Control method of industrial robot, electronic device, and storage medium

Abstract

The application provides a control method of an industrial robot, an electronic device and a storage medium. The control method of the industrial robot comprises: packaging a basic control model of the industrial robot, and building a hybrid model framework; based on a large model, task planning is performed on the industrial robot to obtain a high-level control model; a virtual agent of a work scene is built in a simulator, and the high-level control model is trained based on the simulator; and based on the trained high-level control model, the industrial robot is controlled to perform a work task. The application controls and programs the industrial robot for specific industrial tasks through a large model, realizes flexible use of the robot in an industrial manufacturing environment, enables the industrial robot to perform multiple tasks, enables the industrial robot to have certain thinking and reasoning capabilities, has strong flexibility and reusability, effectively improves production efficiency, reduces system cost, and improves user experience.

Description

Technical Field

[0001] This application relates to the field of industrial robot technology, specifically to a control method, electronic device, and storage medium for an industrial robot. Background Technology

[0002] Industrial robots are multi-jointed manipulators or multi-degree-of-freedom machine devices used in the field of engineering machinery. They can perform various industrial manufacturing tasks by receiving instructions or pre-set programs and relying on their own power and control capabilities. Based on their functions, they are divided into spot welding robots, handling robots, and assembly robots, etc. Industrial robots are mainly composed of mechanical structure systems, drive systems, sensing systems, robot-environment interaction systems, and human-machine interaction and control systems.

[0003] In conceiving and developing this application, the applicant identified at least the following problems: Existing industrial robots have a clearly defined division of labor and can only perform a single task. Industrial robots need to be customized according to specific scenarios, with complex programs pre-programmed, lacking a certain level of thinking and reasoning ability, and exhibiting poor flexibility and reusability. Summary of the Invention

[0004] To alleviate the above problems, this application provides a control method for an industrial robot, including:

[0005] The basic control model of industrial robots is encapsulated to build a hybrid model framework;

[0006] Based on the large model, task planning is performed on the industrial robot to obtain a high-level control model;

[0007] A virtual agent for the work scenario is built in the simulator, and the high-level control model is trained based on the simulator;

[0008] Based on the trained high-level control model, the industrial robot is controlled to perform work tasks.

[0009] Optionally, the basic control model includes a visual language model, a robot control model, and a motion planning model.

[0010] Optionally, the step of encapsulating the basic control model of the industrial robot and building a hybrid model framework includes:

[0011] Multiple basic control models are encapsulated into callable interfaces, and multiple functional functions of the industrial robot are defined based on the callable interfaces, so that the large model can call them for the functions that need to be implemented.

[0012] Optionally, the step of performing task planning on the industrial robot based on a large model to obtain a high-level control model includes:

[0013] Create a prompt script to correspond to the role positioning of the industrial robot and define the task of the industrial robot;

[0014] Based on the multiple functional functions, and according to the task of the industrial robot, the target functional function corresponding to the tool to be used is determined;

[0015] The high-level control model is established based on the implementation order of the target functional functions.

[0016] Optionally, the step of determining the functional function corresponding to the required tool based on the plurality of functional functions and according to the task of the industrial robot includes:

[0017] Using the thought chain capability of the large model, the task of the industrial robot is broken down into multiple operation steps;

[0018] The various operational steps are programmed separately to call the corresponding functional functions.

[0019] Optionally, the virtual agent for the work scenario includes at least one of the following:

[0020] The corresponding virtual agent for the industrial robot;

[0021] The virtual agent for the items that the industrial robot needs to interact with.

[0022] Optionally, the step of building a virtual agent for the work scenario in the simulator and training the high-level control model based on the simulator includes:

[0023] Based on the high-level control model corresponding to the task of the industrial robot, the target functional functions are called in the order of their implementation.

[0024] Based on the target functional function, the robot virtual agent is controlled to perform interactive actions with the item virtual agent;

[0025] Collect interaction data of the interaction actions, and when the interaction data is abnormal, optimize the call to the target functional function.

[0026] Optionally, the industrial robot includes a handling robot, and the task includes handling items; the step of collecting interaction data of the interactive actions and adjusting the call to the target functional function when the interaction data is abnormal includes at least one of the following:

[0027] When the robot virtual agent of the handling robot cannot recognize the virtual agent of the item, the image recognition model in the hybrid model framework is modified;

[0028] When the robot virtual agent of the transport robot lacks a target interactive action, a target functional function corresponding to the target interactive action is added so that the target interactive action can be called accordingly.

