Work method of work robot of humanoid structure and work robot

Abstract

This invention discloses a humanoid-shaped work robot and its dynamic operation method. The robot includes a central controller, a 360-degree rotatable rotating part, and at least two wheel arms driven by the rotating part, each wheel arm carrying multiple work modules. The core of the method lies in: the central controller acquiring the spatial pose parameters and type identifiers of each work module in real time, and combining them with predefined work direction, coordination distance, and module work reference distance parameters; dynamically identifying a set of work module combinations that meet process requirements by matching real-time data with expected work combination templates; optimizing candidate operation planes and transfer planes, and finally synthesizing a task space layout diagram identifying multiple operation units; based on this layout diagram, the central controller intelligently allocates work modules or wheel arms to specific units and generates adaptive work instruction sequences to control the robot to perform assembly line operations, while monitoring efficiency in real time and performing closed-loop adjustments. This system achieves autonomous perception, intelligent planning, and collaborative control of the robot's multi-module resources in a dynamic environment.

Description

Technical Field

[0001] This application relates to humanoid robots, specifically to a method for operating a humanoid operational robot and the operational robot itself. Background Technology

[0002] As the manufacturing industry undergoes a profound transformation towards flexibility and intelligence, production scenarios are evolving towards small-batch, multi-variety, and rapid line changeover models. This places higher demands on the operational capabilities of industrial robots: they not only need to complete complex processes such as welding and assembly, but also adapt to dynamically changing production environments and tasks. Traditional robot systems based on fixed workstations and pre-programmed procedures lack flexibility. Once the production line layout or the work object changes, a significant amount of time is required for re-teaching and debugging, severely restricting production efficiency and response speed.

[0003] To enhance adaptability, "mobile operating robots" integrating mobile platforms and multiple work units have become an important development direction. Such robots (like the system described in this invention with a wheeled chassis, humanoid torso, and multiple arms) can move freely within the workshop and perform diverse tasks by carrying different functional modules, theoretically significantly expanding their working range and flexibility. However, this high degree of physical freedom also brings unprecedented complexity in control and planning. The spatial poses, combinations, and collaborative relationships of the robot's various work modules can change constantly, forming a high-dimensional, dynamic, and strongly coupled system state. How to enable such systems to autonomously, efficiently, and reliably complete assembly line operations involving multiple processes presents a systemic challenge to existing technologies.

[0004] First, dynamic spatial perception and unified modeling are lacking. Existing mobile robot technology focuses on environmental map construction for navigation and obstacle avoidance, lacking a model that integrates the real-time precise poses of all robot operation modules, their specific functional types, and the inter-module collaborative constraints (such as distance and direction) required to complete the process. The robot cannot perceive its complete operational capability configuration and its spatial relationships in real time and accurately, resulting in a lack of reliable input for advanced task planning.

[0005] Secondly, the dynamic identification capability of process-oriented work units is insufficient. Faced with multiple available work modules, the system needs to automatically identify which modules can be combined and in what spatial relationship to form an effective work unit (e.g., welding and cooling modules require specific spacing for coordination). Existing methods mostly rely on pre-configured static settings and cannot dynamically match the embedded process knowledge base based on the real-time position and orientation of the modules, thus making it difficult to automatically reorganize the optimal work unit in changing environments.

[0006] Third, there is a lack of semantic understanding and generation mechanisms for the workspace. Robots need to understand "where to perform tasks." In flexible scenarios, the operational plane and the logistics plane may not be fixed. Existing technologies struggle to automatically infer and generate a virtual work cell layout that conforms to process logic from the identified combinations of work modules and their spatial orientation. Without this dynamic work map, the robot's motion planning lacks accurate spatial reference.

[0007] Fourth, the system suffers from weak resource scheduling and closed-loop execution capabilities in dynamic environments. When the operating environment and the system's own state change dynamically, how to optimally allocate tasks to various operational modules whose positions and attitudes change in real time, and generate precise control commands, constitutes a complex dynamic resource allocation problem. Existing scheduling algorithms struggle to handle such constraints. Furthermore, the lack of real-time monitoring and closed-loop feedback adjustment mechanisms for operational efficiency during execution makes it unable to cope with execution deviations, resulting in poor system robustness.

[0008] In summary, the existing technologies are relatively fragmented, making it difficult to support highly integrated, freely mobile robots to achieve truly environment-driven autonomous operation. Therefore, there is an urgent need for a systematic solution to achieve seamless integration across the entire chain, from multi-dimensional real-time perception to spatial cognition through the fusion of process knowledge, and then to dynamic optimization decision-making and closed-loop control. Summary of the Invention

[0009] In view of this, the present invention provides a method for operating a humanoid robot and the robot itself to solve the above-mentioned technical problems.

[0010] A method for operating a humanoid robot includes a robot system comprising a central controller, at least one rotatable (360-degree rotation) unit mounted on the humanoid structure, and at least two wheeled arms driven by the rotatable unit, each wheeled arm configured to carry one or more operating modules. The method includes:

[0011] (i) The central controller acquires the module spatial pose parameters of each of the work modules, the spatial pose parameters including its position and orientation in the robot coordinate system and / or the world coordinate system;

[0012] (ii) Provide a module type identifier for each of the job modules, and at least one of a module job pointing parameter and a job module collaboration distance parameter for each job module, wherein the module job pointing parameter represents the default or current job direction vector of the job module, and the job module collaboration distance parameter represents the expected job spacing between the module and the associated job module.

[0013] (iii) Provide one or more module operation reference distance parameter values, each reference distance parameter value is associated with a specific module type identifier and corresponds to the nominal distance between the operation end of the operation module of that type and a physical reference plane or operation object when the operation is performed;

[0014] (iv) The central controller dynamically determines a task space layout diagram of the current task environment based on at least one of the module operation reference distance parameter, the module spatial pose parameter, the module type identifier, the module operation pointing parameter, and the inter-module coordination distance parameter. The task space layout diagram at least identifies multiple operational units defined by physical or logical boundaries.

[0015] (v) The central controller assigns the multiple work modules or the wheel arms in which they are located to the set operation units according to the task space layout diagram, and generates a corresponding work instruction sequence to control the robot to perform assembly line operations including flipping, inspection, welding and packaging.

