System for automatically manufacturing robot

Abstract

The invention provides a system for automatically manufacturing a robot, and the system comprises an interaction unit which is used for receiving an input instruction of a decision maker, guiding the decision maker to provide a complete instruction when the input instruction is incomplete, and then generating an initial manufacturing instruction; the information work station is used for performing robot manufacturing task reasoning based on multiple learning modes according to the initial manufacturing instruction to form robot manufacturing strategies including a robot manufacturing scheme and a morphological model and structural parameters of the robot, and is further used for adjusting the robot manufacturing strategies according to manufacturing logs of manufactured robots and sending the robot manufacturing strategies to the robot manufacturing system; until the manufactured robot meets the performance test requirement; and the manufacturing and testing work station is used for automatically carrying out material transportation, part preparation and assembly and robot performance testing according to the generated robot manufacturing strategy, generating a manufacturing log and feeding back the manufacturing log to the information work station and the interaction unit. According to the method, the automation of the whole process from robot design to manufacturing is achieved, the intelligent degree is high, and the self-adaptive capacity is high.

Description

Technical Field

[0001] The present invention relates to the field of robot technology, and in particular to a system for automatically manufacturing robots by robots. Background Art

[0002] With the rapid development of robotics technology, robots are increasingly used in industrial manufacturing, medical health, service industry and other fields. However, the manufacturing process of traditional robots usually requires a lot of manual intervention, including material preparation, parts processing, assembly and debugging, which is not only time-consuming and labor-intensive, but also limits the large-scale customized production capacity of robots. At the same time, the development of modern intelligent manufacturing has put forward higher requirements for the degree of automation, hoping to achieve the management and execution of complex manufacturing processes through the intelligence and autonomy of robots themselves. In recent years, the concept of "Robots Making Robots" has gradually emerged, that is, to achieve the manufacturing of robots themselves by using robotics technology. This concept not only helps to improve production efficiency, but also promotes the high intelligence and adaptability of robot manufacturing systems. However, the current methods and devices for automatic robot manufacturing still face many technical challenges, such as how to efficiently coordinate the collaborative manufacturing of multi-robot systems, how to achieve adaptive assembly of robots with complex structures, and how to ensure the accuracy and reliability of devices during the manufacturing process. Summary of the invention

[0003] The present invention aims to solve one of the technical problems existing in the prior art at least to a certain extent.

[0004] To this end, the purpose of the present invention is to provide a system for automatically manufacturing robots by robots, which can achieve full process autonomy from design to manufacturing by combining modern artificial intelligence, mechanical control and automation technology, and has a high degree of intelligence and strong adaptability.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] The present invention provides a system for automatically manufacturing robots by a robot, comprising:

[0007] An interaction unit, used to receive an input instruction from a decision maker, and to guide the decision maker to provide a complete instruction when the input instruction is incomplete, and then generate an initial manufacturing instruction, wherein the initial manufacturing instruction is relevant information of the robot to be manufactured described in a natural language, including the type of robot, the performance requirements of the robot, and the target task to be completed by the robot;

[0008] An information workstation, for performing robot manufacturing task reasoning based on a plurality of learning modes for the initial manufacturing instructions issued by the interaction unit, forming a robot manufacturing strategy, wherein the robot manufacturing strategy includes a robot manufacturing plan and a robot morphological model and structural parameters; and for adjusting the robot manufacturing strategy according to the manufacturing log of the manufactured robot until the manufactured robot meets the performance test requirements, thereby terminating the current robot manufacturing task;

[0009] The manufacturing and testing workstation is used to automatically carry out material transportation, parts preparation and assembly, and robot performance testing according to the robot manufacturing strategy generated by the information workstation, generate a manufacturing log and feed it back to the information workstation and the interaction unit.

[0010] In some embodiments, the information workstation includes:

[0011] A priori knowledge learning unit, which is used to extract prior knowledge about the entire manufacturing process of the robot from existing technologies including articles, web pages, and patents through deep learning methods and regularly update the prior knowledge;

[0012] A performance evolution learning unit is used to build a simulation environment according to the robot performance requirements in the initial manufacturing instructions, and to achieve the evolution of the robot in the simulation environment by changing the relevant parameters of the robot within a set range and using reinforcement learning training to improve the performance of the robot for the initial robot model, and obtain the evolved robot parameters, wherein the evolved robot parameters include the robot's structural parameters, material parameters, control parameters and morphological model; each type of robot has a set parameter range;

[0013] An active learning unit, configured to generate a robot manufacturing strategy based on deep learning according to the evolved robot parameters output by the performance evolution learning unit and in combination with the prior knowledge output by the prior knowledge learning unit, wherein the robot manufacturing plan in the robot manufacturing strategy includes a software preparation plan, a hardware preparation and operation plan, a molding-integration operation plan and a test plan, and the test plan includes a material performance test plan and a work performance test plan;

[0014] If the manufacturing log of the robot manufactured in the previous round shows that the robot performance does not meet the standard, the active learning unit fine-tunes the robot manufacturing strategy according to the manufacturing log of the robot manufactured in the previous round to generate a new robot manufacturing strategy; if the number of times the active learning unit fine-tunes the robot manufacturing strategy exceeds a set threshold and the robot performance still does not meet the standard, the information workstation re-reasons the robot manufacturing task.

[0015] In some embodiments, the prior knowledge learning unit comprises a retrieval module and a mining module; the retrieval module first identifies keywords from the initial manufacturing instructions through natural language processing technology, and then uses the keywords to retrieve relevant information from the prior art; the mining module performs in-depth analysis on the retrieved relevant information based on a large language model to extract the robot preparation method, morphological structure and performance data information related to the initial manufacturing instructions.

[0016] In some embodiments, the simulation environment constructed by the performance evolution learning unit includes a training task scenario corresponding to the robot's execution of a target task and a verification task scenario corresponding to a performance test of the robot.

[0017] In some embodiments, for each type of robot to be manufactured, a sample set consisting of multiple groups of successful manufacturing strategies is used to train the large model to obtain the corresponding active learning unit.

[0018] In some embodiments, the manufacturing and testing workstation includes a control platform and a 3D printing platform, a 3D data generation platform, a robot software operating platform, a robot hardware operating platform, a robot molding-integration operating platform, a robot testing operating platform and a mobile robot connected thereto;

[0019] The control platform is used to send the robot manufacturing strategy output by the information workstation to the corresponding platform, monitor its task execution status, generate control instructions for the mobile robot according to the task execution status, generate a manufacturing log according to the task execution results fed back by each platform and the mobile robot, and feed it back to the information workstation and the interaction unit;

[0020] The mobile robot, in response to the control instructions issued by the control platform, includes taking corresponding materials from the material warehouse and transporting them to the corresponding operation platform according to the current robot manufacturing process, and interacting with each operation platform and the 3D printing platform during the robot manufacturing process to achieve object transportation and auxiliary operation tasks;

[0021] The 3D data generation platform has a first large model and a second large model therein; the first large model generates three-dimensional structural data of molds, robot hardware and connectors that meet the accuracy and functional requirements based on the robot morphology model and structural parameters in the current robot manufacturing strategy, and sends the data to the 3D printing platform; the second large model generates three-dimensional structural data of test supplies required for performance testing of the manufactured robot entity according to the test plan in the current robot manufacturing strategy, and sends the data to the 3D printing platform; the three-dimensional structural data generated by the first large model has a priority printing level compared to the three-dimensional structural data generated by the second large model;

[0022] The 3D printing platform is used to perform 3D printing according to various three-dimensional structure data generated by the 3D data generation platform;

[0023] The robot software operating platform is used to complete the processing of robot software test samples and software parts according to the software preparation plan in the current robot manufacturing strategy and with the assistance of the mobile robot;

[0024] The robot hardware operation platform is used to complete the processing of robot hardware test samples and hardware parts according to the hardware preparation and operation plan in the current robot manufacturing strategy and with the assistance of the mobile robot;

[0025] The robot forming-integration operation platform is used to complete the forming and integration of robot software and hardware parts according to the forming-integration operation scheme in the current robot manufacturing strategy and with the assistance of the mobile robot;

[0026] The robot testing operation platform is used to complete the material performance test of the software test samples processed by the robot software operating platform for each type of robot software parts according to the test plan in the current robot manufacturing strategy and with the assistance of the mobile robot, to test the material performance of the hardware test samples processed by the robot hardware operating platform for each type of robot hardware parts, to perform working performance tests on the robot entities obtained by the robot forming-integration operation platform, and to generate a performance test report.