[0029] When the interaction efficiency of the robot virtual agent of the handling robot is lower than a preset threshold, the prompt script is modified so that the large model can redetermine the high-level control model corresponding to the task of the industrial robot.

[0030] This application also provides an electronic device, including a processor and a storage medium interconnected, wherein:

[0031] The storage medium is used to store computer programs;

[0032] The processor is used to read and run the computer program to implement the industrial robot control method described above.

[0033] This application also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the industrial robot control method described above.

[0034] The industrial robot control method, electronic device, and storage medium provided in this application encapsulate the basic control model of the industrial robot to build a hybrid model framework. Based on the large model, task planning is performed on the industrial robot to obtain a high-level control model. A virtual agent of the work scene is built in a simulator, and the high-level control model is trained based on the simulator. Based on the trained high-level control model, the industrial robot is controlled to perform work tasks. This application uses a large model to program the control of the industrial robot for specific industrial tasks, realizing the flexible use of the robot in the industrial manufacturing environment. This enables the industrial robot to perform multiple tasks, giving it a certain degree of thinking and reasoning ability, strong flexibility and reusability, effectively improving production efficiency, reducing system costs, and improving user experience. Attached Figure Description

[0035] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0036] Figure 1 This is a flowchart of an industrial robot control method according to an embodiment of this application.

[0037] The realization of the objectives, functional features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. The accompanying drawings have illustrated specific embodiments of this application, which will be described in more detail below. These drawings and textual descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0038] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0039] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, components, features, and elements with the same names in different embodiments of this application may have the same meaning or different meanings, the specific meaning of which must be determined by its interpretation in that specific embodiment or further in conjunction with the context of that specific embodiment.

[0040] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0041] First Embodiment

[0042] On one hand, this application provides a control method for an industrial robot. Figure 1 This is a flowchart of an industrial robot control method according to an embodiment of this application.

[0043] like Figure 1 As shown, in one embodiment, the control method for an industrial robot includes:

[0044] S10: Encapsulate the basic control model of industrial robots and build a hybrid model framework.

[0045] For example, an industrial robot refers to a multi-jointed manipulator or multi-degree-of-freedom machine device applied in the field of engineering machinery. It can perform various industrial manufacturing tasks by receiving instructions or pre-set programs and relying on its own power and control capabilities. Based on their functions, they are divided into spot welding robots, handling robots, and assembly robots, etc. Industrial robots are mainly composed of mechanical structure systems, drive systems, perception systems, robot-environment interaction systems, and human-machine interaction and control systems. Optionally, a hybrid model framework can be built using multiple models of various sizes and with different functions, such as robot control algorithm models and visual inspection models, and embedded into the industrial robot system to process multimodal information. The larger model calls the smaller models to collaboratively complete tasks in industrial manufacturing scenarios.

[0046] S20: Based on the large model, perform task planning for the industrial robot to obtain a high-level control model.

[0047] For example, a large model is embedded in a robot. Industrial production managers input natural voice commands, and the large model can analyze the true intentions and perform task decomposition and planning. Through the functional interfaces encapsulated in the hybrid model framework, the corresponding basic control model is called, thereby programming the control instructions that the machine can understand.

[0048] S30: Build a virtual agent for the work scenario in the simulator, and train the high-level control model based on the simulator.

[0049] For example, since training industrial robots in the real world is costly, robots can be trained in simulators. These simulators create environments that recreate manufacturing scenarios, containing various interactive objects. The industrial robots also have virtual agents within the simulator. These virtual agents can be trained in the simulator, and by evaluating their performance in the virtual environment, the hybrid model framework, functional functions, and the output of the large model can be optimized. Once the results are satisfactory through user testing, the hybrid model framework and the large natural language model can be deployed to a real-world robotic system.

[0050] S40: Based on the trained high-level control model, control the industrial robot to perform work tasks.

[0051] For example, the trained high-level control model is a mature industrial robot control model that can be deployed in the control system of industrial robots. By understanding the user's natural language commands, it automatically decomposes tasks and controls the industrial robot to execute corresponding work steps. Production managers can directly input natural language task commands, which are then broken down by the large natural language model. The step-by-step plan can be simply programmed, and the commands for each step can be implemented by calling different models.

