Multi-scene robot unified control and intelligent cooperation method and system

Abstract

The invention discloses a unified control and intelligent cooperation method and system for robots in multiple scenes, and belongs to the technical field of robot control. The method comprises the steps of performing abstract decomposition on working scenes of a plurality of robots, and constructing a working scene model; based on the working scene model, constructing a task model for the to-be-executed task; according to dynamic heterogeneous characteristics of the robots, the execution capacity of each robot is determined and mapped to a capacity space, a robot capacity model is constructed, and the dynamic heterogeneous characteristics are used for representing the motion capacity, operation precision, sensing range and cruising ability of the robots; and based on the working scene model, performing matching analysis on the task model and the robot capability model, generating a corresponding target task for each robot, and issuing the target task. According to the method, different scenes can be quickly adapted, repeated work of customized development is reduced, the development cost is reduced, and the cross-scene universal management and control capability is improved.

Description

Technical Field

[0001] This application belongs to the field of robot control technology, and in particular relates to a method and system for unified management and intelligent collaboration of robots in multiple scenarios. Background Technology

[0002] With the rapid improvement of individual robots' core capabilities in environmental perception, autonomous decision-making, and precise execution, as well as the improvement of communication infrastructure such as 5G / 6G and edge computing, the application of robots is evolving from single-machine autonomy to swarm intelligence.

[0003] Currently, in multi-robot collaborative systems, a central controller plans an independent path for each robot.

[0004] However, due to the significant differences in the task requirements of robots in different application scenarios, the above solutions are difficult to maintain unified control while taking into account scenario adaptability. This leads to the need for repetitive work for different scenarios, which increases development costs and makes the system lack cross-scenario general control capabilities. Summary of the Invention

[0005] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a method and system for unified control and intelligent collaboration of robots in multiple scenarios, which can quickly adapt to different scenarios, reduce repetitive work in customized development, lower development costs, and improve cross-scenario general control capabilities.

[0006] Firstly, this application provides a method for unified management and intelligent collaboration of robots in multiple scenarios, applied to a control terminal, wherein the control terminal is communicatively connected to multiple robots, and the method includes: The working scenarios of the multiple robots are abstracted and decomposed to construct a working scenario model; Based on the aforementioned work scenario model, a task model is constructed for the task to be performed. Based on the dynamic heterogeneous characteristics of the robots, the execution capabilities of each robot are determined and mapped to the capability space to construct a robot capability model. The dynamic heterogeneous characteristics are used to characterize the robot's motion capability, operation accuracy, perception range, and endurance. Based on the work scenario model, a matching analysis is performed on the task model and the robot capability model, and a corresponding target task is generated and issued for each robot.

[0007] According to one embodiment of this application, the step of abstracting and decomposing the working scenarios of the plurality of robots and constructing a working scenario model includes: Obtain scene data for the work scenario; The scene features of the work scene are extracted from the scene data. The scene features include the geometric information, topological relationships and semantic information of the work scene. The scene features are integrated and analyzed using a scene knowledge graph, the work scene is divided into multiple functional areas, and the key features of each functional area are extracted. Based on the key features of the functional areas, the work scenario model is constructed.

[0008] According to one embodiment of this application, the step of extracting scene features of the work scene from the scene data includes: Entity-relationship extraction is performed on the point cloud data of the work scenario at the semantic layer to generate semantic elements; Based on the topology layer, a topology map is constructed using the semantic elements as topology nodes and the relationships between the semantic elements as edges. Based on the geometric layer, the topological nodes are mapped into a raster map with an uncertainty ellipsoid; The semantic elements, the topological map, and the grid map are aligned hierarchically through a cross-layer attention mechanism to obtain the scene features.

[0009] According to one embodiment of this application, constructing a task model for the task to be performed based on the work scenario model includes: The tasks to be executed are parsed and classified to generate task stages; Perform correlation analysis on the task stages to determine the relationships between each task stage, including dependencies, execution order, and coordination requirements; Based on the division of functional areas in the work scenario model, determine the task type of the corresponding task step in each functional area; Based on the functional area and the task type, a boundary analysis is performed on each task stage to determine the boundary conditions of the task stage. The boundary conditions include the start condition, end condition, execution range, input and output parameters, and interaction interface with other task stages. Based on the aforementioned relationships and boundary conditions, the task model is constructed for the task to be executed.

[0010] According to one embodiment of this application, the step of determining the execution capabilities of each robot and mapping them to a capability space based on the dynamic heterogeneous characteristics of the robots, and constructing a robot capability model, includes: Based on the dynamic heterogeneous characteristics of each robot, the capability indicators of the robot are determined, including motion speed, operation accuracy, perception range, and endurance. Each of the aforementioned capability indicators is quantitatively evaluated to obtain a quantitative value for the capability indicator; Based on the quantized value, a multidimensional capability vector is constructed, and the multidimensional capability vector is mapped to the capability space to obtain the robot capability model. The dimensions of the capability space correspond one-to-one with the dynamic heterogeneous features.

[0011] According to one embodiment of this application, before constructing the robot capability model, the method further includes: Obtain the interaction logs of each robot, which include behavioral data and environmental feedback information of the robot during task execution; The interaction log is segmented using a time-sliding window, and the dynamic performance data of the robot in each time-sliding window is extracted. Motion features, accuracy features, environmental adaptability features, energy consumption features, and perception ability features are extracted from the dynamic performance data and then dimensionality reduced to obtain the dynamic heterogeneous features.

[0012] According to one embodiment of this application, the step of performing a matching analysis on the task model and the robot capability model based on the work scenario model, and generating and issuing a corresponding target task for each robot, includes: Based on the functional areas in the work scenario model and the task steps in the task model, determine the task types of the task steps that each robot can execute in each functional area. The similarity between the task requirements of the task stage and the capability indicators in the robot capability model is calculated to obtain the matching degree of each robot in performing each task stage. Based on the degree of matching, the robot and the task stage are matched to generate a task allocation scheme; The uncertainty in the process of the robot executing the task allocation scheme is calculated using a weak common linear copositive function; Based on the aforementioned uncertainty, the task allocation scheme is optimized and adjusted, and the target tasks in the task allocation scheme are distributed to the corresponding robots.

[0013] According to one embodiment of this application, after generating and issuing a corresponding target task for each of the robots, the method further includes: Based on the robustness requirements of the target task, an event strategy for abnormal situations based on an event-driven mechanism is constructed. The event strategy is used to perform real-time matching of dynamic changes in the task stages of the target task. Based on the event strategy, if the abnormal situation is determined to occur, the dynamic heterogeneous characteristics and working scenario of the robot are updated, and the task allocation scheme is evaluated to obtain the evaluation result. Based on the evaluation results, the task components in the target task are reallocated, the execution order is changed, or the execution parameters are optimized.

[0014] Secondly, this application provides a unified management and intelligent collaboration system for robots in multiple scenarios. The system includes a control terminal and multiple robots, and the robots are communicatively connected to the control terminal. The control terminal is used to abstract and decompose the working scenarios of the multiple robots and construct a working scenario model. Based on the aforementioned work scenario model, a task model is constructed for the task to be performed. Based on the dynamic heterogeneous characteristics of the robots, the execution capabilities of each robot are determined and mapped to the capability space to construct a robot capability model. The dynamic heterogeneous characteristics are used to characterize the robot's motion capability, operation accuracy, perception range, and endurance. Based on the work scenario model, a matching analysis is performed on the task model and the robot capability model, and a corresponding target task is generated and issued for each robot. The robot is used to perform the corresponding target task.

