Industrial robot programming teaching method and system

Abstract

The application discloses an industrial robot programming teaching method and system, and belongs to the technical field of programming teaching, and the method specifically comprises the following steps: collecting programming behavior data and industrial robot operation data in real time; generating learning state data of students according to the programming behavior data and the industrial robot operation data; the learning state data of the students comprises programming behavior, knowledge mastery degree and learning style of the students; combining a preset dynamic knowledge graph; determining a knowledge point node embedding offset according to the learning state data of the students; updating the current dynamic knowledge graph according to the knowledge point node embedding offset; determining a current knowledge difficulty point of the students based on the current dynamic knowledge graph; and pushing difficult point analysis data to a learning terminal of the students based on the current knowledge difficulty point. The application realizes precise knowledge feedback for individual differences, and effectively improves the pertinence and efficiency of industrial robot programming teaching.

Description

Technical Field

[0001] This invention belongs to the field of programming teaching technology, specifically a method and system for teaching programming in industrial robots. Background Technology

[0002] With the development of intelligent manufacturing and industrial automation, industrial robots are increasingly being used in welding, handling, assembly, painting, and other fields. To meet the urgent demand of enterprises for robot programming talent, various vocational education institutions, higher education institutions, and corporate training programs have all introduced robot programming courses.

[0003] Existing teaching methods for industrial robot programming mainly rely on centralized instruction, task-driven training, and simulation operations. These methods have the following shortcomings: a lack of real-time personalized feedback, making it difficult for teachers to promptly identify each student's specific programming difficulties; difficulty in accurately identifying cognitive obstacles; and low efficiency in utilizing teaching resources. Therefore, there is an urgent need for a new teaching method for industrial robot programming.

[0004] For example, Chinese patent CN116312147B discloses a programming teaching method and system based on a teaching robot. The method includes: an electronic device sending a programming file to the teaching robot; the teaching robot querying a set of picking strategies based on the target waste category of the target object to obtain an original picking strategy; executing the original picking strategy according to a reference gripping force; and, if the electronic device determines that the teaching robot has failed to pick up the target object, outputting a first manual operation interface to obtain the target gripping force when the teaching robot successfully picks up the target object. The method also involves determining the target gripping force range based on the target gripping force; the teaching robot obtaining the target gripping force range from the electronic device; updating the reference gripping force range in the original picking strategy based on the target gripping force range; and saving the updated original picking strategy in the picking strategy set. This method can enhance the success rate of the teaching robot in picking up waste and comprehensively improve the user's programming and hands-on abilities.

[0005] The drawbacks of the aforementioned patents are: lack of personalized feedback, difficulty in predicting students' current knowledge difficulties, and low teaching efficiency. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes a teaching method and system for industrial robot programming.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A teaching method for industrial robot programming, comprising:

[0009] Real-time collection of programming behavior data and industrial robot operation data;

[0010] Based on the programming behavior data and the industrial robot operation data, student learning status data is generated, which includes the student's programming behavior, knowledge mastery level and learning style.

[0011] Based on the student learning status data, the embedded offset of knowledge point nodes is determined by combining the preset dynamic knowledge graph. The embedded offset of knowledge point nodes is used to adjust the node association strength of the dynamic knowledge graph. The dynamic knowledge graph is based on the industrial robot programming knowledge graph and is constructed using a dynamic temporal graph neural network.

[0012] Update the current dynamic knowledge graph based on the embedding offset of the knowledge point node;

[0013] Based on the current dynamic knowledge graph, the student's current knowledge difficulties are determined, and difficulty analysis data is pushed to the student's learning terminal based on the current knowledge difficulties.

[0014] Specifically, generating student learning status data based on the programming behavior data and the industrial robot operation data includes:

[0015] The programming behavior data is arranged in chronological order and broken down into the smallest analyzable units to form structured tabular data;

[0016] The process verification method is used to identify logical errors. Based on the error pattern library, similar logical errors are aggregated to obtain logical error types, including: syntax error type, logical vulnerability type and device adaptation error type. The error pattern library is constructed based on the industrial robot knowledge base.

[0017] Programming tasks are broken down into knowledge units and mapped to an industrial robot programming knowledge graph to dynamically assess students’ knowledge mastery.

[0018] Students' behavioral characteristics are extracted, and their learning styles are analyzed based on these extracted characteristics.

