Robot design method and system based on agent collaboration

Abstract

The invention relates to the technical field of robot design, and provides a robot design method and system based on agent collaboration. The method comprises the steps that a design agent generates a plurality of initial robot models based on robot demand information of a user; the simulation agent constructs a virtual scene matched with the robot demand information; a verification agent performs performance verification on each initial robot model in the virtual scene, and feeds back abnormal event information generated in the verification process of each initial robot model to the design agent; the design agent generates a target robot model based on the abnormal event information corresponding to each initial robot model; and the deployment agent generates manufacturing information of the entity robot required by the user based on the target robot model, guides production of the entity robot based on the manufacturing information, and continuously monitors the entity robot after the entity robot runs. The method is helpful for improving the design efficiency of the robot.

Description

Technical Field

[0001] This invention relates to the field of robot design technology, and in particular to a robot design method and system based on intelligent agent collaboration. Background Technology

[0002] As robot applications become increasingly complex and diverse, adopting a modular approach to robot design, development, and deployment has become an important industry trend. This approach aims to quickly build robots that meet specific needs by combining functional modules with standard interfaces, thereby improving design flexibility and reducing development costs.

[0003] However, the existing modular robot development process usually relies on engineers' human experience and discrete toolchains. The various stages from design, simulation, verification to implementation are isolated from each other and lack effective collaboration and feedback mechanisms, which in turn affects the robot's design efficiency and performance in later use. Summary of the Invention

[0004] This invention provides a robot design method and system based on intelligent agent collaboration to improve robot design efficiency and performance.

[0005] One aspect of this invention is to provide a robot design method based on agent collaboration, comprising: The design agent generates functional indicator information and design constraint information for robot design based on the user's robot requirement information, and generates multiple initial robot models based on the functional indicator information, the design constraint information and a preset robot module agent knowledge base. The simulated intelligent agent parses the task scenario in the robot's requirement information and constructs a virtual scenario that matches the task scenario; The verification agent performs performance verification on each of the initial robot models in the virtual scene, records abnormal event information during the entire verification process, and feeds back the abnormal event information of each of the initial robot models to the design agent. The design agent determines the robot model to be optimized from the initial robot models based on the abnormal event information corresponding to each initial robot model, and optimizes the robot model to be optimized based on the abnormal event information corresponding to the robot model to be optimized to obtain the target robot model. The deployed intelligent agent generates manufacturing information for the physical robot required by the user based on the target robot model, guides the production of the physical robot based on the manufacturing information, and continuously monitors the physical robot after it is in operation.

[0006] Furthermore, the design agent generates multiple initial robot models based on the functional indicator information, the design constraint information, and a preset robot module agent knowledge base, including: In the robot module agent knowledge base, match module agents for each functional indicator in the functional indicator information to obtain the module agent set corresponding to each functional indicator; Multiple combinations of module intelligent agents are generated based on the sets of intelligent agents described above; The design constraint information is broadcast to each of the module agent combinations, so that each module agent in each module agent combination can interact and negotiate based on the design constraint information, generate an interaction negotiation result, and feed the interaction negotiation result back to the design agent; the interaction negotiation result is either satisfying the design constraint information or not satisfying the design constraint information. For each of the aforementioned module agent combinations, when the interaction negotiation result fed back by the module agent combination is that the design constraint information is satisfied, the design agent combines each module agent in the module agent combination into an initial robot model.

[0007] Furthermore, the module agents in each of the aforementioned module agent combinations interact and negotiate based on the design constraint information to generate an interaction negotiation result, including: For each of the aforementioned module agent combinations, the module agents in the module agent combination interact and negotiate based on the design constraint information through the A2A protocol to obtain the interaction negotiation result corresponding to the module agent combination.

[0008] Furthermore, the simulated intelligent agent constructs a virtual scene that matches the task scenario, including: The simulation software is invoked via the MCP protocol, so that the simulation software can construct a virtual scene that matches the task scenario.

[0009] Furthermore, the verification agent performs performance verification on each of the initial robot models in the virtual scene, and records abnormal event information throughout the verification process, including: Based on the robot requirement information, multiple test cases are generated, and control scripts corresponding to each test case are generated. For each initial robot model, the initial robot model is imported into the virtual scene, and each of the test cases is executed on the initial robot model to drive the initial robot model to perform tasks in the virtual scene. When an exception occurs during the initial robot's task execution, the exception event type is recorded to generate exception event information corresponding to the initial robot model.