[0027] In some embodiments, the robotic software operating platform includes a No. 1 robotic arm, a No. 1 visual work chamber, a No. 1 sample rack work chamber, a liquid distribution work chamber, a solid distribution work chamber, a stirring work chamber, a vacuum drying work chamber, an electronic scale work chamber, a spin coating work chamber and a capping work chamber, wherein the No. 1 sample rack work chamber includes test tubes, beakers, culture dishes and molds printed by the 3D printing platform.

[0028] In some embodiments, the robot hardware operating platform includes a No. 2 robotic arm, a No. 2 visual work chamber, a No. 2 sample rack work chamber, a circuit construction work chamber, an electrode processing work chamber and a screw assembly work chamber that cooperate with each other, wherein the No. 2 sample rack work chamber includes various hardware accessories, electrical materials and nano silver wires required for making robot hardware.

[0029] In some embodiments, the robotic molding-integrated operating platform includes a No. 3 robotic arm, a No. 3 visual work chamber, a No. 3 sample rack work chamber, a demolding work chamber, a laser cutting work chamber, a hydraulic molding work chamber, a dispensing work chamber and an assembly work chamber that cooperate with each other, wherein the No. 3 sample rack work chamber includes a culture dish, a molding mold and connectors printed by the 3D printing platform.

[0030] In some embodiments, the robot testing operation platform includes a No. 4 robot arm, a No. 4 visual workstation, a power supply workstation, a multi-channel voltage workstation, an arbitrary waveform generator, an LCR digital bridge, a Shore hardness measurement workstation, an electronic universal testing machine, a conductivity test workstation, a multi-task operation platform and a software loading end that cooperate with each other; wherein, the software material performance test is completed by the Shore hardness measurement workstation, the electronic universal testing machine and the conductivity test workstation; the hardware material performance test is completed by the No. 4 visual workstation, the power supply workstation, the multi-channel voltage workstation, the arbitrary waveform generator and the LCR digital bridge; the robot's working performance test is first performed by the software loading end to load the software required to perform the target task to the robot entity, and then the No. 4 visual workstation, the power supply workstation, the multi-channel voltage workstation, the arbitrary waveform generator, the LCR digital bridge and the multi-task operation platform complete the specific working performance item test.

[0031] Compared with the prior art, the present invention has the following characteristics and beneficial effects:

[0032] The present invention utilizes modern artificial intelligence, mechanical control and automation technologies to realize the full process autonomy of the robot from design to manufacturing and testing. It can work 24 / 7 without interruption, greatly improve production efficiency, reduce labor costs, and can be monitored in real time. The data in the production process can be used to optimize the process. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 A structural block diagram of a system for automatically manufacturing robots by a robot provided in an embodiment of the present invention.

[0035] Figure 2 This is a schematic structural diagram of a voxel robot capable of grabbing eggs and its four voxel modules to be manufactured according to an embodiment of the present invention. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0037] On the contrary, the present application covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present application as defined by the claims. Further, in order to make the public have a better understanding of the present application, some specific details are described in detail in the detailed description of the present application below. Those skilled in the art can fully understand the present application without the description of these details.

[0038] See also Figure 1 According to a first aspect of the present invention, a system for automatically manufacturing robots by a robot is provided, comprising:

[0039] The interaction unit 100 is used to receive the input instructions of the decision maker, and when the input instructions are incomplete, guide the decision maker to provide complete instructions and then generate initial manufacturing instructions, wherein the initial manufacturing instructions are relevant information of the robot to be manufactured described in natural language, including the type of robot, the performance requirements of the robot and the target task to be completed by the robot;

[0040] The information workstation 200 is used to perform robot manufacturing task reasoning based on multiple learning modes for the initial manufacturing instructions issued by the interaction unit 100, and form a robot manufacturing strategy, which includes a robot manufacturing plan and a robot morphological model and structural parameters; and to adjust the robot manufacturing strategy according to the manufacturing log of the manufactured robot until the manufactured robot meets the performance test requirements, thereby terminating the current robot manufacturing task;

[0041] The manufacturing and testing workstation 300 is used to automatically carry out material transportation, parts preparation and assembly, and robot performance testing according to the robot manufacturing strategy generated by the information workstation 200, generate a manufacturing log and feed it back to the information workstation 200 and the interaction unit 100.

[0042] In some embodiments, the interactive unit 100 uses a web dialog box or a chat tool (such as WeChat) with parsing and reasoning capabilities. The decision maker's input instructions can be in the form of voice, text, or video, which are related information of the robot to be manufactured described in natural language, and can be a description of the type and performance of the robot to be manufactured, including the specific type of the robot and the target tasks and performance indicators to be performed. The interactive unit 100 parses and infers the decision maker's input instructions to form initial manufacturing instructions, such as requiring the manufacture of a voxel robot that can grab eggs, a soft robot that can climb over a mouse, or an EIT sensor that can measure human breathing. When the interactive unit 100 determines that the decision maker's input instructions are incomplete, it will automatically guide the decision maker to complete the instruction information.

[0043] In some embodiments, the information workstation 200 serves as the core of the system and performs inference of robot manufacturing solutions based on multiple learning modes. Specifically, the information workstation 200 includes:

[0044] A priori knowledge learning unit 210 is used to extract priori knowledge about the entire manufacturing process of the robot from existing technologies such as articles, web pages, and patents through a deep learning method and to regularly update the priori knowledge;

[0045] The performance evolution learning unit 220 is used to build a simulation environment according to the robot performance requirements in the initial manufacturing instructions, and to improve the performance of the robot by changing the relevant parameters of the robot within a set range (each type of robot has a set parameter range) for the initial robot model and using reinforcement learning training to achieve the evolution of the robot in the simulation environment, and obtain the evolved robot parameters, which include the robot's structural parameters, material parameters, control parameters and morphological model;

[0046] The active learning unit 230 is used to generate a robot manufacturing strategy based on deep learning according to the evolved robot parameters output by the performance evolution learning unit 220 and the prior knowledge output by the prior knowledge learning unit 210. The robot manufacturing plan in the robot manufacturing strategy includes a software preparation plan, a hardware preparation and operation plan, a molding-integration operation plan, and a robot performance test plan. The robot performance test plan includes a material performance test plan and a work performance test plan.

[0047] If the manufacturing log of the robot manufactured in the previous round (including various monitoring data during the robot manufacturing process and material performance and working performance test reports of the robot entity) shows that the robot performance does not meet the standards, the manufacturing and testing workstation 300 sends the manufacturing log to the information workstation 200 through the interactive unit 100, and first uses the active learning unit 230 to analyze the reasons for the failure of this round of manufacturing tasks based on the manufacturing log, and fine-tunes the robot manufacturing strategy to generate a new robot manufacturing strategy; if the number of times the active learning unit 230 fine-tunes the robot manufacturing strategy exceeds the set threshold, that is, the robot performance still does not meet the standards after multiple rounds of fine-tuning, it is necessary to use the information workstation 200 to re-infer the robot manufacturing task to ensure that the performance of the robot entity is consistent with the performance of the robot in the simulation environment.

[0048] Furthermore, the prior knowledge learning unit 210 includes a retrieval module and a mining module. The retrieval module first identifies keywords from the initial manufacturing instructions through natural language processing technology, and then uses these keywords to retrieve relevant information from external data sources (such as technical literature, patents, web pages, etc.). The mining module conducts in-depth analysis of the retrieved relevant information based on a large language model (such as BERT, GPT, etc.), and extracts information such as robot preparation methods, morphological structures and performance data related to the initial manufacturing instructions. Through the analysis of the large language model, the mining module can identify key information such as robot manufacturing processes, material requirements, robot design structures and performance evaluation in the relevant information, and integrate and output this information to provide data support for the subsequent active learning unit 230 to generate a production plan. The entire process of the prior knowledge learning unit 210 acquiring prior knowledge is continuously optimized through an adaptive learning mechanism to ensure that the knowledge base can be updated according to new technological advances and literature to provide the latest and most relevant prior knowledge.

