to concrete pouring, as well as the corresponding dimensional tolerances, density levels, and other quality requirements for each stage. To dynamically align the offline digital design information with the actual physical production scene, the system utilizes a multi-threaded laser scanner or a robot-mounted depth camera positioned at the top of the target processing workshop to acquire spatial point cloud data of the workshop.

[0026] Spatial point cloud data refers to a collection of massive spatial coordinate points reflecting the three-dimensional features of an object's surface, obtained through laser ranging or speckle structured light principles. This data can capture in real-time the geometric changes of fixed workstations within a target processing workshop, the distribution of obstacles on the ground, and the stacking patterns of unstructured materials in the environment. By acquiring spatial point cloud data, the system can identify physical constraints not reflected in the static model of the target processing workshop, thus providing realistic environmental envelope constraints for subsequent robot path planning. Combining the geometric data of the prefabricated components to be processed with process data can automatically derive the operational sequence and accuracy criteria for the prefabricated components. Introducing spatial point cloud data provides a digital twin scenario with physical safety guarantees for the execution of these operational sequences. This multi-source data fusion processing solves the problem of the disconnect between process documents and the actual on-site environment in traditional production, ensuring that the generated sub-tasks both conform to design specifications and are adapted to the real-time operating conditions of the target processing workshop.

[0027] S102. Generate a production task based on the geometric data and process data, and divide the production task into multiple sub-tasks, each of which corresponds to a work parameter. To translate the design intent from the front end into executable overall instructions on the production floor, the system generates production tasks based on the aforementioned geometric and process data. Here, "geometric data" typically originates from Building Information Modeling (BIM) or Computer-Aided Design (CAD) files, containing physical spatial attributes such as the three-dimensional dimensions, shape contours, and internal topology of the prefabricated components to be processed, clarifying "what to do." "Process data," on the other hand, refers to pre-defined production stages (such as the sequence of processes like formwork installation, reinforcement, and casting) and quality requirements (such as dimensional tolerances and surface flatness standards), derived from the Manufacturing Execution System (MES) or process planning documents, clarifying "how to do it" and "what standards to achieve." The reason for combining these two is that a single geometric model cannot guide the processing sequence of mobile robots, while a purely technical process lacks specific spatial coordinates and physical boundaries. Only by deeply integrating physical form with processing logic can a complete guiding basis be provided for automated production. In practical implementation, the system extracts feature points, lines, and surfaces from the geometric data through a data parsing module. Using a data fusion algorithm, it maps and binds these spatial features to the production stages and quality requirements specified in the process data, generating a structured, comprehensive digital work order—a "production task"—containing the entire lifecycle manufacturing information of the component. This approach breaks down the data barriers between design and manufacturing, establishing a unique and accurate data foundation for the entire automated processing, ensuring a unified data source for subsequent robot operations. For example, when producing a precast concrete column, the system fuses the column's length, width, height, and the three-dimensional coordinates of the embedded parts (geometric data) with requirements such as "reinforcing steel bars must be tied before formwork erection" and "the positional error of the embedded parts does not exceed 2 millimeters" (process data), generating a comprehensive production task work order covering all spatial and process requirements from material preparation to final forming of the column.

[0028] After generating the overall production task based on geometric and process data, in order to achieve refined scheduling of complex production processes and precise execution of robots, the macroscopic production task needs to be further broken down into multiple specific target tasks, and parameters that can guide actual operations need to be generated for each task. This process not only relies on the reasonable division of tasks according to production stages, but also requires a deep integration of the morphological evolution of prefabricated components during processing and the resulting dynamic changes in the characteristics of the workstation environment, thereby transforming these physical and environmental evolutions into specific operational parameters. Specifically, step S102 can be implemented through the following steps: dividing the production task into multiple target tasks according to the production stages; constructing intermediate geometric models of the prefabricated components to be processed in each of the production stages based on the geometric data; determining the environmental feature evolution data before and after the execution of each target task according to the production stages, wherein the environmental feature evolution data is used to characterize the point cloud feature changes of the fixed workstation due to the increase in components; and associating the environmental feature evolution data as operational parameters with the corresponding target tasks to obtain each sub-task containing the operational parameters.

[0029] After acquiring the overall production task, in order to translate the macro-level production instructions into specific operational units that mobile robots can understand and execute, the system divides the production task into multiple target tasks based on the production stages. Here, the production stages refer to the standardized processing steps preset in the process data, arranged according to time sequence and process dependencies. In practice, the system analyzes the process flow chart or time node table in the process data to slice the massive production task covering the entire precast component manufacturing cycle, breaking it down into independent task modules corresponding to each specific process—the target tasks. This decouples the complex precast component production process into a series of logically clear and well-defined execution units, providing a foundation for subsequent resource allocation and robot scheduling. For example, when producing a precast concrete wall panel, the macro-level production task is divided into multiple progressively advancing target tasks such as "bottom formwork cleaning," "reinforcing mesh placement," "side formwork assembly," and "concrete pouring."

[0030] After clarifying the target tasks for each stage, in order to enable the mobile robot to accurately perceive the actual physical form of the work objects it faces at different stages, the system constructs intermediate geometric models of the prefabricated components to be processed at each of the production stages based on the geometric data. The intermediate geometric model refers to the three-dimensional digital representation of the prefabricated component at a specific process node during its evolution from raw materials to the final product. It is generated by a 3D modeling engine through reverse decomposition or forward layer-by-layer calculation of the final component's geometric data. Specifically, the system combines the process requirements of each production stage and uses algorithms such as 3D Boolean operations to peel off or add features from the final geometric data according to the process progress, thereby generating a 3D model specific to each stage. This approach provides the robot with highly accurate local spatial physical boundary references, effectively avoiding interference and collisions between the robot and components that are not yet formed or partially formed when performing tasks. For example, when performing the target task of "side mold assembly," the intermediate geometric model constructed by the system only includes the bottom mold and the placed steel reinforcement frame, but not the unpoured concrete, thus guiding the robot to accurately identify the currently operable space area.

[0031] Based on the intermediate geometric models constructed at each stage, the system can further determine the environmental feature evolution data before and after the execution of each target task according to the production stage. This environmental feature evolution data is used to characterize the point cloud feature changes of the fixed workstation due to component increments. Here, the environmental feature evolution data is a dynamic spatial difference index. Its principle is to compare the intermediate geometric models of two adjacent production stages, calculate the changes in the local space occupied by entities at the workstation due to the execution of the current target task (such as material addition, component assembly, etc.), and convert this into a difference set in point cloud data format. In specific implementation, the system performs spatial difference calculations between the intermediate model after task execution and the model before execution, extracts the surface contours of the newly added entities, and maps them to the spatial point cloud data coordinate system of the target processing workshop, forming the point cloud regions and feature points expected to change. The reason for extracting this evolution data is that mobile robots rely heavily on point clouds for navigation and obstacle avoidance when moving and operating around fixed workstations. Predicting and quantifying these environmental changes caused by production in advance can prevent the robot from losing its positioning or colliding due to sudden changes in environmental features. For example, before and after the "reinforcing mesh placement" task is performed, the environmental feature evolution data will accurately identify that the originally empty area above the fixed work station will have a new grid-like point cloud obstacle feature.

