Artificial intelligence knowledge and skill education system and method

Abstract

According to the artificial intelligence knowledge and skill education system and method provided by the invention, through a vivid virtual learning environment constructed by the virtual reality module, a learner can throughout the mind of practice operation of artificial intelligence knowledge and skill, just like being in a real working scene or experimental environment, and the knowledge and skill education efficiency is improved. The interestingness, the participation degree and the concentration degree of learning are greatly improved, knowledge and skills can be better understood and mastered, knowledge and skill information sharing of different learning resource subjects is achieved, the learning effect is improved, perceptual mastering of practical skills can be rapidly achieved at low cost, and the interestingness, the participation degree and the concentration degree of learning are improved. And reference is provided for connection of school education and social practice education. According to the method, individuation and self-adaptive learning deepening are realized, individual differences of learners in a virtual reality environment are fully considered, abundant learner features are analyzed by using a neural network, an individualized learning path which is more practical is generated, and a virtual scene and a learning task are dynamically optimized according to a learning condition.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence education, and specifically to a knowledge and skill education system and method for artificial intelligence. Background Art

[0002] With the rapid development of artificial intelligence technology, the requirements for the knowledge and skills of relevant professionals are also increasing day by day. Traditional knowledge and skill training methods often have many limitations. For example, the teaching content is not updated in a timely manner, it is difficult to conduct personalized training according to the characteristics of different learners, and there is a lack of accurate evaluation of training effects. Especially for those involving practical hands-on operations, due to various reasons of experimental resources, it is impossible to put profound theoretical knowledge into practical operations. In addition, with the increasing changes of the technological revolution, emerging technologies are constantly updated and deepened. How to effectively connect the teaching and education places with society both in theory and practice is also a problem worthy of exploration.

[0003] Artificial intelligence has unique advantages in processing complex data, simulating the human learning process, and adaptive adjustment. Applying it to the field of knowledge and skill training is expected to overcome the deficiencies of existing training methods and will improve the efficiency and quality of training. At the same time, virtual reality technology can realize the portrayal and simulation of the real world and achieve a real-world-like experience at a low cost. Therefore, exploring the combination of artificial intelligence and virtual reality technology to improve the effect of teaching and education has practical significance. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a knowledge and skill education system and method for artificial intelligence to solve the deficiencies of the existing technologies described in the background art.

[0005] To solve the above technical problem, the embodiments of the present invention provide the following technical solutions: A knowledge and skill education system for artificial intelligence, including:

[0006] A data collection and integration module, used for data collection and preprocessing, determining the scope and sources of artificial intelligence knowledge and skill data and virtual reality-related data to be collected, obtaining raw data through corresponding data collection technologies, and using data preprocessing technologies to process different types of data; labeling and storing the preprocessed data according to the knowledge and skill classification system and virtual reality resource classification, and constructing a learning resource database;

[0007] A neural network model construction module, used for neural network model construction and training, selecting a suitable neural network architecture according to the specific training objectives and the characteristics of knowledge and skill types in the virtual reality environment, and customizing and constructing the model in combination with relevant improvement mechanisms, and training and validating the constructed model, and adjusting the hyperparameters according to the preset performance indicators;

[0008] The personalized learning path planning module is used to collect the extended personal information of learners, including virtual reality-related features, as well as the original basic information, learning preferences, and historical learning data, and extract the learner feature vectors using data analysis and machine learning algorithms;

[0009] Based on the collected and preprocessed data, a complete artificial intelligence knowledge and skill map is constructed, associating knowledge points and skill points with virtual reality learning links, and sorting out a more practical and relevant knowledge and skill structure framework;

[0010] Combined with the learner feature vectors and the updated knowledge and skill map, the intelligent path planning algorithm is used to generate a personalized learning path exclusive to each learner and integrating virtual reality learning links;

[0011] The virtual reality module is used for virtual reality scene setting and resource pushing. According to the virtual reality training scenes planned in the personalized learning path, the virtual scene construction sub-module is used to build the corresponding virtual learning environment, and the relevant resources of the built virtual scene, not limited to the entrance link and operation instructions, together with other traditional learning resources, are pushed to the learners according to the learning path planning;

[0012] The training execution module is used to perform operational learning according to the planned learning tasks after the learner enters the virtual reality learning environment, interact with virtual objects and devices using the interactive functions in the virtual scene, complete virtual operation tasks and explore learning content; the interactive guidance sub-module provides real-time operation tips and guidance information for the learner, and real-time monitors various performance data of the learner during the virtual reality learning process, and transmits the performance data to the evaluation and feedback module in a timely manner;

[0013] The evaluation and feedback module is used for evaluation feedback and path dynamic adjustment. Through multi-dimensional evaluation methods, comprehensively considering learning outcomes, learning behaviors, and special dimensions in the virtual reality environment, a neural network model is used to comprehensively evaluate the learner's knowledge and skill mastery level, generate a feedback report according to the evaluation results, feedback the evaluation results to the personalized learning path planning module, trigger the dynamic adjustment of the subsequent learning path, and optimize the learning task arrangement, resource recommendation, and difficulty setting in the virtual reality scene.

[0014] Furthermore, the data collection and integration module includes:

[0015] The data collection sub-module, in addition to collecting the knowledge and skill-related data from multiple sources, also expands the collection of data related to virtual reality scene construction, including virtual environment model materials, virtual interaction action data, and multimedia materials suitable for virtual reality display, uses 3D scanning equipment to obtain 3D model data of real objects, and uses a motion capture system to record the action information of the human body in virtual interaction;

[0016] A data preprocessing sub-module performs conventional preprocessing operations such as cleaning, denoising, and normalization on the collected raw data, and performs format conversion, spatial coordinate calibration, and virtual scene adaptation processing on virtual reality-related data.

[0017] Furthermore, the neural network model construction module includes:

[0018] A model construction sub-module constructs a model that combines a convolutional neural network and a recurrent neural network for processing data such as images and sequences according to the perception and interaction requirements in the virtual scene; at the same time, constructs a behavior analysis model based on a deep neural network;

[0019] A model training sub-module extracts the preprocessed training data set from the data storage module, which includes virtual reality-related data and traditional knowledge and skill data, and inputs the data into the corresponding neural network model for training according to the set training strategy;

[0020] A model update and storage sub-module updates the existing neural network model when new learning data, data generated by virtual reality scene updates, or changes in training task requirements occur.