[0010] Furthermore, the design agent determines the robot model to be optimized from among the initial robot models based on the abnormal event information corresponding to each initial robot model, including: Each initial robot model is assigned a specific number of abnormal events from its corresponding abnormal event information. The initial robot model corresponding to the anomaly information with the fewest types of anomalies is identified as the robot model to be optimized.

[0011] Furthermore, the method also includes: If there are multiple exception event information with the fewest exception event types, the initial robot model with the fewest occurrences of exception events among the exception event information with the fewest exception event types is determined as the robot model to be optimized.

[0012] Furthermore, the design agent optimizes the robot model to be optimized based on the abnormal event information corresponding to the robot model to be optimized, to obtain the target robot model, including: For each abnormal event type in the abnormal event information corresponding to the robot model to be optimized, root cause analysis is performed on the abnormal event type to determine the cause of the abnormal event type and the associated module agent in the robot model to be optimized, and the associated module agent is optimized based on the cause.

[0013] Furthermore, the deploying agent generates manufacturing information for the physical robot required by the user based on the target robot model, including: The target robot model is analyzed to generate component information, which includes at least a physical assembly drawing and the components of the manufactured robot; the component information includes at least the component model and the component supplier.

[0014] Another aspect of this invention is to provide a robot design system based on agent collaboration, including designing agents, simulating agents, verifying agents, and deploying agents; The design agent generates functional indicator information and design constraint information for robot design based on the user's robot requirement information, and generates multiple initial robot models based on the functional indicator information, the design constraint information and a preset robot module agent knowledge base. The simulated intelligent agent parses the task scenario in the robot's requirement information and constructs a virtual scenario that matches the task scenario; The verification agent performs performance verification on each of the initial robot models in the virtual scene, records abnormal event information during the entire verification process, and feeds back the abnormal event information of each of the initial robot models to the design agent. The design agent determines the robot model to be optimized from the initial robot models based on the abnormal event information corresponding to each initial robot model, and optimizes the robot model to be optimized based on the abnormal event information corresponding to the robot model to be optimized to obtain the target robot model. The deployed intelligent agent generates manufacturing information for the physical robot required by the user based on the target robot model, guides the production of the physical robot based on the manufacturing information, and continuously monitors the physical robot after it is in operation.

[0015] This embodiment provides a robot design method and system based on agent collaboration. The method includes: a design agent generating functional indicator information and design constraint information for robot design based on user robot requirement information, and generating multiple initial robot models based on the functional indicator information, the design constraint information, and a preset robot module agent knowledge base; a simulation agent parsing the task scenario in the robot requirement information and constructing a virtual scenario matching the task scenario; a verification agent performing performance verification on each of the initial robot models in the virtual scenario, recording abnormal event information throughout the verification process, and feeding back the abnormal event information of each initial robot model to the design agent; the design agent determining the robot model to be optimized among the initial robot models based on the abnormal event information corresponding to each initial robot model, and optimizing the robot model to be optimized based on the abnormal event information corresponding to the robot model to be optimized, to obtain a target robot model; The method deploys an intelligent agent to generate manufacturing information for the physical robot required by the user based on the target robot model, guides the production of the physical robot based on this information, and continuously monitors the physical robot after it starts operating. This method, on the one hand, constructs an automated and intelligent robot design process by introducing intelligent agents with clearly defined roles in design, simulation, verification, and deployment, thereby transforming traditional, experience-dependent, and discrete design activities into a system-level collaborative and continuous process, thus improving overall design efficiency. On the other hand, by employing a strategy of generating multiple initial design schemes in parallel and competitively selecting and optimizing them based on objective verification data, it systematically explores the design space and makes rational decisions based on performance feedback, thereby improving the overall performance of the final design scheme. Furthermore, by deploying intelligent agents to achieve the connection from digital model to physical entity and the continuous collection of operational data, it establishes a design-deployment-feedback closed loop covering the entire robot lifecycle, thus helping to support the continuous iteration and optimization of robot products. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating the robot design method based on agent collaboration provided in this application embodiment; Figure 2 A schematic block diagram of the structure of a robot design system based on agent collaboration provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be noted that when a component is described as "fixed to / set on" another component, it can be directly on the other component or it can be in a middle component. When a component is described as "connected to" another component, it can be directly connected to the other component or it may be in a middle component.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0021] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other. This invention provides a robot design method based on agent collaboration. Figure 1 A flowchart of a robot design method based on agent collaboration provided in an embodiment of the present invention is shown below. Figure 1 As shown, the robot design method based on agent collaboration provided in this embodiment of the invention includes steps S1 to S5.