[0016] Furthermore, the dynamic determination of the task space layout diagram also includes the following steps performed by the central controller:

[0017] Construct a pose dataset for the multiple job modules, wherein the pose dataset contains the module spatial pose parameters and the module type identifier for each job module;

[0018] Multiple expected job combination templates are provided. Each expected job combination template is associated with at least one specific module type identifier and contains the expected value of at least one of the module job pointing parameter and the collaboration distance parameter between job modules. The expected job combination templates are predefined based on historical job data, process specifications or optimal job practices.

[0019] The central controller identifies one or more actual sets of job module combinations by matching the job modules in the pose dataset with the multiple expected job combination templates.

[0020] Furthermore, the central controller determines an expected spatial orientation for each of the identified sets of job module combinations. This spatial orientation is derived based on the module job pointing parameters in the expected job combination template associated with the combination and is used to characterize the preferred job orientation of the combination as a whole in space.

[0021] Furthermore, the central controller derives multiple candidate operating planes and multiple candidate conveying planes based on the identified set of work module combinations and their expected spatial orientation. Each candidate operating plane is associated with a fixed work area for performing assembly, welding, or inspection, and each candidate conveying plane is associated with a material flow path. Each plane is associated with at least one combination in the set of work module combinations and is defined by its spatial parameters.

[0022] Furthermore, determining the set of job module combinations in the pose dataset based on multiple expected job combination templates also includes:

[0023] Select a target template from the multiple expected job combination templates;

[0024] Based on the module type identifier, an initial subset of modules in the pose dataset is matched with the selected target template. This initial subset of modules contains modules of the type and number required by the target template.

[0025] Calculate one or more actual spatial distance values ​​between modules in the initial module subset;

[0026] The calculated one or more actual spatial distance values ​​are compared with the expected values ​​of the inter-module collaboration distance parameters defined in the target template to evaluate their matching degree; and

[0027] Based on whether the matching degree is within the preset tolerance range, it is decided whether to formally allocate the initial module subset and add it to the job module combination set;

[0028] Alternatively, determining the set of job module combinations in the pose dataset based on multiple expected job combination templates further includes:

[0029] Select a target template from the multiple expected job combination templates;

[0030] Based on the module type identifier, an initial subset of modules in the pose dataset is matched with the selected target template;

[0031] Based on the spatial pose parameters of each module in the initial module subset, calculate the actual spatial orientation parameters of the entire subset.

[0032] The calculated actual spatial orientation parameters are compared with the expected values ​​of the module operation pointing parameters defined in the target template to evaluate their directional consistency; and

[0033] Based on whether the directional consistency is within the preset angle tolerance range, it is determined whether to formally allocate the initial module subset and add it to the job module combination set.

[0034] Furthermore, the process by which the central controller determines at least one candidate operating plane and at least one candidate transport plane based on the set of combined work modules also includes:

[0035] For each combination in the set of combinations, a plane normal vector is calculated based on the spatial pose parameters of its member modules to define a plane in space; and

[0036] Based on the module operation pointing parameters or expected combination space orientation in the expected operation combination template corresponding to the combination, determine whether the plane defined by the combination is a candidate operation plane or a candidate transfer plane. The operation plane is usually related to the assembly surface at a horizontal or specific tilt angle, while the transfer plane is usually related to the horizontal transport direction.

[0037] The present invention also provides a humanoid work robot, the work robot comprising a wheeled chassis, a humanoid structure, a rotating part capable of 360-degree rotation, and at least two wheel arms driven by the rotating part, the wheel arms carrying multiple work modules, and the central controller comprising:

[0038] The data acquisition and reception module is used to acquire the module spatial pose parameters of each of the work modules through the pose sensor mounted on the arm or the robot body, and to receive, through the robot's communication interface, at least one of the module type identifier pre-stored or real-time reported by each of the work modules, as well as the module operation pointing parameter and the cooperative distance parameter between work modules determined based on the installation configuration or real-time sensing data of the work modules.

[0039] The parameter receiving module is used to receive one or more module operation reference distance parameter values ​​from a local policy library or a cloud process database. Each module operation reference distance parameter value is associated with a module type identifier and corresponds to the nominal distance between the working end of the operation module of that type and the physical reference surface or the operation object.

[0040] The layout parsing module is used to dynamically determine the task space layout diagram of the current task environment based on at least one of the module job reference distance parameters, the module spatial pose parameters, the module type identifier, the module job pointing parameters, and the inter-module coordination distance parameters, using a built-in layout parsing algorithm. This task space layout diagram identifies multiple operation units defined by physical or logical boundaries.

[0041] The task planning and allocation module is used to allocate the operation module or the wheel arm to a specific operation unit through the motion planning and task scheduling module according to the task space layout diagram, and generate an adaptive operation instruction sequence to control the robot to perform assembly line operations.

[0042] Furthermore, the layout parsing module includes:

[0043] The data preprocessing unit is used to construct a pose dataset of the multiple work modules based on the real-time data acquired by the data acquisition and receiving module. The pose dataset includes the module spatial pose parameters and module type identifier of each work module. It is also used to receive multiple expected work combination templates from the local policy library or cloud process database connected to the parameter receiving module. Each expected work combination template is associated with at least one of the module type identifiers and includes the expected value of at least one of the module work pointing parameters and the work module inter-cooperative distance parameters.

[0044] The combination matching and recognition unit is connected to the data preprocessing unit. It is used to match the job modules in the pose dataset with the multiple expected job combination templates by running the combination matching algorithm to identify the actual job module combination set. It is also used to calculate an expected combination spatial orientation for the combination based on the module spatial pose parameters of the member job modules of each combination in the identified job module combination set through geometric relationships.

[0045] A candidate plane generation unit, connected to the combination matching and identification unit, is used to derive multiple candidate operation planes and multiple candidate transport planes based on the set of job module combinations and their expected combination spatial orientation using a plane fitting algorithm. Each candidate operation plane or candidate transport plane is associated with at least one combination in the set of job module combinations.

[0046] A planar optimization and filtering unit, connected to the candidate planar generation unit, is used to perform the following optimization and filtering operations:

[0047] The normal vectors of each candidate operation plane are calculated by vector operations, and their directional similarity is compared by calculating the dot product or cosine of the angle between the normal vectors. Based on a preset directional parallelism tolerance threshold, candidate operation planes with directional similarity lower than the threshold are marked or temporarily excluded.