[0021] Furthermore, the personalized learning path planning module includes:

[0022] A learner feature analysis sub-module is used to expand and collect multi-dimensional personal information of learners, incorporate features related to virtual reality learning, combine the original basic information, learning preferences, historical learning data, and learning behavior characteristics, and use data analysis and machine learning algorithms for comprehensive analysis to extract the expanded feature vector of learners;

[0023] A knowledge and skill graph construction sub-module is used to improve the artificial intelligence knowledge and skill graph, sort out the logical relationships between knowledge points and skill points, and perform associative mapping with practical operation links, virtual resource utilization, etc. in the virtual reality scene;

[0024] A path planning sub-module combines the expanded feature vector of learners and the updated knowledge and skill graph, and uses the path planning method of intelligent path planning algorithms to generate a personalized learning path integrating virtual reality learning links for each learner, and represents the learning path as an ordered task sequence, where each task corresponds to the learning of specific knowledge points or skill points and the corresponding virtual reality training scenarios, interactive operation tasks, and learning resource recommendations.

[0025] Furthermore, the virtual reality module includes:

[0026] The virtual scene construction submodule uses the collected virtual environment model materials, multimedia materials and other data to build a variety of virtual reality learning scenes according to the needs of artificial intelligence knowledge and skills training and the planned learning path;

[0027] The interactive design submodule develops a variety of interactive methods suitable for virtual reality environments, and establishes an interactive feedback mechanism. When learners operate, they are informed of the operation results in a timely manner through visual, auditory and other multi-sensory feedback;

[0028] The virtual scene management submodule is responsible for real-time management and updating of virtual scenes, and dynamically adjusts the content, difficulty level and presentation effect of virtual scenes according to learners' learning progress, personalized learning paths and evaluation results fed back by the system.

[0029] Furthermore, the training execution module includes:

[0030] The learning resource push submodule selects and integrates the corresponding learning resources from the data storage module according to the learning path generated by the personalized learning path planning module, including virtual reality scene-related resources and traditional text materials, video tutorials, and code examples, and pushes the learning content to learners through 3D projection or virtual reality head-mounted devices, while dynamically adjusting the order and content of resource push according to the learners' learning progress and real-time feedback;

[0031] The interactive guidance submodule provides learners with more targeted interactive guidance tools and prompt information in a virtual reality environment;

[0032] The training process monitoring submodule monitors the learners' performance data in the virtual reality learning process in real time and issues early warning for abnormal situations that occur during the learning process.

[0033] Furthermore, the evaluation feedback module includes:

[0034] The multi-dimensional evaluation submodule is used to use a variety of evaluation methods and neural network models to conduct a more comprehensive and integrated evaluation of learners' knowledge and skills from different perspectives in a virtual reality environment, including: direct evaluation based on learning outcomes, comparing learners' completion results in virtual operation tasks with standard answers and best practice cases, and calculating corresponding error indicators; indirect evaluation based on learning behavior, using deep learning models to deeply analyze learners' behavioral characteristics in the virtual reality learning process and explore potential learning problems and advantages;

[0035] Feedback generation submodule, which is used to generate detailed and personalized feedback reports for learners based on multi-dimensional evaluation results;

[0036] The path dynamic adjustment sub-module is used to feedback the evaluation results to the personalized learning path planning module, triggering the dynamic adjustment of the learner's subsequent learning path. Specifically, according to the learner's mastery in the virtual environment, for the virtual operation skills and knowledge and skill points that have been proficiently mastered, the learning progress is accelerated, the task difficulty in the virtual scene is optimized, and more challenging practice sessions are added; for the parts with weak mastery, special virtual training tasks are added, corresponding basic learning resources are supplemented, and the learning order is adjusted.

[0037] The present invention also provides an artificial intelligence knowledge and skill education method, including the following steps:

[0038] S1. Data collection and preprocessing: Determine the scope and sources of the artificial intelligence knowledge and skill data and virtual reality related data to be collected, obtain the original data through corresponding data collection technologies, and use data preprocessing technologies to process different types of data; label and store the preprocessed data according to the knowledge and skill classification system and virtual reality resource classification, and construct a learning resource database;

[0039] S2. Neural network model construction and training: According to the specific training objectives and the characteristics of knowledge and skill types in the virtual reality environment, select a suitable neural network architecture, and customize the model construction in combination with relevant improvement mechanisms, and train and verify the constructed model, and adjust the hyperparameters according to the preset performance indicators;

[0040] S3. Personalized learning path planning: Collect the learner's extended personal information, including virtual reality related features and the original basic information, learning preferences, and historical learning data, and use data analysis and machine learning algorithms to extract the learner feature vector;

[0041] Based on the collected and preprocessed data, construct a complete artificial intelligence knowledge and skill map, associate and map knowledge points, skill points with virtual reality learning links, and sort out a more practical and relevant knowledge and skill structure framework;

[0042] Combined with the learner feature vector and the updated knowledge and skill map, use the intelligent path planning algorithm to generate a personalized learning path exclusive to each learner and integrating virtual reality learning links;

[0043] S4. Virtual reality scene setting and resource push: According to the virtual reality training scenes planned in the personalized learning path, use the virtual scene construction sub-module to build the corresponding virtual learning environment, and push the built virtual scene related resources, not limited to entrance links and operation instructions, together with other traditional learning resources, to the learners according to the learning path planning;

[0044] S5. Training Execution: After the learner enters the virtual reality learning environment, they perform operational learning according to the planned learning tasks, interact with virtual objects and devices using the interaction functions in the virtual scenario to complete virtual operation tasks and explore learning content. The interactive guidance sub-module provides real-time operation tips and guidance information to the learner, and monitors various performance data of the learner during the virtual reality learning process in real time, and transmits the performance data to the evaluation and feedback module in a timely manner.