[0032] After obtaining accurate predictions of dynamic environmental changes, to enable the mobile robot to perform operations with prior knowledge of these changes, the system associates the environmental feature evolution data as operational parameters with the corresponding target task, resulting in sub-tasks containing these operational parameters. These operational parameters not only include traditional action commands (such as gripping force and movement speed) but also innovatively incorporate the aforementioned dataset representing dynamic changes in the spatial environment, giving the parameters a feedforward guidance function based on environmental perception. Specifically, the system uses data encapsulation technology to bind and structurally encapsulate the calculated environmental feature evolution data with the basic process commands of the target task, generating a complete and self-consistent sub-task data package. This allows the robot, upon receiving a sub-task, not only to clearly understand the specific production actions to be performed but also to anticipate how its actions will alter the surrounding point cloud environment. This enables the robot to dynamically adjust its pose and subsequent movement strategies in real time during or after the operation, significantly improving its adaptability and operational safety in complex and constantly changing production environments. For example, when the robot receives the "place embedded part" subtask containing the operation parameters, after completing the placement action, it will automatically adjust the retraction trajectory of its robotic arm according to the environmental feature evolution data caused by the protrusion of the embedded part indicated in the operation parameters, so as to perfectly avoid the embedded part that has just been placed.

[0033] S103. Obtain real-time status information of the first execution unit and the second execution unit, wherein the first execution unit is a mobile operation robot equipped with a replaceable end effector, and the second execution unit is a fixed workstation; After generating subtasks containing operational parameters, to ensure accurate and efficient allocation of these subtasks to appropriate physical execution resources and avoid resource conflicts and waiting during production, the system needs to acquire real-time status information of the first and second execution units. The first execution unit is a mobile robot (e.g., a wheeled robot, tracked robot, or humanoid robot) equipped with replaceable end effectors. A mobile robot is an automated device capable of autonomously navigating and performing physical operations within the target processing workshop. Specifically, when the first execution unit is a humanoid robot, it adapts to the unstructured ground of the workshop using a bipedal or multi-legged structure and uses its two arms to collaboratively complete complex technological actions. The second execution unit is a fixed workstation, which is a pre-defined physical area within the target processing workshop used to support prefabricated components and perform specific production stage operations. Real-time status information refers to the physical and logical operating parameters of the first and second execution units at the current moment. This real-time status information is data detected in real-time by underlying hardware sensors and uploaded to the control center. It serves as the basis for global resource scheduling, utilizing IoT technology to achieve real-time mapping from the physical world to the digital world. In practice, the system uses onboard sensors, positioning modules, and equipment control interfaces deployed on the first execution unit to collect real-time data on its current spatial coordinates, remaining battery power, current idle status, and the type of end effector currently mounted. Simultaneously, the system uses visual monitoring equipment and IoT sensors installed on the second execution unit to collect data on its occupancy, the production progress of the prefabricated components currently being processed, and the material preparation status of surrounding materials. This real-time data collection provides the most accurate underlying data support for subsequent task and resource matching, enabling the system to dynamically adjust scheduling strategies based on actual site conditions, significantly improving the overall collaborative operation efficiency of the target processing workshop. For example, when the system is about to assign a subtask that requires concrete vibration, the system will obtain the real-time status information of all first execution units, filter out the mobile operation robot that has sufficient power, is in an idle state and is closest to a certain second execution unit. At the same time, the system will also obtain the real-time status information of all second execution units to confirm that the precast component at a certain fixed work station has completed concrete pouring and there is no interference from other equipment around it. This prepares the data for accurately assigning the vibration task to the matching first and second execution units.

[0034] S104. Based on the production stage, determine the task type of each sub-task, and determine the end effector of the first execution unit that matches the sub-task according to the task type, and associate the end effector with the first execution unit to obtain the target first execution unit; Task type refers to the specific action classification and interaction mode performed by the mobile robot in physical space. The task type originates from the preset action primitive library of the underlying robot control system. Its purpose is to clarify the physical contact and operation mode between the first execution unit and the prefabricated component to be processed. The principle is to parse the abstract process flow into the robot's kinematic and dynamic operation paradigm through semantic mapping and feature extraction. In specific implementation, the system first extracts the production stage information contained in each sub-task, such as the mold-laying stage, the reinforcement stage, or the pouring stage. Subsequently, the system calls the pre-constructed process action mapping matrix and compares the extracted production stage with the standard action template in the matrix. The process action mapping matrix stores the basic physical action logic that must be included in different production stages. The system uses a semantic parsing algorithm to decompose and convert the production stage into corresponding physical operation attributes, thereby determining the task type corresponding to each sub-task. This method of converting the production stage into a task type successfully breaks down the data barrier between upper-level process planning and lower-level equipment control, making the abstract process requirements concrete into executable machine instruction classifications, providing a direct basis for accurate matching of the end effector. For example, when the system parses the production stage of a subtask as "embedded part fixing stage", the system will perform matching calculations through the process action mapping matrix to determine the task type of the subtask as "high current welding and precise positioning and gripping", thereby clarifying the specific physical interaction actions that the underlying mobile operation robot needs to perform.

[0035] After acquiring the real-time status information of the execution unit, in order to enable the general-purpose mobile robot to perform specific process operations, the system determines the task type of each sub-task based on the production stage, and determines the end effector of the first execution unit that matches the sub-task according to the task type. The system then associates and configures the end effector with the first execution unit to obtain the target first execution unit. Specifically, the system first parses the production stage to which the sub-task belongs (e.g., rebar tying stage or formwork assembly stage), converting it into specific physical operation attributes, i.e., task type (e.g., clamping, welding, spraying, etc.). Since the first execution unit adopts a modular design, it only provides a mobile chassis and basic power, and must be equipped with a specific end effector to complete the actual work. Therefore, the system needs to assign appropriate end tools to sub-tasks of different task types and complete the association configuration of driver programs and kinematic parameters at the software control level, thereby instantiating the basic mobile robot into a target first execution unit that can directly execute the current sub-task. However, in the complex production sites of prefabricated components, simply matching the end effector based on basic process functional requirements is often insufficient. This is because the geometric volume of different models or structures of end effectors, after being mounted on the first execution unit, may obstruct the robot's onboard sensors, leading to perceptual conflicts with the complex obstacle environment surrounding the second execution unit and resulting in blind-spot collision risks during robot operation. Therefore, to maximize the robot's perceptual integrity and operational safety within the target processing workshop while meeting basic operational requirements, the process of determining the end effector of the first execution unit matching the sub-task based on the task type requires further in-depth optimization and screening by combining the physical envelope characteristics of the equipment with the spatial point cloud environment of the site. Figure 2 This is a flowchart illustrating the matching of the end effector in an embodiment of this application. The following is a summary of the process. Figure 2 Step S104 will be further explained.

