198 results about "Training evaluation" patented technology

The invention belongs to the technical field of analog simulation, and particularly discloses an intelligent agent-based dynamic confrontation simulation system, which comprises a data acquisition module, an intelligent analysis module, a decision generation module, an intelligent agent behavior self-adaption module, a training evaluation module and a multi-mode man-machine interaction module, physiological, action, voice and eye movement data of trainees are collected in real time through a multi-modal sensor, and a tactical intention is recognized and a dynamic three-dimensional battlefield situation thermodynamic diagram is generated in combination with virtual battlefield environment parameters; an agent coping strategy is generated based on reinforcement learning and a decision tree, and an agent is driven to carry out real-time confrontation; and performing multi-dimensional quantitative evaluation on the whole training process through a training evaluation module, and performing closed-loop optimization on an agent decision and strategy library based on an evaluation result. The method supports various natural interaction modes such as voice, gestures and eye movement, remarkably improves the fidelity, intelligence and training efficiency of simulation training, and is suitable for the field of military training and tactical drilling.

The invention relates to the technical field of occupational simulation training evaluation, and discloses a system for college student occupational scene simulation training evaluation. The system comprises a task analysis unit, a dynamic behavior modeling unit, a multi-mode interaction management unit, a state evolution tracking unit, a strategy adaptation unit and a comprehensive evaluation generation unit. The system receives the initial scene description and generates a structured task framework, and behavior parameters are configured for student roles; constructing a multi-modal interaction atlas by fusing aligned voice, action and sight line data in real time; identifying a key decision node, and dynamically tracking a team state evolution path; matching and adjusting strategies in a behavior rule base according to the evolution sequence, and outputting an adaptive intervention scheme; and finally generating a multi-dimensional structured evaluation report. According to the system, deep insight and dynamic and objective intelligent evaluation of complex interaction behaviors in the simulation training process are realized.

The invention relates to the technical field of model construction, can be applied to business scenes such as financial science and technology, and discloses a double-track evaluation driven optimization processing method, device and equipment and a medium, and the method comprises the steps: constructing a double-track evaluation system, analyzing training data, generating optimized training data, and building a target model; generating a capability index by utilizing a double-track evaluation system and multi-role agent simulation, and generating an alignment analysis result in combination with reinforcement learning performance; obtaining real service feedback indexes to form a training evaluation index set, and determining an attribution link; and generating training adjustment information based on the attribution link and driving the target model to be retrained to obtain an optimized target model for processing the service input data and outputting a service processing result. According to the method, the evaluation result and the service feedback are linked to realize a training closed loop, so that the capability gap of the model can be identified and corrected, and the compliance, the reliability and the application effect of the model in the actual service are improved.

The invention discloses a practical training evaluation method and system based on multi-modal data fusion. The method comprises the following steps: firstly, constructing a practical training task topology based on a knowledge graph, mapping elements such as knowledge points into nodes, and defining a relationship through directed edges; in response to task starting, dynamically deploying a multi-modal data fusion agent pre-loaded with a targeted model and a rule; in the operation process, the agent synchronously collects and processes original data from the machine vision and sensor network in real time, and a standardized feature flow is generated; then, on-line space-time correlation and reasoning are carried out on the multi-modal features based on a fusion strategy, and a task scene understanding model is dynamically constructed and maintained; after the task is completed, packaging a program code, a report and process abstract data exported by the model to form an enhanced practical training result package; and finally, multi-dimensional automatic comparison is carried out by calling the rule base and the case model, and an evaluation result is generated. According to the invention, intelligent perception and comprehensive evaluation of the whole practical training operation process are realized.

The invention relates to the technical field of rehabilitation training, and particularly provides a knee bending training evaluation method and system based on WiFi signals, and the method comprises the steps: collecting actual channel state information peak values and knee bending action time information corresponding to multiple knee bending actions; for each knee bending action, calculating an actual knee bending angle according to a corresponding actual channel state information peak value, a preset channel state information peak value corresponding to the preset knee bending angle and a preset channel state information initial value, and then calculating an actual knee bending speed according to the actual knee bending angle and knee bending action time information; evaluating knee bending action stability according to all actual knee bending angles; acquiring a knee bending training score according to the knee bending action stability, all the actual knee bending angles and all the actual knee bending speeds; according to the method, non-contact and accurate evaluation of knee bending training without image acquisition can be realized.