[0045] S6. Evaluation Feedback and Dynamic Path Adjustment: Through multi-dimensional evaluation methods, comprehensively considering learning outcomes, learning behaviors, and special dimensions in the virtual reality environment, a neural network model is used to comprehensively evaluate the learner's knowledge and skill mastery. According to the evaluation results, a feedback report is generated, and the evaluation results are fed back to the personalized learning path planning module to trigger dynamic adjustment of the subsequent learning path, optimizing the learning task arrangement, resource recommendation, and difficulty setting in the virtual reality scenario.

[0046] Further, the intelligent path planning algorithm specifically adopts the DQN reinforcement learning algorithm, uses a multi-layer perceptron neural network MLP to approximately represent the Q-value function, denoted as Q(s,a;θ), where θ is the neural network parameter, the input of the network is the state s, and the output is the Q-value estimate for each action a∈A. The mean square error is used as the loss function to measure the difference between the predicted Q-value and the target Q-value, and its formula is:

[0047] L(θ) = E[(y - Q(s,a;θ)) 2

[0048] where y is the target Q-value, which is calculated based on the target value of the Bellman equation.

[0049] Further, the comprehensive evaluation of the learner's knowledge and skill mastery using the neural network model specifically includes:

[0050] Construct an evaluation index system, including knowledge understanding, skill operation, problem-solving ability, knowledge transfer ability, and learning behavior performance;

[0051] According to different evaluation dimensions and the characteristics of specific evaluation indicators, select a matching neural network model and architecture;

[0052] Collect the learner's learning outcome data and process data, and perform data preprocessing, and label the preprocessed data;

[0053] Divide the preprocessed labeled data into a training set, a validation set, and a test set according to a certain proportion, use the training set to train the selected neural network model, and use the validation set to monitor the performance of the model and adjust and optimize the hyperparameters of the model according to the validation results; ​

[0054] After the learner completes a certain stage of learning tasks, various types of data corresponding to the learner are collected and preprocessed, and then input into each trained neural network model. The model calculates based on the input data and outputs the predicted values corresponding to each evaluation index.

[0055] The beneficial effects of the above technical solutions of the present invention are as follows:

[0056] 1. Through the realistic virtual learning environment constructed by the virtual reality module of the present invention, learners can fully immerse themselves in the practical operation of artificial intelligence knowledge and skills, as if in a real work scenario or experimental environment, greatly improving the fun, participation and concentration of learning, helping to better understand and master knowledge and skills, realizing the sharing of knowledge and skill information among different learning resource entities such as enterprises, education bases, laboratories, schools, etc., enhancing the learning effect, and enabling the perceptual mastery of practical skills to be quickly achieved at a relatively low cost, providing a reference for the connection between school education and social practice education.

[0057] 2. The present invention realizes the deepening of personalized and adaptive learning. It fully considers the individual differences of learners in the virtual reality environment, uses neural networks to analyze rich learner characteristics, generates a more practical personalized learning path, and dynamically optimizes the virtual scene and learning tasks according to the learning situation, realizing an all-round adaptive adjustment from learning content to learning environment, ensuring that the training process fits the individual characteristics and learning progress of learners.

[0058] 3. The present invention realizes the improvement of the interactivity, fun and effectiveness of knowledge and skill training by integrating the advantages of virtual reality and neural networks, and cultivates the practical ability and innovative thinking of learners in the field of artificial intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is the principle block diagram of the artificial intelligence knowledge and skill training system based on neural networks of the present invention;

[0060] Figure 2 is the principle block diagram of the data acquisition and integration module of the artificial intelligence knowledge and skill training system based on neural networks of the present invention;

[0061] Figure 3 is the principle block diagram of the neural network model construction and management module of the artificial intelligence knowledge and skill training system based on neural networks of the present invention;

[0062] Figure 4 is the principle block diagram of the personalized learning path planning module of the artificial intelligence knowledge and skill training system based on neural networks of the present invention;

[0063] Figure 5Block diagram of the virtual reality module of the artificial intelligence knowledge and skill training system based on neural network according to the present invention;

[0064] Figure 6 Block diagram of the training execution module of the artificial intelligence knowledge and skill training system based on neural network according to the present invention;

[0065] Figure 7 Block diagram of the evaluation and feedback module of the artificial intelligence knowledge and skill training system based on neural network according to the present invention;

[0066] Figure 8 Flowchart of the artificial intelligence knowledge and skill training method according to the present invention. Detailed implementation manners

[0067] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0068] The artificial intelligence knowledge and skill training system of the present invention adds a virtual reality module to the original infrastructure. The overall architecture mainly consists of the following six core modules. Each module works in coordination to form a complete training closed-loop as follows Figure 1 As shown, it includes a data collection and integration module 101, a neural network model construction and management module 102, a personalized learning path planning module 103, a virtual reality module 104, a training execution module 105, and an evaluation and feedback module 106.

[0069] As Figure 2 shown, the data collection and integration module 101 includes:

[0070] A data collection sub-module 1011: In addition to collecting artificial intelligence knowledge and skill-related data from multiple channels such as academic databases, professional textbooks, online course platforms, experimental data, school education bases, enterprise cases, and learners' historical learning records, it also expands the collection of data related to virtual reality scene construction, such as virtual environment model materials (3D models, texture maps, etc.), virtual interaction action data (gestures, motion capture data, etc.), and multimedia materials suitable for virtual reality display (3D videos, spatial audio, etc.). Appropriate collection technologies are adopted, such as 3D scanning devices to obtain 3D model data of real objects, and motion capture systems to record the motion information of the human body in virtual interactions, etc.

[0071] Data preprocessing sub-module 1012: Perform conventional preprocessing operations on the collected raw data, such as cleaning, denoising, and normalization. For virtual reality-related data, further perform processing such as format conversion, spatial coordinate calibration, and adaptation to the virtual scene. For example, adjust the 3D model to a unified coordinate system to ensure its correct display in the virtual environment; filter and extract key frames from the motion capture data so that it can accurately drive the actions of virtual characters. The preprocessed data is classified and labeled according to dimensions such as knowledge domain, skill type, difficulty level, and virtual reality resource type, and stored in a distributed database for convenient and quick retrieval and invocation.

