Production control method and control system based on humanoid robot

Abstract

The invention discloses a production control method and system based on a humanoid robot, and belongs to the technical field of robot control, and the method comprises the steps: obtaining a task completion score of the humanoid robot; according to the task completion score, distributing and screening the sub-tasks to obtain a task execution queue; predicting the task queue according to a first prediction model to obtain a first score; selecting a task execution queue by maximizing the first score; and obtaining a production control strategy according to the task execution queue, and performing production control on the humanoid robot. Completing evaluation according to historical tasks, and performing primary matching on the stations and the subtasks; the task queue is scored through the first prediction model, the task queue is subjected to secondary screening through the first score, it is ensured that each humanoid robot is in the execution condition of a single subtask, the execution conditions of all humanoid robots in the queue are in the optimal state, and task distribution and execution are planned as a whole.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot control, in particular to a production control method and control system based on humanoid robots. BACKGROUND

[0002] With the mechanization and automation of industrial production, industrial robots are widely used in production lines of various products due to their multi-joint and multi-degree-of-freedom mechanical hands, such as grabbing, rotating and placing products through the mechanical hands of industrial robots.

[0003] However, the current industrial robots are usually fixed on the production line and operate on the products after positioning by the positioning mechanism on the production line. The current industrial robots can only be used on one production line after installation and are difficult to be reused on other production lines, resulting in low utilization rate of industrial robots and high configuration cost of production lines. In particular, for products with seasonal or periodic production, the production line may have a certain stop state, and the assembled industrial robots cannot be installed and used on other production lines.

[0004] CN119036462A discloses a production control adjustment method based on multi-control mode switching of industrial robots, which evaluates the effectiveness of mode switching by analyzing robot operation data and evaluates before and after mode switching of industrial robots, but the work capacity of industrial robots is affected by the mode, and multiple work capacities cannot be realized.

[0005] A humanoid robot is a humanoid robot similar to human form, usually having a torso, two arms and legs, and usually configured with a visual detection module, having the ability to imitate human behavior to a certain extent, and can perform certain operation tasks and realize multiple work capacities. A production line is usually configured with one or more stations, and each station can be configured with at least one humanoid robot, so the tasks performed by the humanoid robots need to be allocated and controlled.

[0006] CN120038760A discloses a logistics robot control method and system based on artificial intelligence, which intelligently selects an experience area according to the current logistics state to allocate tasks and optimize control strategies, and completely converts the optimal experience area into a collective knowledge base, so that different logistics robots can share the experience of success and failure, forming a stronger collaborative work capacity of logistics robots. The work methods of industrial robots and logistics robots are different, and the task allocation method cannot be applied to industrial production.

[0007] Therefore, it is necessary to design a production control method and control system based on humanoid robots to realize the overall allocation of multiple humanoid robots. SUMMARY

[0008] To address the aforementioned technical problems in the prior art, this invention provides a production control method and control system based on humanoid robots. The method evaluates the humanoid robot and the tasks it performs based on historical data, and assigns tasks according to the evaluation; and controls the humanoid robot.

[0009] This invention discloses a production control method based on a humanoid robot, comprising the following steps: collecting historical data of the humanoid robot; obtaining a task completion score for the humanoid robot based on the historical data; allocating and filtering workstations and subtasks of production tasks according to the task completion score to obtain at least one task execution queue; predicting the task queues according to a first prediction model to obtain a first score; selecting a task execution queue by maximizing the first score; obtaining a production control strategy based on the task execution queues; and performing production control on the humanoid robot according to the production control strategy.

[0010] Preferred methods for obtaining task completion scores include: Historical data is cleaned and normalized, and a dataset is constructed. Select the first modeling indicator from the dataset, and build a scoring model by weighting the first modeling indicator; Based on the scoring model, the dataset is filtered and evaluated to obtain a task completion score.