[0019] Specifically, the step of determining the embedding offset of knowledge point nodes based on the preset dynamic knowledge graph and the student learning status data includes:

[0020] Based on the industrial robot knowledge base, construct an industrial robot programming knowledge graph;

[0021] Based on the knowledge graph of industrial robot programming, a dynamic knowledge graph is constructed using a dynamic temporal graph neural network. The network performs hierarchical temporal feature aggregation through a time decay function and a double loop propagation mechanism to update the edge weights of the dynamic knowledge graph.

[0022] Based on students' learning status data, generate knowledge point node embedding offsets.

[0023] Specifically, generating knowledge point node embedding offsets based on students' learning status data includes:

[0024] Based on students' learning status data, establish a mapping rule between the knowledge point node embedding offset and the actual learning status data.

[0025] Based on the above mapping rules, the embedding offset of each knowledge point node is calculated for each student.

[0026] Specifically, according to the above mapping rules, calculating the embedding offset of each student's knowledge point node includes:

[0027] The initial knowledge point node embedding offset is calculated using mapping rules and a dynamic knowledge graph.

[0028] For each knowledge point node, the initial embedding offset is averaged with the initial embedding offsets of its directly adjacent predecessor and successor nodes.

[0029] Repeat the steps of calculating the initial knowledge point node embedding offset and neighborhood until the maximum change of the knowledge point node is lower than the preset threshold.

[0030] The embedding offset of the knowledge point node output by iteration is normalized and used as the final embedding offset of the knowledge point node.

[0031] Specifically, the step of calculating the initial knowledge point node embedding offset using mapping rules and a dynamic knowledge graph includes:

[0032] The performance differences of students on tasks related to knowledge points are statistically analyzed, and the performance differences are weighted and summed to obtain a measure of the difficulty of the knowledge points.

[0033] Based on students' learning styles, a style sensitivity coefficient is preset. The difficulty level of knowledge point nodes is multiplied by the style sensitivity coefficient to obtain the initial knowledge point node embedding offset.

[0034] Specifically, updating the current dynamic knowledge graph based on the knowledge point node embedding offset includes: applying the calculated knowledge point node embedding offset to the dynamic knowledge graph, recalculating the distance and weight between nodes, and forming a dynamically differentiated knowledge graph.

[0035] An industrial robot programming teaching system is used to implement the aforementioned industrial robot programming teaching method, including: a data acquisition module, a state data generation module, an offset determination module, an update module, and a difficulty identification and resolution module;

[0036] The data acquisition module is used to collect programming behavior data and industrial robot operation data in real time;

[0037] The status data generation module is used to generate student learning status data based on the programming behavior data and the industrial robot operation data.

[0038] The offset determination module is used to determine the embedding offset of knowledge point nodes based on the student learning status data, in conjunction with a preset dynamic knowledge graph.

[0039] The update module is used to update the current dynamic knowledge graph according to the embedding offset of the knowledge point node;

[0040] The difficulty identification and resolution module is used to determine the student's current knowledge difficulties based on the current dynamic knowledge graph, and to push difficulty analysis data to the student's learning terminal based on the current knowledge difficulties.

[0041] Specifically, the state data generation module includes: a disassembly unit and a state data generation unit;

[0042] The decomposition unit is used to arrange the programming behavior data in chronological order, decompose it into the smallest analyzable unit, and form structured tabular data;

[0043] The status data generation unit is used to analyze students' programming behavior based on structured tabular data, and to assess students' knowledge mastery and learning style.

[0044] Specifically, the offset determination module includes: a mapping rule establishment unit and an embedded offset unit;

[0045] The mapping rule establishment unit is used to establish a mapping rule with the embedding offset of knowledge point nodes based on the student's learning status data.

[0046] The embedding offset unit is used to calculate the initial embedding offset according to the above mapping rules, and then perform neighborhood averaging and iteration.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] This invention proposes an industrial robot programming teaching method and system. By constructing a dynamic knowledge graph and introducing multi-dimensional learning behavior analysis and cognitive obstacle inference mechanisms, it can collect students' programming process data in real time, dynamically adjust the correlation paths between knowledge points, and automatically predict students' current knowledge difficulties. The teaching content presented on the teacher's end or electronic whiteboard is highly matched with the students' difficulties, realizing intelligent, differentiated, and precise programming teaching, significantly improving teaching efficiency, students' mastery quality, and the utilization rate of teaching resources. Attached Figure Description