[0022] Step S1: The intelligent agent generates functional indicator information and design constraint information for robot design based on the user's robot requirement information, and generates multiple initial robot models based on the functional indicator information, the design constraint information and the preset robot module intelligent agent knowledge base.

[0023] The functional indicator information refers to the description of the core capabilities required of the robot, and the design constraint information refers to the description of the technical restrictions that need to be followed during the robot design process. For example, the functional indicator information may include movement, visual perception, and grasping, and the design constraint information may include maximum load, motion accuracy, and total cost limit.

[0024] For example, the design agent uses natural language processing technology to semantically parse the user's robot requirement information, transforming the unstructured robot requirement information into structured data that can be processed by the machine, thereby generating functional indicator information and design constraint information. Subsequently, the design agent accesses a preset robot module agent knowledge base, and generates multiple initial robot models that theoretically meet the user's robot design requirements based on the functional indicator information, the design constraint information, and the preset robot module agent knowledge base. Step S1 replaces the traditional human-led requirement parsing and functional module selection process, thereby shortening the early design cycle and improving the efficiency of robot design.

[0025] In some embodiments, the design agent generates multiple initial robot models based on the functional indicator information, the design constraint information, and a preset robot module agent knowledge base, including: Step S121: Match module agents for each functional indicator in the functional indicator information in the robot module agent knowledge base to obtain the module agent set corresponding to each functional indicator.

[0026] For example, each module agent in the robot module agent knowledge base encapsulates its own functional attributes, performance parameters, and interface specifications. For each functional indicator in the functional indicator information, the designed agent searches the robot module agent knowledge base. By comparing the functional indicator with the functional attributes of each module agent, all module agents that functionally meet the requirements are selected for the functional indicator, forming a set of module agents corresponding to the functional indicator. This achieves an effective association between design requirements and underlying module agent resources.

[0027] Step S122: Generate multiple module agent combinations based on each set of module agents.

[0028] For example, Cartesian set operations are performed on each of the said modular agent sets to obtain a combination of multiple modular agents.

[0029] Step S123: Broadcast the design constraint information to each of the module intelligent agent combinations, so that each module intelligent agent in each of the module intelligent agent combinations can interact and negotiate based on the design constraint information, generate an interaction negotiation result, and feed back the interaction negotiation result to the design intelligent agent; the interaction negotiation result is either satisfying the design constraint information or not satisfying the design constraint information.

[0030] For example, the design agent simultaneously sends design constraint information to each generated module agent combination. For each module agent combination, the module agents within the combination interact and negotiate based on the design constraint information via an A2A protocol. This negotiation process includes, but is not limited to, assessing resource supply and demand within the combination, verifying interface compatibility, and functional coordination logic. Through information exchange and collaborative computation, the module agents within the group jointly determine whether the combination as a whole can satisfy all design constraint information, and feed this collective judgment result back to the design agent as the interaction negotiation result.

[0031] Step S124: For each of the module intelligent agent combinations, when the interaction negotiation result fed back by the module intelligent agent combination is that the design constraint information is satisfied, the design intelligent agent combines each module intelligent agent in the module intelligent agent combination into an initial robot model.

[0032] For example, the design agent receives the interactive negotiation results from the combination of module agents. For combinations whose feedback satisfies the design constraints, the design agent determines that they possess the basic conditions to constitute a feasible design solution; subsequently, the design agent virtually integrates the module agents within the combination to generate an initial robot model that includes geometric structure, electrical connections, and control logic.