[0036] S201. Select a set of candidate end effectors from the preset tool library that meet the tool function parameters corresponding to the task type; After clarifying the task types of each subtask, to ensure that the first execution unit has the hardware foundation to perform specific physical operations and to narrow the data processing range for subsequent spatial blind zone calculations, the system will select a set of candidate end effectors from the preset tool library that meet the tool function parameters corresponding to the task type. The preset tool library refers to a pre-established digital database storing the specifications of all available end effectors in the target processing workshop. It consists of hardware ledger information pre-entered by equipment managers and serves as a retrieval space for matching physical execution tools. It uses a relational data structure to map and bind tool numbers to functional attributes. The set of candidate end effectors is a summary combination of all available end effectors that initially possess the physical capability to perform the current subtask. It is the output of the system's function matching algorithm and provides candidate objects for subsequent calculation of the perception conflict index. It is a set selection based on functional conditions. In specific implementation, the system extracts the tool function parameters contained in the task type, such as the required gripping force, welding temperature, or spraying range. Subsequently, the system inputs the tool function parameters into the retrieval engine of the preset tool library, which traverses the function tags of all end effectors in the preset tool library. The system extracts all end effectors whose function labels and tool function parameters match perfectly and combines them to form a set of candidate end effectors.

[0037] Filtering the set of candidate end effectors based on the task type can eliminate ineffective tools with mismatched functions before performing complex spatial geometry calculations, significantly reducing the system's computing power consumption and improving the overall operational efficiency of the target first execution unit configuration process. For example, when the system determines that the task type of a certain subtask is "large volume material gripping", the system will, based on the operational requirements of "large volume material gripping", remove small suction cups and welding guns from the preset tool library, filter out all robotic grippers with large opening size and high load capacity, and combine all the filtered robotic grippers into a set of candidate end effectors for subsequent construction and analysis of self-occlusion blind zone models.

[0038] S202. Obtain the geometric envelope data of each candidate end effector in the candidate end effector set, and construct a self-occlusion blind zone model of the candidate end effector in the mounted state by combining the sensor installation position of the first execution unit. After selecting the set of candidate end effectors, in order to accurately assess the impact of physical occlusion on the field of view of airborne sensors caused by different candidate end effectors after being installed in the first execution unit, the system will obtain the geometric envelope data of each candidate end effector in the set of candidate end effectors, and combine it with the sensor installation position of the first execution unit to construct a self-occlusion blind zone model of the candidate end effectors in the mounted state.

[0039] In practice, the system first retrieves the geometric envelope data of each candidate end effector from the equipment database. The geometric envelope data is a set of spatial boundary coordinates generated from the analysis of 3D design drawings. It is primarily used to accurately represent the maximum physical volume occupied by the tool in digital space, relying on a spatial bounding polygon algorithm to abstract the volume of complex mechanical structures. Simultaneously, the system reads the sensor mounting position coordinates and the sensor's field of view parameters from the first actuator. Subsequently, the system unifies the geometric envelope data and sensor mounting position coordinates into the same robot base coordinate system, simulating the relative spatial pose of the candidate end effectors in their mounted state.

[0040] Based on the unified spatial pose, the system employs a ray tracing algorithm to emit simulated sensing rays outward from the sensor's installation location. It calculates the spatial intersection of the simulated sensing rays and the geometric envelope data, extracting the spatial regions inaccessible to the simulated sensing rays due to obstruction by the geometric envelope data, and constructing a self-occlusion blind zone model. This self-occlusion blind zone model is a dataset of invisible 3D spatial regions generated based on the field-of-view occlusion relationship. Its main function is to quantify and evaluate the blinding range caused by the instrument's own volume on the sensing system, relying on spatial geometric intersection calculations and shadow projection operations.

[0041] Acquiring geometric envelope data and constructing a self-occlusion blind zone model can accurately reproduce the perception limitations of physical devices in a virtual environment in advance, avoiding collisions caused by obstructed vision during actual operation of the robot, and providing a reliable quantitative data foundation for subsequent assessment of perception conflicts.

[0042] For example, when the set of candidate end effectors includes a large dual-arm clamping tool, the system will extract the geometric envelope data of the dual-arm clamping tool and combine it with the sensor installation position of the lidar on the top of the first execution unit. The system will then use ray tracing to calculate the self-occlusion blind zone model of the fan-shaped area directly in front of the dual-arm clamping tool. This will clarify that when the dual-arm clamping tool is mounted, it will cause the first execution unit to be unable to perceive obstacles within a specific angle directly in front of it.

[0043] S203. Based on the spatial point cloud data, identify the spatial distribution characteristics of obstacles within a preset range of each of the second execution units; After constructing the self-occlusion blind zone model of the alternative end effector in the mounted state, in order to accurately assess whether the blind zone will overlap with potential hazardous areas in the actual working environment, the system needs to identify the spatial distribution characteristics of obstacles within a preset range of each second execution unit based on spatial point cloud data.

[0044] In practice, the system first extracts global spatial point cloud data of the target processing workshop and uses the physical center coordinates of each second execution unit as the origin, extending outward by a set distance to define a three-dimensional preset range. Subsequently, the system performs pass-through filtering and planar segmentation on the spatial point cloud data within the preset range, separating the standard point cloud belonging to the ground and the prefabricated components to be processed, while retaining discrete point clouds representing non-target objects. The system further applies a density clustering algorithm to the retained discrete point clouds to calculate their geometric center, boundary contours, and spatial density, thereby generating obstacle spatial distribution features. These obstacle spatial distribution features are a structured dataset used to mathematically represent the physical location and size of obstacles around the workstation. They are derived from the original three-dimensional coordinates through spatial clustering and feature extraction, primarily used to provide a quantitative environmental reference map for subsequent perception conflict calculations.

[0045] Identifying the spatial distribution characteristics of obstacles based on spatial point cloud data can transform complex physical working environments into computer-readable digital spatial features, allowing for early detection of environmental interference around fixed workstations and providing indispensable environmental data support for the subsequent accurate selection of the safest end effector.

[0046] For example, when the second execution unit is a fixed rebar tying station, the system will capture the spatial point cloud data within a three-meter radius around the rebar tying station, and identify the three-dimensional outline and position coordinates of the temporary scaffolding or waste bins in the vicinity through cluster analysis, forming the corresponding spatial distribution characteristics of obstacles, so as to determine whether the robot's visual blind spot just covers the location of the scaffolding or waste bins.

[0047] S204. Calculate the perception conflict index of the obstacle spatial distribution characteristics between each self-occlusion blind zone model and the corresponding second execution unit; After constructing self-occlusion blind zone models of candidate end effectors in their mounted state and identifying the spatial distribution characteristics of obstacles within a preset range of the second execution unit, in order to accurately quantify and evaluate the comprehensive impact of tool blind zones on the actual environmental perception and obstacle avoidance of the mobile robot, the system needs to calculate the perception conflict index between each occlusion blind zone model and the spatial distribution characteristics of obstacles of each second execution unit. Specifically, this may include the following steps: projecting the self-occlusion blind zone model onto the target spatial point cloud data within a preset range of the second execution unit to obtain the blind zone coverage space; identifying target positioning feature points and physical obstacles located within the blind zone coverage space in the obstacle spatial distribution characteristics; calculating the feature loss rate of the target positioning feature points within the blind zone coverage space and the volume ratio of the physical obstacles within the blind zone coverage space; and weighted summing the feature loss rate and the volume ratio to obtain the perception conflict index.