[0052] This embodiment uses a large model to program the control of an industrial robot for specific industrial tasks, enabling flexible use of the robot in industrial manufacturing environments. This allows the industrial robot to perform multiple tasks, possesses a certain level of thinking and reasoning ability, and has strong flexibility and reusability, effectively improving production efficiency, reducing system costs, and enhancing user experience.

[0053] Optionally, the basic control model includes a visual language model, a robot control model, and a motion planning model.

[0054] For example, mature visual language models, robot control models, motion planning models, etc., can be encapsulated into callable interfaces for use in programming large natural language models.

[0055] Optionally, the step of encapsulating the basic control model of the industrial robot and building a hybrid model framework includes:

[0056] Multiple basic control models are encapsulated into callable interfaces, and multiple functional functions of the industrial robot are defined based on the callable interfaces, so that the large model can call them for the functions that need to be implemented.

[0057] A hybrid model framework is built, utilizing different models to achieve different functions. For example, calling a visual detection algorithm can complete object recognition in a handling task, while calling a robot control algorithm can achieve grasping in a handling task. For instance, mature visual language models, robot control models, and motion planning models are encapsulated into callable interfaces. Then, commonly used functional functions for industrial robots are defined based on this. For example, a high-level function `move_to(x, y, z)` can be defined to perform navigation, where x, y, and z represent the position of the destination point. Internally, the function calls motion planning and low-level robot control commands (such as movement direction, movement distance, and rotation).

[0058] Optionally, the step of performing task planning on the industrial robot based on a large model to obtain a high-level control model includes:

[0059] Create a prompt script to correspond to the role positioning of the industrial robot and define the task of the industrial robot;

[0060] Based on the multiple functional functions, and according to the task of the industrial robot, the target functional function corresponding to the tool to be used is determined;

[0061] The high-level control model is established based on the implementation order of the target functional functions.

[0062] For example, a suitable prompt script can be designed. The main content of the prompt script can include: the robot's role positioning (e.g., handling robot, welding robot, inspection robot, packaging robot, maintenance robot, repair robot, etc.), the robot's task (e.g., moving the material numbered 1 on shelf A to location B), and the tools that can be used (i.e., the functional functions defined when building the hybrid model framework, which can be understood as the robot's capabilities).

[0063] The large model used in this scheme can be ChatGPT, or other generative language large models such as Wenxin Yiyan.

[0064] Optionally, the step of determining the functional function corresponding to the required tool based on the plurality of functional functions and according to the task of the industrial robot includes:

[0065] Using the aforementioned large natural language model, the task of the industrial robot is broken down into multiple operation steps;

[0066] The various operational steps are programmed separately to call the corresponding functional functions.

[0067] Strong logical reasoning is one of the core capabilities of large language models' "emergent intelligence," and the key to this reasoning ability lies in the Chain of Thought (CoT) capability. A Chain of Thought breaks down a logical reasoning problem into multiple steps, allowing for a more clear logical chain in the generated result, providing interpretability and showing how the answer was arrived at. The main idea of ​​a Chain of Thought is to present the large language model with a small number of example prompts, explaining the reasoning process within these examples. The large language model also displays the reasoning process when responding to the prompts. This explanation of reasoning often leads to more accurate results. For example, a large natural language model can leverage its Chain of Thought capability to break down a task into a robot-achievable step-by-step plan. In other words, the programming of the large natural language model based on a function library can be understood as a series of instructions calling perception / action algorithms. For the example above, the large model might output:

[0068] move_to(x_a,y_a,z_a), get_object('1'), move_to(x_b,y_b,z_b)

[0069] Optionally, the virtual agent for the work scenario may include:

[0070] A robot virtual agent corresponding to the industrial robot; and / or an item virtual agent corresponding to the items that the industrial robot needs to interact with.

[0071] For example, since training industrial robots in the real world is costly, robots can be trained in simulators to create an environment that recreates the manufacturing scenario. This environment contains various interactive objects, and the industrial robots also have virtual agents in the simulator.

[0072] Optionally, the step of building a virtual agent for the work scenario in the simulator and training the high-level control model based on the simulator includes:

[0073] Based on the high-level control model corresponding to the task of the industrial robot, the target functional functions are called in the order of their implementation.

[0074] Based on the target functional function, the robot virtual agent is controlled to perform interactive actions with the item virtual agent;

[0075] Collect interaction data of the interaction actions, and when the interaction data is abnormal, optimize the call to the target functional function.