[0072] As Figure 3 shown, the neural network model construction and management module 102 includes:

[0073] Model construction sub-module 1021: Construct diverse neural network models according to the different objectives and content characteristics of artificial intelligence knowledge and skills training in the virtual environment. On the original basis, for the perception and interaction requirements in the virtual scene, such as recognizing the gestures, action intentions of learners in the virtual environment, and operation behaviors on virtual objects, construct a model that combines a convolutional neural network (CNN) and a recurrent neural network (RNN) to process data of types such as images and sequences. At the same time, considering the complexity of training effect evaluation in the virtual environment, construct a behavior analysis model based on a deep neural network to analyze the operation behaviors, decision-making processes, and knowledge application situations of learners from multiple dimensions. Determine hyperparameters such as the number of layers, the number of neurons, and the type of activation function of each network model, and perform random initialization.

[0074] Model training sub-module 1022: Extract the preprocessed training data set from the data storage module, which includes virtual reality-related data and traditional knowledge and skills data, and input the data into the corresponding neural network model for training according to the set training strategy (such as an optimization algorithm using adaptive learning rate adjustment). During the training process, monitor the key performance indicators of the model in real time, such as the loss function value, accuracy, recall rate, and specific virtual interaction indicators (such as gesture recognition accuracy, action matching degree, etc.), and judge whether the training is completed according to the preset stop conditions (such as reaching the specified number of training rounds, and each performance indicator converging to a certain threshold, etc.).

[0075] Model Update and Storage Sub-module 1023: When new learning data (including data generated by virtual reality scene updates) or changes in training task requirements occur, the existing neural network model is updated. Through technical means such as transfer learning and online learning, while retaining the effective knowledge of the original model, new data information is incorporated, and the model parameters are adjusted to adapt to the new training requirements. At the same time, version management is carried out for various trained models, and information such as different versions of the models, their related parameters, and performance indicators is stored in the model repository for easy backtracking and selection.

[0076] As Figure 4 shown, the personalized learning path planning module 103 includes:

[0077] Learner Feature Analysis Sub-module 1031: Further expand the collection of learners' multi-dimensional personal information, incorporating features related to virtual reality learning, such as proficiency in using virtual reality devices, spatial perception ability in the virtual environment, and adaptability to virtual interaction methods (such as handle operations, gesture interactions, etc.). Combining the original basic information, learning preferences, historical learning data, and learning behavior characteristics, these information are comprehensively analyzed using data analysis and machine learning algorithms (such as feature extraction methods in deep learning, clustering analysis, etc.) to extract key feature vectors that are more comprehensive and can accurately represent the individual differences of learners, providing a more sufficient basis for subsequent personalized planning.

[0078] Knowledge and Skill Graph Construction Sub-module 1032: Based on the collected and integrated data, the artificial intelligence knowledge and skill graph is improved, not only sorting out the logical relationships between knowledge points and skill points, but also associating and mapping them with practical operation links and virtual resource utilization in the virtual reality scene. For example, for the skill point of model training in machine learning, it corresponds to the specific steps and related virtual model resources for building and operating a virtual experimental platform in the virtual environment, forming a more practical and relevant directed acyclic graph structure, providing a knowledge framework basis that is more suitable for virtual reality learning for planning reasonable learning paths.

[0079] Path Planning Sub-module 1033: Combining the augmented feature vectors of learners and the updated knowledge and skill graph, using path planning methods based on intelligent algorithms such as reinforcement learning (e.g., Deep Q-Network combined with Deep Deterministic Policy Gradient, DQN+DDPG) or evolutionary algorithms, generate personalized learning paths for each learner that incorporate virtual reality learning sessions. Represent the learning path as an ordered task sequence, where each task corresponds to the learning of specific knowledge points or skill points, as well as the recommendation of corresponding virtual reality training scenarios (such as virtual laboratories, virtual project development environments, etc.), interactive operation tasks (such as completing virtual experiment operations, conducting virtual code debugging, etc.), and learning resources (such as virtual operation guides, 3D case explanations, etc.), ensuring that learners gradually master artificial intelligence knowledge and skills in the virtual environment in an order that conforms to their own abilities and learning patterns, and improving their practical operation capabilities.

[0080] Such as Figure 5 shown, the virtual reality module 104 includes:

[0081] Virtual Scene Construction Sub-module 1041: According to the requirements of artificial intelligence knowledge and skill training and the planned learning path, use the data such as virtual environment model materials and multimedia materials collected to construct diverse virtual reality learning scenes. For example, for the learning of machine learning algorithms, create a virtual laboratory scene with virtual computer equipment, dataset display stands, and interactive algorithm model demonstration devices; for the field of computer vision, construct a visual recognition test site containing various virtual objects and scenes. Through professional virtual reality development tools (such as Unity, Unreal Engine, etc.), perform scene construction, model integration, and environmental effect settings such as lighting and sound effects, so that learners can obtain a realistic immersive experience after entering the virtual scene.

[0082] Interaction Design Sub-module 1042: Develop various interaction methods suitable for the virtual reality environment to enhance the participation and operation convenience of learners in the virtual scene. Design interaction operations based on handles, gesture recognition, eye tracking, etc., enabling learners to interact with the objects in the virtual scene, such as grasping, moving, operating virtual devices, and viewing the detailed information of virtual objects. At the same time, establish an interaction feedback mechanism. When learners perform operations, promptly inform learners of the operation results through multi-sensory feedback such as vision and hearing (such as the prompt sound effect of successful operation, the animation effect of object state changes, etc.), enhancing the realism and interest of the interaction.

[0083] Virtual Scenario Management Sub-module 1043: Responsible for the real-time management and update of virtual scenarios. According to the learner's learning progress, personalized learning path, and the evaluation results feedback by the system, dynamically adjust the content, difficulty level, and presentation effect of the virtual scenario. For example, when the learner has mastered a certain basic virtual operation skill, add more complex operation tasks and virtual objects in subsequent scenarios; or adjust the rendering quality, lighting effects, etc. of the virtual scenario according to the learner's learning efficiency to ensure that the virtual learning environment always matches the learner's learning state and provides a good learning experience.

