Virtual reality-based emergency rescue drill method and system

Abstract

This application discloses a virtual reality-based emergency rescue drill method and system, relating to the field of emergency rescue simulation drills. The method includes responding to drill task requests, loading standardized operating procedures, generating a virtual environment with dynamic interactive elements based on key operating nodes; collecting multi-dimensional drill interaction data from drill participants in real time; calling a compliance evaluation set to perform real-time matching and verification of the multi-dimensional drill interaction data, generating matching results containing dynamic compliance scores and driving the virtual environment to produce dynamic responses; and summarizing the matching results and situational response results to generate a quantitative evaluation report. This application provides a realistic, efficient, and scientific evaluation method for emergency rescue drills, effectively improving the practical response and decision-making capabilities of drill participants and the effectiveness of rescue efforts.

Description

Technical Field

[0001] This application relates to the field of emergency rescue simulation exercises, and in particular to an emergency rescue exercise method and system based on virtual reality. Background Technology

[0002] With the development of the information age, technologies such as big data have been gradually applied in the field of disease control, promoting the development of disease control work towards intelligence and precision. However, in terms of training disease control personnel, the traditional "teacher lectures, trainees listen" model and on-site demonstration teaching are insufficient to meet the repeated training needs for operating complex equipment and high-risk operations. Large-scale on-site emergency drills are not only resource-intensive and costly, but also have limited scenarios that cannot cover all the emergency situations that rescue personnel may encounter in real life, thus limiting the effectiveness of training.

[0003] In recent years, the application of virtual reality technology has provided a new technological approach for emergency rescue training. Existing VR emergency training systems can simulate rescue scenarios to a certain extent and support trainees in practicing basic operations. However, these systems still have significant technical shortcomings in terms of exercise evaluation.

[0004] Existing assessment methods generally employ a "rigid scoring" mechanism, which involves a simple "right / wrong" binary judgment of trainees' operations based on a pre-set list of procedures. This scoring method has the following prominent problems: First, the scoring criteria are rigid, failing to distinguish between critical errors and non-critical deviations, resulting in assessment results that do not accurately reflect trainees' actual handling abilities; second, it fails to identify the nature and severity of errors, failing to reflect the impact of different errors on achieving the task objectives; third, it lacks process-oriented diagnosis and personalized feedback, making it impossible to identify trainees' error patterns and provide targeted guidance during drills.

[0005] The aforementioned technical limitations make it difficult for the system to provide diagnostic feedback with educational value. Therefore, there is an urgent need for a VR emergency drill method that can achieve dynamic, intelligent, and multi-dimensional assessment to overcome the drawbacks of the overly rigid scoring system in existing technologies. Summary of the Invention

[0006] Firstly, this application provides an emergency rescue drill method based on virtual reality, employing the technical solution described below: A virtual reality-based emergency rescue drill method includes the following steps: In response to the exercise task request, the corresponding standardized operation process is loaded, and a virtual environment containing dynamic interactive elements is generated based on the key operation nodes of the standardized operation process; the dynamic interactive elements are semantically associated with the key operation nodes. The system collects multi-dimensional interactive data of the drills in real time, which is generated by the drill participants performing virtual rescue operations in the virtual environment. The multi-dimensional interactive data includes at least operational behavior data, semantic interaction data, and scene state change data. For each key operation node, the compliance assessment set corresponding to the key operation node is invoked to perform real-time matching and verification of the multi-dimensional exercise interaction data. Based on the obtained verification results, a matching result containing dynamic compliance scores is generated, and the virtual environment is driven to generate a dynamic response based on the matching results. The matching results of each key operation node and the situational response results of the dynamic response are summarized to generate a quantitative evaluation report for the trainees.

[0007] By adopting the above technical solutions, a complete closed-loop VR training method was constructed, encompassing scene generation, data collection, intelligent response, and quantitative evaluation. This method solves the problems of process disconnect, delayed feedback, and subjective evaluation in traditional training. Through semantic association, this method ensures deep integration of the virtual environment with professional processes. Real-time analysis and dynamic response create an immediate teaching intervention loop, ultimately outputting an objective evaluation report based on full-process data. This achieves a high degree of simulation, intelligence, and standardization in emergency rescue skills training.

[0008] Preferably, the step of responding to the exercise task request, loading the corresponding standardized operation process, and generating a virtual environment containing dynamic interactive elements based on the key operation nodes of the standardized operation process specifically includes the following steps: Receive and respond to pre-selected exercise task requests, and load the corresponding standardized operation process from a pre-set standardized process database according to the exercise task request; The standardized operating procedure is analyzed, and preset key operating nodes are extracted from the standardized operating procedure; According to the exercise task request, a corresponding virtual environment is loaded from the virtual scenario database. The virtual environment contains multiple dynamic interactive elements, including at least one of virtual characters, virtual facilities, and virtual environment objects. Configure at least one state variable to represent the current state and a semantic label to identify the function category for each dynamic interactive element, and establish a logical relationship between the dynamic interactive element and the corresponding key operation node based on the semantic label.

[0009] By adopting the above technical solutions, the standardized operating procedures for the corresponding exercise tasks can be accurately loaded, key operating nodes can be identified, and a virtual environment containing a variety of dynamic interactive elements can be constructed to provide a suitable scenario for the exercise. Configuring state variables and semantic labels for dynamic interactive elements and establishing logical relationships with key operating nodes can make the virtual environment more interactive and relevant, facilitating accurate monitoring and evaluation of the exercise personnel's operations in the future.

[0010] Preferably, the real-time acquisition of multi-dimensional interactive data generated by personnel performing virtual rescue operations in the virtual environment specifically includes the following steps: Listen to and record the interaction events between the trainees and the dynamic interactive elements in the virtual environment, and generate operational behavior data; The virtual character in the virtual environment is driven by an artificial intelligence model. The virtual character is pre-set with a background knowledge base and behavioral logic evaluation items corresponding to the exercise task. The system receives the voice text from the trainees and inputs it into the artificial intelligence model. Based on the artificial intelligence model, the system performs real-time analysis of the voice text, combines the background knowledge base of the virtual character and the behavioral logic evaluation items, generates voice feedback corresponding to the voice text, and records the semantic interaction data. Monitor and record the changes in the values ​​of the state variables of each dynamic interactive element in the virtual environment. When the change of any state variable reaches a preset threshold or is triggered by a preset interactive event, generate scene state transition data.

[0011] By adopting the above technical solutions, the interaction events between the trainees and dynamic interactive elements are monitored and recorded to generate operational behavior data, which allows for a comprehensive understanding of the trainees' operational status. Calling on artificial intelligence models to drive virtual characters and generate voice feedback for the trainees' speech makes the exercise more interactive and realistic, and also allows for the acquisition of the trainees' semantic communication information. Monitoring and recording changes in the state variables of dynamic interactive elements allows for real-time understanding of changes in the virtual environment's state, providing multi-dimensional exercise interaction data for subsequent evaluation, thus achieving a more dynamic, intelligent, and multi-dimensional assessment.