[0076] For example, a task command is input to the virtual agent, the large language model is programmed according to a predefined function library, and the virtual agent completes the task command in a virtual environment by executing the program. Optionally, the virtual agent can be trained in a simulator, and the hybrid model framework, function functions, and output of the large model can be optimized by evaluating the agent's task performance in the virtual environment.

[0077] Users continuously send various tasks to the virtual agent in industrial scenarios, observe the virtual agent's task execution, and continuously optimize it until it passes user testing. Finally, the optimized hybrid model framework and large model are deployed to the actual robot system.

[0078] Optionally, the industrial robot includes a handling robot, and the task includes handling items; the step of collecting interaction data of the interactive actions and adjusting the call to the target functional function when the interaction data is abnormal may include:

[0079] When the robot virtual agent of the transport robot cannot recognize the virtual agent of the item, the image recognition model in the hybrid model framework is modified.

[0080] For example, taking the item handling task mentioned above as an example, if the user finds that the virtual agent cannot correctly identify all items numbered 1 in get_object('1'), then the image recognition model in the hybrid model framework should be optimized or replaced with a model that has better recognition performance.

[0081] Optionally, the industrial robot includes a handling robot, and the task includes handling items; the step of collecting interaction data of the interactive actions and adjusting the call to the target functional function when the interaction data is abnormal may include:

[0082] When the robot virtual agent of the transport robot lacks a target interactive action, a target functional function corresponding to the target interactive action is added so that the target interactive action can be called accordingly.

[0083] For example, if a user finds that the virtual agent stops working after reaching location B and does not perform the material placement operation, then the function library needs to be expanded to add a function for placing materials.

[0084] Optionally, the industrial robot includes a handling robot, and the task includes handling items; the step of collecting interaction data of the interactive actions and adjusting the call to the target functional function when the interaction data is abnormal may include:

[0085] When the interaction efficiency of the robot virtual agent of the handling robot is lower than a preset threshold, the prompt script is modified so that the large model can redetermine the high-level control model corresponding to the task of the industrial robot.

[0086] For example, if a user finds that the virtual agent is inefficient in executing the task planning given by the large model, they can fine-tune the large model by designing a Prompt to make it output a better task planning scheme. It should be noted that the preset threshold for the efficiency of interactive actions can be flexibly set according to the specific work task, and this application does not limit it in this regard.

[0087] Second Embodiment

[0088] This application also provides an electronic device, including a processor and a storage medium interconnected, wherein:

[0089] The storage medium is used to store computer programs;

[0090] The processor is used to read and run the computer program to implement the industrial robot control method described above.

[0091] In one embodiment, an industrial robot based on embodied intelligence, controlled by an electronic device, can perform various industrial manufacturing tasks such as navigation, handling, and spot welding. It has certain human-computer interaction capabilities and autonomous decision-making capabilities, and can be reused in multiple different industrial manufacturing scenarios.

[0092] Optionally, a large natural language model can be embedded into the robot. Industrial production managers input natural voice commands, and the large model can analyze the true intent, decompose and plan the tasks, and program them into machine-understandable control instructions. Optionally, a hybrid model framework can be built using multiple models of various sizes and with different functions, such as robot control algorithm models and visual inspection models, and embedded into the industrial robot system to process multimodal information. The large natural language model calls the smaller models to collaboratively complete tasks in industrial manufacturing scenarios.

[0093] Optionally, industrial robots need to have basic mobility functions, be equipped with robotic arms, and have sensors, cameras, etc. that are generally equipped with robots.

[0094] 1. First, build the hybrid model framework using electronic devices:

[0095] Mature visual language models, robot control models, and motion planning models are encapsulated into callable interfaces. Then, based on these interfaces, commonly used functional functions for industrial robots are defined for use by larger models. For example, a high-level function `move_to(x, y, z)` is defined to perform navigation, where x, y, and z represent the position of the target point. Internally, the function calls motion planning and low-level robot control commands (movement direction, movement distance, rotation).

[0096] A hybrid model framework is built, with different models performing different functions. For example, a visual detection algorithm is called to complete object recognition in the handling task, and a robot control algorithm is called to achieve grasping in the handling task.

[0097] 2. Electronic devices utilize large-scale natural language models for task planning:

[0098] Design a suitable prompt. The main contents of the prompt include: the robot's role (assuming it is a handling robot), the robot's task (e.g., moving the material numbered 1 on shelf A to location B), and the tools that can be used (i.e., the functional functions defined when building the hybrid model framework, which can be understood as the robot's capabilities).