The invention discloses an energy consumption optimal processor frequency prediction method and system based on a support vector machine. The method comprises the steps of obtaining original data of performance monitoring counter values, running time, processor energy consumption, processor frequency and processor power of different programs under different parallelism degrees, different scales and different processor frequencies; constructing a data set and a target vector for the original data according to preset self-defined operation time and energy consumption constraint conditions; designing a target function; performing parameter tuning and training evaluation on the support vector machine model; and performing processor frequency prediction on the target program by using the trained support vector machine model to obtain the processor frequency with optimal energy consumption. The method aims at solving the problem that energy consumption is low due to improper processor frequency setting during program running, it is guaranteed that the optimal processor frequency for application program running can be obtained before the application program is loaded through energy consumption optimal processor frequency prediction based on the support vector machine, and the program running energy efficiency is improved.

The invention provides a scoring method and system for weapon equipment training based on virtuality and reality, and belongs to the technical field of weapon equipment training, and the method comprises the steps: obtaining multi-source data in a target training task executed by a student, the multi-source data comprises student operation behavior data, weapon equipment state data, virtual environment data and student physiological monitoring data; based on the multi-source data, calculating to obtain a basic capability score value; analyzing historical training data of the trainees on the same or similar target training tasks to generate ability development factors; acquiring historical norm data matched with the attributes of the trainees and the attributes of the target training tasks, and calculating group performance reference information based on the historical norm data; based on the basic ability score value and the group performance reference information, calculating to obtain a standardized relative performance score; and generating a comprehensive capability evaluation report based on the basic capability score value, the capability development factor and the standardized relative performance score. According to the invention, the scientificity and effectiveness of training evaluation can be improved.

The invention discloses an enterprise training evaluation system and method fusing RAG and an intelligent agent, and belongs to the technical field of training evaluation of artificial intelligence. The system comprises a computing node, a storage unit, a network interaction unit, a document vectorization processing module, a dynamic question setting module, a real-time semantic scoring module, a training effect visualization module and a learning recommendation module. The method comprises the steps of document preprocessing, dynamic question setting, real-time scoring, effect visualization and learning recommendation. Through deep fusion of RAG retrieval and agent decision, personalized question setting, multi-dimensional instant scoring, long-term knowledge state tracking and early warning and personalized learning recommendation based on student knowledge states are realized, and an evaluation-feedback-optimization closed loop is formed. According to the method, the question setting pertinence, the scoring accuracy (up to 92.3%) and the training efficiency (compressing the training time by 40%) are remarkably improved, and the method is suitable for various enterprise training scenes.

The invention discloses a neurosurgical operation virtual simulation training system based on multi-modal data fusion, and relates to the technical field of operation simulation. The data acquisition and preprocessing module acquires data such as images and force feedback, and standardized storage is performed after filtering and registration; the multi-modal data fusion module performs weighted fusion on the data by using an attention algorithm to generate a three-dimensional model; the virtual operation scene construction module generates a high-fidelity environment through ray tracing and a physical engine; the surgical operation interaction simulation module realizes accurate interaction through positioning tracking and force feedback; the training evaluation and feedback module quantifies operation performance and provides guidance; and the system management and iteration module guarantees operation and supports data updating and optimization. According to the method, a high-reality scene is constructed through multi-modal fusion, interaction and multi-dimensional evaluation help skill improvement, dynamic difficulty and personalized paths are adaptive to different users, collaborative training and privacy data collaboration are supported, operation training standardization is promoted, and the risk of errors in an operation is reduced.

The invention discloses a badminton player action posture analysis system and method based on multi-sensor fusion, and relates to the technical field of athletic training auxiliary systems.The system comprises a wearable sensing subsystem, an environment sensing subsystem and a central processing and feedback subsystem; the wearable sensing subsystem is arranged on an inertia measurement unit and a pressure sensing insole of a body to collect movement inertia and plantar pressure data; the environment perception subsystem collects global videos through multiple cameras. The central processing and feedback subsystem performs synchronous processing on multi-source data, adopts a hierarchical fusion algorithm, restrains inertia integral drift by utilizing a plantar contact state, reconstructs a three-dimensional skeleton posture and a motion trail in combination with visual key points and skeleton restraint, drives a digital twin model, compares with a standard motion model, and performs three-dimensional motion control on the three-dimensional skeleton posture and the motion trail. Quantitative evaluation results such as joint angle deviation, time sequence difference and pressure center track are generated, and visual feedback is performed through a display device and an augmented reality terminal for badminton training evaluation and technical deviation correction.