[0012] Preferably, the step of calling the compliance assessment set corresponding to each key operation node to perform real-time matching and verification of the multi-dimensional exercise interaction data specifically includes the following steps: The compliance evaluation set corresponding to the current key operation node is invoked to perform real-time feature matching on the operation behavior data and the semantic interaction data, generating violation event feature information and basic violation value. The compliance evaluation set includes time-series evaluation items, object evaluation items and quality evaluation items. Extract the node target achievement vectors corresponding to all the key operation nodes from the preset historical exercise database, and construct a set of compliance status feature vectors for each compliance evaluation item; Perform statistical correlation operations on the compliance status feature vector set and the node target achievement vector in the vector space to solve the correlation contribution vector of each compliance evaluation item feature vector relative to the node target achievement vector. Based on the normalization operation result of the correlation contribution vector, calculate the constraint strength coefficient of each compliance evaluation item. When any current operation event is detected to violate the compliance evaluation item, the corresponding violation event is mapped to a violation feature scalar, and a scalar product operation is performed with the constraint strength coefficient corresponding to the currently violated compliance evaluation item to obtain the current basic violation value.

[0013] By adopting the above technical solutions, operational behavior data and semantic interaction data are matched and verified in real time to generate violation events and basic violation values. This enables timely detection of violations in the operations of trainees. The achievement degree of node objectives is extracted from the historical exercise database, and then the contribution of each evaluation item to the achievement of the node objective is calculated through vector space statistical correlation to obtain the constraint strength coefficient. The importance of evaluation items is automatically discovered from the data, so that the evaluation weight is based on objective laws rather than subjective presets. This allows for a more accurate measurement of the compliance of trainees' operations, providing a reliable basis for the subsequent generation of dynamic compliance scores and evaluation reports. This overcomes rigid scoring systems and achieves a more intelligent evaluation.

[0014] Preferably, the method further includes the following steps: When an operation event that does not belong to the expected event of the current key operation node is detected, the operation event is classified; if the operation event belongs to the expected event of the subsequent key operation node, the operation event is recorded as a leading event and associated with the corresponding key operation node; otherwise, it is recorded as an irrelevant event and points are deducted immediately. Based on the real-time monitored scenario state change data, the target achievement status of the current key operation node is determined, and the current achievable optimal target is determined in combination with the predefined target state hierarchy. Based on the current achievable optimal target, the dynamic weight coefficients of each compliance evaluation item in the compliance evaluation set are dynamically adjusted. When the key operation node corresponding to the advanced event is activated, the advanced event deduction value of the advanced event is obtained by combining the first timestamp of the advanced event and the second timestamp of the key operation node being activated.

[0015] By adopting the above technical solutions, operational events that are not expected events of the current critical operation node can be classified and processed, accurately distinguishing between premature events and irrelevant events and recording and deducting points accordingly. The delayed evaluation mechanism for premature events avoids the rigid process of deducting points for every mistake. It can determine the current achievable optimal goal based on scenario state change data and predefined target state levels, and dynamically adjust the weight coefficients of compliance evaluation items, so that the evaluation focus can intelligently shift according to the consequences of the student's mistakes, reflecting a teaching orientation rather than a punishment orientation.

[0016] Preferably, the step of dynamically adjusting the dynamic weight coefficients of each compliance evaluation item in the compliance evaluation set based on the currently achievable optimal goal specifically includes the following steps: Each target level in the target state hierarchy is decomposed into a set of key states that must be achieved. The set of key states that the current achievable optimal target is determined. The target relevance coefficient of each compliance evaluation item is obtained by combining the state variables of all dynamic interactive elements associated with each compliance evaluation item. The target relevance coefficient of each compliance evaluation item is calculated based on the weight ratio of its corresponding state variable in the set of key states. Based on the target relevance coefficient and the constraint strength coefficient of each compliance evaluation item, the dynamic weight coefficient of each compliance evaluation item is generated by normalized weighted calculation.

[0017] By adopting the above technical solution, the goal is decomposed into a set of key states that must be achieved and weight coefficients are assigned, thus making the logic of weight adjustment transparent. By decomposing the goal into a set of states that must be achieved and calculating the relevance of the evaluation rules based on the weight ratio of each state, the evaluation focus can be dynamically adjusted according to the currently achievable optimal goal, realizing goal-driven intelligent evaluation. This provides accurate weight input for subsequent weighted scoring and provides an interpretable mathematical basis for why this rule is important now, thereby enhancing the credibility of the system.

[0018] Preferably, generating matching results that include dynamic compliance scores specifically includes the following steps: Using the dynamic weighting coefficient, all the obtained basic violation values ​​are weighted to obtain a weighted violation value. The total violation value of the node is obtained by accumulating all the weighted violation values, the instant deductions, and the advanced event deductions within the current key operation node. By combining the full score of the node corresponding to the current achievable optimal goal, a dynamic compliance score for the key operation node is calculated and generated. The matching result is output by summarizing the dynamic compliance score, the records of the advanced events, and the violation events.

[0019] By adopting the above technical solution, taking into account basic violations, immediate deductions, and the impact of premature events, and referring to the dynamically adjusted full score, the final score can accurately reflect how well the trainees have made corrections given the existing errors. This enables differentiated and contextualized final scores for trainees, overcoming the drawbacks of overly rigid scoring systems in existing technologies.

[0020] Preferably, the step of driving the virtual environment to generate a dynamic response based on the matching result specifically includes the following steps: The matching results acquired in real time are encoded, and the violation events in the matching results are used to generate a multi-dimensional error feature vector sequence according to the evaluation item type and the severity of the violation; The multi-dimensional error feature vector sequence is analyzed to calculate the frequency and cumulative intensity of each error dimension within a preset time window; the error dimensions corresponding to the frequency and cumulative intensity exceeding a preset threshold are marked as the current high-frequency error dimensions. Based on the statistical values ​​corresponding to the current high-frequency error dimensions, template matching is performed in the preset response evaluation item templates. Each template defines a set of matching conditions and a parameter adjustment strategy. Based on the severity of the violation and the statistical value, the corresponding parameter adjustment strategy is executed on the successfully matched template, an environmental parameter adjustment instruction is calculated and generated, and the corresponding environmental parameters are adjusted in real time according to the environmental parameter adjustment instruction. The adjustment content, trigger template and execution result are recorded as the situation response result.