[0048] To establish an accurate mapping between the theoretical equipment occlusion model and the real physical working environment, the system projects the self-occlusion blind zone model onto the target spatial point cloud data within a preset range of the second execution unit, obtaining the blind zone coverage space. The blind zone coverage space refers to a subset of three-dimensional space in the local environment of the actual workstation that is completely undetectable by onboard sensors due to the physical occlusion of the end effector. It is generated through spatial coordinate system transformation and point cloud intersection operations, used to visually represent the specific physical area where the line of sight is obstructed in the real environment base map. In specific implementation, the system uses the global positioning coordinates of the first execution unit within the target processing workshop and the preset working pose to construct a homogeneous transformation matrix from the robot base coordinate system to the workshop global coordinate system. Through this transformation matrix, the system transforms the three-dimensional boundary coordinates of the self-occlusion blind zone model to the global coordinate system and performs a spatial Boolean intersection operation with the target spatial point cloud data within the preset range of the second execution unit, extracting the point cloud set and spatial envelope falling inside the blind zone boundary, thereby generating the blind zone coverage space. This approach transforms the abstract blind spots of the equipment itself into concrete spatial data closely integrated with the actual working environment, providing precise spatial boundary ranges for subsequent assessments of which specific environmental elements are hidden within the blind spots. For example, when a mobile robot is preparing to work at a prefabricated wall panel assembly station, the system projects the self-occluding blind spot model of the robotic gripper onto the point cloud data surrounding the assembly station through coordinate system transformation, thereby accurately delineating a real space directly below the robotic gripper, measuring half a meter on each side and one meter in height, as the blind spot coverage area.

[0049] After identifying the specific blind spot coverage space, in order to determine whether any environmental elements crucial for robot navigation, positioning, and obstacle avoidance are hidden within this blind spot, the system identifies target positioning feature points and physical obstacles located within the blind spot coverage space from the obstacle spatial distribution characteristics. Target positioning feature points refer to reference points in spatial point cloud data that possess significant geometric features and are used to assist mobile robots in high-precision spatial positioning, such as the edge corners of prefabricated components or reflective labels on workstations. These are extracted from the point cloud using feature extraction algorithms to ensure the absolute accuracy of the robot's end effector. Physical obstacles are spatial obstructions with a risk of physical collision generated by clustering in the aforementioned steps. In practice, the system invokes a spatial inclusion detection algorithm, traversing the coordinates of all elements in the obstacle spatial distribution characteristics, and determining one by one whether the three-dimensional coordinates of each target positioning feature point and the boundary vertices of each physical obstacle are located inside the spatial polygon of the blind spot coverage space. The system marks and extracts feature points and obstacles whose coordinates fall within the blind spot coverage space, forming a set of occluded elements. This operation can accurately pinpoint key positioning references and potential collision hazards that are "unseen" by the system due to tool obstruction, refining the macroscopic blind spot analysis to specific environmental entities. For example, by comparing coordinates, the system identifies that two reference corner points (target positioning feature points) used for visual alignment at the assembly station and a temporarily placed toolbox (physical obstacle) are all located within the blind spot coverage space defined above by the robotic gripper.

[0050] After identifying the specific environmental elements that are obstructed, the system further calculates the feature loss rate of the target localization feature points within the blind zone coverage space and the volume ratio of the physical obstacles within the blind zone coverage space to quantitatively measure the negative impact of this obstruction on the robot's operation. The feature loss rate is a percentage value reflecting the degree of missing localization information, calculated by dividing the number of feature points in the blind zone by the total number of feature points required for the operation; it is used to assess the risk of local localization failure. The volume ratio is the ratio of the space volume occupied by obstacles within the blind zone to the total volume of the blind zone or the total volume of obstacles; it is calculated using voxelized meshes and is used to quantify the severity of hidden collision risks. During implementation, the system first counts the number of target localization feature points within the blind zone coverage space and divides it by the total number of feature points required for the normal execution of the subtask to obtain the feature loss rate. Simultaneously, the system uses voxel integration to calculate the point cloud envelope volume of physical obstacles within the blind zone coverage space and divides it by the total volume of the blind zone coverage space to obtain the volume ratio. By calculating these two quantitative indicators, the system can objectively and accurately digitally express the magnitude of the perception limitations posed by the current end effector from two independent dimensions: positioning accuracy decay and obstacle avoidance failure. For example, if the assembly operation relies on a total of 10 reference corner points, and the blind zone contains 3 of them, the system will calculate the feature loss rate as 30%. At the same time, if the total volume of the space covered by the blind zone is 0.5 cubic meters, and the volume of the toolbox hidden inside is 0.1 cubic meters, the system will calculate the volume ratio of physical obstacles as 20%.

[0051] After obtaining independent quantitative indicators, in order to comprehensively evaluate the overall perception interference of different candidate end effectors for final selection, the system will perform a weighted summation of the feature loss rate and the volume ratio to obtain the perception conflict index. The perception conflict index is a dimensionless scalar used to comprehensively evaluate the overall negative impact of the end effector on the airborne perception system. A higher value indicates a higher operational risk caused by perception limitations. It is calculated using a linear weighted fusion mechanism in multi-objective decision-making and is mainly used to provide a single optimization objective function for automatic tool optimization. In specific implementation, the system will dynamically retrieve preset weight coefficients based on the process quality requirements and safety level of the current sub-task. For example, for assembly tasks with extremely high precision requirements, the system will assign a higher weight to the feature loss rate; while for handling tasks with limited space, it will assign a higher weight to the volume ratio. The system multiplies the feature loss rate by its corresponding positioning weight coefficient, multiplies the volume ratio by its corresponding safety weight coefficient, and finally adds the two together to output the final perception conflict index. This weighted summation method not only unifies multi-dimensional risk indicators into a single intuitive evaluation score, but also allows the evaluation process to flexibly adapt to the focus of different process stages. This ensures that the final selected end effector has the optimal perception field of view and the highest operational safety for the current specific task. For example, the system multiplies the 30% feature loss rate by a weight of 0.6 and the 20% volume ratio by a weight of 0.4, and adds them together to obtain the perception conflict index of the candidate robotic gripper at the current workstation as 0.26, which can then be compared with the indices of other tools.

[0052] S205. The candidate end effector with the smallest perceived conflict index is determined as the end effector.