[0084] As Figure 6 shown, the Training Execution Module 105 includes:

[0085] Learning Resource Push Sub-module 1051: According to the learning path generated by the personalized learning path planning module, screen and integrate corresponding learning resources from the data storage module, including virtual reality scenario-related resources (such as the entry link to enter the virtual scenario, scene operation instructions, etc.) and traditional text materials, video tutorials, code examples, etc., and push them to the learner in a suitable form (such as presenting in a virtual reality head-mounted device, online document, multimedia playlist, etc.) to facilitate their learning activities. At the same time, dynamically adjust the order and content of resource push according to the learner's learning progress and real-time feedback to ensure that the learning resources are closely matched with the learner's current learning stage.

[0086] Interaction Guidance Sub-module 1052: In the virtual reality environment, provide more targeted interaction guidance tools and prompt information for the learner. For example, when the learner enters a new virtual operation scenario, inform the learner of the operation steps and objectives through voice prompts of the virtual assistant, floating operation guides, etc.; during the learner's complex operation process, provide real-time guidance information such as operation skills and error-prone point reminders to help the learner successfully complete the learning task. By recording information such as the learner's operation behaviors and question contents during the virtual reality interaction process, further enrich the learner characteristic data to provide a basis for subsequent evaluation feedback and path adjustment.

[0087] Training Process Monitoring Sub-module 1053: Real-time monitor various performance data of the learner during the virtual reality learning process. In addition to the conventional learning duration, question answering accuracy, etc., focus on virtual interaction-related data, such as the accuracy of gesture operations, the success rate of operating virtual objects, the exploration path in the virtual scenario, etc., collect these data in real-time and transmit them to the evaluation feedback module. At the same time, give early warning prompts for abnormal situations that occur during the learning process (such as no effective operation in the virtual scenario for a long time, frequent operation mistakes that prevent the learning from progressing, etc.) to remind the learner to adjust the learning state or seek help.

[0088] As Figure 7As shown in the figure, the evaluation feedback module 106 includes:

[0089] The multi-dimensional evaluation sub-module 1061: Utilize a variety of evaluation methods and neural network models to conduct a more comprehensive and integrated evaluation of the learner's knowledge and skills mastery from different perspectives in a virtual reality environment. One is the direct evaluation based on learning outcomes, comparing the completion results of the learner in virtual operation tasks (such as whether the parameter settings of virtual experiments are correct, whether the running effect of virtual projects meets the expectations, etc.) with the standard answers and best practice cases, and calculating the corresponding error indicators; the other is the indirect evaluation based on learning behaviors, using deep learning models (such as CNN to analyze operation gesture images, RNN to model operation sequences, and Transformer architecture to understand the behavior logic of learners in virtual scenarios, etc.) to deeply analyze the behavior characteristics of learners in the virtual reality learning process (such as operation standardization, rationality of problem-solving ideas, utilization efficiency of virtual resources, etc.), and mining potential learning problems and advantages. In addition, dimensions such as the development of the learner's spatial perception ability in the virtual environment and the adaptation to complex interaction tasks are also considered to achieve a more three-dimensional evaluation.

[0090] The feedback generation sub-module 1062: Generate a detailed and personalized feedback report for the learner according to the multi-dimensional evaluation results. The report content covers the overall evaluation of the learner's current learning stage in virtual reality learning, the analysis of the mastery of each knowledge and skill point in the virtual environment, the pointing out of specific advantages and deficiencies, and targeted improvement suggestions (such as recommending virtual operation tasks for further practice, supplementary relevant theoretical knowledge for learning, etc.). The feedback report is presented to the learner in a visual form (such as popping up a prompt interface in the virtual reality head-mounted device, pushing messages on the mobile terminal, etc.), facilitating the learner to intuitively understand their learning status and formulate an improvement plan.

[0091] The path dynamic adjustment sub-module 1063: Feed back the evaluation results to the personalized learning path planning module to trigger the dynamic adjustment of the learner's subsequent learning path. According to the learner's mastery in the virtual environment, for the virtual operation skills and knowledge and skill points that have been proficiently mastered, accelerate the learning progress, optimize the task difficulty in the virtual scenario, and add more challenging practice links; for the parts with weak mastery, add special virtual training tasks, supplement the corresponding basic learning resources, and adjust the learning order to ensure that the learner can firmly master each key knowledge and skill point, realizing the dynamic optimization of the learning path in the virtual reality environment.

[0092] The training method of the above-mentioned artificial intelligence knowledge and skill training system based on neural network of the present invention, the specific process mainly includes the following detailed steps, and its process schematic diagram is as Figure 8 shown.

[0093] Step S1: Data collection and preprocessing (corresponding to the data collection and integration module of the system)

[0094] Determine the scope and sources of artificial intelligence knowledge and skills data and virtual reality related data to be collected, comprehensively covering multiple channels such as academic literature, textbooks, online courses, experimental data, school education bases, enterprise cases, learners' historical data, as well as virtual environment model materials, interactive action data, multimedia materials, etc.

[0095] Obtain the original data through corresponding data collection technologies (such as web crawlers, 3D scanning, docking with motion capture systems, etc.), and use data preprocessing technologies (cleaning, normalization, feature extraction, adaptation to virtual scenes, etc.) to process different types of data (text, images, audio, 3D models, etc.) so that they meet the input requirements of the neural network and can be correctly displayed and applied in the virtual reality environment.

[0096] Label and store the preprocessed data according to the knowledge and skills classification system and virtual reality resource classification, build a rich learning resource database, and at the same time establish a data indexing mechanism for quick query and invocation.

[0097] Step S2: Neural network model construction (corresponding to the neural network model construction and management module of the system)

[0098] According to the specific training objectives and the characteristics of knowledge and skills types in the virtual reality environment, select a suitable neural network architecture (such as a combination of CNN and RNN for virtual interaction behavior analysis, deep neural network for behavior analysis and evaluation, etc.), and perform model customization construction in combination with relevant improvement mechanisms (such as attention mechanism, multi-modal fusion mechanism, etc.).