[0049] Furthermore, the process of the performance evolution learning unit 220 evolving the robot performance according to the initial manufacturing instruction includes:

[0050] The performance evolution learning unit 220 builds a training task scenario corresponding to the robot's target task in the first simulation platform (such as genesis) according to the robot performance requirements in the initial manufacturing instructions, initializes the robot model and sets the range of various robot parameters required for evolutionary learning. It is worth noting that robots of the same category to be manufactured have the same set of evolutionary learning parameters, which need to be pre-stored in the performance evolution unit 220. For example, for voxel robots, the evolutionary learning parameters include control parameters (specifically including: single inflation / inhalation volume, inflation / inhalation position, inflation / inhalation strategy), structural parameters (specifically including: module assembly method, module size, number of modules, module wall thickness, module connectivity domain, module shape) and material parameters (specifically including: silicone hardness, silicone tensile properties, silicone Young's modulus);

[0051] According to the parameters of evolutionary learning, the robot's structural parameters, material parameters and control parameters are evolved by reinforcement learning within the set range to ensure that the robot achieves the optimal target performance in the simulation environment;

[0052] A verification task scenario corresponding to the performance test of the robot is built in a second simulation platform (such as ABAQUS or COMSOL) different from the first simulation platform, and the performance of the robot obtained in the training phase is simulated and tested. If the simulation test fails, the performance evolution learning unit 220 performs a new round of evolutionary learning until the simulation test passes, and the evolutionary learning is terminated to obtain the evolved parameters. The performance evolution learning unit 220 transmits the evolved parameters obtained, including the structure / characteristics / control parameters and morphological model of the robot, to the active learning unit 230 as one of the basic data of the active learning unit 230.

[0053] In a specific embodiment of the present application, in response to the decision maker's input instruction: "Please help me make a voxel soft robot that can cross the mouse", the performance evolution learning unit 220 builds a mouse and an initial voxel soft robot in a simulation environment, and changes the structural parameters of the voxel soft robot (such as the size, quantity, shape, wall thickness of each small module of the voxel soft robot, as well as the connectivity and arrangement of the modules, etc.), the material properties of each module (such as elastic modulus, etc.) and the control parameters of the voxel soft robot (such as the filling / inhalation volume, filling / inhalation method, filling / inhalation position, etc. of the active module), and uses reinforcement learning training to continuously optimize the performance of the voxel soft robot crossing the mouse. Once the task is achieved in the simulation environment, the evolutionary learning task is terminated, and the structure, material and control parameters of the evolved robot and the morphological model are output in the form of a list.

[0054] Furthermore, the active learning unit 230 outputs the robot manufacturing strategy according to the evolved robot parameters output by the performance evolution learning unit 220 and the prior knowledge output by the prior knowledge learning unit 210. For each type of robot, a sample set consisting of multiple groups (e.g., 100 groups) of successful manufacturing strategies is used, and based on the prior knowledge generated by the prior knowledge learning unit 210, the large model is trained to obtain the corresponding active learning unit.

[0055] It is understandable that in the embodiment of the present invention, the evolutionary parameters output by the evolutionary learning unit 220 through reinforcement learning in the simulation environment ensure that the corresponding robot meets the needs of the decision maker under ideal conditions, but the actual robot produced may deviate from the ideal robot, and the fine-tuning of the active learning unit 230 promotes the preparation of a real robot that meets the needs, and the active learning unit 230 realizes the generation from robot parameters to manufacturing strategies, which makes it easier for the manufacturing and testing workstation 300 to implement the manufacturing plan. In addition, the active learning unit 230 also summarizes and analyzes the reasons for failure from the previous round of failed manufacturing tasks, and generates a new robot manufacturing strategy to ensure the smooth progress of the robot manufacturing task.

[0056] In some embodiments, the manufacturing and testing workstation 300 is used as the execution part of the system of the embodiment of the present invention, and is used to execute the robot manufacturing strategy issued by the information workstation 200. Specifically, the manufacturing and testing workstation 300 includes a control platform (the control platform is Figure 1 3D printing platform 320, 3D data generation platform 330, robot software operation platform 340, robot hardware operation platform 350, robot molding-integration operation platform 360, robot testing operation platform 370 and mobile robot 310 are connected thereto; wherein,

[0057] The control platform is used to send the robot manufacturing strategy output by the information workstation 200 to the corresponding platform, monitor its task execution status, generate control instructions for the mobile robot 310 according to the task execution status, generate a manufacturing log according to the task execution results fed back by each platform and the mobile robot, and feed it back to the information workstation 200 and the interaction unit 100;

[0058] The mobile robot 310 responds to the control instructions issued by the control platform, including taking the corresponding materials from the material warehouse and transporting them to the corresponding operation platform according to the current robot manufacturing process, and interacting with the four operation platforms and the 3D printing platform 320 during the robot manufacturing process to realize the object transportation and auxiliary operation tasks;

[0059] The 3D data generation platform 330 is provided with a first large model and a second large model, and the three-dimensional structure data of the required items are generated by the first large model and the second large model respectively and sent to the 3D printing platform 320 for printing; the first large model generates the three-dimensional structure data of the mold, robot hardware and connectors that meet the accuracy and functional requirements based on the robot morphology model and structure parameters in the current robot manufacturing strategy, and then sends the generated three-dimensional structure data to the 3D printing platform 320; the second large model generates the three-dimensional structure data of the test supplies required for the performance test of the manufactured robot entity according to the test plan in the current robot manufacturing strategy, and sends the result to the 3D printing platform 320 for printing; the 3D data generation platform 330 of the embodiment of the present invention realizes efficient automation from robot morphology model design to actual printing; the three-dimensional structure data generated by the first large model has a priority printing level compared with the three-dimensional structure data generated by the second large model;

[0060] The 3D printing platform 320 is used to perform 3D printing according to various three-dimensional structure data generated by the 3D data generation platform 330;

[0061] The robot software operating platform 340 is used to complete the processing of robot software test samples and software parts according to the software preparation plan in the current robot manufacturing strategy and with the assistance of the mobile robot 310;

[0062] The robot hardware operation platform 350 is used to complete the processing of robot hardware test samples and hardware parts according to the hardware preparation and operation plan in the current robot manufacturing strategy and with the assistance of the mobile robot 310;

[0063] A robot molding-integration operation platform 360 is used to complete the molding and integration of robot software and hardware parts according to the molding-integration operation scheme in the current robot manufacturing strategy and with the assistance of the mobile robot 310;

[0064] The robot test operation platform 370 is used to complete the material performance test of the software test samples processed by the robot software operation platform 340 for each type of robot software parts (including the construction of material performance test scenarios and the execution of specific test items) according to the test plan in the current robot manufacturing strategy and with the assistance of the mobile robot 310, to test the material performance of the hardware test samples processed by the robot hardware operation platform 350 for each type of robot hardware parts (including the construction of material performance test scenarios and the execution of specific test items), to perform working performance tests on the integrated robot entities (including the construction of working performance test scenarios and the execution of specific test items), to generate performance test reports (including material performance test reports and working performance test reports), and to feed back to the control platform.

[0065] Furthermore, the mobile robot 310 automatically performs material transportation and task coordination by receiving control instructions issued by the control platform. The mobile robot 310 plans the action path according to the control instructions, accurately obtains the required materials and tools from the material warehouse, and transports them to the corresponding operating platform. The mobile robot 310 also collaborates with the 3D printing platform 320 and various operating platforms to provide necessary material support, monitor the progress of the task in real time, and feedback data to the control platform through sensors and communication modules to ensure smooth work. In addition, the mobile robot 310 has intelligent path planning and obstacle avoidance capabilities, can automatically adjust the driving route according to environmental changes, and finally recycle waste materials after completing the task and prepare for the next task, which greatly improves the automation and efficiency of the robot manufacturing process.