[0053] After calculating the perception conflict index for all candidate end effectors, the system needs to select the end effector with the lowest perception conflict index from the candidate end effector set to minimize interference with the onboard sensor field of view of the mobile robot. The perception conflict index is a comprehensive quantitative evaluation standard that integrates the risk of lost positioning features and the risk of physical collisions; a smaller value indicates a lower negative impact on environmental perception after the corresponding tool is mounted. In practice, the system inputs the perception conflict indices of all candidate end effectors into the numerical comparison module, which sorts all the indices in ascending order. The candidate end effector with the lowest perception conflict index at the top is the optimal solution for achieving the best field of view and the highest operational safety under the current subtask. The system then officially confirms the top-ranked candidate end effector as the final mounted end effector and generates the corresponding hardware replacement command, which is sent to the mobile robot. By selecting the candidate end effector with the lowest perceived conflict index, the system can minimize the potential for positioning failures and collisions caused by blind spots from a physical hardware perspective, while meeting the requirements of the manufacturing process, thus ensuring the smooth and safe execution of prefabricated component production tasks. For example, if, for a specific assembly sub-task, the system calculates the perceived conflict indices of a suction cup gripper, a two-finger robotic gripper, and a multi-finger flexible gripper to be 0.45, 0.26, and 0.68 respectively, the system will directly lock in the lowest value of 0.26 through numerical comparison and determine the two-finger robotic gripper corresponding to 0.26 as the final end effector to execute the assembly sub-task.

[0054] After determining the optimal end effector, to ensure the first execution unit can accurately identify and drive the end effector to perform specific production actions, the system associates the end effector with the first execution unit, thus obtaining the target first execution unit. Association configuration refers to binding the physical parameters and communication protocols of a specific tool to the robot's control logic within the mobile robot's control system. Originating from the field of robot kinematics control, this primarily updates the robot's tool center point coordinates and dynamic load model, ensuring the robotic arm can still accurately execute spatial motion trajectories after attaching a new tool. In practice, the system packages the end effector's three-dimensional dimensions, mass distribution data, and electrical interface communication control commands into a configuration data package and sends it to the first execution unit's main control computer. Upon receiving the configuration data package, the main control computer automatically refreshes its internal kinematic solution matrix, establishing a hardware driver connection and software parameter mapping between the end effector and the first execution unit. By executing these steps, the system successfully transforms a mobile robot that was originally in a general standby state into a dedicated work device with specific process execution capabilities and fully compatible hardware and software parameters—the target first execution unit—thus laying a solid equipment foundation for subsequent path planning and actual processing operations. For example, after the system selects the two-finger robotic gripper, it will send the weight, gripping stroke and control code of the two-finger robotic gripper to the mobile operation robot numbered one. After receiving the data and updating its own control system, the mobile operation robot number one becomes the target first execution unit specifically responsible for the prefabricated wall panel assembly task.

[0055] S105. Based on the job parameters and the real-time status information, determine the execution combination of each of the sub-tasks, wherein the execution combination includes at least one target first execution unit and one second execution unit; After configuring the first execution unit, to ensure that subtasks can be carried out safely and efficiently at suitable workstations, the system needs to determine the execution combination of each subtask based on work parameters and real-time status information. Execution combination is essentially a collaborative work binding relationship between the mobile robot and the fixed workstation, aiming to allocate the most suitable hardware resources and space for a specific production action. Specifically, this may include the following steps: Based on the job parameters and the real-time status information, a set of candidate second execution units that meet the job conditions of the sub-task is selected. Construct the dynamic job envelope of the target first execution unit, and extract the neighborhood dynamic point cloud data of each of the candidate second execution units in the set of candidate second execution units within the target range from the spatial point cloud data; The dynamic operation envelope is placed into the neighborhood dynamic point cloud data of each of the candidate second execution units for virtual spatiotemporal interference verification; The alternative second execution unit that has passed the virtual spatiotemporal interference verification is combined with the target first execution unit to form the execution combination.

[0056] Before conducting specific spatial interference checks, the system needs to filter out a set of candidate second execution units that meet the operational conditions of the sub-task based on operational parameters and real-time status information. Operational parameters include the processing technology indicators and material requirements of the current sub-task, while real-time status information reflects whether each fixed workstation is currently idle or in standby mode and the health status of the equipment. To eliminate workstations that are faulty, under maintenance, or performing other tasks, and to avoid resource scheduling conflicts, the system matches the process requirements in the operational parameters with the inherent attributes of all second execution units in the workshop, while simultaneously filtering out non-idle workstations based on real-time status information. After condition comparison, all second execution units that meet the hardware processing capabilities and are currently available are grouped together to form a set of candidate second execution units, thus providing a candidate pool for subsequent precise matching. For example, if the sub-task is wall panel rebar tying, the system will read the tying size requirements in the operational parameters and, combined with real-time status information, filter out three rebar tying stations in the workshop that match the dimensions and are currently unused, forming a set of candidate second execution units.

[0057] After obtaining the set of candidate second execution units, in order to accurately quantify the spatial range occupied by the mobile robot during actual work and the real-time environmental changes around the workstation, the system constructs a dynamic work envelope of the target first execution unit and extracts the neighborhood dynamic point cloud data of each candidate second execution unit within the target range from the spatial point cloud data. The dynamic work envelope originates from the field of robot kinematics collision avoidance and represents the maximum three-dimensional spatial volume swept by the robot when executing a specific motion trajectory. It is mainly used to accurately simulate the robot's physical positioning in a virtual digital environment. Neighborhood dynamic point cloud data refers to the set of three-dimensional spatial points that change over time within a certain range around a fixed workstation, mainly reflecting the movement of personnel, materials, or temporary equipment on site. In specific implementation, the system calculates the spatial sweep model of the target first execution unit throughout the entire motion cycle using forward kinematics based on the motion commands in the work parameters, generating a dynamic work envelope. Simultaneously, the system extracts point cloud information within a preset radius centered on the candidate second execution unit from the spatial point cloud data to form neighborhood dynamic point cloud data. By extracting the two types of spatial data mentioned above, the system can realistically reconstruct the dynamic physical boundaries of the work site in digital space, providing high-precision data support for subsequent safety inspections. For example, the system will calculate a hemispherical dynamic work envelope formed by the robotic arm of a mobile robot equipped with a gripper swinging its mechanical arm when performing a gripping action, and simultaneously capture the dynamic point cloud data of the neighborhood within a two-meter radius around the No. 1 binding platform.

[0058] After completing the extraction and construction of spatial data, in order to predict and avoid potential physical collision accidents in actual production, the system will place the dynamic operation envelope into the neighborhood dynamic point cloud data of each candidate second execution unit for virtual spatiotemporal interference verification. Specifically, this may include the following steps: based on the operation parameters of the sub-task, analyze the joint motion trajectory of the target first execution unit during the execution of the sub-task, and map each joint motion trajectory to a space occupancy set that changes over time; extract dynamic feature points from the neighborhood dynamic point cloud data, and analyze the predicted motion trajectory of the dynamic feature points within the execution cycle of the sub-task based on the displacement vector of the continuous frame point cloud of the neighborhood dynamic point cloud data; perform spatiotemporal overlap detection on the space occupancy set, the predicted motion trajectory, and the static obstacle point cloud in the neighborhood dynamic point cloud data, and calculate the minimum safe distance under different preset time steps; if the minimum safe distance is greater than a preset safety threshold within the execution cycle of the sub-task, then the virtual spatiotemporal interference verification is passed.