[0015] According to one embodiment of this application, the robot is further used for: During the execution of the target task, an interaction log is generated, which includes the robot's behavioral data and environmental feedback information during the execution of the target task.

[0016] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application.

[0017] This application provides a method and system for unified management and intelligent collaboration of robots in multiple scenarios, which has the following advantages over existing technologies: (1) By accurately abstracting and decomposing the work scenario, meticulously constructing the task model, quantitatively presenting the robot capability model, and performing matching analysis based on these models, the system has achieved optimized configuration and intelligent collaboration of robot resources in multiple scenarios. It can cope with the significant differences in the task requirements of robots in different application scenarios, quickly adapt to different scenarios, reduce repetitive work in customized development, improve collaboration efficiency and resource utilization, significantly improve the efficiency and accuracy of task execution, enhance the system's adaptability to complex and ever-changing work scenarios, and dynamically adjust the robot's task allocation and execution strategy according to the needs of different scenarios to ensure that each robot can play its maximum role within its capabilities, reduce development costs, and improve the general management and control capabilities across scenarios.

[0018] (2) Through the scene knowledge graph, the control end can conduct in-depth analysis and understanding of various elements in the work scene and their interrelationships, thereby more accurately dividing functional areas and extracting key features. By extracting and analyzing the key features of functional areas, the control end can build a more accurate and comprehensive work scene model, laying a solid foundation for subsequent robot task allocation and intelligent collaboration. Attached Figure Description

[0019] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart illustrating the method for unified management and intelligent collaboration of robots in multiple scenarios provided in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the unified management and intelligent collaboration system for robots in multiple scenarios provided in the embodiments of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0021] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0022] The following description, in conjunction with the accompanying drawings, details the method and system for unified management and intelligent collaboration of robots in multiple scenarios provided in this application, through specific embodiments and application scenarios.

[0023] It is understandable that the unified management and intelligent collaboration method for robots in multiple scenarios is applied to the control end, which is connected to multiple robots for communication.

[0024] The control unit is a central control unit, such as a server or computing platform, used to manage and coordinate multiple robots of the same or different types. It is responsible for all calculations, decisions and scheduling, enabling multiple robots to work collaboratively in a variety of different work scenarios, such as warehouses, hospitals or factory workshops.

[0025] Robots are automated devices with different functional characteristics and execution capabilities, such as handling robots, nursing robots, or assembly robots, which can complete corresponding tasks according to the instructions issued by the control terminal.

[0026] The control unit establishes a two-way communication link with the robot via a wireless network or Bluetooth, which is used to issue commands and receive robot status information.

[0027] like Figure 1 As shown, this method for unified management and intelligent collaboration of robots in multiple scenarios includes: Step 110: Abstract and decompose the working scenarios of the multiple robots to construct a working scenario model.

[0028] The working environment of robots is complex and varied, and may include a variety of different environmental factors and task requirements.

[0029] Abstract decomposition is the digital modeling of the work scene. The work scene model is a digital abstraction of the robot's work scene, which can accurately reflect the geometric structure, functional division and semantic information in the scene.

[0030] In actual execution, in order to effectively cope with the complexity of the work scenario, the control end performs in-depth abstraction and decomposition of the work scenario. The control end collects point cloud data, image information and sensor readings of the work scenario. These data come from various sensors and drone equipment installed in the scenario, such as LiDAR and cameras. By extracting and analyzing these data, the control end identifies the key features in the work scenario and further integrates geometric data such as spatial coordinates and obstacle distribution, topological relationships such as regional connectivity and path reachability, and semantic features such as object categories and task relevance to form a multi-dimensional scene representation. Finally, the work scenario is divided into multiple functional areas, each with its unique key features. These features together constitute the work scenario model.

[0031] Step 120: Based on the work scenario model, construct a task model for the task to be performed.

[0032] The task to be performed is a specific work item that needs to be completed by the collaboration of multiple robots. It is determined by the actual needs obtained by the control terminal in response to the received user instructions.

[0033] A task model is a digital description of a task to be performed, used to determine key information such as the task's objectives, constraints, execution steps, and required resources.

[0034] In actual execution, the control unit combines the functional area divisions and key features in the work scenario model to perform a detailed analysis of the task to be executed. For example, if the task to be executed is to move goods in a warehouse, the control unit constructs a task model containing information such as the distribution of shelves and aisles in the warehouse work scenario model, as well as key features such as the weight and size of the goods, based on the functional area information in the warehouse work scenario model. This model includes information such as the starting point and ending point of the handling, the planning of the handling path, and the type and number of handling robots required.

[0035] The constructed task model can be closely integrated with the work scenario, ensuring that the robot can perform tasks in the optimal way, thereby improving the efficiency and accuracy of task execution.

[0036] Step 130: Based on the dynamic heterogeneous characteristics of the robot, determine the execution capabilities of each robot and map them to the capability space to construct a robot capability model. The dynamic heterogeneous characteristics are used to characterize the robot's motion capability, operation accuracy, perception range, and endurance.

[0037] Dynamic heterogeneous features are the differentiated and dynamically changing characteristics of a robot in terms of motion, operation, perception, and endurance. They can determine the robot's performance and adaptability in different work scenarios and different tasks to be performed, including but not limited to the robot's motion ability, operation accuracy, perception range, and endurance.

[0038] Motion capability characterizes a robot's maximum moving speed, acceleration, and turning agility; operational accuracy characterizes a robot's error range and repeatability when performing tasks; perception range characterizes a robot's sensor type, detection distance, and angular resolution; and endurance characterizes a robot's battery capacity, energy efficiency, and continuous working time. Execution capability describes the actual effectiveness a robot can achieve in a specific task.

[0039] Capability space is a multi-dimensional virtual region, with each dimension corresponding to a dynamic heterogeneous feature of the robot.

[0040] Motion capability, operational precision, perception range, and endurance are uniformly quantified to obtain the execution capability of each robot. The execution capability is then mapped to various dimensions of the capability space to construct a robot capability model. The robot capability model is a digital representation of the robot's execution capability in the capability space, used to describe the advantages and disadvantages of each robot in different dynamic heterogeneous characteristics.

[0041] In actual operation, the control unit collects relevant parameters and performance indicators for each robot, such as the maximum load and moving speed of the handling robot, and the operating accuracy and sensing sensitivity of the nursing robot. These relevant parameters and performance indicators are uniformly quantified to obtain the execution capability, and the execution capability is converted into coordinate values ​​in the capability space to construct a robot capability model for multiple robots. This model can intuitively display the potential capabilities and limitations of different robots when performing various tasks, comprehensively and accurately reflect the actual capability status of each robot, ensure that each robot can play its maximum role within its capability range, improve the operating efficiency and task completion quality of the entire system, and achieve optimal allocation of robot resources.

[0042] Step 140: Based on the work scenario model, perform a matching analysis on the task model and the robot capability model, generate a corresponding target task for each robot, and issue it.

[0043] Matching analysis is a process in which the control unit, based on the information provided by the work scenario model, comprehensively compares and analyzes the key information in the task model, such as the objectives, constraints, execution steps, and required resources, with the advantages and disadvantages of each robot in the robot capability model in terms of different dynamic heterogeneous characteristics.

[0044] The target task is a specific work instruction tailored to each robot by the control end based on the matching analysis results. It not only takes into account the actual needs and characteristics of the work scenario, but also combines the unique capabilities and advantages of each robot.