[0021] By adopting the above technical solutions, the matching results are encoded to clearly present the judgment item type and severity of the violation; focusing on common mistakes made by trainees, the statistical values ​​of high-frequency errors are used for template matching, enabling the virtual environment to accurately adapt to the corresponding judgment items based on the error situation; environmental parameter adjustment instructions are generated based on the matching results and judgment items, and environmental parameters are adjusted in real time, which can dynamically change the virtual environment and realize personalized guidance "tailored to individual needs". At the same time, recording the situation response results helps to comprehensively evaluate the exercise situation, realizing intelligent environmental feedback based on error patterns, and accurately reflecting the actual handling ability of trainees.

[0022] Preferably, the process of generating the parameter adjustment strategy includes: When the input feature of the successfully matched template is the occurrence frequency, a linear mapping is performed based on the occurrence frequency to calculate the parameter adjustment amount used to adjust the time pressure feedback parameter or visual saliency parameter in the virtual environment. When the input feature of the successfully matched template is the cumulative intensity, proportional clamping mapping is performed based on the cumulative intensity to calculate the parameter adjustment amount used to adjust the visual saliency parameter or operational accuracy requirement parameter in the virtual environment. The parameter adjustment amounts are all based on the weight coefficients of the current key operation nodes obtained from historical exercise data statistics, which are used as multiplicative mapping factors.

[0023] By adopting the above technical solutions, through linear mapping and proportional clamping mapping, error characteristics are transformed into specific executable environmental parameter adjustment quantities, making system feedback quantifiable, traceable, and optimizable. This enables the quantitative generation of response strategies and personalized feedback based on errors. Furthermore, by using historical data weights as mapping factors, the response intensity is linked to the importance of nodes, forming a complete intelligent closed loop that includes perception, analysis, and response.

[0024] Secondly, this application provides an emergency rescue drill system based on virtual reality, which adopts the following technical solution: An emergency rescue drill system based on virtual reality includes the following modules: The virtual scene construction module is used to respond to the exercise task request, load the corresponding standardized operation process, and generate a virtual environment containing dynamic interactive elements based on the key operation nodes of the standardized operation process; the dynamic interactive elements are semantically associated with the key operation nodes. The multidimensional data acquisition module is used to collect multidimensional exercise interaction data generated by the participants performing virtual rescue operations in the virtual environment in real time. The multidimensional exercise interaction data includes at least operation behavior data, semantic interaction data, and scene state change data. The intelligent dynamic evolution module is used to call the compliance evaluation set corresponding to the key operation node for each key operation node to perform real-time matching and verification of the multi-dimensional exercise interaction data, generate a matching result containing dynamic compliance score based on the obtained verification result, and drive the virtual environment to generate a dynamic response based on the matching result. The exercise result output module is used to summarize the matching results of each key operation node and record the situational response results of the dynamic response, and generate a quantitative evaluation report for the exercise personnel.

[0025] In summary, this application contains at least one of the following beneficial effects: (1) This application loads standardized operating procedures in response to exercise task requests and generates a virtual environment containing dynamic interactive elements based on key operation nodes. This makes up for the problem that traditional training methods cannot meet the repeated training needs of complex equipment and high-risk operations, as well as the problem that existing VR emergency training systems have a single scenario. It can provide trainees with richer and more realistic exercise scenarios. (2) This application collects multi-dimensional exercise interaction data in real time and calls the compliance evaluation set for real-time matching and verification. Based on the verification results, it generates matching results containing dynamic compliance scores, which can overcome the drawbacks of rigid scoring standards in existing technologies, distinguish between critical errors and non-critical deviations, and identify the nature and severity of errors. (3) This application drives the virtual environment to generate dynamic responses based on the matching results, and summarizes the matching results and situational response results to generate a quantitative assessment report, which can provide process diagnosis and personalized feedback, realize dynamic, intelligent and multi-dimensional assessment, and provide diagnostic feedback with teaching value. Attached Figure Description

[0026] Figure 1 This is a flowchart of the method in this embodiment; Figure 2 This is a flowchart of the dynamic weight calculation method in this embodiment; Figure 3 This is a structural diagram of the system in this embodiment. Detailed Implementation

[0027] This application provides a method and system for emergency rescue drills based on virtual reality. To make the purpose, technical solution and advantages of this application clearer, the implementation method of this application will be further described in detail below.

[0028] The following describes in further detail an embodiment of an emergency rescue drill method based on virtual reality, in conjunction with the accompanying drawings.

[0029] This application discloses an emergency rescue drill method based on virtual reality, the process of which is as follows: Figure 1 As shown, it includes the following steps: S1. In response to the exercise task request, load the corresponding standardized operation process, and generate a virtual environment containing dynamic interactive elements based on the key operation nodes of the standardized operation process. The dynamic interactive elements are semantically associated with the key operation nodes. Specifically, this includes the following steps: S11. Before the exercise begins, the system creates case study materials for the simulation exercise, such as a case study of pneumonia of unknown cause. After the case study is successfully created, key operation nodes are added.

[0030] Commanders select the participants to be involved in the simulation exercise on the backend management terminal, add them to the corresponding node, assign roles to the participants, and set the simulation operations that the participants need to complete and the corresponding scores for those operations.

[0031] S12. Receive and respond to the pre-selected exercise task request, which includes the selected exercise task code. Load the corresponding standardized operation process from the pre-set standardized process database according to the exercise task code.

[0032] The standardized process database stores various strict technical standards used in disease control work, such as the "Disinfection Technical Specifications", "Personal Protective Equipment Donning and Doffing Procedures", and "Laboratory Biosafety Operating Procedures".

[0033] S13. Analyze the standardized operating procedures and extract the preset key operating nodes and the necessary virtual facilities, operating objectives and compliance standards corresponding to each key operating node from the standardized operating procedures.

[0034] S14. Load the corresponding virtual environment from the virtual scenario database according to the exercise task code. The virtual environment contains multiple dynamic interactive elements that are logically associated with the standardized operation process. The dynamic interactive elements include at least one of virtual characters, virtual facilities, and virtual environment objects.

[0035] S15. Configure at least one state variable to represent the current state and one semantic label to identify the function category for each dynamic interactive element, such as: sampling target, wounded, disinfection equipment, etc. The wounded refer to the virtual NPC character configured in the virtual environment.

[0036] Logical relationships between dynamic interactive elements and their corresponding key operation nodes are established based on semantic tags. For example, if object A is associated with process node X, then the action of operating object A should occur at node X.

[0037] For example, in a specific implementation, the state variables include vital signs parameters, operability status, and disinfection status, while the key operation nodes include injury assessment nodes, facility usage nodes, and disinfection treatment nodes.

[0038] The dynamic interactive elements are semantically associated with key operation nodes. Specifically, the semantic association includes: associating the vital signs parameters of virtual wounded patients with injury assessment nodes, associating the operable status of virtual instruments with instrument usage nodes, and associating the contamination / cleanliness status of virtual environmental objects with disinfection and treatment nodes.