[0011] Preferably, the first modeling metric is selected from: health status, task completion status, task completion rate, power consumption, network health status, and yield rate; The scoring model is expressed as follows: ; in, A Represented as weighting coefficients, S1mt This refers to humanoid robots executing strategies. m Execute the task t The first rating, Q Represented as the normalized value of the first modeling metric. i This is represented by the index of the first modeling indicator. n This represents the total number of the first modeling indicator.

[0012] Preferably, methods for obtaining the task execution queue include: Obtain the subtasks of the task and their corresponding workstations; Based on the first score and the type of subtask, the multiple humanoid robots are sorted in descending order; Take the first N humanoid robots and combine the execution status of their subtasks to obtain a task execution queue, where N is a natural number greater than 1.

[0013] Preferably, the training method of the first prediction model comprises: cleaning the historical data and normalizing the index values to obtain a training set and a validation set; calculating the similarity between the indexes; constructing a minimum connected graph with the similarity as the edge and the index as the vertex; training the minimum connected graph based on the graph neural network GNN to obtain the first prediction model.

[0014] Preferably, the production control strategy comprises a plurality of sub-strategies, and the humanoid robot completes the sub-tasks through the sub-strategies; The sub-strategies include a visual detection sub-strategy, a sorting sub-strategy, a carrying sub-strategy, a product online sub-strategy, and a product offline sub-strategy.

[0015] Preferably, the product online and offline detection method comprises: issuing a navigation path of the station to the humanoid robot; controlling the humanoid robot to enter the corresponding station according to the navigation path; issuing an online sub-strategy to the humanoid robot, the online sub-strategy comprising a product template, a product source template, and a product detection position template; scanning the surrounding environment through the visual unit of the humanoid robot; matching the product source through the product source template; matching the product to be detected through the product template; and matching the target detection position through the product detection position template; clamping the product from the product source and placing the product on the detection position through the arm and the end effector of the humanoid robot; detecting the product through the external detection tool; and collecting the first detection result and the detection process through the visual unit of the humanoid robot; issuing a product offline sub-strategy to the humanoid robot, the product offline sub-strategy comprising a product placement template; and matching the product placement target position through the product placement template; after the detection is completed, clamping the product through the arm and the end effector of the humanoid robot and placing the product on the product placement target position.

[0016] Preferably, the humanoid robot communicates with the external detection tool through the server, and the communication method comprises: collecting a first image of the detection position through the visual unit of the humanoid robot; matching the first image with a first template of placing the product on the detection position to obtain the state of the detection position; if the state of the detection position is that the product has been correctly placed, issuing a detection instruction to the external detection tool through the server; Receive the second detection result of the external detection tool, and combine the first detection result to make a detection report.

[0017] Preferably, a socket is arranged on one side of the station; a corresponding plug is arranged on the humanoid robot, and the method for connecting the plug to the socket comprises the following steps: After the humanoid robot reaches the station, a second image is obtained by scanning through a vision unit; The second image is matched with a station template to obtain the coordinates of the socket; The plug is gripped by the arm and end effector of the humanoid robot and inserted into the socket corresponding to the coordinates, so that the humanoid robot is provided with electric energy and network connection through the socket and the plug.

[0018] The application also provides a control system for implementing the production control method, comprising a first evaluation module, a queue construction module, a second evaluation module and a control module; The first evaluation module is used for obtaining a task completion score of the humanoid robot according to historical data; The queue construction module is used for distributing and screening the stations and sub-tasks of the production task according to the task completion score to obtain at least one task execution queue; The second evaluation module is used for predicting the task queue according to a first prediction model to obtain a first score; and selecting the task execution queue by maximizing the first score; The control module is used for obtaining a production control strategy according to the task execution queue; and controlling the humanoid robot according to the production control strategy.