[0045] For example, in a complex factory environment, if a robot possesses strong mobility and operational precision but relatively weak perception range and endurance, the control unit might assign it tasks requiring high precision and relatively fixed paths, such as assembling precision parts. Simultaneously, the control unit ensures that the assigned task is within the robot's capabilities, preventing the task's difficulty from being too high or too low, which could negatively impact the robot's execution efficiency and task completion quality. After generating the target task, the control unit sends the task instructions to the corresponding robot via a two-way communication link. Upon receiving the task instructions, the robot begins executing the corresponding target task according to the specific requirements, thus achieving efficient unified control and intelligent collaboration of robots across multiple scenarios.

[0046] In actual execution, the control end obtains the specific details of the current work scenario from the work scenario model, such as functional area division and key features; for the tasks to be executed in the task model, it analyzes the various capability requirements required for each task link in the task to be executed, such as the motion capability and operation precision required for the handling task; it performs a detailed match between the capability requirements and the robot capability model, selects robots whose capabilities can meet the requirements of the task, combines the multiple task links corresponding to each robot into a target task, and sends it to the corresponding robot.

[0047] During the robot's execution of the target task, an event strategy based on an event-driven mechanism is constructed to address abnormal situations, according to the robustness requirements of the target task. This event strategy is used to perform real-time matching of dynamic changes in the task stages within the target task. When an abnormal situation is identified based on the event strategy, such as the appearance of new obstacles in the work environment or changes in task requirements, the robot's dynamic heterogeneous characteristics and the real-time changes in the work environment are reassessed to obtain an evaluation result. Based on the evaluation result, the task stages in the target task are reallocated, the execution order is changed, or the execution parameters are optimized to ensure that the robot can complete various tasks efficiently and accurately.

[0048] For example, in an assembly task in a factory workshop, the task model determines the assembly accuracy requirements, assembly speed requirements, and required tool operation capabilities. Through matching analysis, the control unit discovers that a certain assembly robot has advantages in operational accuracy and tool operation capabilities, and that its motion capabilities and endurance can meet the requirements of the assembly task in the work scenario. Therefore, the control unit will assign this assembly target task to that assembly robot.

[0049] Through matching analysis, the control unit can generate target tasks best suited to the capabilities of each robot and accurately distribute these tasks. This not only ensures that the robots can complete tasks efficiently and accurately but also avoids resource waste and inefficiency caused by unreasonable task allocation, thereby achieving efficient operation of unified management and intelligent collaboration of robots in multiple scenarios.

[0050] The unified management and intelligent collaboration method for robots in multiple scenarios provided in this application embodiment achieves optimized allocation and intelligent collaboration of robot resources in multiple scenarios through precise abstraction and decomposition of work scenarios, detailed construction of task models, quantitative presentation of robot capability models, and matching analysis based on these models. It can cope with the significant differences in task requirements of robots in different application scenarios, quickly adapt to different scenarios, reduce repetitive work in customized development, and improve collaboration efficiency and resource utilization. It not only significantly improves the efficiency and accuracy of task execution, but also enhances the system's adaptability to complex and ever-changing work scenarios. The control end can dynamically adjust the task allocation and execution strategy of robots according to the needs of different scenarios, ensuring that each robot can play its maximum role within its capabilities, reducing development costs, and improving cross-scenario general management and control capabilities.

[0051] In some embodiments, the abstract decomposition of the working scenarios of the plurality of robots and the construction of a working scenario model includes: Obtain scene data for the work scenario; The scene features of the work scene are extracted from the scene data. The scene features include the geometric information, topological relationships and semantic information of the work scene. The scene features are integrated and analyzed using a scene knowledge graph, the work scene is divided into multiple functional areas, and the key features of each functional area are extracted. Based on the key features of the functional areas, the work scenario model is constructed.

[0052] Understandably, scene data includes various types of information in the work scene. This information can be obtained through real-time data from various sensors installed in the scene, such as lidar, cameras, temperature sensors, and pressure sensors. These sensors can accurately capture physical parameters and environmental changes in the work scene. Alternatively, it can be obtained through image information and spatial coordinate data collected by drones during flight. Drones can comprehensively scan the work scene from different heights and angles, obtaining information about areas that are difficult for traditional sensors to cover.

[0053] Scene features are key elements for in-depth understanding and digital representation of work scenes, covering multiple levels such as geometric information, topological relationships, and semantic information of work scenes.

[0054] Geometric information includes the spatial layout and object shapes in the work environment, such as the arrangement of shelves, the width and direction of aisles, etc., which serve as the basic data for the robot's path planning and obstacle avoidance.

[0055] Topological relationships include the connectivity and accessibility between functional areas in a work environment. For example, which areas are adjacent and which paths are unobstructed. This is crucial for robots to perform cross-regional tasks.

[0056] Semantic information is used to describe objects and areas in a work scene with specific meanings and uses. For example, an area is a goods storage area, and an object is a goods to be moved. This enables robots to understand task requirements more accurately and perform corresponding operations.

[0057] Scene knowledge graphs are structured knowledge representations built on scene features. They are used to integrate and analyze scene features and can associate and fuse multi-dimensional data such as geometric information, topological relationships, and semantic information. They are organized and associated in the form of graphs to form a structured knowledge system that can comprehensively reflect the characteristics of the work scene.

[0058] Functional areas are different zones divided according to the specific purpose and characteristics of a work scenario. Each functional area has its unique key characteristics, which together define the role and function of that area in the work scenario. For example, in a warehouse work scenario, functional areas such as goods storage area, picking area, packing area, and loading and unloading area can be divided. Each area has its specific spatial layout, topological relationship, and semantic information.

[0059] Key features are unique to each functional area and represent the characteristics and functions of that area. For example, key features of a goods storage area may include shelf type, storage capacity, and types of goods; key features of a picking area may involve picking efficiency and picking route planning.

[0060] In actual execution, the control end acquires scene data of the work environment through various sensors and drones. This data contains rich information about the work environment. In-depth extraction of this scene data reveals scene features such as geometric information, topological relationships, and semantic information. Scene knowledge graphs are then used to integrate and analyze these features, linking and fusing multi-dimensional data such as geometric information, topological relationships, and semantic information to present the overall picture of the work environment in an intuitive graphical form. Through scene knowledge graphs, the hierarchical relationships and association rules between various functional areas in the work environment can be clearly displayed. For example, in a factory workshop, the material flow relationships and production process sequence between the raw material storage area, processing area, assembly area, and finished product storage area can be determined. When dividing the work environment into multiple functional areas, various factors in the scene features are comprehensively considered, and reasonable divisions are made based on the functional characteristics and key features of different areas. For example, a hospital scene can be divided into outpatient area, inpatient area, operating area, and testing area. Through the analysis of the scene knowledge graph, the control terminal divides the work scene into multiple functional areas, such as the goods storage area and picking area mentioned earlier, and further extracts the key features of each functional area. Based on these key features, the control terminal constructs a work scene model. The work scene model not only comprehensively reflects the geometric structure and functional zoning of the work scene, but also deeply includes semantic information.

[0061] In this embodiment, through the scene knowledge graph, the control terminal can conduct in-depth analysis and understanding of various elements in the work scene and their interrelationships, thereby more accurately dividing functional areas and extracting key features. By extracting and analyzing the key features of functional areas, the control terminal can construct a more accurate and comprehensive work scene model, laying a solid foundation for subsequent robot task allocation and intelligent collaboration.