[0049] Figure 1 A flowchart of an industrial robot programming teaching method provided by the present invention;

[0050] Figure 2 The flowchart for generating student learning status data provided by this invention;

[0051] Figure 3 This is a schematic diagram illustrating the calculation of the embedding offset of knowledge point nodes provided by the present invention;

[0052] Figure 4 This is a diagram illustrating the architecture of an industrial robot programming teaching method according to the present invention. Detailed Implementation

[0053] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0055] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0056] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0057] Example 1

[0058] Please see Figure 1The present invention provides an embodiment of an industrial robot programming teaching method, comprising the following specific steps:

[0059] An electronic device sends a programming file to a teaching robot, the programming file including a control program written according to the programming teaching content;

[0060] Step S1: During the teaching process, collect programming behavior data and industrial robot operation data in real time;

[0061] Specifically, when students operate in the virtual programming environment, the following data is recorded in real time: programming behavior data and industrial robot operation data;

[0062] Programming behavior data includes: instruction operation data, recording the types of programming instructions selected by students, such as motion instructions, logic judgment instructions, sensor call instructions, etc., instruction input order, instruction parameter settings, such as motion speed, coordinate values, number of loops, etc.; debugging behavior data, which statistics include the number of times students debug programs, debugging time points, the location and content of code modifications during debugging, and feedback information on program execution results after debugging, such as whether it was successful and the type of error; and learning interaction data, which monitors students' posts in the software's online discussion area, the frequency of asking questions, access records of learning materials, including access duration, number of accesses, specific material types, and interaction behavior with teachers and classmates, such as the number of times they request guidance from teachers and their activity level in group discussions.

[0063] By establishing a real-time data transmission channel between the software and the actual industrial robot equipment, relevant data on the operation of the industrial robot is acquired. The industrial robot operation data includes: robot status data, which collects information such as the robot's real-time position, movement speed, joint angles, and torque magnitude to analyze the performance of the student-written program on the actual equipment; and sensor data, which receives data from the equipment's sensors (such as vision sensors and force sensors) to determine whether the student program correctly processes the sensor signals and whether it makes reasonable use of the sensor data to achieve the task objectives.

[0064] Step S2: Based on the programming behavior data and the industrial robot operation data, generate student learning status data, combine it with a preset dynamic knowledge graph, determine the knowledge point node embedding offset based on the student learning status data, update the current dynamic knowledge graph based on the knowledge point node embedding offset, and determine the student's current knowledge difficulties based on the current dynamic knowledge graph.

[0065] The specific steps of step S2 are as follows:

[0066] Step S201: Construct an industrial robot programming knowledge graph and establish an error pattern library;

[0067] In this embodiment, the knowledge system of industrial robot programming is sorted out, and knowledge points (such as kinematic principles, sensor applications, programming syntax rules, etc.) are constructed as knowledge graph nodes. The nodes are connected by logical relationships (such as association and application) to form a structured knowledge graph.

[0068] We collected a large number of student programming error cases, analyzed error types (such as syntax errors, logic errors, path planning errors, etc.), and combined them with relevant knowledge points in the knowledge graph to establish a mapping relationship between error patterns and knowledge points, i.e., an error pattern library. For example, the error of coordinate values ​​being out of range corresponds to the knowledge points of coordinate system establishment and coordinate calculation.

[0069] Step S202: Analyze students' programming behavior, knowledge mastery, and learning style based on the preprocessed programming behavior data and industrial robot operation data;

[0070] The specific steps of step S202 are as follows:

[0071] Step S2021: Arrange the preprocessed programming behavior data in chronological order, break it down into the smallest analyzable units, and form structured tabular data;

[0072] For example, a complete programming instruction can be broken down into instruction type (such as motion control) + parameter settings (speed, coordinates) + execution result (success / error); debugging behavior records include debugging time + modified code location + error message content, etc.

[0073] Step S2022: Use the process verification method to identify logical errors. Based on the error pattern library, aggregate similar logical errors to obtain logical error types, including: syntax error type, logical vulnerability type, and device adaptation error type.

[0074] In this embodiment, the process verification method includes basic rule checking, sequence rationality analysis, and parameter validity verification. Basic rule checking verifies whether the programming instructions follow the basic logic of industrial robot programming, such as whether motion instructions are used after defining the coordinate system and whether initialization configuration is completed before calling sensors. Sequence rationality analysis analyzes whether the instruction combination meets the task objective. For example, in a welding task, it checks whether the instruction sequence of moving to the welding point → starting welding → moving to the next welding point is correct; if it is reversed, it is marked as a logical error. Parameter validity verification compares whether the instruction parameters are within the allowable range of the equipment, such as whether the movement speed exceeds the robot's maximum speed limit and whether the coordinate values ​​exceed the workspace.