[0059] In conducting specific virtual spatiotemporal interference verification, to accurately grasp the robot's spatial occupancy at every moment during task execution, the system analyzes the joint motion trajectories of the target first execution unit during the execution of the sub-task based on the sub-task's operational parameters, and maps each joint motion trajectory into a time-varying spatial occupancy set. The operational parameters include specific motion instructions and processing paths; the target first execution unit is the mobile robot configured with its end effector. The spatial occupancy set, derived from robot kinematics and computer graphics, refers to the three-dimensional spatial voxel array occupied by each mechanical joint of the robot at different time points, primarily used to quantify the robot's physical occupancy on the time axis in a digital twin environment. In practice, the system parses the motion code in the operational parameters, uses the robot's forward kinematics equations to calculate the three-dimensional coordinate changes of each joint throughout the entire motion cycle, generates continuous joint motion trajectories, and then converts these trajectories into voxel blocks in a three-dimensional mesh according to the time sequence, forming a dynamically changing spatial occupancy set. In this way, the system achieves a high-precision digital three-dimensional representation of the robot's dynamic movements, providing fundamental self-occupancy data for subsequent interference detection. For example, when a robot is performing a task of spraying paint on a wall panel, the system calculates the motion trajectory of the six joints of its robotic arm and converts it into a series of cylindrical voxels that the robotic arm occupies in virtual space every second.

[0060] After clarifying the robot's dynamic space occupancy, the system also needs to grasp the real-time dynamic changes of the surrounding environment. Therefore, it extracts dynamic feature points from the neighborhood dynamic point cloud data and analyzes the displacement vectors of the point clouds in consecutive frames of the neighborhood dynamic point cloud data to predict the motion trajectory of the dynamic feature points within the execution cycle of the subtask. The neighborhood dynamic point cloud data reflects the environmental conditions around the workstation, while dynamic feature points refer to the set of point clouds reflected by objects in the environment whose positions are changing (such as walking workers or moving AGVs). The predicted motion trajectory is the future movement path calculated based on historical displacements, used to predict the direction of dynamic obstacles in the environment. In practice, the system compares the neighborhood dynamic point cloud data of multiple consecutive frames, filters out dynamic feature points whose coordinates have changed significantly, calculates the displacement vectors (i.e., velocity and direction) of these points between adjacent frames, and then uses kinematic prediction models such as Kalman filtering to deduce the predicted motion trajectory of these dynamic feature points throughout the entire execution cycle of the subtask. This operation enables the system to predict the future position of moving objects in the working environment in advance, effectively preventing unforeseen collisions between the robot and surrounding moving personnel or equipment. For example, if the system detects a material transport vehicle moving next to the binding station, it will extract the dynamic feature points on its surface and, based on its current speed and direction of travel, calculate the predicted trajectory of the material transport vehicle in the next twenty seconds.

[0061] After acquiring the robot's spatial occupancy set and the predicted motion trajectory of the environment, the system can conduct in-depth safety checks. It performs spatiotemporal overlap detection on the spatial occupancy set, the predicted motion trajectory, and the static obstacle point cloud in the neighborhood dynamic point cloud data, calculating the minimum safe distance at different preset time steps. The static obstacle point cloud represents fixed entities in the environment, such as load-bearing columns and fixed machine tools. Spatiotemporal overlap detection is an algorithm that determines whether multiple three-dimensional objects intersect at the same time and space coordinates, used to quantify the proximity of the robot to its surrounding environment. In practice, the system divides the entire task's timeline into multiple extremely short preset time steps (e.g., 0.1 seconds). At each time step, the robot's spatial occupancy set, the predicted motion trajectory of dynamic objects, and the static obstacle point cloud are placed in the same virtual three-dimensional coordinate system. The shortest Euclidean distance between the robot's outer envelope and all surrounding obstacle point clouds is calculated, thus determining the minimum safe distance at that moment. This fine-grained detection method achieves full-time, all-round collision risk quantification assessment. For example, the system uses a time step of 0.1 seconds to calculate that when the action is executed for 5 seconds, the shortest distances from the end of the robotic arm to the fixed column (static obstacle) and the material cart (dynamic obstacle) are 0.6 meters and 0.9 meters respectively. Therefore, the minimum safe distance at that moment is 0.6 meters.

[0062] After calculating the distance at each time step, the system needs to make a final safety judgment based on this quantitative data. If the minimum safe distance is greater than the preset safety threshold throughout the subtask's execution cycle, the virtual spatiotemporal interference check is passed. The preset safety threshold is the minimum allowable distance limit set to ensure absolute safety, typically taking into account the robot's maximum braking distance, system communication latency, and sensor measurement errors. The virtual spatiotemporal interference check is the final checkpoint to determine whether the currently selected fixed workstation is truly safe and usable. In practice, the system iterates through all the minimum safe distance values ​​calculated throughout the entire subtask's execution cycle and compares them one by one with the preset safety threshold. As long as the minimum safe distance at any time point does not fall below the threshold, it means that the robot will not have any form of contact or dangerous approach to the environment during the entire operation, and the system determines that the workstation has successfully passed the virtual spatiotemporal interference check. This judgment mechanism ensures the absolute safety of the final generated execution combination in both physical space and time dimensions, avoiding collision accidents in actual production. For example, if the system sets a preset safety threshold of 0.3 meters, and if the thousands of minimum safety distances calculated by the system remain above 0.4 meters throughout the entire two-minute spraying task cycle, the system confirms that the spraying station has passed the verification and that the robot can be safely dispatched there to carry out the work.

[0063] After rigorous virtual space security checks, to ultimately determine the execution location and equipment for sub-tasks, the system combines the candidate second execution units that have passed virtual spatiotemporal interference verification with the target first execution unit into an execution combination. Essentially, an execution combination is a collaborative work binding relationship between a mobile robot and a fixed workstation, aiming to allocate the most suitable and absolutely safe hardware resources and space for a specific production action. The system marks the candidate second execution units that have not overlapped spatially and have successfully passed the safety test as available, selects the optimal workstation from them, and sends the network address and spatial coordinates of the optimal workstation to the main control module of the target first execution unit, completing the logical binding at the hardware and software level. After establishing the execution combination, the system completely breaks down the collaborative barriers between mobile devices and fixed workstations, enabling the target first execution unit to clearly define its mobile destination and safely carry out processing operations, ensuring the continuity of the entire prefabricated component production process. For example, if interference is detected at the No. 1 tying platform and the No. 2 tying platform has ample space around it and passes the verification, the system will bind the No. 2 tying platform to the target first execution unit and formally generate an execution combination specifically for performing the current rebar tying sub-task.

[0064] S106. Based on the spatial point cloud data and the operation parameters, determine the movement path of each target first execution unit; After determining the execution combination consisting of the target first execution unit and the second execution unit, in order for the mobile robot to safely and efficiently traverse the workshop to reach the designated fixed workstation for processing, the system needs to determine the movement path of each target first execution unit based on the spatial point cloud data and the operation parameters. Specifically, this may include the following steps: Based on the spatial point cloud data, the target processing workshop is rasterized to generate a global navigation map containing obstacle information and passable areas. Determine the work stop point of the target first execution unit at the second execution unit based on the work parameters; Obtain the current position of the target first execution unit and map the geometric dimensions of the target first execution unit to a motion collision envelope; In the global navigation map, a collision-free trajectory from the current position to the work stop point is generated based on the mobile collision envelope, thus obtaining the mobile path.