[0062] In some embodiments, extracting scene features of the work scene from the scene data includes: Entity-relationship extraction is performed on the point cloud data of the work scenario at the semantic layer to generate semantic elements; Based on the topology layer, a topology map is constructed using the semantic elements as topology nodes and the relationships between the semantic elements as edges. Based on the geometric layer, the topological nodes are mapped into a raster map with an uncertainty ellipsoid; The semantic elements, the topological map, and the grid map are aligned hierarchically through a cross-layer attention mechanism to obtain the scene features.

[0063] Understandably, the geometry layer uses a grid map to represent the physical spatial layout of the scene, the topology layer is used to describe the connection relationships between various regions in the scene, and the semantic layer is used to give the scene and regions specific semantic meanings.

[0064] Point cloud data is a collection of three-dimensional coordinate data of a work scene obtained by sensors such as LiDAR, which can accurately reflect the spatial position and shape of objects in the scene.

[0065] Entity-relation extraction is a semantic process that identifies and extracts meaningful entity objects and their relationships from point cloud data in three-dimensional coordinates. Through entity-relation extraction, the control end can transform the raw point cloud data into a set of semantic elements rich in semantic information.

[0066] Semantic elements are objects with clear semantics obtained after entity-relation extraction. They not only contain basic information about objects in the scene, such as type and attributes, but also reveal the spatial relationships and logical connections between objects. For example, in a work scene, entities such as shelves, workbenches, and transport vehicles are identified, as well as the adjacent and containment relationships between entities.

[0067] Topological nodes are determined based on semantic elements and are used to represent key locations or areas in a work scenario.

[0068] The relationships between semantic elements are used to define the topological structure among various semantic elements in a work scenario. For example, the adjacency relationship between shelves and aisles determines the feasible path for robots when transporting goods; while the containment relationship between different functional areas reflects the hierarchical division and spatial organization of the work scenario.

[0069] Edges are used to determine the adjacency, inclusion, and intersection relationships between semantic elements, clearly defining the functional boundaries and interaction methods of different areas in a work scenario. A topology map is a graph structure constructed with semantic elements as topological nodes and the relationships between semantic elements as edges, which can present the connection relationships and spatial layout logic of various areas in a work scenario.

[0070] The uncertainty ellipsoid is used to characterize the range of uncertainty of each grid position in a grid map, and can reflect the positioning deviation caused by factors such as sensor measurement errors and dynamic changes in the environment.

[0071] Cross-layer attention mechanism is a mechanism that can align and fuse features across different layers of information. By analyzing the semantic information carried by semantic elements, the connection relationships between elements in the topological map, and the spatial layout presented by the grid map, it can automatically identify and emphasize the most critical parts of information at different layers that comprehensively describe scene features. It learns and captures the intrinsic connections and mutual influences between layers, and organically combines semantic information, topological relationships, and geometric layout to obtain comprehensive and accurate scene features.

[0072] In actual execution, the control end uses semantic elements as topological nodes and the relationships between topological nodes as edges to construct a topological map, which shows the connection relationships between various functional areas in the work scenario and also reflects the spatial layout and accessibility between these functional areas.

[0073] In the geometry layer, by mapping topological nodes to a grid map with an uncertainty ellipsoid, the geometry layer not only preserves the physical spatial information of the scene, but also reflects the uncertainty of node positions and the positional deviation caused by sensor measurement errors or dynamic changes in the environment through the uncertainty ellipsoid. This improves the accuracy and robustness of the robot's understanding of the scene and quantifies the reliability of spatial positioning.

[0074] To further integrate semantic, topological, and geometric information, the control unit employs a cross-layer attention mechanism to perform hierarchical alignment of semantic elements, topological maps, and grid maps. This mechanism dynamically adjusts the weight distribution of semantic, topological, and geometric features to achieve alignment and fusion of the three layers, addressing the semantic gap between different data levels. At the semantic layer, it focuses on the logical consistency of entity relationships; at the topological layer, it strengthens the spatial continuity of regional connections; and at the geometric layer, it optimizes the reliability of positioning accuracy. Ultimately, it generates a feature representation of the work scene that integrates multi-dimensional features. As scene features, the resulting scene features contain both high-level semantic understanding and low-level spatial accuracy.

[0075] For example, when temporary obstacles appear in the work scene, the cross-layer attention mechanism can quickly correct the grid uncertainty of the geometry layer, synchronously update the reachability relationship of the topology layer, and adjust the dynamic division of functional areas in the semantic layer to ensure that scene features always reflect the latest environmental state.

[0076] Through a cross-layer attention mechanism, the weight of topological relationships can be enhanced to optimize path selection in path planning tasks, while the priority of semantic information can be increased to accurately locate targets in object recognition tasks. This dynamic weight adjustment ensures the comprehensiveness and relevance of scene feature extraction. The resulting scene features contain precise spatial location information and integrate rich semantic relationships and topological structures, providing multi-dimensional environmental perception support for robot task execution.

[0077] For example, when a robot perceives its environment using LiDAR, each grid cell in the grid map can represent a spatial region, and the ellipsoid model describes the possible deviation range of the center position of that region, providing the robot with a more reliable spatial positioning reference.

[0078] A grid map discretizes the physical space of a work scene into regular grid cells, each representing a specific area within the scene. The confidence interval for spatial positioning is quantified using an uncertainty ellipsoid. This preserves the scene's geometric information, such as obstacle distribution and aisle width, while also expressing the impact of sensor measurement errors and dynamic environmental changes on positioning accuracy through ellipsoidal parameters such as major and minor axes and orientation. For example, in a dynamic warehouse scenario, changes in the position of moving shelves can cause the uncertainty ellipsoids of adjacent grid cells to overlap. In this case, a cross-layer attention mechanism enhances the weight of accessibility in the topology layer while weakening the positioning reliability of disturbed grid cells in the geometry layer, ensuring that robot path planning balances spatial continuity and positioning robustness.

[0079] For example, a high-precision semantic recognition result can reduce the geometric grid uncertainty of its corresponding region.

[0080] In this embodiment, the grid-ellipsoid composite representation method can more accurately depict the spatial uncertainty in unstructured environments compared to traditional two-dimensional grid maps, providing a more reliable environmental perception basis for robot obstacle avoidance and navigation.

[0081] In some embodiments, constructing a task model for the task to be performed based on the work scenario model includes: The tasks to be executed are parsed and classified to generate task stages; Perform correlation analysis on the task stages to determine the relationships between each task stage, including dependencies, execution order, and coordination requirements; Based on the division of functional areas in the work scenario model, determine the task type of the corresponding task step in each functional area; Based on the functional area and the task type, a boundary analysis is performed on each task stage to determine the boundary conditions of the task stage. The boundary conditions include the start condition, end condition, execution range, input and output parameters, and interaction interface with other task stages. Based on the aforementioned relationships and boundary conditions, the task model is constructed for the task to be executed.

[0082] Understandably, parsing and classification involves the control end using machine learning (ML) technology to break down the specific content and requirements of the task to be executed into a series of interconnected and executable task steps.

[0083] In the context of task parsing and classification, for example, by training a machine learning-based sequence labeling model, the system learns to parse and classify tasks to be executed into task segments from a large amount of labeled historical task data.

[0084] The trained sequence labeling model analyzes the word sequence in the text of the task to be performed, identifies key elements such as action, executor, object, and location, and aggregates fragments describing the same action according to the label sequence to form a structured task segment.

[0085] A task stage is a series of sub-tasks with clear objectives and execution steps, broken down from a task to be performed. These sub-tasks are interconnected and proceed sequentially, together forming the complete task execution process. Each task stage carries a specific function and is an indispensable part of the task execution process. For example, in a logistics sorting task, task stages may include multiple stages such as receiving goods, barcode scanning, sorting and storing, packaging and sealing, and outbound transportation, each with its own clear operational requirements and execution standards.