[0039] This allows the system to automatically identify the target to be operated on and monitor its state changes based on this relationship during process execution. For example, after semantic association, all objects in the scene whose semantics are marked as disinfectant products, such as hand sanitizer and alcohol-based hand gel, have their state variable "whether it is used" automatically associated with the "hand disinfection" node. Whenever the process reaches this node, it will automatically check whether the used state of all associated objects is true. If it is false, it means that disinfection has not been carried out, which triggers a violation judgment and subsequent evolution.

[0040] S2. Real-time collection of multi-dimensional exercise interaction data generated by personnel performing virtual rescue operations in the virtual environment. The multi-dimensional exercise interaction data includes at least operational behavior data, semantic interaction data, and scene state transition data, specifically including the following steps: S21. Listen to and record the interaction events between the trainees and the dynamic interactive elements in the virtual environment, and generate operation behavior data including node identifiers, target object identifiers, and operation timestamps.

[0041] This step relies on a VR interaction event listener. When the trainee interacts with any virtual object with interactive components (grabbing, pressing, placing, etc.), the listener captures the interaction event. Based on semantic tags, it identifies the dynamic interactive elements involved in the interaction event and their associated key operation nodes, generating operation behavior data.

[0042] S22. The AI ​​model drives the virtual characters in the virtual environment. This AI model employs TTS, a large language model (LLM), and a large speech model, enabling it to perform natural language understanding and generation. It can comprehend complex semantic information and generate responses that conform to contextual logic. The virtual characters are pre-set with a background knowledge base and behavioral logic evaluation items corresponding to the training tasks.

[0043] As mentioned earlier, virtual characters are NPCs configured in a virtual scene. Their semantic tags can be wounded or residents, and the background knowledge base and behavioral logic evaluation items corresponding to different semantic tags are also different. Generally speaking, dialogues with resident characters contain epidemiological investigation semantics, while dialogues with wounded characters contain comforting semantics and inquiries about injuries.

[0044] S23. Receive the voice text from the trainees and input it into the artificial intelligence model. The AI ​​model analyzes the voice text in real time, combines it with the virtual character's background knowledge base and behavioral logic evaluation items, generates voice feedback corresponding to the voice text, and records the semantic interaction data. Semantic interaction data includes the dialogue behavior data of both parties, including the speaker's intent, extracted business entities, and sentiment labels. For example, the semantic interaction data for the wounded character is: {Intent: Inquire about symptoms, Entities: [Fever, Cough]}.

[0045] This process integrates a speech recognition engine, providing high-precision speech-to-text capabilities. It employs the Volcano Engine's Doubao speech large model synthesis technology (TTS, Text-to-Speech) to generate voices that closely resemble human speech, making the virtual scene closely resemble the real operation scene and ensuring the accuracy and response speed of voice interaction.

[0046] S24. Monitor and record the numerical changes of state variables of each dynamic interactive element in the virtual environment. When the change of any state variable reaches a preset threshold or is triggered by a preset interactive event, generate scene state transition data, which includes the state variable identifier, the value before the change, the value after the change, and the reason for the change. Each record of scene state transition data is associated with the event identifier of the specific operation behavior data or semantic interaction data that triggered the change. This allows the system to not only know the reason for the change, but also which operation caused it.

[0047] Based on a unified high-precision timestamp, data packets generated within the same time period and belonging to the same logical context are aligned and associated, encapsulated into standardized multimodal time-series data packets, and sent to the subsequent dynamic response engine and evaluation analysis engine.

[0048] S3. During the execution of the virtual rescue operation, for each key operation node, the compliance assessment set corresponding to the key operation node is invoked to perform real-time matching and verification of the multi-dimensional exercise interaction data. Based on the verification results, a matching result containing a dynamic compliance score is generated, and the virtual environment is driven to generate a dynamic response based on the matching result. The specific steps include the following: S31. Call the compliance evaluation set corresponding to the current key operation node, perform real-time feature matching on the operation behavior data and semantic interaction data, and generate violation event feature information and basic violation value. The compliance evaluation set includes time-series evaluation items, object evaluation items and quality evaluation items.

[0049] In one specific implementation, a sequence of expected operation events is maintained for the current critical operation node based on the timing evaluation criteria.

[0050] Specifically, the timing evaluation item defines the necessary execution order of key operation nodes; the object evaluation item defines the set of semantic labels of the target objects that each node is allowed to operate on; and the quality evaluation item defines the quantification threshold range of parameters such as spatial position, force, and angle of the operation.

[0051] S32. Using a single key operational node as the analysis unit, and based on sample data from the historical exercise database, construct a feature vector system with unified dimensions and standardized values. Extract the node target achievement vectors corresponding to all key operational nodes, and construct a compliance status feature vector set for each compliance evaluation item. The compliance status feature vector is a binary vector encoding representing the compliance status, and the node target achievement vector, as the target dependent variable vector for vector operations, is also a binary vector encoding.

[0052] Based on causal discovery algorithms (such as information gain, Bayesian causal inference, Granger causality test, etc.), statistical correlation operations are performed on the compliance status feature vector set and the node target achievement vector in the vector space to solve the correlation contribution of each compliance evaluation item to the node target achievement, forming a correlation contribution vector.

[0053] Perform L1 norm normalization on the correlation contribution vector to map each correlation contribution scalar to the [0,1] interval, eliminate the influence of dimensions, and obtain the constraint strength coefficient vector. The constraint strength coefficient is positively correlated with the correlation contribution of the corresponding compliance evaluation item to the achievement of the node target.

[0054] The constraint strength coefficient quantifies the degree to which compliance with the compliance criterion contributes to achieving the target of this node, or the probability of failure due to violation of the criterion.

[0055] Analysis of a large number of complete exercise records obtained from the historical exercise database, taking the "hand hygiene" node as an example, each record includes whether the "rubbing for ≥15 seconds" (quality assessment item) was followed, whether the "wetting hands first" (timing assessment item) was followed, whether the "use of disinfectant" (object assessment item) was followed, and the final disinfection effect (node ​​target achievement degree). In this embodiment, compliance status and node target achievement degree are both binary variables, with compliance status being compliance or violation; node target achievement degree being success or failure.

[0056] In a specific implementation, statistical association calculations can be based on causal discovery algorithms, selecting methods such as probability difference, information gain, or the absolute value of regression coefficients. Taking information gain as an example: The information gain formula is: IG = H(Y) - H(Y|X); Here, H(Y) is the entropy (uncertainty) of the result Y, and H(Y|X) is the conditional entropy of the result Y after knowing whether the criterion X complies.