[0075] The grammar error category categorizes spelling errors and misuse of symbols (such as missing semicolons) and analyzes frequently occurring error-prone instruction keywords, such as common formatting errors in for loop statements; the logic flaw category identifies functional anomalies caused by incorrect condition judgments and chaotic nested loops, recording typical error cases, such as infinite loops caused by not setting a loop termination condition; the device adaptation error category addresses errors that occur during actual operation, such as motor overload and sensor signal loss, and associates them with possible causes in programming, such as load calculation errors and improper sensor data processing.

[0076] Step S2023: Break down the programming task into knowledge units, match them with the industrial robot programming knowledge graph, and dynamically assess the students' knowledge mastery.

[0077] Specifically, programming tasks are broken down into knowledge units and mapped to an industrial robot programming knowledge graph, including kinematics-related, control logic, and sensor applications. Kinematics-related tasks identify instructions involving forward and inverse kinematics calculations and trajectory planning, and are associated with basic kinematics knowledge nodes. Control logic tasks address the use of conditional statements (if-else) and loop statements (while), corresponding to PLC programming logic nodes. Sensor applications tasks analyze the code for sensor data acquisition, filtering, and threshold judgment, mapping it to sensor principles and applications nodes.

[0078] The dynamic assessment of students' knowledge mastery employs a learning trajectory tracking method, which includes: initial exposure: when students first attempt a task involving a certain type of knowledge point, if the completion rate is below 60% and errors are concentrated in core concepts, it is judged as not mastering; gradual improvement: through multiple practices, if the error rate decreases and students can independently solve similar problems, it is adjusted to preliminary mastery; deep understanding: when students can flexibly apply knowledge points to solve complex tasks, such as multi-sensor fusion control, and have strong ability to apply knowledge to other situations, it is marked as mastery; forgetting warning: if students do not use a certain knowledge point for a period of time and the error rate rises in subsequent exercises, a knowledge consolidation reminder is triggered.

[0079] Step S2024: Extract students' behavioral characteristics and analyze their learning styles based on the extracted behavioral characteristics.

[0080] In this embodiment, behavioral features are extracted from the interaction data, which includes students' access to materials, questions, and debugging activities. The behavioral features include: material preferences, question patterns, and debugging habits. Material preferences are determined by statistically analyzing the duration and frequency of students' access to different types of materials (video tutorials, illustrated manuals, code examples) to assess their learning media preferences. Question patterns are analyzed by examining the content of students' questions in the discussion forum, distinguishing between basic help-seeking questions (such as how to set up a coordinate system) and extended exploration questions (such as how to optimize a path to reduce energy consumption). Debugging habits are observed by observing students' debugging behavior, distinguishing between repeated trial-and-error (frequent minor code modifications) and systematic troubleshooting (locating the problem before making changes).

[0081] Based on behavioral characteristics, students' learning styles are divided into four categories: visually oriented, who prefer learning through video demonstrations and have a high acceptance of dynamic visuals; logical reasoning, who tend to read technical documents and are good at understanding programming logic from the perspective of principles; practical exploration, who are accustomed to accumulating experience through repeated debugging and trial and error and have a strong interest in practical tasks; and social interaction, who rely on online discussions and teacher guidance and make significant progress in collaborative learning.

[0082] Step S203: Based on the analysis results of students' programming behavior, knowledge mastery, and learning style, analyze the difficulties encountered by students in the current learning process.

[0083] The specific steps of step S203 are as follows:

[0084] Step S2031: Based on the industrial robot programming knowledge graph, a dynamic knowledge graph is constructed using a dynamic temporal graph neural network. The network performs hierarchical temporal feature aggregation through a time decay function and a double loop propagation mechanism to update the edge weights of the dynamic knowledge graph.