[0019] Compared with the prior art, the application has the beneficial effects that: the stations and sub-tasks are matched once according to the historical task completion evaluation; the task queue is scored by the first prediction model, and the task queue is screened twice by the first score, so that the execution of each humanoid robot in a single sub-task and the execution of all humanoid robots in the queue are in the best state, and the task distribution and execution are overall planned. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a flowchart of the production control method based on the humanoid robot of embodiment 1; Figure 2 is a flowchart of the method for obtaining the execution queue; Figure 3 is a flowchart of the training method of the first prediction model; Figure 4 is an ROC curve diagram of the first prediction model; Figure 5 is a flowchart of the execution method of the product online. Figure 6 is a control system logic block diagram of Example 3; Figure 7 is a structural schematic diagram of a humanoid robot. DETAILED DESCRIPTION

[0021] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0022] The present application will be further described in detail below with reference to the drawings: Example 1 provides a production control method based on a humanoid robot, as shown in Figure 1 , comprising the following steps: Step 101: Collecting historical data of the humanoid robot.

[0023] The indicators of the historical data are selected from the humanoid robot ID, the humanoid robot type, the execution strategy, the humanoid robot priority, the health degree, the task ID, the task type, the task completion degree, the task completion rate, the power consumption, the network health degree, the station ID, the good product rate, etc.

[0024] Step 102: Obtaining a task completion score of the humanoid robot according to the historical data.

[0025] Step 103: Distributing and screening the stations and sub-tasks of the production task according to the task completion score, to obtain at least one task execution queue.

[0026] Step 104: Predicting the task queue according to a first prediction model, to obtain a first score.

[0027] Step 105: Selecting the task execution queue by maximizing the first score.

[0028] Step 106: Obtaining a production control strategy according to the task execution queue.

[0029] Step 107: Controlling the humanoid robot for production according to the production control strategy.

[0030] According to the historical task completion evaluation, the station and the subtask are matched once; the task queue is scored through the first prediction model, and the task queue is screened twice through the first score, so that each humanoid robot is in the execution of a single subtask, and the execution of all humanoid robots in the queue is in the best state, and some humanoid robots are avoided from being assigned to the subtask with poor execution.

[0031] As Figure 2 , the method for obtaining the task completion score in step 102 includes: Step 201: cleaning and normalizing the historical data, and constructing a data set.

[0032] The cleaning includes removing outliers, deleting missing data, etc., and the specific cleaning method is the prior art, which will not be repeated herein. The normalization can adopt Min-Max normalization or Z-Score normalization.

[0033] Step 202: selecting a first modeling index from the data set, and establishing a scoring model through the weighted sum of the first modeling index.

[0034] In one specific embodiment, the first modeling index is selected from: health degree, task completion degree, task completion rate, power consumption, network health degree, and yield.

[0035] The scoring model is represented as: ; Wherein, A is a weight coefficient, which can be an empirical value or can be calculated based on the analytic hierarchy process or entropy weight method, S1 is the first score of the humanoid robot in executing the task through the execution strategy, m is the normalized value of the first modeling index, t is the serial number of the first modeling index, Q is the total number of the first modeling index. The method for calculating the normalized value of the first modeling index is the prior art, which will not be repeated herein. i n Step 203: screening and evaluating the data set according to the scoring model to obtain the task completion score.

[0036] For the task , the task completion score of each humanoid robot under each execution strategy can be calculated, so that the best humanoid robot and execution strategy for a single task are selected.

[0037] t The method for obtaining the task execution queue in step 103 includes:

[0038] Step 103: screening and evaluating the data set according to the scoring model to obtain the task completion score. ​Step 301: Obtain the sub-tasks of the task and the corresponding stations.

[0039] Step 302: Sort the humanoid robots in descending order according to the task completion score and the type of sub-tasks.

[0040] Step 303: Take the top N humanoid robots and combine the execution of the sub-tasks to obtain a task execution queue. Wherein N represents a natural number greater than 1.

[0041] For example, the task has sub-tasks C1, C2 and C3, humanoid robots D1, D2, D3, the top 2 humanoid robots for C1 task completion score are D1, D2, the top 2 humanoid robots for C2 are D2, D3, and the top 2 humanoid robots for C3 are D1 and D3, then the task execution queue exhaustive combination can be: D1, D2, D3; D2, D3, D1.