The invention discloses a pelvic floor muscle repairing instrument, a pelvic floor muscle training indication method, pelvic floor muscle training indication equipment and a medium, and the pelvic floor muscle repairing instrument comprises a main body part which is used for being inserted into or placed in a part corresponding to pelvic floor muscle; the first electrode and the second electrode are arranged on the main body part and are used for electrically stimulating pelvic floor muscles and detecting pelvic floor muscle surface electromyogram signals after the pelvic floor muscles are electrically stimulated; the third electrode is attached to the peripheral muscle of the pelvic floor muscle during use and is used for detecting an electromyographic signal of the peripheral muscle of the pelvic floor muscle; the posture detection unit is arranged on the trunk part and is used for detecting a posture time sequence signal after the action of the trunk part is changed; the controller is electrically connected with the first electrode and the second electrode and is used for performing feature extraction on the three signals in the same time sequence, generating a fusion feature vector, determining a pelvic floor muscle training evaluation result and generating a training feedback instruction according to the pelvic floor muscle training evaluation result; the training feedback instruction is used for instructing the user to train.

The invention relates to the technical field of artificial intelligence, and discloses a man-machine cooperation-based old-age care practical training evaluation system, which comprises a sensor, data preprocessing, feature extraction and fusion, clustering analysis and practical training evaluation. According to the system, capability group division is realized under double constraints of an interactive structure and attribute similarity by using a double-graph diffusion-contour weighted K-Means clustering method; in combination with a GNN-Transformer model, through joint asymmetric loss function optimization formed by cross entropy, focus loss and passive asymmetric q-norm loss, cross-modal and time-sequence-dependent deep modeling is realized, and multi-dimensional evaluation indexes such as standardization, proficiency and humanistic care degree are output; finally, the evaluation result is visualized and fed back through a human-computer interaction interface, a perception-analysis-feedback-improvement collaborative closed loop is formed, and the scientificity and guidance of nursing training of the aged are improved.

The embodiment of the invention provides a person training evaluation method based on a virtual reality simulation environment, an edge computing device and a medium. The method comprises the following steps: acquiring initial multi-modal data evaluated by a tested person in a first virtual reality simulation environment; calculating a first evaluation result based on the skill score corresponding to each piece of initial modal data and the weight corresponding to each piece of initial modal data; adjusting training scene parameters of the first virtual reality simulation environment based on the skill scores corresponding to the initial modal data, the first evaluation result and a preset score threshold, and generating a corresponding second virtual reality simulation environment and a training scheme; the second evaluation result is obtained based on the adjusted training scene of the virtual reality simulation environment, that is, the evaluation result is obtained based on the weight and the skill score corresponding to the at least two modal data of the tested person, and the evaluation result is obtained again by adjusting the training scene parameters, so that the comprehensiveness of obtaining the evaluation result of the person is improved.

The invention discloses an fNIRS adaptive feedback emotion cognition cooperative training device and method, and relates to the technical field of emotion disorder cognition training. The device comprises a data acquisition module used for acquiring an original dual-wavelength light intensity signal from a brain; the data processing module is used for performing multi-stage preprocessing on the original dual-wavelength light intensity signal; the neural feedback module is used for calculating a multi-dimensional evaluation index according to the hemoglobin concentration data and adjusting a difficulty level and a feedback threshold value of a training task by using a dynamic threshold value control algorithm; the training module is used for training participants based on the emotion stimulation task and the cognitive training task; and the evaluation module is used for performing training evaluation according to the comparison result of the comprehensive performance score and the feedback threshold. Brain function equipment is closely combined with cognitive training and emotion stimulation tasks, dynamic evaluation and real-time feedback are added for traditional tasks, and through the brain network analysis technology, accurate evaluation of the user training effect is achieved.