[0033] Understandably, the methods provided in steps S121 to S124, on the one hand, encapsulate functional modules into modular intelligent agents with negotiation capabilities, enabling the compatibility between functional modules to be self-verified during the combination stage, thereby eliminating module combinations that do not meet design constraint information in advance, and thus reducing the computational resource consumption of subsequent simulation verification stages; on the other hand, by generating multiple feasible initial robot models in parallel, diverse candidate solutions are provided for design exploration, thereby avoiding the local optima that traditional serial design processes may fall into, and thus improving the global optimization potential of the final design solution; furthermore, by automating the entire process from module matching and combination generation to compatibility checks, engineers are freed from tedious low-level coordination work, thereby improving overall design efficiency.

[0034] Step S2: The simulated intelligent agent parses the task scenario in the robot requirement information and constructs a virtual scenario that matches the task scenario.

[0035] For example, the simulation agent analyzes the robot's requirements information, extracts descriptions related to the robot's task execution environment, and parses out key task scenario elements, including but not limited to spatial layout, environmental physical attributes, and dynamic entity behavior rules. Subsequently, the simulation agent calls the simulation software via the MCP protocol, enabling the software to construct a virtual scenario matching the task scenario, thus forming a virtual scenario containing geometric models, physical parameters, and dynamic logic. This virtual scenario serves as a verification site for subsequent performance validation, ensuring the consistency between the test environment and the real task scenario, thereby providing a reliable environmental foundation for objectively evaluating the performance of the initial robot model.

[0036] Step S3: The verification agent performs performance verification on each of the initial robot models in the virtual scene, records the abnormal event information of the entire verification process, and feeds back the abnormal event information of each of the initial robot models to the design agent.

[0037] For example, for each initial robot model, the verification agent imports the initial robot model into the virtual scene and drives the initial robot model to perform multiple tasks related to the core functions of the initial robot model in the virtual scene. During the execution of tasks, the verification agent monitors and collects data in real time on the model's functional implementation, performance, and behavioral compliance. Throughout the verification process, the verification agent continuously compares the model's actual behavior with the expected goals. Once any deviation from the expectations is detected, such as functional failure, performance failure, or violation of safety rules, it is recorded as an abnormal event, and the event type of the abnormal event is recorded to obtain the abnormal event information of the initial robot model. The abnormal event information is then fed back to the design agent. Step S3, by testing multiple initial robot models and recording abnormal events, provides a basis for the design agent to conduct subsequent solution selection and targeted optimization.

[0038] In some embodiments, the verification agent performs performance verification on each of the initial robot models in the virtual scene and records abnormal event information throughout the verification process, including: Step S311: Generate multiple test cases based on the robot requirement information, and generate control scripts corresponding to each test case.

[0039] For example, the verification agent parses the robot's requirement information and extracts the defined environmental boundary conditions and core function performance boundary conditions. The environmental boundary conditions refer to the extreme values ​​of the environmental parameters of the task scenario, and the core function boundary conditions refer to the extreme requirements of the performance indicators of each core function. Based on these boundary conditions, the verification agent generates stress test cases to test the performance of the initial robot model under extreme conditions. Simultaneously, within the parameter range defined by the environmental and core function boundary conditions, typical parameter values ​​are selected to generate benchmark test cases to test the performance of the initial robot model under normal conditions. Subsequently, the verification agent writes corresponding control scripts for each generated test case. These control scripts precisely set the configuration parameters of the test environment, the input command sequence of the initial robot model, and the expected criteria for judging the robot model's output and behavior.

[0040] Step S312: For each initial robot model, import the initial robot model into the virtual scene, and execute each of the test cases on the initial robot model to drive the initial robot model to perform tasks in the virtual scene. When an exception occurs during the initial robot's task execution, record the exception event type to generate the exception event information corresponding to the initial robot model.

[0041] For example, for each initial robot model, the verification agent places it in the virtual scene; then, the verification agent loads and executes all control scripts prepared for benchmark testing and stress testing in sequence. During the test execution, the verification agent continuously compares the actual behavior of the initial robot model with the expected criteria defined in the control script. When the actual behavior of the initial robot model does not match the expected criteria, the verification agent determines that an anomaly has occurred and records the anomaly event type according to a predefined anomaly event classification system, ultimately forming the anomaly event information corresponding to the initial robot model.