[0057] First, calculate the overall success rate, let the success rate be p, then... H(Y)=-[p·log2(p)+(1-p)·log2(1-p)]; Calculate the success rates of the compliance group and the violation group separately to obtain their respective conditional entropies. Take the weighted average to obtain H(Y|X) and calculate the information gain IG. The larger the information gain, the more important the compliance judgment item is to the prediction success and the stronger the constraint.

[0058] To make it easier to understand, let's take the "hand hygiene" node as an example. Assume that there are 200 "hand disinfection" records in the database, with an overall success rate of 75%. For the quality assessment item (rubbing for ≥15 seconds): the compliance group had 120 attempts, 110 of which were successful, with a success rate of 91.7%; the violation group had 80 attempts, 40 of which were successful, with a success rate of 50%. The information gain was calculated to be 0.42. For the timing evaluation item (wet hands first): 150 times in the compliance group, 113 times were successful, with a success rate of 75.3%; 50 times in the violation group, 37 times were successful, with a success rate of 74%. The information gain is calculated to be 0.01 (negligible). For the evaluation item (use of disinfectant): the compliance group had 180 attempts, 150 of which were successful, with a success rate of 83.3%; the violation group had 20 attempts, 0 of which were successful, with a success rate of 0%. The information gain was calculated to be 0.58. The information gain mentioned above is the correlation contribution scalar. After performing normalization, the respective constraint strength coefficients are calculated.

[0059] S33. When any current operation event is detected to violate a compliance assessment item, the corresponding violation event is mapped to a violation feature scalar, and a scalar product operation is performed with the constraint strength coefficient corresponding to the currently violated compliance assessment item to obtain the current basic violation value. This achieves quantitative coupling between constraint strength and violation degree, ensuring the scientific nature and distinguishability of the basic violation value.

[0060] Specifically, the detection process is divided into verification of the operation event regarding timing compliance, object compliance, and quality compliance: timing compliance verification matches the event with the head of the expected operation event sequence to determine whether there is a skip or reversal of the order; object compliance verification verifies whether the virtual object operated by the event belongs to the set of operation objects permitted by the current node; quality compliance verification verifies whether the spatial, mechanical, and duration parameters related to the event meet the quality threshold range preset by the current node.

[0061] Based on the above verification results, if a violation of a judgment item occurs, the basic violation value of the current violation event is obtained by calculating the product between the constraint strength coefficient corresponding to the judgment item and the violation severity factor. First, the violation severity factor of the event is calculated. The calculation method of the violation severity factor is different depending on the type of judgment item. The timing evaluation item uses the operation sequence in the standard operating procedure as the standard. The sequence is either correct or incorrect. If the sequence is incorrect, the violation severity factor is 1. The object evaluation item also uses the operation sequence in the standard operating procedure as the standard. If the object is different, the violation severity factor is 1. For quality assessment items, a violation severity factor is calculated based on the degree of deviation between the actual value and the target value of the operational event to determine the severity of the violation.

[0062] For example, if the standard for kneading time is no less than 15 seconds and the actual kneading time is 10 seconds, then the violation severity factor is (target value - actual value) / target value = (15 - 10) / 15 = 0.33. Assuming the constraint strength coefficient is 0.42, then the basic violation value is 0.42 × 0.33 = 0.1419.

[0063] S34. When an operation event that does not belong to the expected event of the current critical operation node is detected, the operation event is classified as follows: If the operation event is an expected event of a subsequent key operation node, the operation event is recorded as an advanced event and associated with the corresponding key operation node, and the evaluation is delayed until the node is activated; otherwise, it is recorded as an irrelevant event and an immediate deduction is made based on its degree of interference.

[0064] In one specific implementation method, the standardized operating procedure for protective clothing is as follows: hand hygiene, wearing a disposable cap, wearing a medical protective mask, wearing inner gloves, wearing a medical protective suit, wearing disposable shoe covers, wearing outer gloves, and wearing goggles.

[0065] If a participant is currently at the node "wearing a disposable cap" but prematurely performs the action of "picking up the protective suit" at the node "wearing medical protective suit", it is considered an expected event and is recorded as an advanced event.

[0066] If a participant performs the "put on medical protective clothing" step and then goes back to complete the "put on disposable cap" step, the system will still perform a normal compliance assessment of all operations within that step (including the subsequent cap-putting action) and generate the basic violation value for the current step.

[0067] When a critical operation node corresponding to an advanced event is activated, the recorded advanced event is included in the evaluation sequence of that node. When evaluating that node, the evaluation is carried out by combining the first timestamp of the advanced event and the second timestamp of the critical operation node being activated, and the advanced event deduction score is obtained to evaluate the impact of the advanced event on achieving the current goal.

[0068] In a specific implementation, when the trainee finally reaches the medical protective suit node, the system activates the evaluation of the medical protective suit node. Since this event is marked as an advanced event, the system performs a cross-node timeliness assessment, calculating a contamination risk deduction value for the advanced operation based on the time difference between the second and first timestamps. This deduction will be added to the total violation value of the current node.

[0069] S35. Based on real-time monitoring of scene state change data, determine the target achievement status of the current key operation node, and determine the current achievable optimal target by combining the predefined target state hierarchy.

[0070] The compliance assessment set is the standard for evaluating specific operations, but its full score and weight will change dynamically according to the current goal to be achieved.

[0071] Taking hand hygiene as an example, suppose the following evaluation criteria exist: R1: Use of disinfectant; R2: Rubbing time ≥ 15 seconds; R3: Rinsing time ≥ 10 seconds; Each evaluation item has a corresponding constraint strength coefficient.

[0072] The system predefines the target state hierarchy for the "hand hygiene" node: The target state hierarchy can be divided into G1 ideal target (extremely low hand pathogens, requires perfect achievement of R1+R2+R3, full score 100 points), G2 good target (low hand pathogens, can be achieved by hand sanitizer + rubbing, R3 can be omitted, full score 70 points) and G3 failed target (high hand pathogens, no effective operation, full score 0 points); the initial system defaults to the perfect target, and all evaluation items are enabled according to the constraint strength coefficient.

[0073] Assuming real-time monitoring of operational behavior data reveals a violation of R1, the system queries scenario state transition data and finds the "Disinfectant Used" state variable is False. The diagnosis is: due to a permanent violation of R1, the ideal goal G1 is unattainable. The system then checks if G2 is the best achievable result in the current state, but since G2 requires R1, it is also unattainable. Therefore, the current achievable optimal goal is G3 (the failed goal), and the maximum score drops to 0.

[0074] Based on the current achievable optimal goal, the constraint strength coefficients of each compliance evaluation item are dynamically adjusted to obtain the dynamic weight coefficients of each compliance evaluation item, so as to reflect the changes in the importance of each evaluation item to achieving the current goal.