The invention relates to the technical field of intelligent training and data analysis, in particular to an intelligent model construction method and system for student training data analysis. Comprising the following steps: constructing an SOP time sequence knowledge graph, obtaining a standard operation specification document and expert-level operation demonstration data, and constructing the SOP time sequence knowledge graph; real-time feature extraction: capturing a multi-modal data stream, and generating a real-time behavior feature vector; state recognition and path deduction: inputting the real-time behavior characteristics into an event classifier to recognize a key event, performing path deduction based on deterministic finite state automaton logic in an SOP time sequence knowledge graph, and determining and outputting a current state pointer; deviation analysis and judgment: extracting a compliance mask based on a current state, and injecting the compliance mask into a decoding process of the attention network; and performing feedback and adjustment to generate a corresponding deviation event signal. Through deep analysis, feedback is more targeted, and the accuracy and fineness of training evaluation are remarkably improved.

The invention relates to the technical field of training evaluation, in particular to an evaluation method and system based on functional actions and physical fitness, and the method comprises the steps: collecting an image when a user does an evaluation action; the action quality score after the user completes the assessment action is calculated, the action and physical fitness of the user are assessed according to the action quality score, and the action quality score is in positive correlation with the action execution score for completing the assessment action and the balance coefficient for completing the assessment action. The balance coefficient reflects the stability of plantar pressure distribution when the user completes the evaluation action. The method not only improves the accuracy and individuation level of posture evaluation, but also can flexibly adjust scoring key points according to different users and action types. Finally, the intelligent mirror has higher adaptability, real-time performance and guidance in scenes such as rehabilitation evaluation and physical fitness detection, and the training effect of the user is remarkably improved.

The invention relates to the technical field of fire safety and simulation training, and discloses a rule-driven fire-fighting real smoke and real fire training simulation system, which comprises a data acquisition and modeling module, a rule driving module, a rendering output module and a training evaluation module. The data acquisition and modeling module is used for acquiring initial fire data and constructing a reference model of smoke diffusion and fire spreading; the rule driving module simulates the non-synchronous evolution process of the fire through a particle simulation algorithm, a time coordination rule and a feedback loop algorithm; the rendering output module is used for integrating the environment rendering data and the multi-factor propulsion mode and generating a training effect evaluation index; and the evaluation optimization module judges whether the simulation precision of the reference model reaches a preset standard or not according to the training effect evaluation index until a simulation result meeting a precision requirement is obtained. According to the invention, the reality and accuracy of fire-fighting simulation training are improved.

To properly evaluate training for a worker who has been trained in the past, while considering practical work.SOLUTION: A training evaluation apparatus 1 includes: a training recognition unit 13 which acquires leaning tendency of a trainee from a history of training that the trainee received; a work recognition unit 14 which acquires work achievement of the trainee from a history of work that the trainee performed; an effect prediction unit 16 which predicts training effect to be obtained by re-training the trainee, by proceeding with a training level of re-training in accordance with the learning tendency of the trainee, the training level according to the work achievement of the trainee being defined as a training level at the beginning of the retraining of the trainee; and a display control unit 18 which displays a result predicted by the effect prediction unit 16 on a screen.SELECTED DRAWING: Figure 1

This application discloses a method and apparatus for constructing a multi-scale adaptive gating MambaPlus network, belonging to the field of artificial intelligence and machine learning technology. The method includes: initializing the network configuration and constructing the basic structure; preprocessing the input data to generate a standard dataset; mapping the input data to the hidden space via an input mapping layer, and extracting backbone features from the Mamba backbone; constructing at least two parallel scale branches in the hidden space to obtain a multi-scale feature set; inputting the backbone features and multi-scale features into an adaptive gating module, dynamically allocating weights and adaptively fusing them through a hierarchical gating mechanism to generate fused features; further enhancing the features through cross-scale attention and feedforward enhancement, and then superimposing the residuals to generate the final discriminative features; finally, completing category prediction and model training evaluation. This application, while retaining the advantages of Mamba's long-range dependency modeling, addresses the problems of insufficient utilization of multi-scale information, poor adaptive feature fusion, and low robustness in complex scenarios.

The invention discloses a multi-sensor motion evaluation method and system for special training of lower limbs. The method comprises the following steps: step S100, collecting data of various sensors of the lower limbs; step S200, based on the collected data, carrying out gait and support state identification; step S300, judging the state of foot varus and valgus; step S400, calculating a foot stability index; step S500, calculating a lower limb exercise ability index; step S600, carrying out special exercise evaluation; and step S700, generating training feedback. Compared with an optical system, the core advantages of the system in training are reflected in that the system can be used for real training instead of a forced performance type test; the device is suitable for long-distance and continuous special training; the technical stability and injury risk can be directly reflected; and the training evaluation cost and the use threshold are obviously reduced.