[0066] Furthermore, the 3D printing platform 320, after receiving the three-dimensional data from the 3D data generating platform 330, performs printing using a resin material, and the three-dimensional structure data generated by the first large model has a priority printing level.

[0067] Furthermore, the 3D data generation platform 330 generates the three-dimensional structure data of the required object through the first large model and the second large model respectively and sends them to the 3D printing platform for printing. Since there are large differences in the input and output of the first large model and the second large model, in order to reduce the difficulty of training the large model, the 3D data generation platform 330 uses a large model to process the two types of input data (i.e., the robot's morphological model and structural parameters, and the test plan) respectively, and outputs the corresponding three-dimensional structure data.

[0068] Furthermore, in each operating platform, the corresponding manufacturing process is completed based on machine vision:

[0069] The robot software operating platform 340 includes a No. 1 robotic arm, a No. 1 visual work chamber, a No. 1 sample rack work chamber, a liquid distribution work chamber, a solid distribution work chamber, a stirring work chamber, a vacuum drying work chamber, an electronic scale work chamber, a spin coating work chamber and a capping work chamber, wherein the No. 1 sample rack work chamber includes test tubes, beakers, culture dishes and 3D printing preparation molds, etc.;

[0070] The robot hardware operation platform 350 includes a No. 2 robotic arm, a No. 2 visual work chamber, a No. 2 sample rack work chamber, a circuit construction work chamber, an electrode processing work chamber and a screw assembly work chamber, wherein the No. 2 sample rack work chamber includes various hardware accessories, electrical materials and nano silver wires, etc.;

[0071] The robot molding-integrated operation platform 360 includes a No. 3 robotic arm, a No. 3 visual work chamber, a No. 3 sample rack work chamber, a demoulding work chamber, a laser cutting work chamber, a hydraulic molding work chamber, a dispensing work chamber and an assembly work chamber, wherein the No. 3 sample rack work chamber includes a culture dish, a molding mold and assembly parts, etc. The assembly parts are connectors printed by the 3D printing platform 320;

[0072] The robot test operation platform 370 includes a No. 4 robot arm, a No. 4 visual workstation, a power workstation, a multi-channel voltage workstation, an arbitrary waveform generator, an LCR digital bridge, a Shaw hardness measurement workstation, an electronic universal testing machine, a conductivity test workstation, a multi-task operation platform and a software loading end that cooperate with each other; wherein, the soft material performance test is completed by the Shaw hardness measurement workstation, the electronic universal testing machine, and the conductivity test workstation; the hard material performance test is completed by the No. 4 visual workstation, the power workstation, the multi-channel voltage workstation, the arbitrary waveform generator and the LCR digital bridge; the robot's working performance test is firstly performed by the software loading end to load the software required to perform the target task to the robot entity, and then the No. 4 visual workstation, the power workstation, the multi-channel voltage workstation, the arbitrary waveform generator, the LCR digital bridge and the multi-task operation platform complete the specific working performance project test;

[0073] Each of the robotic arms and work chambers, the electronic universal testing machine and the 3D printing platform 320 in the above-mentioned operating platforms are connected to the control platform through a serial port. The control platform can obtain data and control these devices in real time based on the code; in addition, each operating platform is provided with a control operation program, in which messages can be set to be sent.

[0074] The working process of the manufacturing and testing workstation 300 of this embodiment includes the following steps (the execution order of the following steps can be adjusted according to actual conditions):

[0075] Step (1) Material preparation and 3D printing (thread 1 and thread 2 are carried out simultaneously)

[0076] Thread 1: The mobile robot 310 takes the corresponding materials from the material bin to complete the material preparation according to the production plan generated by the information workstation 200;

[0077] Thread 2: The first large model in the 3D data generation platform 330 generates the three-dimensional structural data of the molds, robot hardware and connectors required in the robot manufacturing process based on the morphological model and structural parameters in the current robot manufacturing strategy, and sends it to the 3D printing platform 320 for printing; then, the second large model generates the three-dimensional structural data of the test supplies according to the test plan in the production plan of the current robot manufacturing strategy, and sends it to the 3D printing platform 320 for printing.

[0078] Step (2) Feeding and testing sample processing (thread 1 and thread 2 are carried out simultaneously)

[0079] Thread 1: The mobile robot 310 transports the materials in the material bin and places them on the robot software operating platform 340 and the robot hardware operating platform 350; then, the mobile robot 310 transports the molds, hardware, connectors, and testing supplies printed by the 3D printing platform 320 and places them on the robot software operating platform 340, the robot hardware operating platform 350, the robot molding-integration operating platform 360, and the robot testing operating platform 370;

[0080] Thread 2: After the robot software operating platform 340 has visually inspected the materials and prepared them, it starts processing the test samples of the robot software part. For each type of software part of the robot, a corresponding software test sample is processed for subsequent material performance testing;

[0081] Thread 3: After the robot hardware operating platform 350 has completed the visual inspection of the materials, it starts processing the test samples of the robot hardware part. For each type of hardware part of the robot, a corresponding hardware test sample is processed for subsequent material performance testing;

[0082] Step (3) Transporting test samples

[0083] Thread 1: After the robot software operating platform 340 completes the preparation of the software test sample, it sends a signal to the transport robot 310, and uses the mobile robot 310 to send the software test sample to the robot test operating platform 370 to prepare for the material testing experiment; after the robot hardware operating platform 350 completes the preparation of the hardware test sample, it sends a signal to the transport robot 310 to send the test sample, and uses the mobile robot 310 to send the hardware test sample to the robot test operating platform 370 to prepare for the material testing experiment;

[0084] Step (4) Testing sample material performance

[0085] Thread 1: After the mobile robot 310 transports the software test sample prepared by the robot software operating platform 340 to the test operating platform 370, the software test sample is placed on the corresponding detection platform, and then a signal is sent to the robot test operating platform 370 to test the sample material performance. After the sample material performance test is completed, the robot test operating platform 370 sends a message to the robot software operating platform 340 to start the preparation of the corresponding software parts; correspondingly, the mobile robot 310 transports the hardware test sample prepared by the robot hardware operating platform 350 to the robot test operating platform 370, places the hardware test sample on the corresponding detection platform, and then sends a signal to the robot test operating platform 370 to test the sample material performance. After the sample material performance test is completed, the robot test operating platform 370 sends a message to the robot hardware operating platform 350 to start the processing of the corresponding hardware parts;

[0086] Step (5) Robot parts processing

[0087] Thread 1: After the robot software operating platform 340 has visually inspected the materials and has prepared them, it starts processing the robot software parts;

[0088] Thread 2: After the robot hardware operating platform 350 has completed the visual inspection of the materials, it starts processing the robot hardware parts;

[0089] Step (6) Collecting materials

[0090] Thread 1: After the robot software operating platform 340 and the robot hardware operating platform 350 have finished using the materials, they send a message to the mobile robot 310 to retrieve the unused materials to the material warehouse and return them to the corresponding material platform;

[0091] Step (7) Robotic molding-integration

[0092] Thread 1: After the robot software operating platform 340 and the robot hardware operating platform 350 complete the preparation of the corresponding parts, they send a message to the mobile robot 310, first transport the software parts of the software operating platform 340 to the robot forming-integrated operating platform 360, and then transport the hardware parts of the robot hardware operating platform 350 to the robot forming-integrated operating platform 360, and then transport the connector parts of the 3D printing platform 320 to the robot forming-integrated operating platform 360;

[0093] Step (8) Robot performance test

[0094] Thread 1: After the robot molding-integrated operation platform 360 is completed, a message is sent to the mobile robot 310. The mobile robot 310 transports the integrated robot entity to the software loading end of the robot performance test platform 370 for software loading, and then builds the corresponding test scene and performs work performance testing to generate a performance test report.

[0095] Step (9) Robot entity and manufacturing log

[0096] The robot entity is placed on the robot test operation platform 370. The control platform generates this round of manufacturing logs based on the task execution results (including performance test reports) fed back by each operation platform, and sends them to the interactive platform 100 and the active learning model 230 in the information workstation 200. The interactive platform 100 passes the performance test report in the manufacturing log to the decision maker.