[0085] Specifically, a dynamic temporal graph neural network model is constructed, comprising an input layer, hidden layers, and an output layer. The input layer contains student learning status data, including students' programming behavior, knowledge mastery, and learning styles. The hidden layer is designed with a time decay function and a double loop propagation mechanism to handle the temporal characteristics of the data. The output layer outputs the updated knowledge graph edge weights, and the time decay function is defined as: γ(t) = e^(-t / t). -λ×Δt , where Δt represents the time interval between the current operation and the historical similar operations, λ represents the learning progress adjustment parameter, and the edge weights are dynamically updated according to the time decay function;

[0086] The dual-loop propagation mechanism consists of two loops: the inner loop, which aggregates local temporal features in the knowledge graph at each time step, centered on the knowledge point node involved, and integrates the feature information of the node itself and its neighboring nodes at that time step; and the outer loop, which repeats the inner loop process over time, transmitting and integrating information between different time steps to achieve hierarchical temporal feature aggregation, gradually updating the weights of each edge in the knowledge graph, and enabling the knowledge graph to dynamically reflect the changes in the relationships between knowledge points during the student's learning process.

[0087] Step S2032: Based on the analysis results of students' programming behavior, knowledge mastery, and learning style, generate knowledge point node embedding offsets, adjust the node association strength of the dynamic knowledge graph, and form an individualized dynamic knowledge graph.

[0088] The specific steps of step S2032 include:

[0089] 1. Based on the frequency of instruction usage and error type distribution in programming behavior data, the mastery probability of each knowledge point in knowledge mastery data, and learning preferences and interaction behaviors in learning style data, establish a mapping rule with the embedding offset of knowledge point nodes.

[0090] For example, the mapping rules can be determined based on students' behavioral data during the learning process. For instance, if a student makes multiple errors related to path planning algorithms during programming exercises, such as causing robot collisions due to unreasonable path planning, the initial embedding offset of the path planning algorithm knowledge point node will increase accordingly according to the mapping rules. If a student frequently consults materials on welding programming applications, the initial embedding offset of the welding programming application node will decrease according to the mapping rules, indicating that the student has a good grasp of the knowledge point.

[0091] 2. Based on the above mapping rules, calculate the embedding offset of each student's knowledge point node;

[0092] Based on the above mapping rules, the embedding offset of each student's knowledge point node is calculated. The specific steps include:

[0093] The initial knowledge point node embedding offset is calculated using mapping rules and a dynamic knowledge graph.

[0094] Specifically, the performance differences of the student on tasks related to this knowledge point are statistically analyzed, including: accuracy difference (the gap between the accuracy rate and the average accuracy rate); completion time difference (the deviation from the standard time); and error reproduction difference (the deviation from the average number of retries on this knowledge point). The above three differences are weighted and summarized according to pre-set weights (set by those in the field according to the actual situation) to obtain the difficulty level measure of the node. Based on the student's learning style, such as visualization preference, text preference, hands-on practice preference, etc., a style sensitivity coefficient is predefined for each type of knowledge point. The difficulty level measure is multiplied by the style sensitivity coefficient of the corresponding knowledge point to obtain the initial knowledge point node embedding offset.

[0095] For each knowledge point node, the initial embedding offset is averaged with the initial embedding offsets of its directly adjacent predecessor and successor nodes.

[0096] Repeat the steps of calculating the initial knowledge point node embedding offset and neighborhood until the maximum change of the knowledge point node is lower than the preset threshold.

[0097] The embedding offset of the knowledge point node output by iteration is normalized and used as the final embedding offset of the knowledge point node.

[0098] In this embodiment, the initial offset value is calculated based on local behavior. For example, if a student gets stuck on a question, it will cause the model to overfit to local anomalies if it is not propagated to adjacent knowledge points. In the first round of offset value calculation, each node only considers its own behavioral characteristics and does not consider its position in the entire knowledge path. Through iterative neighborhood smoothing, the state of each node is corrected by its upstream and downstream nodes, and finally a structurally reasonable offset pattern is formed. Multiple rounds of iteration are the modeling of this process, which reflects the estimation of the student's cognitive state, rather than a static score.

[0099] The purpose of normalization is to unify the numerical scale, making it easier to apply to dynamic knowledge graphs.

[0100] 3. The calculated knowledge point node embedding offsets are applied to the dynamic knowledge graph, and the distances and weights between nodes are recalculated to form a dynamic knowledge graph with individual differences.

[0101] Specifically, the weights are updated using embedded offsets to enhance the association between knowledge points that are closely related to the student's current learning status, while weakening the association between irrelevant or already mastered knowledge points. This forms a differentiated knowledge transmission path that conforms to the individual characteristics of the student, highlighting the student's unique knowledge cognitive structure in the learning process.