[0042] In step 104, the first prediction model can be constructed based on machine learning methods, and machine learning methods such as random forest, vector machine, neural network, and graph neural network can be used for training; wherein the classification algorithm such as random forest can train multiple classifications, and each classification corresponds to a score.

[0043] In one specific embodiment, a graph neural network is used to construct the first prediction model. The specific indicators include: humanoid robot ID, humanoid robot type, execution strategy, health degree, task type, task completion degree, task completion rate, power consumption, network health degree, station ID, and yield rate. Among them, the humanoid robot type, execution strategy, humanoid robot priority, and task type are mapped to digital ID; the label is the first score, and in specific implementation, a 5-point system is used, but not limited to this. For example, if there are 5 types of humanoid robots, they are encoded as 1-5 in turn. For example Figure 3 The training method of the first prediction model includes the following steps: Step 401: Clean the historical data and standardize the values of each indicator to obtain a training set and a validation set.

[0044] In specific training, after standardization, there are 1503 rows of data in total.

[0045] Step 402: Calculate the similarity between indicators.

[0046] The Euclidean distance or cosine similarity method can be used to calculate the similarity between two indicators. In specific embodiments, the Euclidean distance is used.

[0047] Step 403: Construct a minimum connected graph with maximum similarity with similarity as edge and indicator as vertex.

[0048] Step 404: training the minimum connected graph based on the graph neural network GNN to obtain a first prediction model.

[0049] Step 405: verifying the first prediction model by using a verification set.

[0050] Figure 4 The verification ROC curve of the first prediction model is shown, wherein the area under the curve AUC of the first curve is 0.86; the area under the curve of the second curve is 0.77; the area under the curve of the third curve is 0.66; the area under the curve of the fourth curve is 0.71; the area under the curve of the fifth curve is 0.76; the area under the curve of the macro average ROC curve is 0.77; the area under the curve of the micro average ROC curve is 0.75, which can meet the industrial production requirements. Wherein, the Cross-Entropy formula is used as the loss function.

[0051] By graphing the indicators and constructing the minimum connected graph, the relationship between the indicators can be intuitively reflected and observed. The first prediction model has good scoring / classification effect and good robustness.

[0052] In step 106, the production control strategy includes a plurality of sub-strategies, which are specific sub-strategies for the humanoid robot to perform sub-tasks at the workstations, such as visual inspection sub-strategy, sorting sub-strategy, carrying / transporting sub-strategy, product online sub-strategy, product offline sub-strategy, etc.

[0053] Embodiment 2 provides specific control sub-strategies. As shown in Figure 5 , the method for visual inspection includes the following steps: Step 501: issuing a navigation path of the workstation to the humanoid robot.

[0054] Step 502: controlling the humanoid robot to enter the corresponding workstation according to the navigation path.

[0055] The position can be detected by visual inspection or feedback from the humanoid robot.

[0056] Step 503: issuing an online sub-strategy to the humanoid robot, wherein the online sub-strategy includes a product template, a product source template, and a product detection position template.

[0057] Step 504: scanning the surrounding environment by the visual unit of the humanoid robot; matching the detection box, detection transmission line, etc. of the product source through the product source template; matching the product to be detected through the product template; and matching the target detection position through the product detection position template.

[0058] Step 505: picking up the product from the product source by the arm and the end effector of the humanoid robot, and placing the product on the detection position.

[0059] Step 506: detecting the product by the external detection tool, and collecting the first detection result and the detection process by the vision unit of the humanoid robot. The external detection tool is used to detect the performance of the product, and can be selected according to the product, such as torque, weight, pressure, rated current, rated voltage, resistance, insulation, etc.

[0060] Step 507: issuing a product offline sub-strategy to the humanoid robot, wherein the product offline sub-strategy includes a product template and a product placement template; and matching the product placement target position by the product placement template.

[0061] Step 508: after the detection is completed, picking up the product by the arm and the end effector of the humanoid robot, and placing the product to the product placement target position.

[0062] The humanoid robot performs the detection sub-tasks of the product online detection and offline detection.