[0097] If the performance test of the robot manufactured in this round fails, the manufacturing and testing workstation 300 sends the manufacturing log to the information workstation 200 through the interactive unit 100, and first uses the active learning model 230 to fine-tune the robot manufacturing strategy to start the next cycle. If the number of times the active learning unit 230 fine-tunes the robot manufacturing strategy exceeds the set threshold, that is, the robot performance still does not meet the standard after multiple rounds of fine-tuning, the information workstation 200 is used to re-infer the robot manufacturing task until a robot that passes the performance test is manufactured, and the robot manufacturing task is terminated.

[0098] The following describes a specific embodiment of the present invention - the manufacture of a voxel robot, and the specific steps are as follows:

[0099] (1) The decision maker inputs the instruction: Please help me make a voxel robot that can grab eggs;

[0100] (2) After receiving the above instructions, the interactive unit 100 parses them to generate initial manufacturing instructions, and sends the instructions to the performance evolution learning unit 220 of the information workstation 200. The performance evolution learning unit 220 establishes a simulation environment with a bionic egg model and a voxel robot initial model; then continuously trains the voxel robot's morphology, structure, inflation and suction strategy, Young's modulus, hardness, etc. of the material used to prepare the voxel robot in the simulation environment; verifies the performance of the trained voxel robot in the egg grabbing operation in the ABAQUS simulation environment; generates a morphological model of the voxel robot that can achieve the target function - grabbing eggs, as well as structure, material and control parameters under the criteria of material saving, high efficiency and easy manufacturing, as the evolved robot parameters. The evolved robot parameters are input to the active learning unit 230. Based on the prior knowledge generated by the prior knowledge learning unit 210, the active learning unit 230 generates a manufacturing strategy, which includes: the preparation materials, preparation steps, ratios and time of the voxel modules of the voxel robot, the material properties of the software (Young's modulus, hardness, etc.), the size / model / position of the inflation hose used, the model / quantity / inflation size / inflation time of the inflation pump, the voxel robot's molding-integration plan (voxel robot integration order), and the testing plan (using the voxel robot to grab eggs).

[0101] See also Figure 2 In this embodiment, the voxel robot to be manufactured contains four types of voxel modules. Figure 2 , which are shown in light blue, green, yellow and red respectively, the first voxel module contains only the submodule numbered 10, the second voxel module contains only the submodule numbered 20, the third voxel module contains submodules numbered 31 to 36, and the fourth voxel module contains submodules numbered 41 to 46. The structure / characteristic / control parameters for the voxel robot generated by the active learning unit 230 are:

[0102] 4,2,4; 20,42,42,20,20,42,42,20,41,32,32,41,41,32,32,41,10,0,0,10,10,0,0,10,31,0,0,41,31,0,0,41; 15,15,15; 1 / 420,520,520,520; 30,43,43,43 / 4; 7,1,7.5,7.5,2; 15,1,7.5,7.5,2; 29,1,6,6,2; 32,1, 6,6,2; 0,0.2; 4,2,4 are the number of voxel units contained in the length, width and height of the voxel robot's outer contour; 20,42,42,20,20,42,42,20,41,32,32,41,41,32,32,41,10,0,0,10,10,0,0,10,31,0,0,41,31,0,0,41 are the arrangement parameters of the voxel modules from top to bottom and from left to right of the voxel robot (corresponding to the numbers of each submodule). There are four types of voxel modules, such as Figure 2 As shown, no position arrangement parameter of the voxel module is set to 0; 15, 15, 15 means that the size of a single voxel module is 15mm*15mm*15mm; 1 means that the wall thickness of the voxel module is 1mm; 420, 520, 520, 520 means that the Young's modulus of the four voxel modules are 420KPa, 520KPa, 520KPa, 520KPa respectively; 30, 43, 43, 43 means that the hardness of the four voxel modules is 30ShaoreA, 43ShaoreA, 43ShaoreA, 43ShaoreA respectively; 4 means that the voxel robot has four opening positions; 7, 1, 7.5, 7.5, 2 means that the first surface of the 7th voxel module is 7 meters away from the left line of the voxel module .5mm, and a 2mm hole is opened at a position 7.5mm away from the right side line of the voxel module; 15,1,7.5,7.5 means that a 2mm hole is opened on the first surface of the 15th voxel module, 7.5mm away from the left side line of the voxel module and 7.5mm away from the right side line of the voxel module; 29,1,6,6,2 means that a 2mm hole is opened on the first surface of the 29th voxel module, 6mm away from the left side line of the voxel module and 6mm away from the upper side line of the voxel module; 32,1,6,6,2 means that a 2mm hole is opened on the first surface of the 32nd voxel module, 6mm away from the left side line of the voxel module and 6mm away from the upper side line of the voxel module; 0 means the standard control strategy; 0.2 means that the peak suction control pressure is 0.2Pa.

[0103] The production plan output by the active learning unit 230 is:

[0104] (2.1) The mobile robot 310 transports the materials: ecoflex 0050A, ecoflex 0050B, yellow pigment, green pigment, red pigment and blue pigment to the software operating platform of the robot 310;

[0105] (2.2) Preparation and testing steps of test samples:

[0106] (2.2.1) Yellow voxel module test sample preparation

[0107] (2.2.1.1) Robot arm No. 1 grabs test tube No. 1 from a clean test tube rack and unscrews the cap of test tube No. 1 in the cap-unscrewing working compartment. Robot arm No. 1 places the unscrewed test tube No. 1 in the electronic scale working compartment, and the electronic scale working compartment is reset to zero;

[0108] (2.2.1.2) Robot arm No. 1 grabs the ecoflex 0050A material bottle and pours 13 g into test tube No. 1;

[0109] (2.2.1.3) Robot arm No. 1 grabs the ecoflex 0050B material bottle and pours 15 g into test tube No. 1;

[0110] (2.2.1.4) Robot arm No. 1 grabs the yellow material bottle and pours 1g into test tube No. 1;

[0111] (2.2.1.5) Robot arm No. 1 grabs test tube No. 1 and screws the cap on the test tube in the cap screwing chamber;

[0112] (2.2.1.6) Robot arm No. 1 transports the capped test tube No. 1 to the mixing work chamber;

[0113] (2.2.1.7) Stir the working chamber at a speed of πrad / s for 5 minutes;

[0114] (2.2.1.8) Robot arm No. 1 unscrews the cap of test tube No. 1 after stirring in the capping chamber, and then pours 5g of the material in test tube No. 1 into test mold 1;

[0115] (2.2.1.9) Robot arm No. 1 places test tube No. 1 on the dirty test tube rack. After cleaning, robot arm No. 1 places test tube No. 1 on the clean test tube rack;

[0116] (2.2.2) Preparation of green voxel module test samples

[0117] (2.2.2.1) Robot arm No. 1 grabs test tube No. 2 from a clean test tube rack and unscrews the cap of test tube No. 2 in the cap-unscrewing working chamber. Robot arm No. 1 places the unscrewed test tube No. 2 in the electronic scale working chamber, and the electronic scale working chamber is reset to zero;

[0118] (2.2.2.2) Robot arm No. 1 grabs the ecoflex 0050A material bottle and pours 15 g into test tube No. 2;

[0119] (2.2.2.3) Robot arm No. 1 grabs the ecoflex 0050B material bottle and pours 15 g into test tube No. 2;

[0120] (2.2.2.4) Robot arm No. 1 grabs the green material bottle and pours 1g into test tube No. 2;

[0121] (2.2.2.5) Robot arm No. 1 grabs test tube No. 2 and screws the cap on the test tube in the cap screwing chamber;

[0122] (2.2.2.6) Robot arm No. 1 transports the capped test tube No. 2 to the mixing chamber;

[0123] (2.2.2.7) Stir the working chamber at a speed of πrad / s for 5 minutes;

[0124] (2.2.2.8) Robot arm No. 1 unscrews the cap of test tube No. 2 after stirring in the capping chamber, and then pours 5g of the material in test tube No. 2 into test mold 2;

[0125] (2.2.2.9) Robot arm No. 1 places test tube No. 2 on the dirty test tube rack. After cleaning, robot arm No. 1 places test tube No. 2 on the clean test tube rack;