[0048] The direction vector of each candidate transmission plane is calculated by the main direction extraction algorithm. The direction alignment is compared by vector comparison. The angle between the candidate transmission plane and the candidate operation plane is calculated and compared by vector angle. According to the preset alignment tolerance threshold and orthogonality tolerance threshold, the candidate transmission plane with insufficient alignment or orthogonality exceeding the allowable range is marked or temporarily excluded.

[0049] The layout element determination unit, connected to the plane optimization and filtering unit, is used to determine the precise position of at least one of the operation plane layout elements and the transmission plane layout elements in the task space layout diagram based on the filtered and retained candidate operation planes and candidate transmission planes, as well as the module operation reference distance parameters of the associated operation modules, and in combination with the module spatial pose parameters, through spatial offset calculation.

[0050] The synthesis unit, connected to the layout element determination unit, is used to determine multiple operation plane layout elements and multiple transport plane layout elements through clustering and merging algorithms, and to synthesize a complete task space layout diagram through a topology construction algorithm based on the spatial relative relationship between the multiple operation plane layout elements and the transport plane layout elements.

[0051] Furthermore, the combined matching and recognition unit also includes:

[0052] The parameter processing subunit is used to receive or read the module job pointing parameters from the local policy library through the communication interface in the data acquisition and receiving module. The module job pointing parameters are used to determine whether each job module in a job module combination is expected to be deployed on an operation plane layout element or a conveying plane layout element. The operation plane layout element corresponds to a vertical or inclined operation plane, and the conveying plane layout element corresponds to a horizontal material flow plane.

[0053] The orientation matching subunit is used to use the module operation orientation parameter as a key matching condition when running the combined matching algorithm, so as to ensure that the identified combination conforms to the predefined operation logic in terms of spatial orientation.

[0054] Furthermore, the task planning and allocation module includes:

[0055] The status monitoring and performance evaluation unit is used to continuously monitor the status of each operation unit in the task space layout diagram and the operation data of the operation modules assigned to them through the pose sensor in the data acquisition and reception module and the feedback sensor of the operation module after the adaptive operation instruction sequence is generated. Based on the monitoring data, the unit dynamically determines whether the actual operation performance of at least one operation module in its assigned operation unit has reached the expected level through the performance evaluation algorithm.

[0056] An adaptive adjustment unit, connected to the status monitoring and performance evaluation unit, is used to adaptively change the operation mode of at least one work module and / or the boom that drives it through a parameter adjustment interface when it is determined that the work performance has not met expectations, according to a preset optimization strategy. The change in operation mode includes adjusting the work speed, the force applied, the accuracy level of the motion trajectory, or enabling a collaborative work strategy with other modules in the same unit.

[0057] The coordinate mapping unit is used to associate and map the task space layout diagram with the real-time spatial pose parameters of each job module acquired by the data acquisition and receiving module after the layout parsing module synthesizes the task space layout diagram.

[0058] A parameter set invocation unit, connected to the coordinate mapping unit, is used to automatically invoke and load a matching dedicated job parameter set from the local policy library for each of the plurality of job modules, based on its mapping position in the task space layout diagram; and

[0059] The real-time control unit, connected to the parameter set calling unit, is used to control the operation mode of the corresponding operation module in real time through the underlying drive controller according to the dedicated operation parameter set during the execution of the assembly line operation, thereby realizing the assembly line operation.

[0060] The layout analysis module dynamically constructs a task space layout diagram identifying multiple operating units based on real-time acquired module spatial pose parameters, module type identifiers, module operation pointing parameters, and collaborative distance parameters between operating modules. It does not rely on pre-set, fixed workstation coordinates but can generate an online understanding model of the work space based on the robot's own module distribution status and environmental perception data. Based on this dynamic layout diagram, the task planning and allocation module can flexibly assign operating modules or arms to the most suitable operating units. The robot can autonomously adapt to temporary adjustments in the production line layout, changes in workstation occupancy, or material position shifts without interrupting production for manual reprogramming, thus significantly enhancing the system's adaptability to unstructured and semi-structured scenarios.

[0061] The expected job combination templates are predefined based on historical data, process specifications, or best practices, and include constraints such as module type, expected job orientation, and coordination distance. The combination matching and recognition unit matches the real-time pose dataset against these templates. The matching process compares the actual spatial distance value with the coordination distance parameter between the expected job modules, or compares the actual spatial orientation parameter with the expected module job orientation parameter, and makes a judgment based on a preset tolerance range. It transforms abstract process knowledge and best practices into computable matching rules, ensuring that the job module combination set identified by the system is not only functionally correct but also strictly conforms to process requirements in terms of spatial geometry.

[0062] After identifying the combination of work modules, the system determines an expected spatial orientation for each combination. Then, the candidate plane generation unit, based on the spatial parameters of each combination, derives candidate operation planes and candidate transport planes through a plane fitting algorithm. It determines the plane type by calculating the plane normal vector and considering the module's work orientation parameters or the combined spatial orientation. This allows for comprehensive consideration of the spatial relationships between different functional areas such as the assembly area, inspection area, and material flow channels, enabling global collaborative planning that includes obstacle avoidance, path optimization, and process integration. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 This is a general flowchart of the method of the system of the present invention;

[0065] Figure 2 This is a flowchart of the method for dynamically determining the task space layout diagram as described in the present invention, executed by the central controller.

[0066] Figure 3 This is a flowchart of a method for determining a set of job module combinations in a pose dataset based on multiple expected job combination templates, as described in this invention.

[0067] Figure 4 This is a flowchart of another method for determining the set of job module combinations in a pose dataset based on multiple expected job combination templates, as described in this invention.