[0086] Association analysis involves in-depth analysis of the task segments obtained after parsing and classification to determine the dependencies, execution order, and collaboration requirements between task segments.

[0087] Dependency means that the execution of a certain task depends on the completion of other task stages, which clarifies the sequence and conditional constraints between task stages; execution order specifies the order of each task stage on the timeline; collaboration emphasizes the interaction and cooperation between different task stages in terms of resources, information, etc.

[0088] The division of functional areas is used to determine the task types of each task stage. Different functional areas carry different work scenario characteristics and task requirements. When building a task model, it is necessary to clarify the task types that each task stage in each area should have based on the division of functional areas. For example, in a warehouse scenario, the goods storage area may mainly involve the storage and retrieval of goods, while the picking area focuses more on the fast and accurate picking of goods.

[0089] Task type defines the nature of the task process within each functional area, reflecting the main characteristics and execution requirements of the task within that area. Different task types require robots to possess different combinations of capabilities to complete the task efficiently and accurately. For example, in assembly tasks, robots need high operational precision and tool usage skills; while in handling tasks, robots focus more on mobility and load-bearing capacity.

[0090] By combining task types with the characteristics of functional areas, the control unit can set more precise task type descriptions for each task stage, such as handling tasks, operation tasks, or monitoring tasks. These task types not only reflect the core functions of the task stage but also provide important basis for subsequent task allocation and robot scheduling.

[0091] Boundary analysis defines the nature of task segments within each functional area, reflecting the main characteristics and execution requirements of tasks within that area.

[0092] Boundary conditions include start conditions, end conditions, execution range, input and output parameters, and interaction interfaces with other task stages; The initial conditions define the prerequisites for the start of a task segment, such as a sensor detecting a specific signal or reaching a preset value. The termination conditions specify the completion criteria for a task segment, such as completing an action, reaching a position, or meeting a performance indicator. The execution range defines the spatial or temporal scope of the task segment, ensuring the robot can perform the task within the correct area. Input and output parameters describe the information required and the results generated during task segment execution, providing the robot with clear task instructions and feedback mechanisms. The interaction interfaces with other task segments define the methods and rules for resource and information transfer between different task segments, ensuring the smooth operation of the entire task flow. The task model, constructed based on relationships and boundary conditions, comprehensively and accurately reflects the structure and requirements of the task to be executed. It not only provides the robot with clear task instructions and execution paths but also offers the control unit effective task monitoring and management tools. Through the task model, the control unit can understand the robot's task execution status in real time, promptly identify and resolve problems, and ensure the efficient operation of unified control and intelligent collaboration of the robot across multiple scenarios.

[0093] In actual execution, after clarifying the task type, the control unit needs to perform boundary analysis on each task stage to determine its boundary conditions. Boundary conditions include start conditions, end conditions, execution scope, input / output parameters, and interfaces with other task stages. Start conditions define the necessary conditions for the task stage to begin execution, such as the startup of specific equipment or resource availability. End conditions clarify the criteria for task completion, such as task completion rate or quality achievement. Execution scope defines the specific spatial or temporal range of the task stage, ensuring efficient execution within the defined area. Input / output parameters describe the data inputs and outputs required by the task stage during execution. Interfaces with other task stages define the interface standards and communication protocols for collaborative work, ensuring seamless integration and efficient collaboration between task stages.

[0094] Based on the relationships and boundary conditions, the control unit can construct a task model for the task to be executed. The task model describes the overall framework and execution flow of the task in a structured form, including the division of task stages, the determination of relationships, and the setting of boundary conditions. Through the task model, the control unit can clearly understand the specific content and requirements of the task. Simultaneously, the robot can efficiently and accurately complete various tasks in different scenarios according to the requirements of the task model.

[0095] In actual execution, by analyzing and classifying tasks, the control end can identify each task stage and further analyze the relationships between task stages.

[0096] Based on the functional area division of the work scenario model, the control unit can determine the task type of the corresponding task steps within each functional area. For example, in a warehouse work scenario, the goods storage area may correspond to the tasks of receiving and storing goods, while the picking area corresponds to the tasks of barcode scanning and sorting. This correspondence helps the control unit to more accurately understand the execution requirements and characteristics of tasks in different functional areas, providing a basis for subsequent task allocation and robot scheduling.

[0097] After determining the task type, the control unit performs boundary analysis on each task stage to clarify its boundary conditions. Start and end conditions define the start and end times of the task stage, while the execution scope defines the spatial area of ​​influence of the task stage. Input and output parameters specify the information required and results generated during the execution of the task stage, and the interaction interfaces with other task stages ensure smooth information transfer and collaboration between different task stages.

[0098] Finally, based on the relationships and boundary conditions, the control unit constructs a task model for the task to be executed. The task model describes the various components of the task and their interrelationships in a structured manner, providing clear and accurate guidance for subsequent robot task allocation and intelligent collaboration. By constructing the task model, the control unit can better understand the task requirements, rationally allocate robot resources, and ensure that the task can be completed efficiently and accurately.

[0099] For example, the task model can be defined as: in, This is the j-th task stage; It is a task phase The task type is represented by a semantic label; It is a task phase Required capability vector; These are the task boundary conditions, such as time, space, and resources.

[0100] The robot's capability model is as follows: in, It is the i-th robot Robot capability model The dynamic capability mapping function uses an LSTM+Attention structure. The input is the robot's historical interaction log, and the output is the current capability vector. It's a robot. Real-time capability vector, It is a d-dimensional real vector space, where d is the dimension of the capability space and corresponds one-to-one with the dynamic heterogeneous features.

[0101] Matching degree calculation: in, It is the match score; It is a semantic matching item; It is a boundary matching item. It is the boundary Euclidean distance. It is a capability vector Projection in the boundary subspace.

[0102] In this embodiment, through systematic task parsing, correlation analysis, boundary condition setting, and task model construction, not only is the accuracy and efficiency of task execution improved, but the automation, intelligence, and flexibility of the task execution process are also enhanced. This helps to ensure the stability, controllability, and efficiency of task execution in complex work scenarios, thereby improving the performance and adaptability of the entire system.

[0103] In some embodiments, the step of determining the execution capabilities of each robot and mapping them to a capability space based on the dynamic heterogeneous characteristics of the robots, and constructing a robot capability model, includes: Based on the dynamic heterogeneous characteristics of each robot, the capability indicators of the robot are determined, including motion speed, operation accuracy, perception range, and endurance. Each of the aforementioned capability indicators is quantitatively evaluated to obtain a quantitative value for the capability indicator; Based on the quantized value, a multidimensional capability vector is constructed, and the multidimensional capability vector is mapped to the capability space to obtain the robot capability model. The dimensions of the capability space correspond one-to-one with the dynamic heterogeneous features.

[0104] Understandably, dynamic heterogeneous characteristics are a comprehensive description of the diverse and ever-changing properties exhibited by robots in different working scenarios. They encompass the diverse and ever-changing characteristics of robots in different working scenarios, as well as their basic capabilities in motion, manipulation, and perception. They also reflect key factors such as the robot's adaptability to environmental changes and energy utilization efficiency.

[0105] Capability indicators are specific quantitative standards determined based on the dynamic heterogeneous characteristics of robots, and are used to evaluate the actual performance of robots in terms of movement speed, operational accuracy, perception range, and endurance.