[0126] (2.2.3) Red voxel module test sample preparation

[0127] (2.2.3.1) Robot arm No. 1 grabs test tube No. 3 from a clean test tube rack and unscrews the cap of the test tube in the cap-unscrewing working chamber. Robot arm No. 1 places the uncapped test tube No. 3 in the electronic scale working chamber, and the electronic scale working chamber is reset to zero;

[0128] (2.2.3.2) Robot arm No. 1 grabs the ecoflex 0050A material bottle and pours 15 g into test tube No. 3;

[0129] (2.2.3.3) Robot arm No. 1 grabs the ecoflex 0050B material bottle and pours 15 g into test tube No. 3;

[0130] (2.2.3.4) Robot arm No. 1 grabs the red pigment material bottle and pours 1g into test tube No. 3;

[0131] (2.2.3.5) Robot arm No. 1 grabs test tube No. 3 and screws the cap on the test tube in the cap screwing chamber;

[0132] (2.2.3.6) Robot arm No. 1 transports the capped test tube No. 3 to the mixing chamber;

[0133] (2.2.3.7) Stir the working chamber at a speed of πrad / s for 5 minutes;

[0134] (2.2.3.8) Robot arm No. 1 unscrews the cap of test tube No. 3 after stirring in the capping chamber, and then pours 5g of the material in test tube No. 3 into test mold 3;

[0135] (2.2.3.9) Robot arm 1 places test tube 3 on the dirty test tube rack. After cleaning, robot arm 1 places test tube 3 on the clean test tube rack.

[0136] (2.2.4) Preparation of blue voxel module test samples

[0137] (2.2.4.1) Robot arm No. 1 grabs test tube No. 4 from a clean test tube rack and unscrews the cap of the test tube in the cap-unscrewing working chamber. Robot arm No. 1 places the uncapped test tube No. 4 in the electronic scale working chamber, and the electronic scale working chamber is reset to zero;

[0138] (2.2.4.2) Robot arm No. 1 grabs the ecoflex 0050A material bottle and pours 15 g into test tube No. 4;

[0139] (2.2.4.3) Robot arm No. 1 grabs the ecoflex 0050B material bottle and pours 15 g into test tube No. 4;

[0140] (2.2.4.4) Robot arm No. 1 grabs the blue material bottle and pours 1g into test tube No. 4;

[0141] (2.2.4.5) Robot arm No. 1 grabs test tube No. 4 and screws the cap on the test tube in the cap screwing chamber;

[0142] (2.2.4.6) Robot arm No. 1 transports the capped test tube No. 4 to the mixing chamber;

[0143] (2.2.4.7) Stir the working chamber at a speed of πrad / s for 5 minutes;

[0144] (2.2.4.8) Robot arm No. 1 unscrews the cap of test tube No. 4 after stirring in the capping working chamber, and then pours 5g of the material in test tube No. 4 into test mold 4;

[0145] (2.2.4.9) Robot arm No. 1 places test tube No. 4 on the dirty test tube rack. After cleaning, robot arm No. 1 places test tube No. 4 on the clean test tube rack.

[0146] (2.2.5) 20 minutes later, the mobile robot 310 transports the test mold 1, the test mold 2, the test mold 3, and the test mold 4 to the demolding work chamber of the robot molding-integrated operation platform 360;

[0147] (2.2.6) The demoulding work chamber of the robot molding-integrated operation platform 360 demoulds the module test sample in the mold;

[0148] (2.2.7) The mobile robot 310 transports the test samples of each module to the Shaw hardness measurement work chamber and the electronic universal testing machine of the robot test operation platform 370 to measure the Young's modulus and hardness, and stores the data in the performance report. If it is consistent with 420KPa, 520KPa, 520KPa, 520KPa, 30ShaoreA, 43ShaoreA, 43ShaoreA, 43ShaoreA (the error within ±1 indicates consistency), the voxel module will be prepared; if it is inconsistent, the performance report will be sent to the interaction unit 100, and the manufacturing and testing workstation 300 will stop operating. The interaction unit 100 sends the parameters after the previous round of evolution to the active learning unit 230, starts the prediction of the next round of active learning, and generates a new robot manufacturing strategy;

[0149] It is worth noting that if the requirements are not met after 10 cycles, the interaction unit 100 will resend the decision maker's requirements to the information workstation 200 to re-reason the robot manufacturing task.

[0150] (2.3) Preparation and testing steps of voxel modules:

[0151] (2.3.1) Yellow voxel module preparation

[0152] (2.3.1.1) Robot arm No. 1 grabs test tube No. 1 from a clean test tube rack and unscrews the cap of test tube No. 1 in the cap-unscrewing working compartment. Robot arm No. 1 places the unscrewed test tube No. 1 in the electronic scale working compartment, and the electronic scale working compartment is reset to zero;

[0153] (2.3.1.2) Robot arm No. 1 grabs the ecoflex 0050A material bottle and pours 13 g into test tube No. 1;

[0154] (2.3.1.3) Robot arm No. 1 grabs the ecoflex 0050B material bottle and pours 15 g into test tube No. 1;

[0155] (2.3.1.4) Robot arm No. 1 grabs the yellow material bottle and pours 1g into test tube No. 1;

[0156] (2.3.1.5) Robot arm No. 1 grabs test tube No. 1 and screws the cap on the test tube in the cap screwing chamber;

[0157] (2.3.1.6) Robot arm No. 1 transports the capped test tube No. 1 to the mixing chamber;

[0158] (2.3.1.7) Stir the working chamber at a speed of πrad / s for 5 minutes;

[0159] (2.3.1.8) Robot arm No. 1 unscrews the cap of the stirred test tube No. 1 in the capping chamber, and then pours 20g of the material in test tube No. 1 into mold No. 1;

[0160] (2.3.1.9) Robot arm No. 1 places test tube No. 1 on the dirty test tube rack. After cleaning, robot arm No. 1 places test tube No. 1 on the clean test tube rack;

[0161] (2.3.2) Preparation of green voxel module

[0162] (2.3.2.1) Robot arm No. 1 grabs test tube No. 2 from a clean test tube rack and unscrews the cap of test tube No. 2 in the cap-unscrewing working chamber. Robot arm No. 1 places the uncapped test tube No. 2 in the electronic scale working chamber, and the electronic scale working chamber is reset to zero;

[0163] (2.3.2.2) Robot arm No. 1 grabs the ecoflex 0050A material bottle and pours 15 g into test tube No. 2;

[0164] (2.3.2.3) Robot arm No. 1 grabs the ecoflex 0050B material bottle and pours 15 g into test tube No. 2;

[0165] (2.3.2.4) Robot arm No. 1 grabs the green material bottle and pours 1g into test tube No. 2;

[0166] (2.3.2.5) Robot arm No. 1 grabs test tube No. 2 and screws the cap on the test tube in the cap screwing chamber;

[0167] (2.3.2.6) Robot arm No. 1 transports the capped test tube No. 2 to the mixing chamber;

[0168] (2.3.2.7) Stir the working chamber at a speed of πrad / s for 5 minutes;

[0169] (2.3.2.8) Robot arm No. 1 unscrews the cap of test tube No. 2 after stirring in the capping chamber, and then pours 20g of the material in test tube No. 2 into mold No. 2;

[0170] (2.3.2.9) Robot arm 1 places test tube 2 on the dirty test tube rack. After cleaning, robot arm 1 places test tube 2 on the clean test tube rack.