[0068] Figure 5 This is a schematic diagram of the system framework principle of the present invention;

[0069] Figure 6 The robot system provided by the present invention. Implementation

[0070] To better understand the structure, functional features, and advantages of the present invention, preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings:

[0071] See attached document Figure 1-5A method for operating a humanoid robot includes a robot system comprising a central controller; a control platform 2; a humanoid structure mounted on the control platform 2; a wheel set 1 at the bottom of the control platform; at least one rotatable rotating part mounted on the humanoid structure; and at least two wheel arms 4 driven by the rotating part, each wheel arm 4 configured to carry one or more operating modules 1. The operating method includes: (i) obtaining module spatial pose parameters of each operating module from the central controller, the spatial pose parameters including its position and orientation in the robot coordinate system and / or world coordinate system; (ii) providing a module type identifier for each operating module, and at least one of a module operation pointing parameter and an inter-module cooperative distance parameter for each operating module, wherein the module operation pointing parameter represents the default or current operation direction vector of the operating module, and the inter-module cooperative distance parameter represents the module's relationship with associated operations. (iii) Providing one or more module operation reference distance parameter values, each reference distance parameter value being associated with a specific module type identifier and corresponding to the nominal distance between the working end of the module of that type and a physical reference plane or working object when performing the operation; (iv) The central controller dynamically determines the task space layout diagram of the current task environment based on at least one of the module operation reference distance parameters, the module spatial pose parameters, the module type identifier, the module operation pointing parameters, and the cooperation distance parameters between operation modules, the task space layout diagram identifying at least a plurality of operation units defined by physical boundaries or logical boundaries; and (v) The central controller assigns the plurality of operation modules or the arm in which they are located to the set operation units based on the task space layout diagram, and generates corresponding operation instruction sequences to control the robot to perform assembly line operations including flipping, inspection, welding, and packaging.

[0072] In the above embodiments, step (iv) further includes dynamically determining the task space layout diagram: calculating the spatial interference risk index between each operation module in real time based on the module spatial pose parameters. The risk index includes at least the minimum distance between modules, the number of predicted intersection points of motion trajectories, and the overlap rate of the module motion envelope volume. The risk index is compared with a preset safety threshold. If any index exceeds the threshold, a high-risk area is dynamically marked in the task space layout diagram. When generating the operation instruction sequence in step (v), an avoidance trajectory is automatically inserted or the operation sequence is adjusted to avoid interference. At the same time, the high-risk area information is alerted to the operator in real time through the human-machine interface, and at least one manual intervention suggestion is provided.

[0073] Following step (v), during the execution of the work instruction sequence, dynamic feedback data during the actual work process is continuously collected using force sensors, vision sensors, or acoustic sensors integrated on the work module. Based on the dynamic feedback data, the current work quality assessment index is calculated in real time, including at least one of weld seam deviation, packaging seal integrity, and detection false alarm rate. If the quality assessment index is lower than the preset standard for multiple consecutive work cycles, adaptive replanning is triggered to dynamically adjust the boundary definitions of relevant operation units in the task space layout diagram or modify the parameter set in the work instruction sequence until the quality index recovers to the standard.

[0074] Furthermore, the dynamic determination of the task space layout diagram also includes the following steps performed by the central controller: constructing a pose dataset of the plurality of work modules, the pose dataset containing the module space pose parameters and the module type identifier of each work module; providing a plurality of expected work combination templates, each expected work combination template being associated with at least one specific module type identifier and containing an expected value of at least one of the module work pointing parameters and the work module inter-cooperation distance parameters, the expected work combination templates being predefined based on historical work data, process specifications, or optimal work practices; and the central controller identifying one or more actual work module combination sets by matching the work modules in the pose dataset with the plurality of expected work combination templates.

[0075] In the above embodiments, during the matching process, based on the environmental characteristics description of the current task, the top N template versions with the highest scene matching degree are selected from the cloud as candidates; through a lightweight simulation engine, the execution effect of each candidate template under the current actual module pose is quickly simulated in a virtual environment, and the predicted job efficiency is calculated based on the simulation results; finally, the template with the highest comprehensive score is selected as the actual matching basis, and the job data is anonymized and fed back to the cloud for continuous optimization and iteration of the template.

[0076] The matching step further includes: analyzing the historical trajectory, extracting the motion stability, repeatability accuracy, and trend drift features of the module; during matching, assigning lower matching confidence weights to modules with poor motion stability or obvious drift, and prioritizing combinations composed of high-confidence modules in the combined recognition; at the same time, automatically suggesting maintenance checks or calibration for low-confidence modules, and temporarily assigning them to operation units with lower accuracy requirements before their pose stabilizes.

[0077] In the above embodiments, one or more constraint parameters are defined for each expected job combination template. The constraints include the acceptable range of distance between modules, the allowable deviation range of orientation, and the flexibility level of job sequence. During matching, in addition to checking whether the basic constraints of the template are met, the degree to which the current module subset satisfies each constraint is calculated, and a comprehensive fit score is generated. The strictness of the constraints can be dynamically adjusted according to the real-time job load.

[0078] Furthermore, the central controller determines an expected spatial orientation for each of the identified sets of job module combinations. This spatial orientation is derived based on the module job pointing parameters in the expected job combination template associated with the combination and is used to characterize the preferred job orientation of the combination as a whole in space.

[0079] The expected spatial orientation of the combination is determined as follows: statistical analysis is performed on the module operation pointing parameters of all operation modules in the combination, and their average direction vector is calculated. This average vector is used as the initial spatial orientation of the combination. Considering the pointing weight of the core module that plays a dominant role in the combination, a higher weight is assigned to the pointing of the core module, and the weighted average direction is recalculated. This weighted average direction is compared with the typical orientation of such combinations in the process database. If the deviation exceeds the historical range, manual confirmation or automatic correction to the closest typical orientation is triggered.

[0080] In the above embodiments, a dynamic priority score is assigned to each identified combination. This score is calculated based on the alignment of the combination's spatial orientation with the current main task direction, the health status of the modules within the combination, and the combination's historical success rate. In subsequent planar derivation and task allocation, the planes and operation units associated with high-priority combinations are processed first to ensure the smooth operation of critical task paths and priority resource allocation. The priority score is dynamically updated during operation based on actual performance feedback, forming a closed-loop optimization. When multiple combinations are identified as spatially adjacent and their expected spatial orientations conflict, an orientation negotiation mechanism is automatically initiated. This mechanism slightly adjusts the pose of the modules within each combination (within the allowable range of arm movement) or adjusts the rotation angle of its respective arm to achieve overall coordination in the final spatial orientation of each combination, reducing mutual occlusion or interference. The negotiation results are presented in a three-dimensional visualization and can only be applied after operator confirmation or system security verification.