[0102] Step S2033: Establish a cognitive impairment inference model, extract the correlation between the current operation and each node in the dynamic knowledge graph, and use a Bayesian network to construct a probabilistic graphical model to infer the difficulties students are currently learning.

[0103] Step S3: Push the analysis data of the difficulties to the students' learning terminals based on the current knowledge difficulties.

[0104] Specifically, based on the difficulties students encounter in their current learning process, targeted teaching content delivery strategies are generated, timely interventions are provided, personalized learning paths are planned for students, and the effectiveness of the interventions is evaluated.

[0105] In this embodiment, a targeted teaching content push strategy is generated based on the data analysis results, including: if a student's mastery probability of a certain knowledge node is lower than the threshold and the student has a passive learning style, the teacher will provide one-on-one explanation to the student, or push a video tutorial containing a detailed explanation of the knowledge node, along with step-by-step guided exercises, and provide a pop-up window with instructions in the programming interface.

[0106] During the subsequent student programming process, real-time intervention is provided based on the problems detected in real time. For example, real-time error prompts are provided. When a programming error is detected, an error prompt window pops up to explain the error type and cause. The erroneous code line is highlighted in different colors in the code editing area, and code correction examples related to the error are provided.

[0107] The evaluation of intervention effectiveness, i.e., the assessment of teaching effectiveness, involves establishing a multi-dimensional evaluation index system, including: academic performance evaluation, comparing students' completion of programming tasks before and after receiving personalized intervention, including indicators such as task completion time, code correctness, and completeness of function implementation; knowledge mastery evaluation, assessing the extent of improvement in knowledge mastery by observing students' application of previously weak knowledge points through regular tests or in new tasks; and learning behavior evaluation, analyzing changes in students' learning interaction behaviors after receiving intervention, such as whether participation in online discussion forums has increased and whether the frequency of access to self-study materials has increased.

[0108] Example 2

[0109] Please see Figure 4 Another embodiment of the present invention provides: an industrial robot programming teaching system, comprising: a data acquisition module, a state data generation module, an offset determination module, an update module, and a difficulty identification and resolution module;

[0110] The data acquisition module is used to collect programming behavior data and industrial robot operation data in real time;

[0111] The status data generation module is used to generate student learning status data based on the programming behavior data and the industrial robot operation data.

[0112] The offset determination module is used to determine the embedding offset of knowledge point nodes based on the student learning status data, in conjunction with a preset dynamic knowledge graph.

[0113] The update module is used to update the current dynamic knowledge graph according to the embedding offset of the knowledge point node;

[0114] The difficulty identification and resolution module is used to determine the student's current knowledge difficulties based on the current dynamic knowledge graph, and to push difficulty analysis data to the student's learning terminal based on the current knowledge difficulties.

[0115] The status data generation module includes: a disassembly unit and a status data generation unit;

[0116] The decomposition unit is used to arrange the programming behavior data in chronological order, decompose it into the smallest analyzable unit, and form structured tabular data;

[0117] The status data generation unit is used to analyze students' programming behavior based on structured tabular data, and to assess students' knowledge mastery and learning style.

[0118] The offset determination module includes: a mapping rule establishment unit and an embedded offset unit;

[0119] The mapping rule establishment unit is used to establish a mapping rule with the embedding offset of knowledge point nodes based on the student's learning status data.

[0120] The embedding offset unit is used to calculate the initial embedding offset according to the above mapping rules, and then perform neighborhood averaging and iteration.