The invention discloses a power plant equipment three-dimensional interactive training system based on a structured knowledge base, and the system comprises a data processing module which is used for collecting the operation data of power plant equipment, and carrying out the preprocessing of the operation data; the knowledge base construction module is used for constructing equipment object-oriented structured knowledge units and fusing the equipment object-oriented structured knowledge units to form a structured knowledge base; the model construction module is used for constructing a three-dimensional interaction model corresponding to the equipment object and binding a corresponding knowledge unit label; the scene training module is used for constructing a scene training process and embedding operation guide information and risk prompt information; the comparison recording module is used for receiving operation behaviors of a user in the three-dimensional interaction model, recording abnormal operations and positioning the abnormal operations to geometric nodes of corresponding structural components; the demonstration calculation module is used for performing similarity semantic calculation and dynamically presenting an error correction demonstration process; and the training evaluation module is used for generating training evaluation data. According to the invention, the recovery degree, error correction efficiency and evaluation capability of training are improved.

The invention discloses a mine rescue multi-user VR practical training interaction method based on cloud computing, and the method comprises the following steps: collecting multi-source sensing data and task prior information, and constructing a rescue task data set; based on a cloud computing platform, constructing an initial task graph by using an improved Grapher model, and generating a global task graph in combination with a time sequence dependency relationship and a resource constraint relationship between nodes; performing decomposition processing on the global task graph by adopting a hypergraph division algorithm to obtain a plurality of local task graphs; binding the local task graph with the role identity information to generate a corresponding role task sequence; based on global progress management of a cloud computing platform, issuing a role task sequence to a VR training environment, and driving a multi-user interaction process; and carrying out real-time monitoring and data recording on the multi-user interaction process in the cloud computing platform, generating a training evaluation result and outputting the training evaluation result to the practical training management end. According to the invention, the dynamic adaptation capability of task scheduling and the response efficiency of multi-user collaborative practical training are improved.

The invention provides a small sample unbalanced industrial Internet of Things detection data amplification method and device, and the method comprises the steps: carrying out the preprocessing of industrial Internet of Things original data, obtaining an industrial Internet of Things data set, distributing an independent generator for each data category, and constructing a network architecture of a discriminator and classifier sharing feature extraction layer; performing adversarial training fusing relative pairing adversarial loss, a dual gradient penalty mechanism and generator feature decoupling loss, and generating sample optimization, including screening samples consistent with a real label by using a discriminator, and removing samples deviating from real distribution based on a feature similarity threshold; and inputting the optimized sample into an intrusion detection model, and carrying out training evaluation. Through a dual optimization strategy, the diversity and quality of generated samples are effectively improved, and the model training stability is enhanced; through a lightweight parallel architecture and a dynamic adversarial training strategy, the model robustness is improved while the generation efficiency is ensured, and the problem of imbalance of intrusion detection data in an industrial Internet of Things environment is solved.

The invention provides a training plan auxiliary generation method and device, electronic equipment and a storage medium, and the method comprises the steps: determining target training plan information according to at least one of all levels of annual training plan information, all levels of monthly training plan information, task on-duty plan information and training plan related information; wherein each stage of annual training plan information comprises at least one of a training subject, training time, a training evaluation standard, a training stage and training posts; the monthly training plan information at each level comprises monthly training plan information of at least one of a training subject, training time, a training evaluation standard, a training stage and training posts. According to the method, the target training plan information comprises at least one of all levels of annual training plan information, all levels of monthly training plan information, task on-duty plan information and training plan related information, deep integration is carried out on multi-dimensional data, the data utilization rate is improved, and therefore the accuracy of training plan generation is improved, and the management efficiency is improved.

This application discloses a training method, device, electronic device, and storage medium based on a large language model. The method includes pre-configuring training modes, which at least include open-ended training and restricted training. The open-ended training and restricted training can run independently or be switched on demand. Responding to the input of the user to be trained, the method selects the training mode and simultaneously conducts training based on a fallback process, outputting the final training evaluation result. The fallback process ensures the continuity of the large language model training process. This application provides a training method including open-ended training, process tree restricted training, and a fallback process. The open-ended training integrates a knowledge base and a large evaluation model; the process tree restricted training includes configurable nodes and eight core functions; and the fallback process has model monitoring, self-interruption, and dialogue switching capabilities, achieving integrated training with flexible interaction, standardized control, and stable operation.