[0075] The adjustment method for dynamic weighting coefficients is as follows: Figure 2 As shown: The target state hierarchy is decomposed into a set of key states that must be achieved. The set of key states that can be achieved at the current optimal target is determined. The state variables of all dynamic interactive elements associated with each compliance evaluation item are combined to obtain the target relevance coefficient of each compliance evaluation item. The target relevance coefficient of each compliance evaluation item is calculated based on the weight ratio of its corresponding state variable in the set of key states, reflecting the importance of the evaluation item to achieving the current target.

[0076] Based on the target relevance coefficient and the constraint strength coefficient of each compliance evaluation item, a normalized weighted calculation is used to generate the dynamic weight coefficient of each compliance evaluation item. The calculation formula is as follows: Current dynamic weight coefficient = (current constraint strength coefficient × current target relevance coefficient) / Σ(constraint strength coefficient of all compliance assessment items × target relevance coefficient).

[0077] In the above embodiment, because G3 is the failure target, the weights of all evaluation items are reduced to near zero (e.g., 0.1), since the node is destined to fail regardless of subsequent operations. The student stopped rinsing after 8 seconds.

[0078] S36. Generate matching results containing dynamic compliance scores, specifically including the following steps: Using dynamic weighting coefficients, all the obtained basic violation values ​​are weighted to obtain a weighted violation value. The total violation value of the node is obtained by accumulating all weighted violation values, immediate deductions, and advanced event deductions within the current key operation node. By combining the full score of the node corresponding to the current achievable optimal goal, a dynamic compliance score for the key operation node is calculated and generated; Dynamic compliance score = maximum score of the node corresponding to the current achievable optimal goal - total violation value of the node; The maximum score for the node corresponding to the current achievable optimal goal is the highest score that the student can theoretically obtain under the current conditions, as diagnosed by the system based on the real-time status.

[0079] It aggregates dynamic compliance scores, records of advanced events, and violation events, and outputs matching results.

[0080] S37. Based on the matching results, drive the virtual environment to generate dynamic responses, specifically including the following steps: S371. Encode the real-time acquired matching results, and generate a multi-dimensional error feature vector sequence of the violation events in the matching results according to the type of evaluation item (time-series evaluation item, quality evaluation item, etc.) and the severity of violation. The severity of violation is the aforementioned violation degree factor.

[0081] S372. Analyze the multi-dimensional error feature vector sequence and calculate the frequency and cumulative intensity of each error dimension within a preset time window (300 seconds). The frequency is calculated by counting the number of non-zero values ​​in each dimension within the window and dividing by the window length. For example, if the time-series dimension has 150 non-zero values ​​within 300 seconds, the frequency is 0.5. The cumulative intensity is calculated by summing the severity of all violations occurring in each dimension within the window.

[0082] S373. Mark the error dimensions corresponding to the frequency of occurrence and cumulative intensity that exceed the preset threshold as the current high-frequency error dimensions. The preset thresholds for frequency of occurrence and cumulative intensity are 0.3 and 50, respectively.

[0083] S374. Based on the statistical values ​​corresponding to the current high-frequency error dimensions, template matching is performed in the preset response evaluation item templates. Each template defines a set of matching conditions and a parameter adjustment strategy. The matching conditions include error dimension requirements and threshold requirements.

[0084] Based on the specific characteristics of the current error dimension, the system selects, combines, and instantiates these templates in real time and dynamically generates mapping table entries. For the same error dimension, different responses will be generated under different frequencies, intensities, and nodes, thus achieving adaptive adjustment.

[0085] Specifically, the process of generating the parameter adjustment strategy includes the following steps: When the input feature of a successfully matched template is the frequency of occurrence, a linear mapping is performed based on the frequency of occurrence to calculate the parameter adjustment amount used to adjust the time pressure feedback parameter or visual saliency parameter in the virtual environment. When the input feature of a successfully matched template is cumulative intensity, proportional clamping mapping is performed based on the cumulative intensity to calculate the parameter adjustment amount used to adjust the visual saliency parameter or the operational accuracy requirement parameter in the virtual environment. The parameter adjustment amounts are all based on the weight coefficients of the current key operation nodes obtained from historical exercise data statistics as multiplicative mapping factors. The weight coefficients of the current key operation nodes indicate the importance of the current node in the entire process.

[0086] Because there is an inherent correlation between different types of error dimensions and teaching intervention methods: frequency of occurrence reflects the habitual biases of trainees, which are suitable for correction by strengthening consequence feedback or visual guidance; cumulative intensity reflects the severity of a single error, which is suitable for intervention by providing decision support, i.e., visual guidance or raising operational requirements.

[0087] In a specific, implementable scenario of casualty rescue, the following parameter adjustment strategy is defined: A. For matching conditions that satisfy: the frequency of occurrence in the time-series dimension > 0.3; the parameter adjustment strategy is: increase the rate of deterioration of the virtual character's condition, and the formula for calculating the parameter adjustment amount is: Adjustment amount = frequency of occurrence × baseline coefficient × criticality weight of the current node.

[0088] B. For matching conditions that meet the following criteria: temporal dimension and cumulative intensity > 50, the parameter adjustment strategy is to increase the intensity of the task urgency visual cue, and the adjustment amount is calculated using the following formula: Adjustment amount = min(cumulative intensity / 100, 1.0) × baseline coefficient × current node criticality weight; 100 is the baseline reference value, representing the intensity level at which a full response should be triggered, and 1.0 is the upper limit value.

[0089] C. For matching conditions that satisfy: object dimension and frequency of occurrence > 0.2, the parameter adjustment strategy is: increase the visual salience threshold of key objects, and the adjustment amount is calculated using the following formula: Adjustment amount = frequency of occurrence × baseline coefficient × difficulty weight of the current node; D. For matching conditions that satisfy: quality dimension and cumulative intensity > 30, the parameter adjustment strategy is to reduce the operational tolerance range, and the adjustment amount is calculated using the following formula: Adjustment amount = min(cumulative intensity / 50, 1.0) × baseline coefficient × current node criticality weight.

[0090] The baseline coefficient is a fixed, predefined normalization factor used to map the input frequency, cumulative intensity, etc., to a reasonable adjustment range, ensuring that the adjustment range is neither too large nor too small. The baseline coefficient varies between different templates and is set during the system design phase based on teaching experience.

[0091] The criticality weight of the current node is the weight coefficient of the current critical operation node, which is used to represent the importance of the node to achieving the final task goal. In a specific implementation method, its calculation method is as follows: Historical exercise records are retrieved from the historical exercise database. Each record contains the completion metrics of each key operation node and the corresponding final task achievement label. Statistical analysis is performed on the historical exercise records to establish a correlation model between the completion metrics of each key operation node and the final task achievement label. Based on the correlation model, the contribution metric of each key operation node to the final task achievement is extracted. The contribution metrics of all key operation nodes are normalized so that the sum of the weight coefficients of all nodes is a preset constant (1 in this embodiment), thus obtaining the weight coefficient of each key operation node.