[0099] The large model used in this approach can be ChatGPT or other generative language large models.

[0100] The large model leverages its thought chain capability to break down tasks into step-by-step plans that the robot can implement. In other words, the large model's programming, based on a function library, can be understood as a series of instructions that call perception / action algorithms. For the example above, the large model might output: move_to(x_a,y_a,z_a), get_object('1'), move_to(x_b,y_b,z_b).

[0101] 3. Electronic devices train robotic agents in embodied intelligence simulators.

[0102] Because training industrial robots in the real world is costly, they can be trained in simulators. Simulators can be used to create environments that recreate manufacturing scenarios, containing various interactive objects. The industrial robots also have virtual agents within the simulator. These virtual agents can be trained in the simulator, and by evaluating their performance in the virtual environment, the hybrid model framework can be optimized, functional functions can be improved or expanded, and the output of larger models can be optimized.

[0103] The specific training optimization process is as follows:

[0104] Input a task command into the virtual agent, and the natural language big data model is programmed according to the predefined function library. The virtual agent executes the program to complete the task command in the virtual environment.

[0105] Taking the task of moving items mentioned in step 2 as an example, if the user finds that the virtual agent cannot correctly identify all materials numbered 1 in get_object('1'), then optimize the image recognition model in the hybrid model framework or replace it with a model with better recognition performance;

[0106] If users find that the virtual agent stops working after reaching location B and does not perform the material placement operation, the function library needs to be expanded to add the function for placing materials.

[0107] If users find that the virtual agent is inefficient in executing the task planning given by the large model, they can fine-tune the large model by designing a Prompt to make it output a better task planning scheme.

[0108] Users continuously send various tasks to the virtual agent in industrial scenarios, observe the virtual agent's task execution, and continuously optimize it until it passes user testing. Finally, the optimized hybrid model framework and natural language big data model are deployed to the actual robot system.

[0109] 4. Electronic devices control industrial robots to accomplish production tasks through a pre-trained natural language model:

[0110] Production managers can directly input natural language task instructions, which will be broken down into tasks by a large model. They can also perform simple programming on the step-by-step plan and implement the instructions for each step by calling different models.

[0111] It should be noted that the large model used in this scheme can be a natural language large model such as ChatGPT or Wenxin Yiyan, or other generative language large models.

[0112] Compared with traditional industrial robots, the industrial robots controlled by electronic devices in this solution can perform a variety of different manufacturing tasks. There is no need to pre-set a fixed robot control program. After embedding a hybrid model framework and defining sufficient robot function functions, the large model can be programmed to control the robot for specific industrial tasks, realizing the flexible use of robots in industrial manufacturing environments.

[0113] Through the electronic device in this embodiment, production managers can directly input natural language task instructions, which are then broken down by a large model. The step-by-step plan can be easily programmed, and instructions for each step can be implemented by calling different models.

[0114] By building a hybrid model framework, different models can perform different functions, such as calling visual detection algorithms to complete object recognition in handling tasks, and calling robot control algorithms to achieve grasping in handling tasks.

[0115] Third Embodiment

[0116] This application also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the industrial robot control method described above.

[0117] The industrial robot control method, electronic device, and storage medium provided in this application encapsulate the basic control model of the industrial robot to build a hybrid model framework; based on a large natural language model, task planning is performed on the industrial robot to obtain a high-level control model; a virtual agent of the work scene is built in a simulator, and the high-level control model is trained based on the simulator; based on the trained high-level control model, the industrial robot is controlled to perform work tasks. This application uses a large natural language model to program the industrial robot for specific industrial tasks, enabling flexible use of the robot in industrial manufacturing environments, allowing the industrial robot to perform multiple tasks, possessing a certain degree of thinking and reasoning ability, strong flexibility and reusability, effectively improving production efficiency, reducing system costs, and enhancing user experience.

[0118] It should be noted that step designations such as S10 and S20 are used in this application for the purpose of more clearly and concisely describing the corresponding content, and do not constitute a substantial limitation on the order. In specific implementation, those skilled in the art may execute S20 first and then S10, etc., but these should all be within the protection scope of this application.

[0119] In the embodiments of the electronic devices and storage media provided in this application, all the technical features of any of the above-described method embodiments may be included. The extended and explanatory content of the specification is basically the same as that of the embodiments of the above methods, and will not be repeated here.