[0081] Furthermore, the central controller derives multiple candidate operating planes and multiple candidate conveying planes based on the identified set of work module combinations and their expected spatial orientation. Each candidate operating plane is associated with a fixed work area for performing assembly, welding, or inspection, and each candidate conveying plane is associated with a material flow path. Each plane is associated with at least one combination in the set of work module combinations and is defined by its spatial parameters.

[0082] In the above embodiments, when deriving candidate operation planes and candidate transport planes: for each combination, not only based on the spatial pose of its member modules, but also combined with its module type identifier, the plane type and spatial attributes (such as plane size and required buffer range) that such combinations usually correspond to are queried from the technology library.

[0083] Using the retrieved attributes as prior knowledge, the plane fitting algorithm is guided to improve the process rationality and dimensional accuracy of the derived plane. For the conveying plane, the main material flow direction arrows on it are additionally derived and marked in the layout diagram. After deriving the candidate planes, spatial topology analysis is performed on all planes to identify the connection points, intersection areas, and height differences between planes. Based on the analysis results, virtual material transfer nodes, safety guardrail areas, and robot track changing suggestion points are automatically added to the layout diagram.

[0084] Furthermore, determining the set of job module combinations in the pose dataset based on multiple expected job combination templates further includes: selecting a target template from the multiple expected job combination templates; matching an initial subset of modules in the pose dataset with the selected target template based on a module type identifier, wherein the initial subset of modules contains modules of the type and number required by the target template; calculating one or more actual spatial distance values ​​between modules in the initial subset of modules; comparing the calculated one or more actual spatial distance values ​​with the expected values ​​of the collaborative distance parameters between job modules defined in the target template to evaluate their matching degree; and deciding whether to formally allocate and add the initial subset of modules to the set of job module combinations based on whether the matching degree is within a preset tolerance range.

[0085] Alternatively, determining the set of job module combinations in the pose dataset based on multiple expected job combination templates further includes: selecting a target template from the multiple expected job combination templates; matching an initial subset of modules in the pose dataset with the selected target template based on a module type identifier; calculating the actual spatial orientation parameter of the subset as a whole based on the spatial pose parameters of each module in the initial subset; comparing the calculated actual spatial orientation parameter with the expected value of the module job pointing parameter defined in the target template to evaluate its directional consistency; and deciding whether to formally allocate and add the initial subset of modules to the set of job modules based on whether the directional consistency is within a preset angle tolerance range.

[0086] In the above, the matching degree evaluation adopts fuzzy logic. Specifically, the difference between the actual spatial distance value and the expected value, and the angle difference between the actual spatial orientation and the expected value are used as fuzzy input variables. Through the preset membership function and fuzzy rule base (such as "if the distance difference is small and the angle difference is small, the matching degree is high"), a comprehensive matching degree value between 0 and 1 is calculated. This matching degree value is not only used for the binary decision of whether to add it to the set, but also serves as the confidence weight of the combination in the subsequent task allocation, affecting the complexity and accuracy requirements of the assigned task.

[0087] When a match based on a target template fails (i.e., the initial module subset fails to be formally added to the set), the system automatically records the reason for the failure (distance deviation or orientation inconsistency); analyzes the pattern of the failure reasons, and if multiple failures are due to the pose abnormality of the same module, the module is marked, and a special diagnostic process for the module is triggered, and the diagnostic results are associated with the matching log.

[0088] Furthermore, the process by which the central controller determines at least one candidate operating plane and at least one candidate transport plane based on the set of work module combinations further includes: for each combination in the set of combinations, calculating a plane normal vector for spatially defining a plane based on the spatial pose parameters of its member modules; and determining whether the plane defined by the combination is a candidate operating plane or a candidate transport plane based on the module operation pointing parameters or the expected combination spatial orientation in the expected work combination template corresponding to the combination, wherein the operating plane is typically associated with a horizontal or specific tilt angle assembly surface, and the transport plane is typically associated with a horizontal transport direction.

[0089] For the endpoint poses of all member modules within a combination, the least median squares method is used for plane fitting to resist the interference of individual module pose outliers; after successful fitting, the goodness of fit of the plane (such as the average distance from the point to the plane) is calculated.

[0090] This invention also provides a humanoid work robot, comprising a wheeled chassis, a humanoid structure, a 360-degree rotatable rotating part, and at least two wheel arms driven by the rotating part. Each wheel arm carries multiple work modules. The central controller includes: a data acquisition and reception module, used to acquire the module spatial pose parameters of each work module via pose sensors mounted on the wheel arms or the robot body, and to receive, through the robot's communication interface, a module type identifier pre-stored or reported in real-time by each work module, and at least one of a module work pointing parameter and a work module collaborative distance parameter determined based on the work module's installation configuration or real-time sensing data; and a parameter receiving module, used to receive one or more module work reference distance parameter values ​​from a local strategy library or a cloud-based process database, for each module... The operation reference distance parameter value is associated with a module type identifier and corresponds to the nominal distance between the operation end of the operation module of that type and the physical reference surface or the operation object; the layout parsing module is used to dynamically determine the task space layout diagram of the current task environment through a built-in layout parsing algorithm based on at least one of the module operation reference distance parameter, the module spatial pose parameter, the module type identifier, the module operation pointing parameter, and the cooperation distance parameter between operation modules. The task space layout diagram identifies multiple operation units defined by physical boundaries or logical boundaries; and the task planning and allocation module is used to allocate the operation module or arm to a specific operation unit through the motion planning and task scheduling module according to the task space layout diagram, and generate an adaptive operation instruction sequence to control the robot to perform pipeline operations.