[0042] Understandably, the method provided in steps S311 to S312 systematically constructs test cases covering normal working conditions and extreme boundary conditions, thereby performing layered verification of the stability and robustness of the initial robot model, and accurately identifying the performance and failure modes of the model under different intensity loads, providing effective feedback for design optimization that focuses on core requirements and critical states.

[0043] Step S4: The design agent determines the robot model to be optimized from the initial robot models based on the abnormal event information corresponding to each initial robot model, and optimizes the robot model to be optimized based on the abnormal event information corresponding to the robot model to be optimized to obtain the target robot model.

[0044] For example, the design agent receives and comprehensively analyzes the abnormal event information of each initial robot model fed back by the verification agent. Based on a preset decision-making strategy, it selects the model with the greatest optimization potential from all initial robot models as the robot model to be optimized. Subsequently, the design agent performs root cause analysis on the abnormal event information corresponding to the robot model to be optimized, locating the key module agent or design parameter causing the abnormality. Focusing on the abnormal event information corresponding to the robot model to be optimized, it performs root cause analysis to locate the associated module agent or design parameter causing the abnormality. Based on this, the design agent initiates optimization operations, which may include adjusting module parameters, replacing module agents, or reconstructing control logic. The optimized model is confirmed as the target robot model. Step S4 realizes the design closed loop. By using the feedback data generated in the verification stage to drive the evolution of the initial robot model, the design scheme can continuously improve the functional correctness and performance of the initial robot model in the iteration, thereby gradually approaching and ultimately meeting the user's needs.

[0045] In some embodiments, the design agent determines the robot model to be optimized from among the initial robot models based on the abnormal event information corresponding to each initial robot model, including: Step S411: Count the types of abnormal events in the abnormal event information corresponding to each initial robot model.

[0046] Step S412: Determine the initial robot model corresponding to the abnormal event information with the fewest abnormal event types as the robot model to be optimized.

[0047] Understandably, the methods provided in steps S411 to S412 prioritize the optimization of models with the fewest types of abnormal events, thereby focusing on design schemes with the fewest problems and relatively sound structures, thus improving the efficiency and success rate of the optimization process and helping to save computing resources.

[0048] In some embodiments, if there are multiple exception event information entries with the fewest exception event types, the initial robot model with the fewest occurrences of exception events among these exception event information entries is determined as the robot model to be optimized. It can be understood that the method provided in this embodiment, by further considering the frequency of exception occurrences when the exception types are the same, helps to identify candidate models with relatively higher operational stability, thereby improving the performance stability of the finally designed physical robot.

[0049] In some embodiments, the design agent optimizes the robot model to be optimized based on the abnormal event information corresponding to the robot model to be optimized, to obtain a target robot model, including: For each abnormal event type in the abnormal event information corresponding to the robot model to be optimized, root cause analysis is performed on the abnormal event type to determine the cause of the abnormal event type and the associated module agent in the robot model to be optimized, and the associated module agent is optimized based on the cause.

[0050] For example, the design agent analyzes abnormal event information of the robot model to be optimized, and uses methods such as fault tree analysis or causal reasoning to trace the root cause for each type of abnormal event. This analysis aims to determine which module agent(s)' attributes, behaviors, or interactions caused the abnormality. After identifying the cause and associated module agents, the design agent implements targeted optimization measures, such as modifying the internal parameters of the associated module agents, matching them with better-performing alternative module agents, or adjusting its interaction protocols with other module agents. This precise optimization method based on root cause analysis avoids blind and trial-and-error modifications, ensuring that optimization measures directly target the source of the problem, thereby effectively improving the quality and reliability of the target robot model.

[0051] Step S5: Deploy the intelligent agent to generate the manufacturing information of the physical robot required by the user based on the target robot model, guide the production of the physical robot based on the manufacturing information, and continuously monitor the physical robot after it is in operation.

[0052] For example, the deployed agent analyzes the target robot model, extracting key data such as its physical structure, parts list, and electrical connections. Based on this data, the deployed agent generates manufacturing information to guide the production of the physical robot. Subsequently, this manufacturing information is used to drive the manufacturing process of the physical robot. After the physical robot is put into operation, the deployed agent establishes a data connection with the physical robot, continuously collecting its operating status, performance indicators, and fault information. Step S5 realizes the transformation path from digital model to physical entity and builds a feedback channel between the physical world and digital design through continuous monitoring, thereby enabling the design closed loop to cover the entire life cycle of the robot, thus providing the possibility for continuous optimization and product iteration based on actual operating data.