[0106] Among them, the motion speed index is used to measure how fast the robot moves during the execution of a task, which affects the efficiency of task completion; the operation accuracy index reflects the accuracy of the robot when performing fine operations; the perception range index is used to characterize the robot's perception ability of its surrounding environment, including the identification of obstacles and the detection of target objects, which is the basis for the robot's autonomous navigation and obstacle avoidance; and the endurance index determines the length of time the robot can work continuously without charging or refueling, which has a key impact on the execution of long-term or long-distance tasks.

[0107] Quantitative evaluation is the process of accurately measuring the capability indicators of each robot. By mapping capabilities of different dimensions to a unified metric space, these capability indicators can be quantitatively evaluated to obtain the specific numerical performance of each robot in different aspects, thereby constructing a multidimensional capability vector. The multidimensional capability vector not only contains the quantitative values ​​of the robot in each capability indicator, but also intuitively shows the robot's advantages and disadvantages in different dynamic heterogeneous characteristics by mapping to the capability space.

[0108] Based on this multidimensional capability vector and mapping result, the control end can construct a robot capability model that accurately reflects the robot's actual capabilities.

[0109] Capability space is a multi-dimensional mathematical space, where each dimension corresponds to a dynamic, heterogeneous characteristic of the robot, such as movement speed or operational precision. By mapping the multi-dimensional capability vector of each robot to the capability space, it is possible to determine the performance and relative position of different robots in each capability dimension.

[0110] In this embodiment, the control terminal can more accurately understand the actual capabilities of each robot and select the most suitable robot from the capability space to perform the task based on the task requirements and the characteristics of the work scenario, thereby achieving optimal allocation of robot resources and intelligent collaboration.

[0111] In some embodiments, prior to constructing the robot capability model, the method further includes: Obtain the interaction logs of each robot, which include behavioral data and environmental feedback information of the robot during task execution; The interaction log is segmented using a time-sliding window, and the dynamic performance data of the robot in each time-sliding window is extracted. Motion features, accuracy features, environmental adaptability features, energy consumption features, and perception ability features are extracted from the dynamic performance data and then dimensionality reduced to obtain the dynamic heterogeneous features.

[0112] It is understandable that interaction logs are data resources generated by the robot during task execution, used to record the robot's behavioral data and environmental feedback information.

[0113] The behavioral data includes multiple aspects such as the robot's movement trajectory, operation actions, and execution time. The environmental feedback information includes parameters of the robot's environment, such as temperature, humidity, and light intensity, as well as the interaction between the robot and the environment, such as whether it encountered obstacles or successfully completed the task. The robot's real-time state is estimated through a positive observer.

[0114] The time-based sliding window can automatically adjust its size based on the characteristics of the data stream in the interaction log. For example, when the robot's movement is stable and the environment changes little, the window automatically increases to optimize processing efficiency; when the environment changes drastically or the robot's behavior changes drastically, the window automatically decreases to respond quickly to changes.

[0115] Dynamic performance data is a collection of real-time states and behavioral characteristics exhibited by a robot when performing a task within a specific time period. It includes data such as the robot's movement trajectory, operation actions, execution efficiency, and interaction with the environment during that time period, reflecting the robot's current capability status and its potential performance in future tasks.

[0116] By sliding the time window sequentially in chronological order, the continuous interaction logs are divided into multiple time-local data segments, each of which reflects the robot's dynamic performance within a certain time period.

[0117] Within each time sliding window, data that has a key impact on the robot's capability assessment is selected from each data segment. This data reflects the robot's real-time status and behavioral characteristics during task execution and serves as dynamic performance data for the robot within that time period.

[0118] By extracting and reducing the dimensions of dynamic performance data, the extracted multidimensional features are mapped to a lower-dimensional space, transforming them into a simpler and more representative feature representation. This allows for the extraction of robot motion features, such as speed, acceleration, and direction of motion; accuracy features, such as operational accuracy and positioning accuracy; environmental adaptability features, such as the ability to adapt to different lighting and temperature environments; energy consumption features, such as energy consumption per unit task; and perception capability features, such as the perception range and accuracy of the surrounding environment. Dimensionality reduction of these features yields the robot's dynamic heterogeneous features, which comprehensively and accurately reflect the diverse and constantly changing characteristics of the robot in different working scenarios.

[0119] In this embodiment, by acquiring the robot's interaction logs and using a time sliding window for segmented processing, the dynamic performance data of the robot during task execution can be effectively extracted. By performing feature extraction and dimensionality reduction on these data, the dynamic heterogeneous features of the robot can be obtained.

[0120] In some embodiments, the step of performing a matching analysis on the task model and the robot capability model based on the work scenario model, generating and issuing corresponding target tasks for each robot, includes: Based on the functional areas in the work scenario model and the task steps in the task model, determine the task types of the task steps that each robot can execute in each functional area. The similarity between the task requirements of the task stage and the capability indicators in the robot capability model is calculated to obtain the matching degree of each robot in performing each task stage. Based on the degree of matching, the robot and the task stage are matched to generate a task allocation scheme; The uncertainty in the process of the robot executing the task allocation scheme is calculated using a weak common linear copositive function; Based on the aforementioned uncertainty, the task allocation scheme is optimized and adjusted, and the target tasks in the task allocation scheme are distributed to the corresponding robots.

[0121] Task requirements are the specific requirements for the robot in terms of capabilities, operations, and resources, as clearly stated in the task model for each task stage. These include key capability indicators such as the robot's required speed, operational accuracy, perception range, and endurance when performing a specific task. For example, in a task requiring rapid goods transport, the task requirements include a high speed to ensure the goods are transported to the designated location promptly and efficiently. In a task involving precision assembly, the task requirements include the robot's operational accuracy to ensure the accuracy and quality of the assembly work.

[0122] In similarity calculation, each data range requirement in the task specifications is compared one by one with the corresponding capability indicators in the robot's capability model. The higher the similarity score, the better the robot matches the task requirements in that capability indicator. Based on the similarity score, the matching degree of each robot in performing each task stage is determined.

[0123] The task allocation scheme is a specific plan that divides the task into multiple target tasks and assigns them to the corresponding robots based on the degree of matching between the robot's capability indicators and the task requirements.

[0124] For example, robots Execution capability is achieved through multidimensional capability vectors. Indicate: Among them, multidimensional capability vector Used to quantify the robot in the capability space Comprehensive execution capabilities; Represents robots The speed ability value of movement, Representation Robot The operational accuracy capability value, Representation Robot Perception range capability value, Representation Robot The battery life value.

[0125] The value of each dimension can be a normalized numerical value (such as between 0 and 1).

[0126] Task Phase The demand is represented as a demand vector. : Among them, the demand vector To ensure the successful completion of the task. Minimum required skill level; Indicates task phase The required speed of movement Indicates task phase The required value for operational accuracy, Demonstration of task phase The required value for the sensing range, Indicates task phase The required range of battery life.

[0127] So, the robot Task execution phase similarity for: Among them, similarity A higher value indicates that the robot is more suitable for performing that task. It is the weight of the k-th capability dimension, used to characterize the different capabilities of the robot in different task stages; Represents robots The ability value in the k-th ability dimension is a multidimensional ability vector. The k-th component; represents the task stage. The demand value in the k-th capability dimension is the demand vector. The kth component.

[0128] Set the semantic layer output to a collection of group entities. and relation set .