[0091] Further, the layout parsing module includes: a data preprocessing unit, used to construct a pose dataset of the multiple work modules based on real-time data acquired by the data acquisition and receiving module, the pose dataset containing module spatial pose parameters and module type identifiers for each work module, and used to receive multiple expected job combination templates from a local policy library or cloud process database connected to the parameter receiving module, each expected job combination template being associated with at least one of the module type identifiers and containing expected values ​​of at least one of the module job pointing parameters and the inter-work module cooperative distance parameters; and a combination matching and recognition unit, connected to the data preprocessing unit, used to convert the pose dataset into a single target by running a combination matching algorithm. The system matches the collected task modules with the multiple expected task combination templates to identify the actual task module combination set. Based on the module spatial pose parameters of the member task modules in each combination within the identified task module combination set, it calculates an expected combination spatial orientation for the combination using geometric relationships. A candidate plane generation unit, connected to the combination matching and identification unit, derives multiple candidate operation planes and multiple candidate transport planes based on the task module combination set and its expected combination spatial orientation using a plane fitting algorithm. Each candidate operation plane or candidate transport plane is associated with at least one combination in the task module combination set. A plane optimization and filtering unit, connected to the candidate plane... The face generation unit performs the following optimization and filtering operations: Calculates the normal vector of each candidate operation plane through vector operations, compares their directional similarity by calculating the dot product or cosine of the angle between the normal vectors, and marks or temporarily excludes candidate operation planes with directional similarity below a preset directional parallelism tolerance threshold; calculates the direction vector of each candidate transmission plane through a main direction extraction algorithm, compares their directional alignment by vector comparison, and compares the angle between the candidate transmission plane and the candidate operation plane by calculating the vector angle, and marks or temporarily excludes candidate transmission planes with insufficient alignment or orthogonality exceeding the allowable range based on preset alignment tolerance thresholds and orthogonality tolerance thresholds; and determines layout elements. A unit, connected to the plane optimization and filtering unit, is used to determine the precise position of at least one of the operation plane layout elements and the transport plane layout elements in the task space layout diagram based on the filtered and retained candidate operation planes and candidate transport planes, as well as the module operation reference distance parameters of the associated operation modules, and in combination with the module spatial pose parameters, through spatial offset calculation; a synthesis unit, connected to the layout element determination unit, is used to determine multiple operation plane layout elements and multiple transport plane layout elements through clustering and merging algorithms, and synthesize the complete task space layout diagram through topology construction algorithm based on the spatial relative relationships between the multiple operation plane layout elements and the transport plane layout elements.

[0092] Furthermore, the combination matching and identification unit further includes: a pointing parameter processing subunit, used to receive or read the module job pointing parameters from the local policy library through the communication interface in the data acquisition and receiving module, wherein the module job pointing parameters are used to determine whether each job module in a job module combination is expected to be deployed on an operation plane layout element or a conveying plane layout element, wherein the operation plane layout element corresponds to a vertical or inclined operation plane, and the conveying plane layout element corresponds to a horizontal material flow plane; and an orientation matching subunit, used to use the module job pointing parameters as a key matching condition when running the combination matching algorithm to ensure that the identified combination conforms to the predefined job logic in terms of spatial orientation.

[0093] Furthermore, the task planning and allocation module includes: a status monitoring and performance evaluation unit, used to continuously monitor the status of each operation unit within the task space layout diagram and the operation data of the operation modules assigned to them through the pose sensor in the data acquisition and reception module and the feedback sensor of the operation module after generating the adaptive operation instruction sequence, and dynamically determine whether the actual operation performance of at least one operation module in its assigned operation unit has reached the expected level based on the monitoring data and through a performance evaluation algorithm; and an adaptive adjustment unit, connected to the status monitoring and performance evaluation unit, used to adaptively change the operation mode of the at least one operation module and / or the arm driving it through a parameter adjustment interface according to a preset optimization strategy when the operation performance is determined not to have reached the expected level, including adjusting the operation speed. The system includes: a force and accuracy level of motion trajectory, or enabling collaborative operation strategies with other modules within the same unit; a coordinate mapping unit, used to associate and map the task space layout diagram with the real-time spatial pose parameters of each operation module acquired by the data acquisition and receiving module after the layout parsing module synthesizes the task space layout diagram; a parameter set calling unit, connected to the coordinate mapping unit, used to automatically call and load a matching dedicated operation parameter set from the local strategy library for each of the multiple operation modules according to its mapping position in the task space layout diagram; and a real-time control unit, connected to the parameter set calling unit, used to control the operation mode of the corresponding operation module in real time through the underlying drive controller according to the dedicated operation parameter set during the execution of the assembly line operation, thereby realizing the assembly line operation.

[0094] The central controller actively collects two types of key information through its data acquisition and reception modules. The first type is the module spatial pose parameters, namely the precise position and orientation of each working module mounted on the end arm in space. This pose data provides the geometric basis for all subsequent spatial calculations. The second type is the attribute and relationship parameters of the working modules, including module type identifiers that identify their functions, module working direction parameters that define their working direction tendency obtained from configuration or sensors, and inter-module cooperative distance parameters that define their cooperative spatial relationships with other modules. Simultaneously, the parameter receiving module imports preset module working reference distance parameter values, which define the standard distance that each type of working module should maintain between its working end arm and the surface of the work object.

[0095] After obtaining the basic data, each template is associated with a specific module type, and the expected spatial relationships (such as distance or orientation) when these modules work together are defined. The real-time constructed pose dataset is compared with multiple expected job combination templates. The matching process is not a simple type check, but a strongly constrained matching that includes spatial relationship verification. Verification methods include: calculating the actual spatial distance values ​​between modules in the candidate module subset and comparing them with the expected values ​​of the collaborative distance parameters between job modules in the template to evaluate whether the matching degree is within a preset tolerance range; or, calculating the actual spatial orientation parameters of the subset and comparing them with the expected values ​​of the module job orientation parameters in the template to evaluate whether the directional consistency is within a preset angle tolerance range. The module subset that passes the verification is formally identified as a job module combination set, and an expected combined spatial orientation is determined for it. This principle realizes the transformation from low-level geometric data to high-level job semantics.

[0096] After identifying the job combinations, each job module combination contains a specific working area in space due to the spatial pose parameters of its member modules and the overall combined spatial orientation. Based on these parameters, the candidate plane generation unit uses a plane fitting algorithm to derive candidate operation planes (for fixed operations such as assembly and welding) or candidate transfer planes (for material flow) for each combination. Next, the plane optimization and screening unit optimizes these candidate planes according to process logic, screening operation planes by calculating plane normal vectors and comparing directional similarity; and screening transfer planes by calculating direction vectors and comparing directional alignment and the angle with the operation plane. Finally, the layout element determination unit uses module job reference distance parameters to calculate the spatial offset of the screened planes, determining their precise position in the global coordinate system. Then, the synthesis unit uses clustering and merging algorithms and topology construction algorithms to synthesize all layout elements into a task space layout diagram that defines multiple operation units.