[0065] To enable mobile robots to navigate autonomously and safely in complex workshop environments, the system needs to perform rasterization processing on the target processing workshop based on the spatial point cloud data, generating a global navigation map containing obstacle information and passable areas. While spatial point cloud data contains highly precise 3D environmental details, its sheer volume and lack of topological structure mean that directly using it for path planning would result in significant computational resource consumption and extremely high latency. Therefore, the system uses rasterization to divide the continuous 3D point cloud space into discrete 2D or 3D grid units, determining the occupied state of each grid based on whether it contains point cloud data. This approach abstracts the complex physical environment into a global navigation map composed of "obstacle information" grids and "passable area" grids, providing lightweight and structured underlying data support for subsequent path search algorithms and significantly improving the efficiency of pathfinding calculations. For example, the system divides the point cloud data of a workshop covering thousands of square meters into two-dimensional grids with an accuracy of 5 centimeters. The grids containing walls and fixed machine tools are marked as impassable obstacles, while open passageways are marked as passable areas.

[0066] After establishing a comprehensive understanding of the workshop environment, the system needs to further define the robot's specific destination, namely, determining the docking point of the target first execution unit at the second execution unit based on the operational parameters. The docking point refers to the optimal spatial coordinates and orientation angle that the mobile robot must occupy near the fixed workstation to successfully complete a specific processing task. Since different sub-tasks (such as welding, painting, or gripping) impose strict physical limitations on the robot arm's extension range, viewing angle, and force angle, the system analyzes the processing trajectory requirements and the working radius of the end effector contained in the operational parameters. Through inverse kinematics solutions, it calculates a docking pose within the safe working area around the fixed workstation that completely covers the processing range without interfering with the workstation itself. This step ensures that the robot can directly commence processing operations upon reaching its destination, avoiding repeated adjustments or exceeding working limits due to poor docking positions. For example, when the operation parameters indicate that a straight line welding of 1 meter long needs to be carried out on the side of a precast component, the system will calculate that the robot should stop 0.8 meters in front of the fixed work position and be oriented parallel to the weld seam to ensure that the robotic arm is in the best force exertion and extension posture.

[0067] Once the destination is determined, to ensure the robot's absolute safety during movement, the system acquires the current position of the target first execution unit and maps its geometric dimensions to a motion collision envelope. The current position is the starting coordinate of the robot's movement, while the motion collision envelope is a virtual geometric boundary introduced to address the robot's "occupancy" problem on the map. In the real physical world, a robot is not a point mass without volume, but a physical device with length, width, and height. If path planning is only based on the center point, collisions are highly likely to occur at corners or narrow passages. Therefore, the system extracts the robot's chassis outline, the onboard end effector, and the geometric dimensions of any potentially extending robotic arms, and appropriately expands these outwards to construct a virtual shell (i.e., the motion collision envelope) that completely encloses the robot. This simplification of complex physical contours into a regular geometric envelope greatly simplifies the mathematical calculations for collision detection while preserving the robot's actual spatial occupancy characteristics. For example, a mobile robot that is 1.2 meters long and 0.8 meters wide will be mapped by the system into a cylindrical motion collision envelope with a radius of 0.7 meters to cover all the physical space it may sweep when it rotates in place or moves.

[0068] After integrating the environmental map, start and end points, and the robot's spatial occupancy characteristics, the system ultimately generates a collision-free trajectory from the current position to the work stop point in the global navigation map based on the motion collision envelope, thus obtaining the movement path. In this calculation process, the system uses the size of the motion collision envelope as a safety margin and performs equidistant expansion on the obstacle boundaries in the global navigation map, transforming the original pathfinding problem into a process of a point mass finding a connected path in the expanded map. The system uses a heuristic search algorithm to continuously expand nodes towards the work stop point within the passable area, selecting a smooth curve that not only avoids all static and dynamic obstacles but also satisfies the robot's kinematic constraints (such as the maximum turning radius). This collision-free trajectory directly constitutes the movement path executable by the robot's chassis controller, completely establishing the execution chain from task issuance to physical movement. For example, using the A* pathfinding algorithm, after avoiding temporarily stacked material pallets and other operating equipment, the system plans an S-shaped movement path with smooth chamfers for the robot, guiding it safely and accurately from the standby area to the designated welding stop point.

[0069] It should be noted that the dynamic operation envelope and the motion collision envelope in this application correspond to different motion states of the robot: the dynamic operation envelope focuses on characterizing the full-cycle three-dimensional sweep volume generated by the robotic arm and its joints waving and extending in space according to the operation parameters when the first execution unit of the target performs a specific sub-task (such as grasping, spraying, welding, etc.) at a fixed operation position, which reflects the dynamic occupancy boundary of the robot in the operation state; while the motion collision envelope focuses on characterizing the physical space volume occupied by the robot as a whole (including the chassis and the end effector in a stored or stable posture) in the global map during the process of the robot moving from the current position to the preset operation position, which reflects the geometric size constraints of the robot in the navigation and driving state. The two have significant differences in three-dimensional modeling logic, the range of mechanical devices included, and the safety judgment criteria for environmental obstacles.

[0070] S107. Control the first execution unit of the target to travel to the preset work position of the second execution unit in the execution combination according to the movement path, and execute each of the sub-tasks according to the work parameters.

[0071] To translate the front-end planning into actual physical manufacturing actions, the system needs to control the first target execution unit to travel along a movement path to the preset working position of the second execution unit in the execution assembly, and execute each sub-task according to the work parameters. The system translates the movement path into control commands for the underlying drive motors, driving the first target execution unit to navigate autonomously within the target processing workshop. After the first target execution unit arrives at the area where the second execution unit is located, it will precisely stop at the preset working position. The preset working position is the optimal spatial coordinates and attitude angle calculated in the early stage based on spatial point cloud data and geometric envelope. The preset working position ensures that the robotic arm of the first target execution unit, when fully extended, just covers the processing area of ​​the prefabricated component to be processed, while avoiding physical interference with the second execution unit. After the first target execution unit is in place, the system controls the first target execution unit to execute each sub-task according to the work parameters. The work parameters include the movement trajectory of each joint of the first target execution unit during the execution of the sub-task, the working voltage or speed of the end effector, and other specific process commands. The work parameters are generated by parsing the previous process data and geometric data to guide the underlying hardware actions. The first execution unit drives the end effector to perform specific processing operations on the prefabricated component according to the operation parameters. The coordinated operation of the first and second execution units enables unmanned and high-precision processing of prefabricated components, significantly improving the overall production efficiency and product quality of the target processing workshop.

[0072] For example, the system issues a control command to the first execution unit of the target, which is equipped with a welding torch. The first execution unit of the target moves smoothly along a path that avoids obstacles to the preset working position next to the second fixed welding station. Then, the first execution unit of the target starts the welding torch to precisely weld the side joint of the prefabricated component to be processed according to the welding speed and weld coordinate trajectory specified in the working parameters, until the current welding sub-task is completed.