[0171] (2.3.3) Preparation of red voxel module

[0172] (2.3.3.1) Robot arm No. 1 grabs test tube No. 3 from a clean test tube rack and unscrews the cap of the test tube in the cap-unscrewing working chamber. Robot arm No. 1 places the uncapped test tube No. 3 in the electronic scale working chamber, and the electronic scale working chamber is reset to zero;

[0173] (2.3.3.2) Robot arm No. 1 grabs the ecoflex 0050A material bottle and pours 15 g into test tube No. 3;

[0174] (2.3.3.3) Robot arm No. 1 grabs the ecoflex 0050B material bottle and pours 15 g into test tube No. 3;

[0175] (2.3.3.4) Robot arm No. 1 grabs the red pigment material bottle and pours 1g into test tube No. 3;

[0176] (2.3.3.5) Robot arm No. 1 grabs test tube No. 3 and screws the cap on the test tube in the cap screwing chamber;

[0177] (2.3.3.6) Robot arm No. 1 transports the capped test tube No. 3 to the mixing chamber;

[0178] (2.3.3.7) Stir the working chamber at a speed of πrad / s for 5 minutes;

[0179] (2.3.3.8) Robot arm No. 1 unscrews the cap of test tube No. 3 after stirring in the capping chamber, and then pours 20g of the material in test tube No. 3 into mold No. 3;

[0180] (2.3.3.9) Robot arm 1 places test tube 3 on the dirty test tube rack. After cleaning, robot arm 1 places test tube 3 on the clean test tube rack.

[0181] (2.3.4) Preparation of blue voxel module

[0182] (2.3.4.1) Robot arm No. 1 grabs test tube No. 4 from a clean test tube rack and unscrews the cap of the test tube in the cap-unscrewing working chamber. Robot arm No. 1 places the uncapped test tube No. 4 in the electronic scale working chamber, and the electronic scale working chamber is reset to zero;

[0183] (2.3.4.2) Robot arm No. 1 grabs the ecoflex 0050A material bottle and pours 15 g into test tube No. 4;

[0184] (2.3.4.3) Robot arm No. 1 grabs the ecoflex 0050B material bottle and pours 15 g into test tube No. 4;

[0185] (2.3.4.4) Robot arm No. 1 grabs the blue material bottle and pours 1g into test tube No. 4;

[0186] (2.3.4.5) Robot arm No. 1 grabs test tube No. 4 and screws the cap on the test tube in the cap screwing chamber;

[0187] (2.3.4.6) Robot arm No. 1 transports the capped test tube No. 4 to the mixing chamber;

[0188] (2.3.4.7) Stir the working chamber at a speed of πrad / s for 5 minutes;

[0189] (2.3.4.8) Robot arm No. 1 unscrews the cap of test tube No. 4 after stirring in the capping chamber, and then pours 20g of the material in test tube No. 4 into mold No. 4;

[0190] (2.3.4.9) Robot arm No. 1 places test tube No. 4 on the dirty test tube rack. After cleaning, robot arm No. 1 places test tube No. 4 on the clean test tube rack.

[0191] (2.3.5) 20 minutes later, the mobile robot 310 transports mold No. 1, mold No. 2, mold No. 3, and mold No. 4 to the demolding work chamber of the robot molding-integrated operation platform 360;

[0192] (2.3.6) The mobile robot 310 selects a 2 mm thin tube from the No. 2 sample rack work chamber of the robot hardware operating platform 350 and places it in the assembly work chamber of the robot molding-integrated operating platform 360;

[0193] (2.3.7) Robotic molding - The demoulding work chamber of the integrated operating platform 360 demoulds the voxel module in the mold, and the robot arm No. 3 transports the voxel module to the dispensing work chamber;

[0194] (2.3.8) The dispensing work chamber of the robot molding-integrated operating platform 360 dispenses glue on the voxel module according to the structural parameters of the voxel robot (4, 2, 4; 20, 42, 42, 20, 20, 42, 42, 20, 41, 32, 32, 41, 41, 32, 32, 41, 10, 0, 0, 10, 10, 0, 0, 10, 31, 0, 0, 41, 31, 0, 0, 41);

[0195] (2.3.9) Robot arm No. 3 transports the voxel module after dispensing to the assembly work chamber, and the assembly work chamber assembles the voxel module and the 2 mm capillary into a voxel robot;

[0196] (2.3.10) The mobile robot 310 transports the voxel robot to the multi-task operation platform of the robot test operation platform 370;

[0197] (2.3.11) Robot arm No. 4 inserts the thin tube of the voxel robot into the pump tube;

[0198] (2.3.12) The software loading end of the robot test operation platform 370 is connected to the voxel robot, and the parameters 0, 0.2. (control mode and pressure) are input to test the working performance of the voxel robot;

[0199] (2.3.13) The No. 4 visual workstation of the robot test operation platform 370 outputs a test report to the interactive unit 100 and the information workstation 200;

[0200] (2.3.14) If the interactive unit 100 evaluates that the performance report meets the decision maker's needs, the operation is terminated; if not, the interactive unit 100 sends the parameters evolved in the previous round to the active learning unit 230 to generate a new robot manufacturing strategy; it is worth noting that after 10 cycles of non-compliance, the interactive unit 100 will resend the decision maker's needs to the performance evolution learning unit 220.

[0201] (3) The information workstation 200 sends the generated voxel robot's morphological model and structural parameters to the first large model of the 3D data generation platform 330, generates the 3D structural data of the mold required for the voxel robot preparation, and then sends it to the 3D printing platform 320. After the 3D printing platform 320 completes the processing, it sends a message to the mobile robot 310, and the mobile robot 310 sends the printed mold to the robot software operating platform 340; the information workstation 200 sends the generated test plan to the second large model of the 3D data generation platform 330, generates the 3D structural data of the test egg (the same as the bionic egg model used in the simulation environment), and then sends it to the 3D printing platform 320. After the 3D printing platform 320 completes the processing, it sends a message to the mobile robot 310, and the mobile robot 310 transports the printed test egg to the robot test operating platform 370.

[0202] (4) The information workstation 200 sends the prepared materials of the voxel robot to the mobile robot 310. The mobile robot 310 transports the prepared materials from the material warehouse to the robot software operating platform 340. The information workstation 200 sends detailed operation information such as material ratio, preparation steps and time to the robot software operating platform 340. The robot software operating platform 340 uses the No. 1 visual work warehouse to detect that the prepared materials are complete and then starts to prepare material test samples. After the robot software operating platform 340 completes the preparation of the test samples, it sends a message to the mobile robot 310. After the mobile robot 310 transports the prepared material test samples to the robot test operating platform 370, it sends a message to the robot test operating platform 370. After the No. 4 visual work warehouse of the robot test operating platform 370 detects the material test samples, it starts to test the Young's modulus, soft and hardness values ​​of the materials, etc., and generates a material performance report test and sends it to the interactive unit 100.

[0203] After the interactive unit 100 determines the material properties, the robot software operating platform 340 will start preparing the voxel module. After the robot software operating platform 340 completes the preparation of the voxel module, it sends a message to the mobile robot 310, and the mobile robot 310 transports the prepared voxel module to the robot molding-integration operating platform 360. At the same time, if the process involves hardware part testing, it can be carried out simultaneously with the soft material testing, and the process and processing method of passing and failing the test are the same as those of the soft material testing.

[0204] (6) The information workstation 200 sends the size / model of the hardware hose used to the mobile robot 310 , and the mobile robot 310 transports the hardware hose to the robot molding-integrated operating platform 360 .

[0205] (7) The information workstation 200 sends the pump model / quantity / filling / filling time, and the voxel robot molding-integration plan (voxel robot integration sequence) to the robot molding-integration operation platform 360. After the No. 3 visual workstation of the robot molding-integration operation platform 360 detects that the materials are ready, the voxel robot molding-integration begins. The No. 3 robot arm and the dispensing workstation first assemble the voxel robot according to the voxel robot integration sequence; the No. 3 robot arm integrates one side of the hard hose on the voxel robot according to the position, and the No. 3 robot arm installs the other side of the hard hose on the pump. After completing the integration operation, the robot molding-integration operation platform 360 sends a signal to the mobile robot 310, and the mobile robot 310 transports the voxel robot to the robot test operation platform 370.

[0206] (8) The information workstation 200 sends the test plan (using a voxel robot to grab eggs) to the robot test operation platform 370. After the No. 4 visual workstation of the robot test operation platform 370 detects the voxel robot, it starts to control the filling size / filling time of the pump to allow the voxel robot to grab the bionic egg for the test. Then the robot test operation platform 370 sends the performance report to the interactive unit 100.

[0207] (9) The interactive unit 100 evaluates that the performance report meets the needs of the decision maker and the manufacturing task is completed.