[0097] After generating the task space layout diagram, the system enters the task execution phase. The task planning and allocation module assigns each work module or its corresponding arm to a specific operating unit in the diagram, generating a corresponding sequence of work instructions. To ensure accurate instruction execution, the coordinate mapping unit maps the virtual coordinates of the layout diagram to the real-time spatial pose parameters of each work module. The parameter set invocation unit automatically retrieves the matching dedicated work parameter set from the strategy library based on the mapped position of each module in the layout diagram. During pipeline operation, the real-time control unit uses these dedicated parameter sets to control the work modules in real-time through the underlying drive controller.

[0098] The status monitoring and performance evaluation unit continuously monitors the status of each operating unit and the module operation data, and dynamically determines whether the operation performance meets expectations through a performance evaluation algorithm. When the performance does not meet expectations, the adaptive adjustment unit will adaptively change the operating mode of the relevant operating module or boom through a parameter adjustment interface according to a preset optimization strategy.

[0099] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Any person skilled in the art can make many possible variations and modifications to the technical solutions of the present invention, or modify them into equivalent embodiments, without departing from the scope of the present invention. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technology of the present invention without departing from the scope of the present invention are within the protection scope of the present invention.

Claims

1. A method for operating a humanoid robot, characterized in that, The system includes a robot system comprising a central controller, at least one rotatable rotating part disposed on the humanoid structure, and at least two wheeled arms driven by the rotating part, each wheeled arm being configured to carry one or more work modules, the work method comprising: (i) The central controller acquires the module spatial pose parameters of each of the work modules, the spatial pose parameters including its position and orientation in the robot coordinate system and / or the world coordinate system; (ii) Provide a module type identifier for each of the job modules, and at least one of a module job pointing parameter and a job module collaboration distance parameter for each job module, wherein the module job pointing parameter represents the default or current job direction vector of the job module, and the job module collaboration distance parameter represents the expected job spacing between the module and the associated job module. (iii) Provide one or more module operation reference distance parameter values, each reference distance parameter value is associated with a specific module type identifier and corresponds to the nominal distance between the operation end of the operation module of that type and a physical reference plane or operation object when the operation is performed; (iv) The central controller dynamically determines a task space layout diagram of the current task environment based on at least one of the module operation reference distance parameter, the module spatial pose parameter, the module type identifier, the module operation pointing parameter, and the inter-module coordination distance parameter. The task space layout diagram at least identifies multiple operational units defined by physical or logical boundaries. (v) The central controller assigns the multiple work modules or the wheel arms in which they are located to the set operation units according to the task space layout diagram, and generates a corresponding work instruction sequence to control the robot to perform assembly line operations including flipping, inspection, welding and packaging.

2. The operation method of the humanoid robot according to claim 1, characterized in that, The dynamic determination of the task space layout diagram also includes the following steps performed by the central controller: constructing a pose dataset of the multiple job modules, wherein the pose dataset contains the module space pose parameters and the module type identifier of each job module; Multiple expected job combination templates are provided. Each expected job combination template is associated with at least one specific module type identifier and contains the expected value of at least one of the module job pointing parameter and the collaboration distance parameter between job modules. The expected job combination templates are predefined based on historical job data, process specifications or optimal job practices. The central controller identifies one or more actual sets of job module combinations by matching the job modules in the pose dataset with the multiple expected job combination templates.

3. The operation method of the humanoid robot according to claim 2, characterized in that, Also includes: The central controller determines an expected spatial orientation for each of the identified task module combination sets. This spatial orientation is derived based on the module task pointing parameters in the expected task combination template associated with the combination and is used to characterize the preferred task orientation of the combination as a whole in space.

4. The method according to claim 3, characterized in that, Also includes: The central controller derives multiple candidate operation planes and multiple candidate transfer planes based on the identified set of work module combinations and their expected spatial orientation. Each candidate operation plane is associated with a fixed work area for performing assembly, welding, or inspection, and each candidate transfer plane is associated with a material flow path. Each plane is associated with at least one combination in the set of work module combinations and is defined by its spatial parameters.

5. The operation method of the humanoid robot according to claim 4, characterized in that, When the central controller determines at least one candidate operating plane and at least one candidate transport plane based on the set of work module combinations, it includes: for each combination in the set of combinations, calculating a plane normal vector for spatially defining a plane based on the spatial pose parameters of its member modules; and determining whether the plane defined by the combination is a candidate operating plane or a candidate transport plane based on the module operation pointing parameters or the expected combination spatial orientation in the expected work combination template corresponding to the combination, wherein the operating plane is usually associated with a horizontal or specific tilt angle assembly surface, and the transport plane is usually associated with a horizontal transport direction.

6. The operation method of the humanoid robot according to claim 2, characterized in that, The step of determining the set of job module combinations in the pose dataset based on multiple expected job combination templates further includes: selecting a target template from the multiple expected job combination templates; Based on the module type identifier, an initial subset of modules in the pose dataset is matched with the selected target template. The initial subset of modules contains modules of the type and number required by the target template. One or more actual spatial distance values ​​between modules in the initial subset of modules are calculated. The calculated one or more actual spatial distance values ​​are compared with the expected values ​​of the collaborative distance parameters between work modules defined in the target template to evaluate their matching degree; And based on whether the matching degree is within a preset tolerance range, decide whether to formally allocate the initial module subset and add it to the job module combination set; or, the step of determining the job module combination set in the pose dataset according to multiple expected job combination templates further includes: selecting a target template from the multiple expected job combination templates; Based on the module type identifier, an initial module subset in the pose dataset is matched with the selected target template; based on the spatial pose parameters of each module in the initial module subset, the actual spatial orientation parameters of the subset as a whole are calculated; the calculated actual spatial orientation parameters are compared with the expected values ​​of the module operation pointing parameters defined in the target template to evaluate their directional consistency. And based on whether the directional consistency is within the preset angle tolerance range, decide whether to formally allocate the initial module subset and add it to the job module combination set.