The invention discloses a satellite orbit single-point state forecasting method based on deep learning. An orbit precision ephemeris is used as training data, a six-dimensional position speed state at a current moment is used as input, a state at a next moment is used as a target value, a prediction network in which LSTM and a one-dimensional CNN are connected in series is constructed, and a CBAM attention module is introduced to realize adaptive weighting of channel and time sequence features. And designing a weighted loss function, carrying out weighted fusion on a mean square error of the model to a real ephemeris, an HPOP equal-step prediction result and a residual mean square error output by the model, and improving prediction rationality through physical constraints. And training and evaluating through the training set, the verification set and the test set, and outputting a predicted ephemeris. According to the method, a historical sequence or TLE is not needed, high-precision and rapid forecasting can be achieved only through a single-point state, and unmodeled perturbative force is implicitly compensated.

The invention discloses a railway modularized intelligent emergency training system based on a TRIZ theory, and belongs to the technical field of railway emergency training. The system is composed of a data acquisition module, a scene simulation module, an intelligent interaction module, a practical training evaluation module and a modular management unit. The data acquisition module acquires railway operation data and historical emergency case data in real time; the scene simulation module constructs a vivid and parameter-adjustable practical training scene by using VR, AR and other technologies; the intelligent interaction module realizes natural interaction and real-time feedback between trainees and scenes; the practical training evaluation module scientifically evaluates the practical training performance of the trainee from multiple dimensions; and the modular management unit flexibly manages and dispatches each functional module. According to the invention, innovative design is carried out based on the TRIZ theory, the authenticity of practical training is improved, the flexibility and expansibility of the system are enhanced, the evaluation scientificity of the practical training effect is improved, and the emergency processing capability of railway workers can be effectively improved.

The invention provides a model training method combining parameter adjustment record screening and a large model and a related product. According to one specific mode of the method, simplified historical parameter adjustment records are obtained through screening from historical parameter adjustment records. And analyzing a change rule of the simplified historical parameter adjustment record by using the large model, outputting a next round of hyper-parameter configuration, and performing a next round of training by using the next round of hyper-parameter configuration. Therefore, the method can reduce the number of tokens for large model analysis, improve the speed of generating the next-round hyper-parameter configuration, enable the next-round hyper-parameter configuration to be explained, and determine the feature distance between two hyper-parameter configurations through the feature sensitivity of each hyper-parameter based on the target machine learning model to the training evaluation result. And parameter adjustment records are screened based on the characteristic distance between the hyper-parameter configurations, so that high-value simplified parameter adjustment records can be obtained. The lower round of hyper-parameter configuration obtained by utilizing a large model and based on a more representative high-value simplified parameter adjustment record is easier to understand and more stable.

The invention discloses a high-voltage electrician practical training evaluation method, device and system based on AI visual identification, and relates to the field of intelligent identification, and the method comprises the steps: responding to a practical training signal, and determining a target practical training host in a plurality of practical training hosts; acquiring a real-time video stream of at least one path of high-voltage electrician practical training scene; performing AI visual identification on the real-time video stream to determine physical states of scoring elements and environmental elements in the practical training scene; performing data analysis on the physical states of the scoring elements and the environmental elements to generate comprehensive evaluation data so as to evaluate the safety wearing compliance of the operator and the correctness of the operation steps, and outputting operation feedback information to the target practical training host; and in response to the received operation signal fed back by the target practical training host, generating a practical training evaluation report. The operation details of the operator in the practical training process can be deeply analyzed, the accuracy of the practical training evaluation result is improved, accurate personalized guidance is conveniently provided for the practical training operator, and the practical training effect is improved.

This invention discloses a method and system for parallel training, evaluation, and control based on a digital twin using an equipment bus. It unifies the time-stamping and state-space mapping of multi-bus observation data, and solves for a unified state estimate based on quality indicators, fusion residuals, and cross-bus physical consistency constraints. This drives the state evolution of the digital twin, executing the Lyapunov descent admission condition under synchronization error and end-to-end delay constraints. Further, multi-scenario parallel simulations are conducted to obtain an upper bound on the risk probability and construct a training evaluation function. When the evaluation is below a threshold or the risk exceeds the limit, only the affected control components are subject to safe feasible region projection correction, and closed-loop control commands are output. This invention improves state estimation accuracy, synchronization stability, risk prediction capability, and training control safety.