[0208] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0209] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A system for automatically manufacturing robots by robots, characterized in that: include: An interaction unit, used to receive an input instruction from a decision maker, and to guide the decision maker to provide a complete instruction when the input instruction is incomplete, and then generate an initial manufacturing instruction, wherein the initial manufacturing instruction is relevant information of the robot to be manufactured described in a natural language, including the type of robot, the performance requirements of the robot, and the target task to be completed by the robot; An information workstation, for performing robot manufacturing task reasoning based on a plurality of learning modes for the initial manufacturing instructions issued by the interaction unit, forming a robot manufacturing strategy, wherein the robot manufacturing strategy includes a robot manufacturing plan and a robot morphological model and structural parameters; and for adjusting the robot manufacturing strategy according to the manufacturing log of the manufactured robot until the manufactured robot meets the performance test requirements, thereby terminating the current robot manufacturing task; The manufacturing and testing workstation is used to automatically carry out material transportation, parts preparation and assembly, and robot performance testing according to the robot manufacturing strategy generated by the information workstation, generate a manufacturing log and feed it back to the information workstation and the interaction unit.

2. The system according to claim 1, characterized in that The information workstation comprises: A priori knowledge learning unit, which is used to extract prior knowledge about the entire manufacturing process of the robot from existing technologies including articles, web pages, and patents through deep learning methods and regularly update the prior knowledge; A performance evolution learning unit is used to build a simulation environment according to the robot performance requirements in the initial manufacturing instructions, and to achieve the evolution of the robot in the simulation environment by changing the relevant parameters of the robot within a set range and using reinforcement learning training to improve the performance of the robot for the initial robot model, and obtain the evolved robot parameters, wherein the evolved robot parameters include the robot's structural parameters, material parameters, control parameters and morphological model; each type of robot has a set parameter range; An active learning unit, configured to generate a robot manufacturing strategy based on deep learning according to the evolved robot parameters output by the performance evolution learning unit and in combination with the prior knowledge output by the prior knowledge learning unit, wherein the robot manufacturing plan in the robot manufacturing strategy includes a software preparation plan, a hardware preparation and operation plan, a molding-integration operation plan and a test plan, and the test plan includes a material performance test plan and a work performance test plan; If the manufacturing log of the robot manufactured in the previous round shows that the robot performance does not meet the standard, the active learning unit fine-tunes the robot manufacturing strategy according to the manufacturing log of the robot manufactured in the previous round to generate a new robot manufacturing strategy; if the number of times the active learning unit fine-tunes the robot manufacturing strategy exceeds a set threshold and the robot performance still does not meet the standard, the information workstation re-reasons the robot manufacturing task.

3. The system according to claim 2, characterized in that The prior knowledge learning unit comprises a retrieval module and a mining module; the retrieval module first identifies keywords from the initial manufacturing instructions through natural language processing technology, and then uses the keywords to retrieve relevant information from the prior art; the mining module conducts in-depth analysis of the retrieved relevant information based on a large language model to extract the robot preparation method, morphological structure and performance data information related to the initial manufacturing instructions.

4. The system according to claim 2, characterized in that The simulation environment constructed by the performance evolution learning unit includes a training task scenario corresponding to the robot's execution of a target task and a verification task scenario corresponding to a performance test of the robot.

5. The system according to claim 2, characterized in that For each type of robot to be manufactured, a sample set consisting of multiple groups of successful manufacturing strategies is used to train the large model to obtain the corresponding active learning unit.

6. The system according to claim 2, characterized in that The manufacturing and testing workstation includes a control platform and a 3D printing platform, a 3D data generation platform, a robot software operating platform, a robot hardware operating platform, a robot molding-integration operating platform, a robot testing operating platform and a mobile robot connected thereto; The control platform is used to send the robot manufacturing strategy output by the information workstation to the corresponding platform, monitor its task execution status, generate control instructions for the mobile robot according to the task execution status, generate a manufacturing log according to the task execution results fed back by each platform and the mobile robot, and feed it back to the information workstation and the interaction unit; The mobile robot, in response to the control instructions issued by the control platform, includes taking corresponding materials from the material warehouse and transporting them to the corresponding operation platform according to the current robot manufacturing process, and interacting with each operation platform and the 3D printing platform during the robot manufacturing process to achieve object transportation and auxiliary operation tasks; The 3D data generation platform has a first large model and a second large model therein; the first large model generates three-dimensional structural data of molds, robot hardware and connectors that meet the accuracy and functional requirements based on the robot morphology model and structural parameters in the current robot manufacturing strategy, and sends the data to the 3D printing platform; the second large model generates three-dimensional structural data of test supplies required for performance testing of the manufactured robot entity according to the test plan in the current robot manufacturing strategy, and sends the data to the 3D printing platform; the three-dimensional structural data generated by the first large model has a priority printing level compared to the three-dimensional structural data generated by the second large model; The 3D printing platform is used to perform 3D printing according to various three-dimensional structure data generated by the 3D data generation platform; The robot software operating platform is used to complete the processing of robot software test samples and software parts according to the software preparation plan in the current robot manufacturing strategy and with the assistance of the mobile robot; The robot hardware operation platform is used to complete the processing of robot hardware test samples and hardware parts according to the hardware preparation and operation plan in the current robot manufacturing strategy and with the assistance of the mobile robot; The robot forming-integration operation platform is used to complete the forming and integration of robot software and hardware parts according to the forming-integration operation scheme in the current robot manufacturing strategy and with the assistance of the mobile robot; The robot testing operation platform is used to complete the material performance test of the software test samples processed by the robot software operating platform for each type of robot software parts according to the test plan in the current robot manufacturing strategy and with the assistance of the mobile robot, to test the material performance of the hardware test samples processed by the robot hardware operating platform for each type of robot hardware parts, to perform working performance tests on the robot entities obtained by the robot forming-integration operation platform, and to generate a performance test report.

7. The system according to claim 6, characterized in that The robot software operating platform includes a No. 1 robotic arm, a No. 1 visual work chamber, a No. 1 sample rack work chamber, a liquid distribution work chamber, a solid distribution work chamber, a stirring work chamber, a vacuum drying work chamber, an electronic scale work chamber, a spin coating work chamber and a capping work chamber, wherein the No. 1 sample rack work chamber includes test tubes, beakers, culture dishes and molds printed by the 3D printing platform.

8. The system according to claim 6, characterized in that The robot hardware operating platform includes a No. 2 robotic arm, a No. 2 visual work chamber, a No. 2 sample rack work chamber, a circuit construction work chamber, an electrode processing work chamber and a screw assembly work chamber, wherein the No. 2 sample rack work chamber includes various hardware accessories, electrical materials and nano silver wires required for making robot hardware.

9. The system according to claim 6, characterized in that The robot molding-integrated operation platform includes a No. 3 robotic arm, a No. 3 visual work chamber, a No. 3 sample rack work chamber, a demolding work chamber, a laser cutting work chamber, a hydraulic molding work chamber, a dispensing work chamber and an assembly work chamber, wherein the No. 3 sample rack work chamber includes a culture dish, a molding mold and connectors printed by the 3D printing platform.

10. The system according to claim 6, characterized in that The robot test operation platform includes a No. 4 robot arm, a No. 4 visual workstation, a power supply workstation, a multi-channel voltage workstation, an arbitrary waveform generator, an LCR digital bridge, a Shore hardness measurement workstation, an electronic universal testing machine, a conductivity test workstation, a multi-task operation platform and a software loading end that cooperate with each other; wherein, the soft material performance test is completed by the Shore hardness measurement workstation, the electronic universal testing machine and the conductivity test workstation; the hard material performance test is completed by the No. 4 visual workstation, the power supply workstation, the multi-channel voltage workstation, the arbitrary waveform generator and the LCR digital bridge; the robot's working performance test is firstly performed by the software loading end to load the software required to execute the target task to the robot entity, and then the No. 4 visual workstation, the power supply workstation, the multi-channel voltage workstation, the arbitrary waveform generator, the LCR digital bridge and the multi-task operation platform complete the specific working performance project test.

Images