[0063] In step 505, the humanoid robot can communicate with the external detection tool through the server: Step 601: collecting a first image of the detection position by the vision unit of the humanoid robot.

[0064] Step 602: matching the first image with a first template of placing the product to the detection position, and obtaining the state of the detection position.

[0065] Step 603: if the state of the detection position is that the product has been correctly placed, issuing a detection instruction to the external detection tool through the server.

[0066] Step 604: receiving a second detection result of the external detection tool, and combining the first detection result to make a detection report.

[0067] The humanoid robot is connected to the server in a wired or wireless manner. In a specific design scheme, a socket is arranged on one side of the workstation, and a corresponding plug is arranged on the humanoid robot.

[0068] After the humanoid robot reaches the workstation, the socket and the plug are scanned and positioned by the vision unit. Specifically, the first coordinates of the socket are marked in the workstation template, the scanned image is aligned with the workstation template, the first coordinates are converted into second coordinates in the scanned image, the plug is pushed to the second coordinates, and the insertion of the plug can be judged according to the feedback pressure of the socket.

[0069] The plug is inserted into the socket corresponding to the coordinates by the arm and end effector of the humanoid robot, and stable power and network connection are provided to the humanoid robot through the socket and the plug; and the humanoid robot can work stably for a long time.

[0070] The plug insertion action and other sub-strategy training can be performed by collecting training data through artificial VR and training the humanoid robot to obtain a corresponding plug insertion action model or strategy.

[0071] In one specific embodiment, the specific product is a caliper, and the method of clamping the product comprises the following steps: Step 701: sequentially arranging a detection point, a first operation point and a second operation point on the production line.

[0072] The detection point is arranged on the production line, and the first operation point and the second operation point are located above the production line, preferably 5-30 cm away from the conveying belt of the production line.

[0073] Step 702: directing the detection direction of the visual unit of the humanoid robot towards the detection point.

[0074] Step 703: moving the end effector to the first operation point.

[0075] Step 704: obtaining the moving speed of the target product.

[0076] The moving speed can be obtained by visual detection, or can be obtained by synchronously reading the moving speed of the conveying belt of the production line.

[0077] Step 705: determining whether the target product reaches the first operation point.

[0078] If yes, step 706 is performed: moving the wrist joint of the arm to the second operation point downstream at the moving speed; and clamping the target product by the end effector. The speed difference between the wrist joint and the end effector and the target product can be reduced, the clamping stability can be improved, and the collision between the end effector and the target product can be avoided.

[0079] If no, the target product is waited to reach the first operation point.

[0080] During the clamping process, the moving position of the wrist joint of the arm and the end effector mounted on the wrist joint can be constrained between the first operation point and the second operation point.

[0081] Step 707: If the position of the target product is at the second operation point and the corresponding clamping operation is not performed, abandon the clamping operation, save and upload the information of the target product and the execution information of the subtask. The missed target product can be clamped by other humanoid robots downstream; if the target product is missed in the entire production task, the number of execution failures of the subtask is counted. Adjust the corresponding execution strategy and execution mode according to the execution failure.

[0082] A plurality of vision units can be provided on the humanoid robot: a first vision unit for detecting products, which can be provided on the chest of the humanoid robot; a second vision unit for tracking target products and target positions, which can be provided on the head of the humanoid robot, and the target product located in the second vision unit is tracked by controlling the rotation of the head to be within a preset target frame.

[0083] Specifically, the method for tracking the target product comprises the following steps: Step 711: Acquire a second image of the target product by the second vision unit.

[0084] Step 712: Obtain a bounding box of the target product in the second image.

[0085] Step 713: Obtain a first center point of the bounding box and a second center point of a preset tracking frame.

[0086] Step 714: Determine whether the first value of the first center point on a coordinate axis is greater than the second value of the second center point on the coordinate axis.

[0087] More specifically, the following determination method can be used: whether the difference between the first value and the second value is greater than a first threshold value.

[0088] If yes, perform step 715: rotate the head in the direction in which the coordinate axis increases.