[0120] This application also provides a computer program product, which includes computer program code. When the computer program code is run on a computer, it causes the computer to perform the methods described in the various possible implementations above.

[0121] This application also provides a chip, including a memory and a processor. The memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that a device with the chip installed performs the methods described in the various possible implementations above.

[0122] It is understood that the above scenarios are merely examples and do not constitute a limitation on the application scenarios of the technical solutions provided in the embodiments of this application. The technical solutions of this application can also be applied to other scenarios. For example, as those skilled in the art will know, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0123] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0124] The steps in the method of this application embodiment can be adjusted, combined, or deleted according to actual needs.

[0125] The units in the device of this application embodiment can be merged, divided, and deleted according to actual needs.

[0126] In this application, the same or similar terms, concepts, technical solutions and / or application scenario descriptions are generally described in detail only when they appear for the first time. When they appear again, they are generally not repeated for the sake of brevity. When understanding the technical solutions and other contents of this application, the same or similar terms, concepts, technical solutions and / or application scenario descriptions that are not described in detail later can be referred to their previous relevant detailed descriptions.

[0127] In this application, the descriptions of the various embodiments have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0128] The technical features of the present application can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of the present application.

[0129] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A control method for an industrial robot, characterized in that, include: The basic control model of industrial robots is encapsulated to build a hybrid model framework; The industrial robot is task-planned based on a large model to obtain a high-level control model. A virtual agent for the work scenario is built in the simulator, and the high-level control model is trained based on the simulator; Based on the trained high-level control model, the industrial robot is controlled to perform work tasks. The step of performing task planning on the industrial robot based on a large model to obtain a high-level control model includes: Create a prompt script to correspond to the role positioning of the industrial robot and define the task of the industrial robot; Based on multiple functional functions, the target functional function corresponding to the tools to be used is determined according to the task of the industrial robot; Based on the implementation order of the target functional functions, the high-level control model is established; The step of determining the functional function corresponding to the required tool based on multiple functional functions and according to the task of the industrial robot includes: The large model is used to break down the task of the industrial robot into multiple operation steps; The various operational steps are programmed separately to call the corresponding functional functions.

2. The control method for an industrial robot according to claim 1, characterized in that, The basic control model includes a visual language model, a robot control model, and a motion planning model.

3. The control method for an industrial robot according to claim 1, characterized in that, The steps for encapsulating the basic control model of the industrial robot and building a hybrid model framework include: Multiple basic control models are encapsulated into callable interfaces, and multiple functional functions of the industrial robot are defined based on the callable interfaces, so that the large model can call them for the functions that need to be implemented.

4. The control method for an industrial robot according to claim 3, characterized in that, The virtual agent for the work scenario includes at least one of the following: The corresponding virtual agent for the industrial robot; The virtual agent for the items that the industrial robot needs to interact with.

5. The control method for an industrial robot according to claim 4, characterized in that, The steps of building a virtual agent for the work scenario in the simulator and training the high-level control model based on the simulator include: Based on the high-level control model corresponding to the task of the industrial robot, the target functional functions are called in the order of their implementation. Based on the target functional function, the robot virtual agent is controlled to perform interactive actions with the item virtual agent; Collect interaction data of the interaction actions, and when the interaction data is abnormal, optimize the call to the target functional function.

6. The control method for an industrial robot according to claim 5, characterized in that, The industrial robot includes a handling robot, and the task includes handling items; the step of collecting interaction data of the interaction actions and adjusting the call to the target functional function when the interaction data is abnormal includes at least one of the following: When the robot virtual agent of the handling robot cannot recognize the virtual agent of the item, the image recognition model in the hybrid model framework is modified; When the robot virtual agent of the transport robot lacks a target interactive action, a target functional function corresponding to the target interactive action is added so that the target interactive action can be called accordingly. When the interaction efficiency of the robot virtual agent of the handling robot is lower than a preset threshold, the prompt script is modified so that the large model can redetermine the high-level control model corresponding to the task of the industrial robot.

7. An electronic device, characterized in that, This includes interconnected processors and storage media, wherein: The storage medium is used to store computer programs; The processor is used to read and run the computer program to implement the industrial robot control method as described in any one of claims 1-6.

8. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the steps of the industrial robot control method as described in any one of claims 1-6.

Images