7. A humanoid work robot, characterized in that, The work robot includes a wheeled chassis, a humanoid structure, a 360-degree rotatable rotating part, and at least two wheel arms driven by the rotating part. The wheel arms are equipped with multiple work modules. The central controller includes a data acquisition and reception module, which is used to acquire the module spatial pose parameters of each work module through a pose sensor mounted on the wheel arm or the robot body, and to receive, through the robot's communication interface, a module type identifier pre-stored or reported in real time from each work module, and at least one of the module work pointing parameters and the work module cooperative distance parameters determined based on the installation configuration or real-time sensing data of the work module. The parameter receiving module is used to receive one or more module operation reference distance parameter values ​​from a local policy library or a cloud process database. Each module operation reference distance parameter value is associated with a module type identifier and corresponds to the nominal distance between the working end of the operation module of that type and the physical reference plane or the operation object. The layout parsing module is used to dynamically determine the task space layout diagram of the current task environment based on at least one of the module operation reference distance parameters, the module spatial pose parameters, the module type identifier, the module operation pointing parameters, and the cooperation distance parameters between operation modules, using a built-in layout parsing algorithm. The task space layout diagram identifies multiple operation units defined by physical boundaries or logical boundaries. The system also includes a task planning and allocation module, which, based on the task space layout diagram, allocates the operation module or the wheel arm to a specific operation unit through the motion planning and task scheduling module, and generates an adaptive operation instruction sequence to control the robot to perform assembly line operations.

8. The humanoid work robot according to claim 7, characterized in that, The layout parsing module includes: a data preprocessing unit, used to construct a pose dataset of the multiple work modules based on real-time data acquired by the data acquisition and receiving module. The pose dataset contains module spatial pose parameters and module type identifiers for each work module. The preprocessing unit is used to receive multiple expected work combination templates from a local policy library or cloud process database connected to the parameter receiving module. Each expected work combination template is associated with at least one module type identifier and contains expected values ​​of at least one of the module work pointing parameters and the inter-work module collaboration distance parameters. The combination matching and recognition unit is connected to the data preprocessing unit. It is used to match the job modules in the pose dataset with the multiple expected job combination templates by running the combination matching algorithm to identify the actual job module combination set. It is also used to calculate an expected combination spatial orientation for the combination based on the module spatial pose parameters of the member job modules of each combination in the identified job module combination set through geometric relationships. A candidate plane generation unit, connected to the combination matching and identification unit, is used to derive multiple candidate operation planes and multiple candidate transport planes based on the set of job module combinations and their expected combination spatial orientation using a plane fitting algorithm. Each candidate operation plane or candidate transport plane is associated with at least one combination in the set of job module combinations. A plane optimization and filtering unit, connected to the candidate plane generation unit, is used to perform the following optimization and filtering operations: calculating the normal vector of each candidate operation plane through vector operations, comparing their directional similarity by calculating the dot product or cosine of the included angle between the normal vectors, and marking or temporarily excluding candidate operation planes with directional similarity lower than a preset directional parallelism tolerance threshold according to a preset directional parallelism tolerance threshold. The direction vector of each candidate transmission plane is calculated by the main direction extraction algorithm. The direction alignment is compared by vector comparison. The angle between the candidate transmission plane and the candidate operation plane is calculated and compared by vector angle. According to the preset alignment tolerance threshold and orthogonality tolerance threshold, the candidate transmission plane with insufficient alignment or orthogonality exceeding the allowable range is marked or temporarily excluded. The layout element determination unit, connected to the plane optimization and filtering unit, is used to determine the precise position of at least one of the operation plane layout elements and the transmission plane layout elements in the task space layout diagram based on the filtered and retained candidate operation planes and candidate transmission planes, as well as the module operation reference distance parameters of the associated operation modules, and in combination with the module spatial pose parameters, through spatial offset calculation. The synthesis unit, connected to the layout element determination unit, is used to determine multiple operation plane layout elements and multiple transport plane layout elements through clustering and merging algorithms, and to synthesize a complete task space layout diagram through a topology construction algorithm based on the spatial relative relationship between the multiple operation plane layout elements and the transport plane layout elements.

9. The humanoid work robot according to claim 8, characterized in that, The combination matching and identification unit further includes: a pointer parameter processing subunit, used to receive or read the module job pointer parameter from the local policy library through the communication interface in the data acquisition and receiving module, wherein the module job pointer parameter is used to determine whether each job module in a job module combination is expected to be deployed on an operation plane layout element or a conveying plane layout element, the operation plane layout element corresponds to a vertical or inclined operation plane, and the conveying plane layout element corresponds to a horizontal material flow plane; The orientation matching subunit is used to use the module operation orientation parameter as a key matching condition when running the combined matching algorithm, so as to ensure that the identified combination conforms to the predefined operation logic in terms of spatial orientation.

10. The humanoid work robot according to claim 7, characterized in that, The task planning and allocation module includes a status monitoring and performance evaluation unit, which, after generating an adaptive work instruction sequence, continuously monitors the status of each operation unit in the task space layout diagram and the operation data of the work modules assigned to them through the pose sensor in the data acquisition and reception module and the feedback sensor of the work module, and dynamically determines, based on the monitoring data, whether the actual work performance of at least one work module in its assigned operation unit has reached the expected level through a performance evaluation algorithm. An adaptive adjustment unit, connected to the status monitoring and performance evaluation unit, is used to adaptively change the operation mode of at least one work module and / or the boom that drives it through a parameter adjustment interface when it is determined that the work performance has not met expectations, according to a preset optimization strategy. The change in operation mode includes adjusting the work speed, the force applied, the accuracy level of the motion trajectory, or enabling a collaborative work strategy with other modules in the same unit. A coordinate mapping unit is used to associate and map the task space layout diagram with the real-time spatial pose parameters of each job module acquired by the data acquisition and receiving module after the layout parsing module synthesizes the task space layout diagram; a parameter set calling unit is connected to the coordinate mapping unit and is used to automatically call and load a matching dedicated job parameter set from the local policy library for each of the plurality of job modules according to its mapping position in the task space layout diagram; and a real-time control unit is connected to the parameter set calling unit and is used to control the operation mode of the corresponding job module in real time through the underlying drive controller according to the dedicated job parameter set during the execution of the pipeline operation, thereby realizing the pipeline operation.

Images