[0099] Determine the hyperparameters of the model, including the number of layers, the number of neurons, activation functions, learning rate, etc., and assign initial parameter values to the model through random initialization.

[0100] Use some representative preprocessed data (including virtual reality related data) to conduct preliminary training and verification on the constructed model, and adjust the hyperparameters according to the preset performance indicators (such as loss value, accuracy, virtual interaction indicators, etc.) to ensure that the model has basic learning and generalization capabilities, especially can effectively play a role when dealing with learning tasks in the virtual reality environment.

[0101] Step S3: Personalized learning path planning (corresponding to the personalized learning path planning module of the system)

[0102] Collect the expanded personal information of learners, including virtual reality-related features as well as the original basic information, learning preferences, historical learning data, etc., and use data analysis and machine learning algorithms to extract the comprehensive feature vectors of learners to characterize their individual differences in the virtual reality learning environment.

[0103] Based on the collected and preprocessed data, construct a complete artificial intelligence knowledge and skill map, associate knowledge points and skill points with virtual reality learning links, and sort out a more practical and relevant knowledge and skill structure framework.

[0104] Combined with the learner feature vectors and the updated knowledge and skill map, use intelligent path planning algorithms to generate personalized learning paths exclusive to each learner and integrated into the virtual reality learning links, and clarify the learning tasks at each stage, the corresponding virtual reality training scenarios, and the recommended learning resources.

[0105] Step S4: Virtual reality scene setting and resource push (add some functions of the virtual reality module and training execution module corresponding to the system)

[0106] According to the virtual reality training scenarios planned in the personalized learning path, use the virtual scene construction sub-module to build the corresponding virtual learning environment. Through operations such as integrating virtual environment model materials, setting interaction functions, and optimizing the environment effect, create a realistic virtual scene that meets the learning task requirements.

[0107] Push the relevant resources such as the entry link and operation instructions of the built virtual scene, together with other traditional learning resources (such as text materials, video tutorials, etc.), to the learners according to the learning path plan, so as to facilitate them to enter the virtual environment to carry out learning activities.

[0108] Step S5: Training execution (training execution module of the corresponding system)

[0109] After the learner enters the virtual reality learning environment, perform operational learning according to the planned learning tasks, interact with virtual objects, devices, etc. using the interaction functions in the virtual scene, complete virtual operation tasks, explore learning content, etc.

[0110] During the learning process, the system provides real-time operation prompts and guiding information for the learners through the interaction guidance sub-module to help them learn smoothly; at the same time, the training process monitoring sub-module monitors the performance data of the learners during the virtual reality learning process in real time, including learning duration, virtual interaction-related data, etc., and transmits these data to the evaluation and feedback module in a timely manner for real-time mastery of the learning situation.

[0111] Step S6: Evaluation and feedback and dynamic path adjustment (evaluation and feedback module of the corresponding system)

[0112] Through a multi-dimensional evaluation method, comprehensively considering learning outcomes, learning behaviors, and special dimensions in the virtual reality environment (such as spatial perception, interaction adaptation, etc.), a neural network model is used to comprehensively evaluate the learner's mastery of knowledge and skills.

[0113] Generate a detailed and personalized feedback report based on the evaluation results, presenting the learner's learning advantages, deficiencies, and improvement suggestions to help the learner clarify the improvement direction.

[0114] Feed back the evaluation results to the personalized learning path planning module, trigger the dynamic adjustment of the subsequent learning path, optimize the learning task arrangement, resource recommendation, and difficulty setting in the virtual reality scenario, etc., to ensure the continuous improvement of learning effects.

[0115] Among them, the intelligent path planning algorithm in the above step S3 specifically adopts the DQN reinforcement learning algorithm, and uses a multi-layer perceptron neural network MLP to approximately represent the Q-value function, denoted as Q(s,a;θ), where θ is the neural network parameter, the input of the network is the state s, and the output is the Q-value estimate for each action a∈A. The mean square error is used as the loss function to measure the difference between the predicted Q-value and the target Q-value, and its formula is:

[0116] L(θ) = E[(y - Q(s,a;θ)) 2

[0117] Among them, y is the target Q-value, which is calculated based on the target value of the Bellman equation.

[0118] The above step S6 uses a neural network model to comprehensively evaluate the learner's mastery of knowledge and skills, specifically including:

[0119] S61. Construct an evaluation index system, including knowledge understanding, skill operation, problem-solving ability, knowledge transfer ability, and learning behavior performance;

[0120] S62. According to different evaluation dimensions and the characteristics of specific evaluation indicators, select a neural network model and architecture that match them;

[0121] S63. Collect the learner's learning outcome data and process data, and perform data preprocessing, and label the preprocessed data;

[0122] S64. Divide the preprocessed labeled data into a training set, a validation set, and a test set according to a certain proportion, use the training set to train the selected neural network model, and use the validation set to monitor the performance of the model and adjust and optimize the hyperparameters of the model according to the validation results;

[0123] ​S65. When the learner completes a certain stage of learning tasks, various types of data corresponding to the learner are collected and preprocessed, and then input into each of the pre-trained neural network models. The models calculate based on the input data and output the predicted values corresponding to each evaluation index.

[0124] In summary, through the realistic virtual learning environment constructed by the virtual reality module in the present invention, learners can fully immerse themselves in the practical operation of artificial intelligence knowledge and skills, as if in a real work scenario or experimental environment, greatly improving the fun, participation, and concentration of learning, helping to better understand and master knowledge and skills, realizing the sharing of knowledge and skill information among different learning resource entities such as enterprises, education bases, laboratories, and schools, enhancing the learning effect, and enabling the perceptual mastery of practical skills to be achieved quickly at a relatively low cost, providing a reference for the connection between school education and social practice education.

[0125] It realizes the deepening of personalized and adaptive learning, fully considers the individual differences of learners in the virtual reality environment, analyzes rich learner characteristics using neural networks, generates a more practical personalized learning path, and dynamically optimizes the virtual scene and learning tasks according to the learning situation, achieving a full range of adaptive adjustments from learning content to learning environment, ensuring that the training process conforms to the individual characteristics and learning progress of learners.