[0092] In one specific implementation, contribution analysis employs a logistic regression model. Using a binary classification (success / failure) as the final task completion label, the basic compliance score of each node is used as feature input to train the logistic regression model. The regression coefficient for each node obtained during training represents the degree to which that node's completion contributes to the probability of task success. The absolute values ​​of all regression coefficients are normalized so that the sum of all node weights is 1, yielding the criticality weight coefficient for each node, representing the node's proportion of influence on the task outcome among all nodes.

[0093] S375. For successfully matched templates, based on the severity of the violation and statistical values, execute the corresponding parameter adjustment strategy and calculate and generate environmental parameter adjustment instructions. Environmental parameters include the intensity of the virtual character's status feedback, the tolerance range of operational difficulty, and the salience level of environmental information.

[0094] S376. Based on the environmental parameter adjustment instructions, the corresponding environmental parameters are adjusted in real time by calling the interface of the virtual reality engine, driving the virtual environment to produce perceptible dynamic changes, and the adjustment content, trigger template and execution result are recorded as the situation response result.

[0095] For example, the instruction "virtual character's condition deterioration rate +0.4" means that the vital signs of the injured person marked in red in the scene decrease at a rate 1.4 times faster than before, simulating the accelerated deterioration of the condition due to the trainee's time delay; the instruction "task urgency visual cue intensity +0.5" means that the flashing frequency of the task countdown panel increases and the color gradually changes from yellow to red, enhancing the trainee's sense of time urgency.

[0096] S38. The process of generating a dynamic response also includes the following steps: Before entering any critical operation node, the state variables of each dynamic interaction element are used to check whether the prerequisite state conditions on which the current critical operation node depends are met. If the conditions are not met, the dynamic interactive elements associated with the current critical operation node will be reset to the minimum required state to start the current critical operation node, and a state compensation record will be generated. The compensation record includes the reason for compensation and the compensation deduction value. The state compensation record will be associated with the preceding node that caused the condition to be unmet.

[0097] In one specific implementation method, the compensation deduction value is calculated as follows: Compensation deduction value = Node criticality weight of preceding node × Node full score × Compensation deduction ratio; The compensation deduction ratio is calculated by multiplying the node criticality weight of the preceding node by a preset scaling factor; and the scaling factor is determined by the ratio of the maximum allowed compensation ratio to the maximum node criticality weight among all nodes.

[0098] At the end of the preceding node, obtain the dynamic compliance score contained in the matching result of the preceding node, deduct the compensation deduction value from it, obtain the final score of the node, and write the final score and compensation record into the final evaluation report.

[0099] If the preceding node is not executed, the dynamic compliance score is set to 0.

[0100] S4. Summarize the matching results of each key operation node and record the situational response results of dynamic responses to generate a quantitative evaluation report for the trainees.

[0101] The report also includes details of the exercise, a task list, and expert evaluations.

[0102] The exercise details include participation time, scores, and other information; the task list records detailed data from the exercise process, such as operation names and scores.

[0103] Expert evaluations are automatically generated based on the trainee's operation process data, evaluating aspects such as the smoothness of operation and the operation score. Secondary editing and report download are allowed.

[0104] Based on the same inventive concept described above, this application also discloses an emergency rescue drill system based on virtual reality, with the following architecture: Figure 3 As shown, the system includes the following modules: The virtual scene construction module is used to respond to exercise task requests, load the corresponding standardized operation process, and generate a virtual environment containing dynamic interactive elements based on the key operation nodes of the standardized operation process; the dynamic interactive elements are semantically associated with the key operation nodes. The multidimensional data acquisition module is used to collect multidimensional exercise interaction data generated by the personnel performing virtual rescue operations in the virtual environment in real time. The multidimensional exercise interaction data includes at least operation behavior data, semantic interaction data, and scene state change data. The intelligent dynamic evolution module is used to call the compliance judgment set corresponding to each key operation node to perform real-time matching and verification of multi-dimensional exercise interaction data for each key operation node. Based on the verification results, it generates a matching result containing dynamic compliance scores and drives the virtual environment to generate dynamic responses based on the matching results. The exercise results output module is used to summarize the matching results of each key operation node and record the situational response results of dynamic responses, and generate a quantitative evaluation report for the exercise participants.

[0105] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in the computer-readable storage medium, which includes, for example, various media capable of storing program code such as: USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk.

[0106] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for emergency rescue drills based on virtual reality, characterized in that, Includes the following steps: In response to the exercise task request, the corresponding standardized operation process is loaded, and a virtual environment containing dynamic interactive elements is generated based on the key operation nodes of the standardized operation process; the dynamic interactive elements are semantically associated with the key operation nodes. The system collects multi-dimensional interactive data of the drills in real time, which is generated by the drill participants performing virtual rescue operations in the virtual environment. The multi-dimensional interactive data includes at least operational behavior data, semantic interaction data, and scene state change data. For each key operation node, the compliance assessment set corresponding to the key operation node is invoked to perform real-time matching and verification of the multi-dimensional exercise interaction data. Based on the obtained verification results, a matching result containing dynamic compliance scores is generated, and the virtual environment is driven to generate a dynamic response based on the matching results. The matching results of each key operation node and the situational response results of the dynamic response are summarized to generate a quantitative evaluation report for the trainees.

2. The emergency rescue drill method based on virtual reality according to claim 1, characterized in that, In response to the exercise task request, the corresponding standardized operation process is loaded, and a virtual environment containing dynamic interactive elements is generated based on the key operation nodes of the standardized operation process. Specifically, this includes the following steps: Receive and respond to pre-selected exercise task requests, and load the corresponding standardized operation process from a pre-set standardized process database according to the exercise task request; The standardized operating procedure is analyzed, and preset key operating nodes are extracted from the standardized operating procedure; According to the exercise task request, a corresponding virtual environment is loaded from the virtual scenario database. The virtual environment contains multiple dynamic interactive elements, including at least one of virtual characters, virtual facilities, and virtual environment objects. Configure at least one state variable to represent the current state and a semantic label to identify the function category for each dynamic interactive element, and establish a logical relationship between the dynamic interactive element and the corresponding key operation node based on the semantic label.

3. The emergency rescue drill method based on virtual reality according to claim 2, characterized in that, The real-time acquisition of multi-dimensional interactive data generated by participants performing virtual rescue operations in the virtual environment includes the following steps: Listen to and record the interaction events between the trainees and the dynamic interactive elements in the virtual environment, and generate operational behavior data; The virtual character in the virtual environment is driven by an artificial intelligence model. The virtual character is pre-set with a background knowledge base and behavioral logic evaluation items corresponding to the exercise task. The system receives the voice text from the trainees and inputs it into the artificial intelligence model. Based on the artificial intelligence model, the system performs real-time analysis of the voice text, combines the background knowledge base of the virtual character and the behavioral logic evaluation items, generates voice feedback corresponding to the voice text, and records the semantic interaction data. Monitor and record the changes in the values ​​of the state variables of each dynamic interactive element in the virtual environment. When the change of any state variable reaches a preset threshold or is triggered by a preset interactive event, generate scene state transition data.