[0089] If no, perform step 716: rotate the head in the direction in which the coordinate axis decreases.

[0090] After the target product is clamped, the method for placing the target product comprises: Step 721: Acquire a third image in a preset area by the second vision unit.

[0091] Step 722: Obtain a template of the target area, match the template with the third image, and obtain a specified position and its placement coordinates.

[0092] Step 723: Move the wrist joint and the end effector above the placement coordinates.

[0093] Step 724: Place the clamped target product to the specified position by the end effector.

[0094] Detecting products by a humanoid robot, and putting the defective target products into a target position.

[0095] Taking automobile caliper assembly as an example, the parts of the automobile caliper include caliper body, piston, sealing ring and brake pad, and the execution method of assembly includes: Step 731: issuing the first part template, the second part template, the first part grabbing strategy, the second part grabbing strategy, the first assembly strategy, and the first assembly template to the humanoid robot.

[0096] Step 732: matching the first part and its coordinates according to the first part template, and controlling one arm of the humanoid robot and its end effector to grab the first part according to the first part grabbing strategy.

[0097] Step 733: matching the second part and its coordinates according to the second part template, and controlling the other arm of the humanoid robot and its end effector to grab the second part according to the second part grabbing strategy.

[0098] Step 734: controlling the two arms of the humanoid robot and their end effectors to assemble the first part and the second part according to the first assembly strategy.

[0099] Step 735: obtaining the sixth image after assembly by the vision unit.

[0100] Step 736: matching the sixth image with the first assembly template to obtain the assembly state.

[0101] Step 737: if the assembly state is good, assembling another part according to the third part template, the third part grabbing strategy, and the second assembly strategy, the second assembly template; until the product assembly is completed.

[0102] The above steps can realize the assembly of 3-8 parts, but are not limited thereto.

[0103] The product carrying method includes: Step 751: obtaining a plurality of product templates.

[0104] Step 752: collecting a fifth image by the vision unit of the humanoid robot.

[0105] Step 753: segmenting the segmentation image of the product in the fifth image.

[0106] Step 754: matching the segmentation image with the product template to obtain the product category corresponding to the segmentation image.

[0107] Step 755: obtaining the carrying strategy, the destination navigation path, the target position template and the offline strategy corresponding to the product category; and sending the carrying strategy and the destination to the humanoid robot.

[0108] A corresponding carrying strategy is preset for various types of products, and the humanoid robot carries the products through the carrying strategy. The carrying strategy includes the carrying method, such as when the weight of the product is 0.1-10KG, one or two end effectors cooperate to carry the product; when the weight of the product is 10-30KG, the end effector cooperates with the arm to carry the product; when the weight of the product is 30-50KG, the end effector, the arm and the chest cooperate to carry the product.

[0109] Step 756: According to the carrying strategy, the humanoid robot carries the product; and according to the destination navigation path, the humanoid robot reaches the destination.

[0110] Step 757: According to the target position template, the delivery target position is matched; according to the offline strategy, the humanoid robot places the product at the target position.

[0111] The delivery action of the offline strategy is trained in advance, such as by collecting artificial VR action data, training a corresponding delivery action model, and controlling the humanoid robot to place the product at the target position through the corresponding delivery action model. The training method is prior art, and the present application will not be repeated.

[0112] Embodiment 3 provides a control system for implementing the above production control method, as shown in Figure 6 The control system includes a first evaluation module 1, a queue construction module 2, a second evaluation module 3 and a control module 4. The first evaluation module 1 is used to obtain the task completion score of the humanoid robot according to the historical data; The queue construction module 2 is used to allocate and screen the stations and sub-tasks of the production task according to the task completion score, and obtain at least one task execution queue; The second evaluation module 3 is used to predict the task queue according to the first prediction model, and obtain a first score; and select the task execution queue by maximizing the first score; The control module 4 is used to obtain the production control strategy according to the task execution queue; and control the humanoid robot according to the production control strategy.

[0113] Among them, the first evaluation module 1, the queue construction module 2, the second evaluation module 3 and the control module 4 are deployed on the server, and the server communicates with the humanoid robot and the production equipment / external detection tool through the communication module.