[0126] Finally, by integrating the advantages of virtual reality and neural networks, it improves the interactivity, fun, and effectiveness of knowledge and skill training, and cultivates the practical ability and innovative thinking of learners in the field of artificial intelligence.

[0127] The above is the preferred implementation manner of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle described in the present invention, several improvements and refinements can still be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An artificial intelligence knowledge and skills education system, characterized in that: include: The data collection and integration module is used for data collection and preprocessing, determining the scope and source of the artificial intelligence knowledge and skills data and virtual reality related data to be collected, obtaining the original data through the corresponding data collection technology, and using the data preprocessing technology to process different types of data; The pre-processed data is labeled and stored according to the knowledge and skills classification system and virtual reality resource classification to build a learning resource database; The neural network model building module is used for neural network model building and training. According to the specific training objectives and the characteristics of knowledge and skills in the virtual reality environment, the appropriate neural network architecture is selected, and the model is customized and built in combination with relevant improvement mechanisms. The built model is trained and verified, and the hyperparameters are adjusted according to the preset performance indicators. The personalized learning path planning module is used to collect learners' extended personal information, including virtual reality related features and original basic information, learning preferences, and historical learning data, and use data analysis and machine learning algorithms to extract learners' feature vectors; Based on the collected and preprocessed data, a complete AI knowledge and skills map is constructed, and knowledge points, skill points and VR learning links are mapped and correlated to form a more practical and relevant knowledge and skills structure framework. Combining learner feature vectors and updated knowledge and skills maps, an intelligent path planning algorithm is used to generate a unique, personalized learning path for each learner that integrates virtual reality learning. The virtual reality module is used for setting up virtual reality scenes and pushing resources. According to the virtual reality training scenes planned in the personalized learning path, the corresponding virtual learning environment is built using the virtual scene construction submodule. The built virtual scenes are not limited to the entry links, related resources of operation instructions, together with other traditional learning resources, and pushed to learners according to the learning path planning; The training execution module is used for learners to operate and learn according to the planned learning tasks after entering the virtual reality learning environment, and to interact with virtual objects and devices using the interactive functions in the virtual scene to complete virtual operation tasks and explore learning content; The interactive guidance submodule provides learners with real-time operation prompts and guidance information, monitors learners' performance data in real time during the virtual reality learning process, and transmits the performance data to the evaluation feedback module in a timely manner; The evaluation feedback module is used for evaluation feedback and dynamic adjustment of paths. Through a multi-dimensional evaluation method, it comprehensively considers learning outcomes, learning behaviors, and special dimensions in the virtual reality environment, and uses a neural network model to comprehensively evaluate the learners' knowledge and skills. A feedback report is generated based on the evaluation results, and the evaluation results are fed back to the personalized learning path planning module to trigger dynamic adjustment of subsequent learning paths and optimize learning task arrangements, resource recommendations, and difficulty settings in virtual reality scenarios.

2. The artificial intelligence knowledge and skills training system based on neural network according to claim 1 is characterized in that: The data collection and integration module includes: The data collection submodule, in addition to collecting the knowledge and skills related data from the above-mentioned multiple sources, also expands the collection of data related to the construction of virtual reality scenes, including virtual environment model materials, virtual interactive action data, and multimedia materials suitable for virtual reality display, using 3D scanning equipment to obtain 3D model data of real objects, and using a motion capture system to record the action information of the human body in virtual interaction; The data preprocessing submodule performs routine cleaning, denoising, and normalization preprocessing operations on the collected raw data, and performs format conversion, spatial coordinate calibration, and virtual scene adaptation processing on virtual reality related data.

3. The artificial intelligence knowledge and skills training system based on neural network according to claim 1 is characterized in that: The neural network model building module includes: The model building submodule builds a model combining convolutional neural network and recurrent neural network to process image, sequence and other types of data in response to the perception and interaction needs in virtual scenes. At the same time, it builds a behavior analysis model based on deep neural network. The model training submodule extracts the preprocessed training data set from the data storage module, including virtual reality related data and traditional knowledge and skills data, and inputs the data into the corresponding neural network model for training according to the set training strategy; The model update and storage submodule updates the existing neural network model when new learning data, data generated by virtual reality scene updates, or changes in training task requirements appear.

4. The artificial intelligence knowledge and skills training system based on neural network according to claim 1 is characterized in that: The personalized learning path planning module includes: The learner feature analysis submodule is used to expand and collect learners' multi-dimensional personal information, incorporate features related to virtual reality learning, combine the original basic information, learning preferences, historical learning data, and learning behavior characteristics, use data analysis and machine learning algorithms for comprehensive analysis, and extract learners' expanded feature vectors; The knowledge and skills map construction submodule is used to improve the AI ​​knowledge and skills map, sort out the logical relationship between knowledge points and skill points, and associate and map them with the practical operation links and virtual resource utilization in the virtual reality scene; The path planning submodule combines the learner's expanded feature vector and updated knowledge and skills map, and uses the path planning method of the intelligent path planning algorithm to generate a personalized learning path for each learner that integrates into the virtual reality learning process. The learning path is represented as an ordered task sequence, and each task corresponds to the learning of specific knowledge points or skill points, as well as corresponding virtual reality training scenarios, interactive operation tasks, and learning resource recommendations.

5. The artificial intelligence knowledge and skills training system based on neural network according to claim 1 is characterized in that: The virtual reality module includes: The virtual scene construction submodule uses the collected virtual environment model materials, multimedia materials and other data to build a variety of virtual reality learning scenes according to the needs of artificial intelligence knowledge and skills training and the planned learning path; The interactive design submodule develops a variety of interactive methods suitable for virtual reality environments, and establishes an interactive feedback mechanism. When learners operate, they are informed of the operation results in a timely manner through visual, auditory and other multi-sensory feedback; The virtual scene management submodule is responsible for real-time management and updating of virtual scenes, and dynamically adjusts the content, difficulty level and presentation effect of virtual scenes according to learners' learning progress, personalized learning paths and evaluation results fed back by the system.