4. The emergency rescue drill method based on virtual reality according to claim 2, characterized in that, For each key operation node, the step of calling the compliance assessment set corresponding to the key operation node to perform real-time matching and verification of the multi-dimensional exercise interaction data specifically includes the following steps: The compliance evaluation set corresponding to the current key operation node is invoked to perform real-time feature matching on the operation behavior data and the semantic interaction data, generating violation event feature information and basic violation value. The compliance evaluation set includes time-series evaluation items, object evaluation items and quality evaluation items. Extract the node target achievement vectors corresponding to all the key operation nodes from the preset historical exercise database, and construct a set of compliance status feature vectors for each compliance evaluation item; Perform statistical correlation operations on the compliance status feature vector set and the node target achievement vector in the vector space to solve the correlation contribution vector of each compliance evaluation item feature vector relative to the node target achievement vector. Based on the normalization operation result of the correlation contribution vector, calculate the constraint strength coefficient of each compliance evaluation item. When any current operation event is detected to violate the compliance evaluation item, the corresponding violation event is mapped to a violation feature scalar, and a scalar product operation is performed with the constraint strength coefficient corresponding to the currently violated compliance evaluation item to obtain the current basic violation value.

5. The emergency rescue drill method based on virtual reality according to claim 4, characterized in that, It also includes the following steps: When an operation event that does not belong to the expected event of the current key operation node is detected, the operation event is classified; if the operation event belongs to the expected event of the subsequent key operation node, the operation event is recorded as a leading event and associated with the corresponding key operation node; otherwise, it is recorded as an irrelevant event and points are deducted immediately. Based on the real-time monitored scenario state change data, the target achievement status of the current key operation node is determined, and the current achievable optimal target is determined in combination with the predefined target state hierarchy. Based on the current achievable optimal target, the dynamic weight coefficients of each compliance evaluation item in the compliance evaluation set are dynamically adjusted. When the key operation node corresponding to the advanced event is activated, the advanced event deduction value of the advanced event is obtained by combining the first timestamp of the advanced event and the second timestamp of the key operation node being activated.

6. The emergency rescue drill method based on virtual reality according to claim 5, characterized in that, The step of dynamically adjusting the dynamic weight coefficients of each compliance evaluation item in the compliance evaluation set based on the currently achievable optimal goal specifically includes the following steps: Each target level in the target state hierarchy is decomposed into a set of key states that must be achieved. The set of key states that the current achievable optimal target is determined. The target relevance coefficient of each compliance evaluation item is obtained by combining the state variables of all dynamic interactive elements associated with each compliance evaluation item. The target relevance coefficient of each compliance evaluation item is calculated based on the weight ratio of its corresponding state variable in the set of key states. Based on the target relevance coefficient and the constraint strength coefficient of each compliance evaluation item, the dynamic weight coefficient of each compliance evaluation item is generated by normalized weighted calculation.

7. The emergency rescue drill method based on virtual reality according to claim 5, characterized in that, The process of generating matching results that include dynamic compliance scores specifically includes the following steps: Using the dynamic weighting coefficient, all the obtained basic violation values ​​are weighted to obtain a weighted violation value. The total violation value of the node is obtained by accumulating all the weighted violation values, the instant deductions, and the advanced event deductions within the current key operation node. By combining the full score of the node corresponding to the current achievable optimal goal, a dynamic compliance score for the key operation node is calculated and generated. The matching result is output by summarizing the dynamic compliance score, the records of the advanced events, and the violation events.

8. The emergency rescue drill method based on virtual reality according to claim 2, characterized in that, The step of driving the virtual environment to generate a dynamic response based on the matching result specifically includes the following steps: The matching results acquired in real time are encoded, and the violation events in the matching results are used to generate a multi-dimensional error feature vector sequence according to the evaluation item type and the severity of the violation; The multi-dimensional error feature vector sequence is analyzed to calculate the frequency and cumulative intensity of each error dimension within a preset time window; the error dimensions corresponding to the frequency and cumulative intensity exceeding a preset threshold are marked as the current high-frequency error dimensions. Based on the statistical values ​​corresponding to the current high-frequency error dimensions, template matching is performed in the preset response evaluation item templates. Each template defines a set of matching conditions and a parameter adjustment strategy. Based on the severity of the violation and the statistical value, the corresponding parameter adjustment strategy is executed on the successfully matched template, an environmental parameter adjustment instruction is calculated and generated, and the corresponding environmental parameters are adjusted in real time according to the environmental parameter adjustment instruction. The adjustment content, trigger template and execution result are recorded as the situation response result.

9. The emergency rescue drill method based on virtual reality according to claim 8, characterized in that, The generation process of the parameter adjustment strategy includes: When the input feature of the successfully matched template is the occurrence frequency, a linear mapping is performed based on the occurrence frequency to calculate the parameter adjustment amount used to adjust the time pressure feedback parameter or visual saliency parameter in the virtual environment. When the input feature of the successfully matched template is the cumulative intensity, proportional clamping mapping is performed based on the cumulative intensity to calculate the parameter adjustment amount used to adjust the visual saliency parameter or operational accuracy requirement parameter in the virtual environment. The parameter adjustment amounts are all based on the weight coefficients of the current key operation nodes obtained from historical exercise data statistics, which are used as multiplicative mapping factors.

10. An emergency rescue drill system based on virtual reality, characterized in that, Includes the following modules: The virtual scene construction module is used to respond to the exercise task request, load the corresponding standardized operation process, and generate a virtual environment containing dynamic interactive elements based on the key operation nodes of the standardized operation process; the dynamic interactive elements are semantically associated with the key operation nodes. The multidimensional data acquisition module is used to collect multidimensional exercise interaction data generated by the participants performing virtual rescue operations in the virtual environment in real time. The multidimensional exercise interaction data includes at least operation behavior data, semantic interaction data, and scene state change data. The intelligent dynamic evolution module is used to call the compliance evaluation set corresponding to the key operation node for each key operation node to perform real-time matching and verification of the multi-dimensional exercise interaction data, generate a matching result containing dynamic compliance score based on the obtained verification result, and drive the virtual environment to generate a dynamic response based on the matching result. The exercise result output module is used to summarize the matching results of each key operation node and record the situational response results of the dynamic response, and generate a quantitative evaluation report for the exercise personnel.

Images