[0114] Figure 7The structure of the specific humanoid robot 6 is shown, including a head 62, a chest 61, a trunk 64 and a leg 68 connected in sequence, a visual unit 63 is arranged on the head 62, arms 65 are arranged on both sides of the chest 61, and end effectors 67 are arranged at the wrist joints 66 at the ends of the arms 65. The humanoid robot is configured with a receiving module, a first execution module and a second execution module.

[0115] The receiving module is used to receive sub-tasks of a production line, receive navigation paths of corresponding workstations, and receive sub-strategies of the sub-tasks, and the above data can be received from a server in a wired or wireless manner. The first execution module is used to enter the corresponding workstations according to the navigation paths. The second execution module is used to execute the sub-tasks according to the sub-strategies.

[0116] The humanoid robot executes an industrial production method, including the following steps: Receiving sub-tasks of a production line, receiving navigation paths of workstations, entering corresponding workstations according to the navigation paths, receiving execution strategies of the sub-tasks, and executing the sub-tasks according to the execution strategies.

[0117] The present application applies humanoid robots to industrial production, utilizes the flexibility and usability of humanoid robots, reduces the comprehensive configuration cost of the production line, and is especially suitable for an industrial production method with multiple production lines and multiple workstations in the production line. Through a task allocation mode, corresponding humanoid robots are matched for corresponding workstations, and corresponding strategies are issued to the humanoid robots. The humanoid robots can adapt to the execution of multiple types of sub-tasks, can replace manual operation, can simulate manual operation mode, and can be used for dangerous operation tasks to avoid personnel injuries.

[0118] In specific embodiments, the humanoid robot is suitable for detection, clamping, sorting, carrying and the like of products or intermediate products, but the specific sub-tasks are not limited thereto.

[0119] It should be noted that the terms "first", "second" and the like in the specification are used to distinguish different messages, devices, modules and the like, and do not represent the order or limit the types of "first" and "second".

[0120] In the present disclosure, it should be understood that terms such as "including" or "having" are intended to indicate the presence of features, numbers, steps, actions, components, parts or combinations thereof disclosed in the specification, and do not exclude the possibility of existence or addition of one or more other features, numbers, steps, actions, components, parts or combinations thereof.

[0121] In addition, it should be further noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0122] The computer system can include a plurality of individual components or devices that are interconnected via a bus and communicate with one another. The system can include a processing unit, a system memory, and a system bus. The processing unit can be any of various processors or microprocessors as are known to those of ordinary skill in the art. The system memory can include read-only memory (ROM), random access memory (RAM), and the like, as are well known. The system bus can include any of various busses as are known to those of ordinary skill in the art, including a memory bus or an I / O bus, or both. The computer system can also include a variety of input / output devices, such as a user interface, a display, etc. The user interface can include any of various devices for allowing a user to interact with the computer system, such as an alphanumeric input device, a cursor control device, etc. The display can include any of various displays as are known to those of ordinary skill in the art, such as a cathode ray tube (CRT), a liquid crystal display (LCD), a plasma display, etc.

[0123] The above merely provides preferred embodiments of the present application and is not intended to limit the present application. The present application can be modified and changed in various ways by those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A production control method based on a humanoid robot, characterized by, The method comprises the following steps: obtaining a task completion score of a humanoid robot according to historical data; allocating and screening workstations and sub-tasks of a production task according to the task completion score, and obtaining at least one task execution queue; predicting the task queue according to a first prediction model, and obtaining a first score; selecting the task execution queue by maximizing the first score; obtaining a production control strategy according to the task execution queue; controlling the humanoid robot according to the production control strategy.

2. The production control method according to claim 1, characterized by, The method for obtaining the task completion score comprises: cleaning and normalizing the historical data, and constructing a data set; selecting a first modeling index from the data set, and establishing a scoring model through the weighted sum of the first modeling index; screening and evaluating the data set according to the scoring model, and obtaining the task completion score.