[0129] Construction graph of topology layer Among them, topology nodes ,side ; In the geometric layer, the x-th grid cell The states of being occupied, idle, or unknown are determined by probability values. express, Used to characterize grid cells The probability of being occupied is a value between 0 (definitely free) and 1 (definitely occupied), grid. Uncertainty through information entropy The measurement is performed in bits. In cross-layer attention mechanisms, attention weights are defined. Represents semantic elements Topology nodes and grid cells The correlation strength among the three is a scalar value between 0 and 1; the higher the value, the stronger the correlation. Attention weights can be generated using a neural network based on a multilayer perceptron (MLP). Perform the calculation: Among them, neural networks Neural networks are used to learn how to calculate the correlation strength between three elements based on the input feature vectors. parameters Acquired through training; It is a feature vector of semantic elements, obtained through word embedding. It is a topology node The feature vectors are used to characterize the topological nodes. Graph network characteristics such as degree and centrality in topological graphs. It is a grid unit The feature vector is a vector composed of the occupancy probability value, entropy value, etc. of the grid cell and its surrounding grid cells.

[0130] Then, through attention weights Coordinate and update information at each level: in, It is a grid unit The updated entropy value. It is a grid unit The old entropy value before the update.

[0131] The weak common linear copositive function is a Lyapunov function used to quantify the uncertainty in robot task execution. By comprehensively considering various uncertainties that the robot may encounter when performing a task, such as environmental changes, equipment failures, and changes in task requirements, these uncertainties are transformed into quantifiable numerical indicators.

[0132] During the matching analysis process, the uncertainty of task allocation can be calculated using information entropy theory. The matching strategy can be dynamically adjusted through entropy increase and entropy decrease factors to improve the rationality and stability of task allocation.

[0133] After generating the task allocation scheme, the uncertainties during robot task execution are calculated using a weakly common linear copositive function. These uncertainties may arise from various factors, such as environmental changes and robot malfunctions. By calculating these uncertainties, the task allocation scheme can be further optimized and adjusted to ensure its feasibility and stability in actual execution.

[0134] For example, when the calculated uncertainty value is high, the control unit may choose strategies such as adding backup robots, adjusting the task execution order, or replanning the task path to reduce the risk of task execution failure.

[0135] Based on the uncertainty value calculated from the weak common linear copositive function, the control end can dynamically optimize and adjust the task allocation scheme to ensure that each robot can execute its corresponding target task in the best way.

[0136] The optimized task allocation scheme distributes the target tasks to the corresponding robots, enabling the robots to understand their task content and execution requirements, and thus begin to execute the corresponding target tasks.

[0137] The task allocation optimization based on uncertainty is as follows: in, It is the optimal subset of robots, which is the set that maximizes the objective function value among all possible robot combinations; It is an optimization operator, indicating the solution of the function. The set of robots that achieves the maximum value; It is a subset of robots. express It is a collection of all robots A subset of them is considered for allocation to perform tasks; It is a weakly common linear copositive function, with the robot subset As input, output a scalar value that quantifies the overall success probability or reliability of the robot team in collaboratively completing the entire task process. This indicates the cost of activating robot i, including economic costs, time costs, or energy consumption; This represents the total cost of activating all robots in the robot subset S.

[0138] In this embodiment, by accurately defining the task requirements, it is ensured that each robot can be assigned to a task that matches its actual capabilities and can be completed efficiently, thereby achieving optimized allocation and intelligent collaboration of robot resources in multiple scenarios.

[0139] In some embodiments, after generating and issuing a corresponding target task for each of the robots, the method further includes: Based on the robustness requirements of the target task, an event strategy for abnormal situations based on an event-driven mechanism is constructed. The event strategy is used to perform real-time matching of dynamic changes in the task stages of the target task. Based on the event strategy, if the abnormal situation is determined to occur, the dynamic heterogeneous characteristics and working scenario of the robot are updated, and the task allocation scheme is evaluated to obtain the evaluation result. Based on the evaluation results, the task components in the target task are reallocated, the execution order is changed, or the execution parameters are optimized.

[0140] Robustness requirements refer to the ability of a target task to resist and adapt to various uncertainties and unforeseen circumstances during the execution of the task.

[0141] Abnormal situations may alter the robot's actual performance in terms of movement speed, operational accuracy, etc., such as increased energy consumption and decreased operational accuracy, as well as affect the geometry or functional zoning of the scene, such as the appearance of new obstacles or deterioration of lighting conditions.

[0142] Event-driven mechanisms are dynamic response modes that trigger corresponding operations and strategies based on specific events or conditions. They can monitor various dynamic changes during the execution of target tasks in real time, such as sudden changes in environmental parameters or abnormalities in robot equipment.

[0143] When constructing an event strategy for abnormal situations based on an event-driven mechanism, the dynamic changes that the target task may encounter in different working scenarios are fully considered, such as sudden changes in environmental parameters, adjustments to task requirements, and fluctuations in the robot's own state. The event strategy is set with multiple trigger conditions to monitor and respond to dynamic changes in the task stages in real time. These trigger conditions can monitor key indicators during the execution of the target task in real time. Once these indicators exceed the preset normal range, it is considered that an abnormal situation has occurred.

[0144] When an anomaly is detected and confirmed by the event-driven mechanism, the control unit quickly activates the pre-built event strategy. Based on the type and severity of the anomaly, it updates the dynamic heterogeneous characteristics of the robot and the working scenario. Then, it conducts a comprehensive evaluation of the original task allocation scheme, including the degree of matching between the current capability status of each robot and the task requirements. It also needs to consider the impact of the anomaly on task execution efficiency, quality, and overall collaboration. Through comprehensive evaluation, the control unit can obtain evaluation results on the feasibility and optimization potential of the task allocation scheme.

[0145] Based on the evaluation results, the control unit will reallocate task components, change their execution order, or optimize execution parameters within the target task. For example, if a robot's operational precision decreases due to an abnormal situation, rendering it unable to complete a fine assembly task, the control unit can adjust its task to a handling task with lower operational precision requirements and reassign it to a robot with higher operational precision to complete the original assembly task. Alternatively, if ambient lighting conditions deteriorate, affecting the robot's perception capabilities, the control unit can adjust the task execution order, prioritizing tasks with higher perception requirements during periods of better lighting conditions.

[0146] In this embodiment, by constructing an event strategy based on an event-driven mechanism, it is possible to monitor and respond to dynamic changes during the execution of the target task in real time. Once an anomaly occurs, the control terminal can quickly update the robot's dynamic heterogeneous characteristics and working scenario, and comprehensively evaluate and optimize the task allocation scheme. This not only enhances the target task's resistance and adaptability to uncertainties, but also ensures that robot resources are always in the optimal configuration state in multiple scenarios.

[0147] This application also provides a unified management and intelligent collaboration system for robots in multiple scenarios.

[0148] like Figure 2 As shown, the unified management and intelligent collaboration system for robots in multiple scenarios includes a control terminal 210 and multiple robots 220, wherein the robots 220 are communicatively connected to the control terminal 210; The control terminal 210 is used to abstract and decompose the working scenarios of the multiple robots 220 and construct a working scenario model. Based on the aforementioned work scenario model, a task model is constructed for the task to be performed. Based on the dynamic heterogeneous characteristics of the robot 220, the execution capabilities of each robot 220 are quantified and mapped to the capability space to construct a capability model of the robot 220. The dynamic heterogeneous characteristics are used to characterize the motion capability, operation accuracy, perception range, and endurance of the robot 220. Based on the work scenario model, a matching analysis is performed on the task model and the robot 220 capability model, and a corresponding target task is generated and issued for each robot 220. The robot 220 is used to perform the corresponding target task.