6. The artificial intelligence knowledge and skills training system based on neural network according to claim 1 is characterized in that: The training execution module comprises: The learning resource push submodule selects and integrates the corresponding learning resources from the data storage module according to the learning path generated by the personalized learning path planning module, including virtual reality scene-related resources and traditional text materials, video tutorials, and code examples, and pushes the learning content to learners through 3D projection or virtual reality head-mounted devices, while dynamically adjusting the order and content of resource push according to the learners' learning progress and real-time feedback; The interactive guidance submodule provides learners with more targeted interactive guidance tools and prompt information in a virtual reality environment; The training process monitoring submodule monitors the learners' performance data in the virtual reality learning process in real time and issues early warning for abnormal situations that occur during the learning process.

7. The artificial intelligence knowledge and skills training system based on neural network according to claim 1 is characterized in that: The evaluation feedback module comprises: The multi-dimensional evaluation submodule is used to use a variety of evaluation methods and neural network models to conduct a more comprehensive and integrated evaluation of learners' knowledge and skills from different perspectives in a virtual reality environment, including: direct evaluation based on learning outcomes, comparing learners' completion results in virtual operation tasks with standard answers and best practice cases, and calculating corresponding error indicators; indirect evaluation based on learning behavior, using deep learning models to deeply analyze learners' behavioral characteristics in the virtual reality learning process and explore potential learning problems and advantages; Feedback generation submodule, which is used to generate detailed and personalized feedback reports for learners based on multi-dimensional evaluation results; The path dynamic adjustment submodule is used to feed back the evaluation results to the personalized learning path planning module, triggering dynamic adjustment of the learner's subsequent learning path. Specifically, according to the learner's mastery of the virtual environment, for the virtual operation skills and knowledge points and skill points that have been mastered, the learning progress is accelerated, the task difficulty in the virtual scene is optimized, and more challenging practical links are added; for the weak parts, special virtual training tasks are added, the corresponding basic learning resources are supplemented, and the learning order is adjusted.

8. A method for teaching knowledge and skills of artificial intelligence, characterized in that: The following steps are involved: S1. Data collection and preprocessing: determine the scope and source of the artificial intelligence knowledge and skills data and virtual reality related data to be collected, obtain the original data through the corresponding data collection technology, and use data preprocessing technology to process different types of data; The pre-processed data is labeled and stored according to the knowledge and skills classification system and virtual reality resource classification to build a learning resource database; S2. Neural network model construction and training. According to the specific training objectives and the characteristics of knowledge and skills in the virtual reality environment, the appropriate neural network architecture is selected, and the model is customized and constructed in combination with relevant improvement mechanisms. The constructed model is trained and verified, and the hyperparameters are adjusted according to the preset performance indicators. S3, personalized learning path planning, collect learners' extended personal information, including virtual reality related features and original basic information, learning preferences, historical learning data, and use data analysis and machine learning algorithms to extract learners' feature vectors; Build an AI knowledge and skills map based on collected and preprocessed data, associate knowledge points, skill points with VR learning links, and sort out a more practical and relevant knowledge and skills structure framework; Combining learner feature vectors and updated knowledge and skills maps, an intelligent path planning algorithm is used to generate a unique, personalized learning path for each learner that integrates virtual reality learning. S4, virtual reality scene setting and resource push: according to the virtual reality training scene planned in the personalized learning path, the corresponding virtual learning environment is built using the virtual scene construction submodule, and the built virtual scene is not limited to the entry link, the operation instructions and other related resources are pushed to the learners together with other traditional learning resources according to the learning path planning; S5, training execution: after entering the virtual reality learning environment, learners perform operations and learning according to the planned learning tasks, use the interactive functions in the virtual scene to interact with virtual objects and equipment, complete virtual operation tasks, and explore learning content; The interactive guidance submodule provides learners with real-time operation prompts and guidance information, monitors learners' performance data in real time during the virtual reality learning process, and transmits the performance data to the evaluation feedback module in a timely manner; S6. Evaluation feedback and dynamic adjustment of paths. Through multi-dimensional evaluation methods, comprehensive consideration of learning outcomes, learning behaviors and special dimensions in the virtual reality environment, a neural network model is used to comprehensively evaluate the learners' knowledge and skills. A feedback report is generated based on the evaluation results, and the evaluation results are fed back to the personalized learning path planning module to trigger dynamic adjustment of subsequent learning paths and optimize the learning task arrangement, resource recommendation and difficulty setting in the virtual reality scene.

9. The artificial intelligence knowledge and skills training method based on neural network according to claim 8 is characterized in that: The intelligent path planning algorithm specifically adopts the DQN reinforcement learning algorithm, and uses a multi-layer perceptron neural network MLP to approximate the Q value function, denoted as Q(s,a;θ), where θ is a neural network parameter, the input of the network is the state s, and the output is the Q value estimate of each action a∈A. The mean square error is used as the loss function to measure the difference between the predicted Q value and the target Q value. The formula is: L(θ)=E[(yQ(s,a;θ)) 2 ] Where y is the target Q value, which is calculated based on the target value of the Bellman equation.

10. The artificial intelligence knowledge and skills training method based on neural network according to claim 8, characterized in that: The use of the neural network model to comprehensively evaluate the learner's knowledge and skills mastery includes: Construct an evaluation index system, including knowledge understanding, skill operation, problem-solving ability, knowledge transfer ability, and learning behavior performance; According to different evaluation dimensions and specific evaluation indicator characteristics, select the matching neural network model and architecture; Collect learners' learning outcome data and process data, perform data preprocessing, and label the preprocessed data; The preprocessed labeled data is divided into training set, validation set and test set according to a certain ratio. The training set is used to train the selected neural network model, and the validation set is used to monitor the performance of the model and adjust and optimize the model's hyperparameters according to the validation results. When a learner completes a certain stage of learning tasks, various types of data corresponding to the learner are collected and preprocessed, and then input into the trained neural network models. The model calculates based on the input data and outputs the predicted values ​​of the corresponding evaluation indicators.

Images