[0053] In some embodiments, the deploying agent generates manufacturing information for the physical robot required by the user based on the target robot model, including: The target robot model is analyzed to generate component information, which includes at least a physical assembly drawing and the components of the manufactured robot; the component information includes at least the component model and the component supplier.

[0054] For example, the deployed agent decomposes the target robot model into a 3D model, generating a physical assembly drawing that details the spatial relationships and assembly sequence of each component. Simultaneously, it compiles a list containing all the components needed to manufacture the physical robot, i.e., component information, which explicitly lists the identification model and recommended supply source for each component. This process enables the automatic conversion from an integrated digital model to production guidance documents, helping to improve the manufacturing efficiency of the physical robot.

[0055] The robot design method based on agent collaboration provided in this application, on the one hand, constructs an automated and intelligent robot design process by introducing intelligent agents with clear division of labor in design, simulation, verification, and deployment, thereby transforming the traditional discrete design activities that rely on human experience into a continuous process of system-level collaboration, thus improving overall design efficiency; on the other hand, by adopting a strategy of generating multiple initial design schemes in parallel and conducting competitive screening and optimization based on objective verification data, the design space is systematically explored and rational decisions are made based on performance feedback, thereby improving the overall performance of the final design scheme; furthermore, by deploying intelligent agents to achieve the connection from digital model to physical entity and the continuous collection of operational data, a design-deployment-feedback closed loop covering the entire life cycle of the robot is established, which helps to support the continuous iteration and optimization of robot products.

[0056] Another embodiment of the present invention also provides a robot design system based on intelligent agent collaboration. Figure 2 A schematic block diagram of a robot design system based on agent collaboration provided in another embodiment of the present invention, as shown below. Figure 2 As shown, the robot design system based on agent collaboration includes: designing the agent and simulating and verifying the agent. The design agent generates functional indicator information and design constraint information for robot design based on the user's robot requirement information, and generates multiple initial robot models based on the functional indicator information, the design constraint information and a preset robot module agent knowledge base. The simulated intelligent agent parses the task scenario in the robot's requirement information and constructs a virtual scenario that matches the task scenario; The verification agent performs performance verification on each of the initial robot models in the virtual scene, records abnormal event information during the entire verification process, and feeds back the abnormal event information of each of the initial robot models to the design agent. The design agent determines the robot model to be optimized from the initial robot models based on the abnormal event information corresponding to each initial robot model, and optimizes the robot model to be optimized based on the abnormal event information corresponding to the robot model to be optimized to obtain the target robot model. The deployed intelligent agent generates manufacturing information for the physical robot required by the user based on the target robot model, guides the production of the physical robot based on the manufacturing information, and continuously monitors the physical robot after it is in operation.

[0057] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the system and each intelligent agent described above can be referred to in the aforementioned embodiments of the robot design method based on intelligent agent collaboration, and will not be repeated here.

[0058] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A robot design method based on agent collaboration, characterized in that, include: The design agent generates functional indicator information and design constraint information for robot design based on the user's robot requirement information, and generates multiple initial robot models based on the functional indicator information, the design constraint information and a preset robot module agent knowledge base. The simulated intelligent agent parses the task scenario in the robot's requirement information and constructs a virtual scenario that matches the task scenario; The verification agent performs performance verification on each of the initial robot models in the virtual scene, records abnormal event information during the entire verification process, and feeds back the abnormal event information of each of the initial robot models to the design agent. The design agent determines the robot model to be optimized from the initial robot models based on the abnormal event information corresponding to each initial robot model, and optimizes the robot model to be optimized based on the abnormal event information corresponding to the robot model to be optimized to obtain the target robot model. The deployed intelligent agent generates manufacturing information for the physical robot required by the user based on the target robot model, guides the production of the physical robot based on the manufacturing information, and continuously monitors the physical robot after it is in operation.