[0121] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0122] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A teaching method for industrial robot programming, characterized in that, include: Real-time collection of programming behavior data and industrial robot operation data; Based on the programming behavior data and the industrial robot operation data, student learning status data is generated, which includes the student's programming behavior, knowledge mastery level and learning style. Based on the student learning status data, the embedded offset of knowledge point nodes is determined by combining the preset dynamic knowledge graph. The embedded offset of knowledge point nodes is used to adjust the node association strength of the dynamic knowledge graph. The dynamic knowledge graph is based on the industrial robot programming knowledge graph and is constructed using a dynamic temporal graph neural network. Update the current dynamic knowledge graph based on the embedding offset of the knowledge point node; Based on the current dynamic knowledge graph, the student's current knowledge difficulties are determined, and difficulty analysis data is pushed to the student's learning terminal based on the current knowledge difficulties. The step of determining the knowledge point node embedding offset based on the student learning status data, combined with a preset dynamic knowledge graph, includes: Based on the industrial robot knowledge base, construct an industrial robot programming knowledge graph; Based on the knowledge graph of industrial robot programming, a dynamic knowledge graph is constructed using a dynamic temporal graph neural network. The network performs hierarchical temporal feature aggregation through a time decay function and a double loop propagation mechanism to update the edge weights of the dynamic knowledge graph. Based on students' learning status data, generate knowledge point node embedding offsets; The step of generating knowledge point node embedding offsets based on students' learning status data includes: Based on students' learning status data, establish a mapping rule between the knowledge point node embedding offset and the actual learning status data. Based on the above mapping rules, calculate the embedding offset of each student's knowledge point node; The step of calculating the knowledge point node embedding offset for each student according to the above mapping rules includes: The initial knowledge point node embedding offset is calculated using mapping rules and a dynamic knowledge graph. For each knowledge point node, the initial embedding offset is averaged with the initial embedding offsets of its directly adjacent predecessor and successor nodes. Repeat the steps of calculating the initial knowledge point node embedding offset and neighborhood until the maximum change of the knowledge point node is lower than the preset threshold. The embedding offset of the knowledge point node output by the iteration is normalized and used as the final embedding offset of the knowledge point node. The calculation of the initial knowledge point node embedding offset using mapping rules and dynamic knowledge graphs includes: The performance differences of students on tasks related to knowledge points are statistically analyzed, and the performance differences are weighted and summed to obtain a measure of the difficulty of the knowledge points. Based on students' learning styles, a style sensitivity coefficient is preset. The difficulty level of knowledge point nodes is multiplied by the style sensitivity coefficient to obtain the initial knowledge point node embedding offset.

2. The industrial robot programming teaching method as described in claim 1, characterized in that, The step of generating student learning status data based on the programming behavior data and the industrial robot operation data includes: The programming behavior data is arranged in chronological order and broken down into the smallest analyzable units to form structured tabular data; The process verification method is used to identify logical errors. Based on the error pattern library, similar logical errors are aggregated to obtain logical error types, including: syntax error type, logical vulnerability type and device adaptation error type. The error pattern library is constructed based on the industrial robot knowledge base. Programming tasks are broken down into knowledge units and mapped to an industrial robot programming knowledge graph to dynamically assess students’ knowledge mastery. Students' behavioral characteristics are extracted, and their learning styles are analyzed based on these extracted characteristics.

3. The industrial robot programming teaching method as described in claim 2, characterized in that, The step of updating the current dynamic knowledge graph based on the knowledge point node embedding offset includes: applying the calculated knowledge point node embedding offset to the dynamic knowledge graph, recalculating the distance and weight between nodes, and forming a dynamic knowledge graph with individual differences.

4. An industrial robot programming teaching system, used to implement the industrial robot programming teaching method according to any one of claims 1-3, characterized in that, include: The module includes a data acquisition module, a status data generation module, an offset determination module, an update module, and a problem identification and resolution module. The data acquisition module is used to collect programming behavior data and industrial robot operation data in real time; The status data generation module is used to generate student learning status data based on the programming behavior data and the industrial robot operation data. The offset determination module is used to determine the embedding offset of knowledge point nodes based on the student learning status data, in conjunction with a preset dynamic knowledge graph. The update module is used to update the current dynamic knowledge graph according to the embedding offset of the knowledge point node; The difficulty identification and resolution module is used to determine the student's current knowledge difficulties based on the current dynamic knowledge graph, and to push difficulty analysis data to the student's learning terminal based on the current knowledge difficulties.

5. The industrial robot programming teaching system as described in claim 4, characterized in that, The status data generation module includes: a disassembly unit and a status data generation unit; The decomposition unit is used to arrange the programming behavior data in chronological order, decompose it into the smallest analyzable unit, and form structured tabular data; The status data generation unit is used to analyze students' programming behavior based on structured tabular data, and to assess students' knowledge mastery and learning style.

6. The industrial robot programming teaching system as described in claim 5, characterized in that, The offset determination module includes: a mapping rule establishment unit and an embedded offset unit; The mapping rule establishment unit is used to establish a mapping rule with the embedding offset of knowledge point nodes based on the student's learning status data. The embedding offset unit is used to calculate the initial embedding offset according to the above mapping rules, and then perform neighborhood averaging and iteration.

Images