[0149] The unified robot management and intelligent collaboration system for multiple scenarios provided in this application embodiment achieves optimized allocation and intelligent collaboration of robot resources in multiple scenarios through precise abstraction and decomposition of work scenarios, detailed construction of task models, quantitative presentation of robot capability models, and matching analysis based on these models. It can cope with the significant differences in task requirements of robots in different application scenarios, quickly adapt to different scenarios, reduce repetitive work in customized development, and improve collaboration efficiency and resource utilization. It not only significantly improves the efficiency and accuracy of task execution, but also enhances the system's adaptability to complex and ever-changing work scenarios. The control end can dynamically adjust the task allocation and execution strategy of robots according to the needs of different scenarios, ensuring that each robot can play its maximum role within its capabilities, reducing development costs, and improving cross-scenario general management and control capabilities.

[0150] In some embodiments, the robot 220 is further configured to: During the execution of the target task, an interaction log is generated, which includes the robot 220's behavior data and environmental feedback information during the execution of the target task.

[0151] The unified robot management and intelligent collaboration system for multiple scenarios provided in this application can realize the various processes implemented in the above embodiments of the unified robot management and intelligent collaboration method for multiple scenarios. To avoid repetition, these processes will not be described again here.

[0152] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

[0153] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0154] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for unified management and intelligent collaboration of robots in multiple scenarios, characterized in that, The method is applied to a control terminal, which is communicatively connected to multiple robots, and includes: The working scenarios of the multiple robots are abstracted and decomposed to construct a working scenario model; Based on the aforementioned work scenario model, a task model is constructed for the task to be performed. Based on the dynamic heterogeneous characteristics of the robots, the execution capabilities of each robot are determined and mapped to the capability space to construct a robot capability model. The dynamic heterogeneous characteristics are used to characterize the robot's motion capability, operation accuracy, perception range, and endurance. Based on the work scenario model, a matching analysis is performed on the task model and the robot capability model, and a corresponding target task is generated and issued for each robot.

2. The method for unified management and intelligent collaboration of robots in multiple scenarios according to claim 1, characterized in that, The abstract decomposition of the working scenarios of the multiple robots to construct a working scenario model includes: Obtain scene data for the work scenario; The scene features of the work scene are extracted from the scene data. The scene features include the geometric information, topological relationships and semantic information of the work scene. The scene features are integrated and analyzed using a scene knowledge graph, the work scene is divided into multiple functional areas, and the key features of each functional area are extracted. Based on the key features of the functional areas, the work scenario model is constructed.

3. The method for unified management and intelligent collaboration of robots in multiple scenarios according to claim 2, characterized in that, The step of extracting scene features of the work scene from the scene data includes: Entity-relationship extraction is performed on the point cloud data of the work scenario at the semantic layer to generate semantic elements; Based on the topology layer, a topology map is constructed using the semantic elements as topology nodes and the relationships between the semantic elements as edges. Based on the geometric layer, the topological nodes are mapped into a raster map with an uncertainty ellipsoid; The semantic elements, the topological map, and the grid map are aligned hierarchically through a cross-layer attention mechanism to obtain the scene features.

4. The method for unified management and intelligent collaboration of robots in multiple scenarios according to claim 1, characterized in that, The step of constructing a task model for the task to be performed based on the work scenario model includes: The tasks to be executed are parsed and classified to generate task stages; Perform correlation analysis on the task stages to determine the relationships between each task stage, including dependencies, execution order, and coordination requirements; Based on the division of functional areas in the work scenario model, determine the task type of the corresponding task step in each functional area; Based on the functional area and the task type, a boundary analysis is performed on each task stage to determine the boundary conditions of the task stage. The boundary conditions include the start condition, end condition, execution range, input and output parameters, and interaction interface with other task stages. Based on the aforementioned relationships and boundary conditions, the task model is constructed for the task to be executed.

5. The method for unified management and intelligent collaboration of robots in multiple scenarios according to claim 1, characterized in that, The step of determining the execution capabilities of each robot and mapping them to a capability space based on the dynamic heterogeneous characteristics of the robots, and constructing a robot capability model, includes: Based on the dynamic heterogeneous characteristics of each robot, the capability indicators of the robot are determined, including motion speed, operation accuracy, perception range, and endurance. Each of the aforementioned capability indicators is quantitatively evaluated to obtain a quantitative value for the capability indicator; Based on the quantized value, a multidimensional capability vector is constructed, and the multidimensional capability vector is mapped to the capability space to obtain the robot capability model. The dimensions of the capability space correspond one-to-one with the dynamic heterogeneous features.

6. The method for unified management and intelligent collaboration of robots in multiple scenarios according to claim 5, characterized in that, Prior to constructing the robot capability model, the method further includes: Obtain the interaction logs of each robot, which include behavioral data and environmental feedback information of the robot during task execution; The interaction log is segmented using a time-sliding window, and the dynamic performance data of the robot in each time-sliding window is extracted. Motion features, accuracy features, environmental adaptability features, energy consumption features, and perception ability features are extracted from the dynamic performance data and then dimensionality reduced to obtain the dynamic heterogeneous features.

7. The method for unified management and intelligent collaboration of robots in multiple scenarios according to claim 1, characterized in that, The process of matching and analyzing the task model and the robot capability model based on the work scenario model, generating and issuing corresponding target tasks for each robot, includes: Based on the functional areas in the work scenario model and the task steps in the task model, determine the task types of the task steps that each robot can execute in each functional area. The similarity between the task requirements of the task stage and the capability indicators in the robot capability model is calculated to obtain the matching degree of each robot in performing each task stage. Based on the degree of matching, the robot and the task stage are matched to generate a task allocation scheme; The uncertainty in the process of the robot executing the task allocation scheme is calculated using a weak common linear copositive function; Based on the aforementioned uncertainty, the task allocation scheme is optimized and adjusted, and the target tasks in the task allocation scheme are distributed to the corresponding robots.

8. The method for unified management and intelligent collaboration of robots in multiple scenarios according to claim 1, characterized in that, After generating and issuing a corresponding target task for each robot, the method further includes: Based on the robustness requirements of the target task, an event strategy for abnormal situations based on an event-driven mechanism is constructed. The event strategy is used to perform real-time matching of dynamic changes in the task stages of the target task. Based on the event strategy, if the abnormal situation is determined to occur, the dynamic heterogeneous characteristics and working scenario of the robot are updated, and the task allocation scheme is evaluated to obtain the evaluation result. Based on the evaluation results, the task components in the target task are reallocated, the execution order is changed, or the execution parameters are optimized.

9. A unified management and intelligent collaboration system for robots in multiple scenarios, characterized in that, It includes a control terminal and multiple robots, the robots being communicatively connected to the control terminal; The control terminal is used to abstract and decompose the working scenarios of the multiple robots and construct a working scenario model. Based on the aforementioned work scenario model, a task model is constructed for the task to be performed. Based on the dynamic heterogeneous characteristics of the robots, the execution capabilities of each robot are determined and mapped to the capability space to construct a robot capability model. The dynamic heterogeneous characteristics are used to characterize the robot's motion capability, operation accuracy, perception range, and endurance. Based on the work scenario model, a matching analysis is performed on the task model and the robot capability model, and a corresponding target task is generated and issued for each robot. The robot is used to perform the corresponding target task.

10. The unified management and intelligent collaboration system for robots in multiple scenarios according to claim 9, characterized in that, The robot is further used for: During the execution of the target task, an interaction log is generated, which includes the robot's behavioral data and environmental feedback information during the execution of the target task.

Images