2. The robot design method based on agent collaboration according to claim 1, characterized in that, The design agent generates multiple initial robot models based on the functional indicator information, the design constraint information, and a preset robot module agent knowledge base, including: In the robot module agent knowledge base, match module agents for each functional indicator in the functional indicator information to obtain the module agent set corresponding to each functional indicator; Multiple combinations of module intelligent agents are generated based on the sets of intelligent agents described above; The design constraint information is broadcast to each of the module agent combinations, so that each module agent in each module agent combination can interact and negotiate based on the design constraint information, generate an interaction negotiation result, and feed the interaction negotiation result back to the design agent; the interaction negotiation result is either satisfying the design constraint information or not satisfying the design constraint information. For each of the aforementioned module agent combinations, when the interaction negotiation result fed back by the module agent combination is that the design constraint information is satisfied, the design agent combines each module agent in the module agent combination into an initial robot model.

3. The robot design method based on agent collaboration according to claim 2, characterized in that, Each module agent in the aforementioned module agent combination interacts and negotiates based on the design constraint information to generate an interaction negotiation result, including: For each of the aforementioned module agent combinations, the module agents in the module agent combination interact and negotiate based on the design constraint information through the A2A protocol to obtain the interaction negotiation result corresponding to the module agent combination.

4. The robot design method based on agent collaboration according to claim 1, characterized in that, The simulated intelligent agent constructs a virtual scene that matches the task scenario, including: The simulation software is invoked via the MCP protocol, so that the simulation software can construct a virtual scene that matches the task scenario.

5. The robot design method based on agent collaboration according to claim 1, characterized in that, The verification agent performs performance verification on each of the initial robot models in the virtual scene and records abnormal event information throughout the verification process, including: Based on the robot requirement information, multiple test cases are generated, and control scripts corresponding to each test case are generated. For each initial robot model, the initial robot model is imported into the virtual scene, and each of the test cases is executed on the initial robot model to drive the initial robot model to perform tasks in the virtual scene. When an exception occurs during the initial robot's task execution, the exception event type is recorded to generate exception event information corresponding to the initial robot model.

6. The robot design method based on agent collaboration according to claim 1, characterized in that, The design agent determines the robot model to be optimized from among the initial robot models based on the abnormal event information corresponding to each initial robot model, including: Each initial robot model is assigned a specific number of abnormal events from its corresponding abnormal event information. The initial robot model corresponding to the anomaly information with the fewest types of anomalies is identified as the robot model to be optimized.

7. The robot design method based on agent collaboration according to claim 6, characterized in that, The method further includes: If there are multiple exception event information with the fewest exception event types, the initial robot model with the fewest occurrences of exception events among the exception event information with the fewest exception event types is determined as the robot model to be optimized.

8. The robot design method based on agent collaboration according to claim 1, characterized in that, The design agent optimizes the robot model to be optimized based on the abnormal event information corresponding to the robot model to be optimized, to obtain the target robot model, including: For each abnormal event type in the abnormal event information corresponding to the robot model to be optimized, root cause analysis is performed on the abnormal event type to determine the cause of the abnormal event type and the associated module agent in the robot model to be optimized, and the associated module agent is optimized based on the cause.

9. The robot design method based on agent collaboration according to claim 1, characterized in that, The deploying agent generates manufacturing information for the physical robot required by the user based on the target robot model, including: The target robot model is analyzed to generate component information, which includes at least a physical assembly drawing and the components of the manufactured robot; the component information includes at least the component model and the component supplier.

10. A robot design system based on agent collaboration, characterized in that, This includes designing intelligent agents, simulating intelligent agents, validating intelligent agents, and deploying intelligent agents; The design agent generates functional indicator information and design constraint information for robot design based on the user's robot requirement information, and generates multiple initial robot models based on the functional indicator information, the design constraint information and a preset robot module agent knowledge base. The simulated intelligent agent parses the task scenario in the robot's requirement information and constructs a virtual scenario that matches the task scenario; The verification agent performs performance verification on each of the initial robot models in the virtual scene, records abnormal event information during the entire verification process, and feeds back the abnormal event information of each of the initial robot models to the design agent. The design agent determines the robot model to be optimized from the initial robot models based on the abnormal event information corresponding to each initial robot model, and optimizes the robot model to be optimized based on the abnormal event information corresponding to the robot model to be optimized to obtain the target robot model. The deployed intelligent agent generates manufacturing information for the physical robot required by the user based on the target robot model, guides the production of the physical robot based on the manufacturing information, and continuously monitors the physical robot after it is in operation.

Images