3. The production control method according to claim 2, characterized by, The first modeling index is selected from the following: health degree, task completion degree, task completion rate, power consumption, network health degree, and good product rate; The scoring model is expressed as: ； wherein, A is represented as a weight coefficient, S1mt is represented as a humanoid robot by executing a policy m executes a task t a first score, Q is represented as a normalized value of the first modeling index, i is represented as a serial number of the first modeling index, n is represented as a total number of the first modeling index.

4. The production control method according to claim 1, characterized by, The method for obtaining the task execution queue comprises: obtaining sub-tasks of the task and corresponding workstations; sorting multiple humanoid robots in descending order according to the task completion score and the type of the sub-tasks; taking the first N humanoid robots, combining the execution of the sub-tasks, and obtaining the task execution queue, wherein N represents a natural number greater than 1.

5. The production control method according to claim 1, characterized by, The training method of the first prediction model comprises: standardizing the index values of the historical data, and constructing a training set; calculating the similarity between the indexes; constructing a minimum connected graph with the maximum similarity by taking the similarity as the edge and the index as the vertex; training the minimum connected graph based on a graph neural network to obtain the first prediction model.

6. The production control method according to claim 1, characterized by, The production control strategy comprises multiple sub-strategies, which are used to control the humanoid robot to complete the sub-tasks; The sub-strategies include a visual detection sub-strategy, a sorting sub-strategy, a carrying sub-strategy, a product online sub-strategy, and a product offline sub-strategy.

7. The production control method according to claim 6, characterized by, The method for product visual detection comprises: issuing a navigation path of a workstation to a humanoid robot; controlling the humanoid robot to enter the corresponding workstation according to the navigation path; issuing an online sub-strategy to the humanoid robot, wherein the online sub-strategy comprises a product template, a product source template, and a product detection position template; scanning the surrounding environment by the visual unit of the humanoid robot, matching the product source by the product source template, matching the product to be detected by the product template, and matching the target detection position by the product detection position template; clipping the product from the product source by the arm and the end effector of the humanoid robot, and placing the product on the detection position; detecting the product by an external detection tool, and collecting the first detection result and the detection process by the visual unit of the humanoid robot; issuing a product offline sub-strategy to the humanoid robot, wherein the product offline sub-strategy comprises a product placement template; and matching the product placement target position by the product placement template; after the detection is completed, clipping the product by the arm and the end effector of the humanoid robot, and placing the product on the product placement target position.

8. The production control method according to claim 7, characterized by, The humanoid robot communicates with the external detection tool through a server, and the communication method comprises: A first image of the detection position is collected by a vision unit of the humanoid robot; The first image is matched with a first template of a preset product placed on the detection position to obtain a state of the detection position; If the state of the detection position is that the product has been correctly placed, a detection instruction is issued to an external detection tool through a server; A second detection result of the external detection tool is received, and a detection report is made in combination with a first detection result.

9. The production control method according to claim 1, characterized by, A socket is arranged on one side of the work station; A corresponding plug is arranged on the humanoid robot, and the plug is connected to the socket by the following method: After the humanoid robot reaches the work station, a second image is obtained by scanning through a vision unit; The second image is matched with a work station template to obtain coordinates of the socket; The plug is gripped by an arm and an end effector of the humanoid robot and is inserted into the socket corresponding to the coordinates, so that the socket and the plug provide electric energy and network connection for the humanoid robot.

10. A control system characterized by, The control system for implementing the production control method according to any one of claims 1-9 comprises a first evaluation module, a queue construction module, a second evaluation module and a control module; The first evaluation module is configured to obtain a task completion score of the humanoid robot according to historical data; The queue construction module is configured to distribute and screen work stations and sub-tasks of production tasks according to the task completion score to obtain at least one task execution queue; The second evaluation module is configured to predict the task queue according to a first prediction model to obtain a first score, and select the task execution queue by maximizing the first score; The control module is configured to obtain a production control strategy according to the task execution queue, and control the humanoid robot according to the production control strategy.

Images