[0073] Please see Figure 3 This is a schematic diagram of the structure of a prefabricated component production system in an embodiment of this application.

[0074] It should be noted that, Figure 3 The structure of a prefabricated component production system shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0075] like Figure 3As shown, a prefabricated component production system includes a central processing unit 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory 302 or a program loaded from a storage section 308 into a random access memory 303, such as performing the methods described in the above embodiments. The random access memory 303 also stores various programs and data required for system operation. The central processing unit 301, the read-only memory 302, and the random access memory 303 are interconnected via a bus 304. An input / output interface 305 is also connected to the bus 304.

[0076] The following components are connected to the input / output interface 305: an input section 306 including audio input devices, push-button switches, etc.; an output section 307 including an LCD display, audio output devices, indicator lights, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the input / output interface 305 as needed. A removable medium 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 310 as needed so that computer programs read from it can be installed into the storage section 308 as needed.

[0077] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit 301, it performs the various functions defined in the present invention.

[0078] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, flash memory, optical fiber, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0079] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0080] Specifically, a prefabricated component production system according to this embodiment includes a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it implements a prefabricated component production method provided in the above embodiment.

[0081] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in a prefabricated component production system described in the above embodiments; or it may exist independently and not assembled into the prefabricated component production system. The storage medium carries one or more computer programs that, when executed by a processor of the prefabricated component production system, cause the prefabricated component production system to implement the prefabricated component production method provided in the above embodiments.

Claims

1. A method for producing precast components, characterized in that, The method includes: Obtain geometric and process data of the prefabricated components to be processed, as well as spatial point cloud data of the target processing workshop. The process data includes production stages and quality requirements. Production tasks are generated based on the geometric data and the process data, and the production tasks are divided into multiple sub-tasks, each of which corresponds to a work parameter. The real-time status information of the first execution unit and the second execution unit is obtained. The first execution unit is a mobile operation robot equipped with a replaceable end effector, and the second execution unit is a fixed work station. Based on the production stage, the task type of each sub-task is determined, and the end effector of the first execution unit that matches the sub-task is determined according to the task type. The end effector is then associated with the first execution unit to obtain the target first execution unit. The execution combination of each subtask is determined based on the job parameters and the real-time status information, and the execution combination includes at least one target first execution unit and at least one second execution unit; Based on the spatial point cloud data and the operation parameters, the movement path of each target first execution unit is determined; The target first execution unit is controlled to travel along the movement path to the preset work position of the second execution unit in the execution combination, and each of the sub-tasks is executed according to the work parameters.

2. The method according to claim 1, characterized in that, The production task is divided into multiple sub-tasks, specifically including: The production task is divided into multiple target tasks according to the production stage; Based on the geometric data, construct the intermediate geometric model of the prefabricated component to be processed at each of the production stages; Based on the production stage, environmental feature evolution data before and after the execution of each target task are determined. The environmental feature evolution data is used to characterize the point cloud feature changes of the fixed workstation due to the increase in components. The environmental feature evolution data is used as the operation parameter and associated with the corresponding target task to obtain each sub-task containing the operation parameter.

3. The method according to claim 1, characterized in that, The step of determining the end effector of the first execution unit that matches the subtask according to the task type specifically includes: Select a set of candidate end effectors from the preset tool library that meet the tool function parameters corresponding to the task type; Obtain the geometric envelope data of each candidate end effector in the candidate end effector set, and combine it with the sensor installation position of the first execution unit to construct the self-occlusion blind zone model of the candidate end effector in the mounted state; Based on the spatial point cloud data, the spatial distribution characteristics of obstacles within a preset range of each of the second execution units are identified; Calculate the perception conflict index between each self-occlusion blind spot model and the corresponding second execution unit for the spatial distribution characteristics of obstacles; The candidate end effector with the smallest perceived conflict index is determined as the end effector.

4. The method according to claim 3, characterized in that, The calculation of the perception conflict index between each of the self-occlusion blind spot models and the corresponding second execution unit regarding the spatial distribution characteristics of obstacles specifically includes: The self-occlusion blind spot model is projected onto the target space point cloud data within the preset range of the second execution unit to obtain the blind spot coverage space; Identify target location feature points and physical obstacles located within the blind zone coverage space in the spatial distribution characteristics of the obstacles; Calculate the feature loss rate of the target location feature points within the blind zone coverage space, and the volume ratio of the physical obstacle within the blind zone coverage space; The perceptual conflict index is obtained by weighting and summing the feature loss rate and the volume ratio.

5. The method according to claim 1, characterized in that, The step of determining the execution combination of each subtask based on the job parameters and the real-time status information specifically includes: Based on the job parameters and the real-time status information, a set of candidate second execution units that meet the job conditions of the sub-task is selected. Construct the dynamic job envelope of the target first execution unit, and extract the neighborhood dynamic point cloud data of each of the candidate second execution units in the set of candidate second execution units within the target range from the spatial point cloud data; The dynamic operation envelope is placed into the neighborhood dynamic point cloud data of each of the candidate second execution units for virtual spatiotemporal interference verification; The alternative second execution unit that has passed the virtual spatiotemporal interference verification is combined with the target first execution unit to form the execution combination.

6. The method according to claim 5, characterized in that, The step of placing the dynamic job envelope into the neighborhood dynamic point cloud data of each of the candidate second execution units for virtual spatiotemporal interference verification specifically includes: Based on the operation parameters of the sub-task, the motion trajectory of each joint of the target first execution unit during the execution of the sub-task is analyzed, and each joint motion trajectory is mapped to a space occupancy set that changes over time. Dynamic feature points are extracted from the neighborhood dynamic point cloud data, and the predicted motion trajectory of the dynamic feature points within the execution cycle of the subtask is analyzed based on the displacement vector of the continuous frame point cloud of the neighborhood dynamic point cloud data. Spatiotemporal overlap detection is performed on the spatial occupancy set, the predicted motion trajectory, and the static obstacle point cloud in the neighborhood dynamic point cloud data to calculate the minimum safe distance under different preset time steps. If the minimum safety distance is greater than the preset safety threshold during the execution cycle of the subtask, then the virtual spatiotemporal interference verification is passed.

7. The method according to claim 1, characterized in that, The step of determining the movement path of each target first execution unit based on the spatial point cloud data and the operation parameters specifically includes: Based on the spatial point cloud data, the target processing workshop is rasterized to generate a global navigation map containing obstacle information and passable areas. Determine the work stop point of the target first execution unit at the second execution unit based on the work parameters; Obtain the current position of the target first execution unit and map the geometric dimensions of the target first execution unit to a motion collision envelope; In the global navigation map, a collision-free trajectory from the current position to the work stop point is generated based on the mobile collision envelope, thus obtaining the mobile path.

8. A precast component production system, characterized in that, The prefabricated component production system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors invoke the computer instructions to cause the prefabricated component production system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are run on the prefabricated component production system, the prefabricated component production system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the prefabricated component production system, the prefabricated component production system performs the method as described in any one of claims 1-7.

Images