A virtual simulation training system for rural tourism emergency management

Abstract

This invention relates to the fields of virtual reality technology and emergency management training technology, specifically to a virtual simulation training system for rural tourism emergency management, comprising a client training terminal and a server. The client training terminal includes: an operation error judgment module, which determines sequential errors through time sequence matching, determines positional errors through screen rays and spatial distance, and identifies critical step errors; a replay control module, which performs single-step rollback replay for sequential or positional errors, resetting only the transformation parameters and script variables of the manipulated object and smoothly returning to their original positions, and performing a full-scene reset replay for critical step errors; and a highlighting module, which applies visual highlighting to the target object and dynamically loads prompt text when a timeout or positional error occurs. The server includes three types of computing resources, deployed as a signaling service cluster, a physical computing and AI inference cluster, and a state synchronization cluster, respectively, to achieve lightweight request processing, offloading of complex calculations, and multi-user collaborative state synchronization.

Description

Technical Field

[0001] This invention relates to the fields of virtual reality technology and emergency management training technology, specifically to a virtual simulation training system for emergency management in rural tourism. Background Technology

[0002] Currently, training for emergency management personnel in rural tourism mainly employs either theoretical lectures or field drills. While theoretical lectures are low-cost and easy to organize, trainees lack immersive experiences, making it difficult to develop realistic stress response capabilities, thus limiting training effectiveness. Field drills, on the other hand, offer a strong sense of immersion, but are time-consuming to organize, costly, have limited participation, and pose certain safety risks, making them difficult to conduct regularly. Therefore, some existing technologies are considering leveraging 3D modeling and virtual reality to provide scene navigation and operational practice functions.

[0003] However, the above systems still have the following problems in practical applications: (1) High resource consumption and long waiting time during retraining: When students make mistakes and need to retrain, the existing systems generally use the method of unloading the entire virtual scene and then reloading it. This method involves releasing all scene resources, rereading model and texture data from the disk, and reinitializing the state of all objects. During continuous training, the CPU peak usage of a single retraining session is too high, and the scene loading time is as long as 5 to 8 seconds. Especially when multiple people are training at the same time, the server needs to independently execute scene reloading for each user, which increases resource consumption exponentially and leads to a significant decrease in overall training efficiency; (2) Single retraining strategy and inability to distinguish error types: The existing systems can only judge whether the operation is "correct" or "incorrect". For errors of different natures such as incorrect sequence, position deviation, and incorrect tool selection, the same full scene reset method is used, which requires a lot of system resources and waiting time. (3) Emergency response in rural tourism often requires teamwork, but most of the existing virtual simulation systems are single-machine mode and cannot support multiple students to operate collaboratively in the same virtual scene. When attempting to support multiple users, the state synchronization mechanism often adopts a full broadcast approach, resulting in high network bandwidth consumption and limited concurrency capabilities, further exacerbating resource competition in re-training scenarios.

[0004] In summary, existing technologies for handling operational errors and retraining in virtual simulation training suffer from high resource consumption, long waiting times, and limited strategies. How to achieve refined error detection and differentiated rapid retraining without significantly increasing system overhead, while simultaneously ensuring efficient resource scheduling in multi-user collaborative scenarios, has become a pressing technical challenge in this field. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention aims to provide a virtual simulation training system for emergency management in rural tourism, solving the problems of insufficient precision in operational judgment, low retraining efficiency, and inability to adapt to the needs of multi-person collaboration in existing technologies.

[0006] The basic solution provided by this invention is a virtual simulation training system for emergency management in rural tourism, comprising a client training terminal and a server. The client training terminal includes: The operation error judgment module is configured to: record the time sequence of user operations and match the current user operation with the preset correct step sequence; if they do not match, it is judged as a sequence error; convert the user's screen click position into a ray in the world coordinate system, calculate the spatial distance between the ray's hit point and the center of the target object, and if the distance exceeds a preset distance threshold, it is judged as a position error; when the user performs the correct operation steps but the selected tool or object does not meet the scene requirements, it is judged as a critical step error. The replay control module is configured to: when a sequence error or position error is determined, execute a single-step rollback replay, which includes: resetting only the transformation parameters and script variables of the object that was erroneously manipulated to the initial state of the step, and moving the object back to its original position through a smooth animation; when a critical step error is determined, execute a full-scene reset replay, which includes: first loading the transition interface, then asynchronously destroying the current scene in the background and re-instantiating the entire virtual scene. The highlighting module is configured to automatically apply a visual highlighting effect to the target object to be operated on in the next step when the user stays in the current step for more than a preset time threshold, or when the position is determined to be incorrect. The corresponding prompt text is also dynamically loaded. The server-side includes three types of computing resources: the first type of computing resources, which deploys a signaling service cluster to handle lightweight client requests; the second type of computing resources, which deploys a physical computing and AI inference cluster to execute complex computing tasks that cannot be efficiently completed locally on the client; and the third type of computing resources, which deploys a state synchronization cluster to perform state synchronization based on a room isolation strategy when multiple users operate the same virtual scene simultaneously.

[0007] The principle of this invention lies in optimizing the identification of operational errors from the traditional binary (right / wrong) to a multi-dimensional classification (sequence error / position error / critical step error), and matching differentiated rework strategies based on the error type. Specifically: the operational error determination module captures the sequential relationship of operation steps through temporal sequence matching and captures the spatial accuracy of user clicks through screen ray and spatial distance calculation, thereby transforming abstract operational behaviors into quantifiable error type labels; the rework control module selects different scene recovery paths based on the error type label: for non-essential errors (sequence error, position error), only the state of a single object is rolled back to avoid resource overload of the entire scene; for essential errors (critical step errors), a full scene reset is performed; the server uses three specifications of computing resources to handle signaling, computation, and state synchronization respectively, decoupling local determination from cloud collaboration and avoiding resource competition caused by rework operations in multi-user scenarios. This mechanism achieves a dynamic balance between error determination accuracy and rework resource consumption.

[0008] The beneficial effects of this invention are as follows: 1. Refined judgment of operation error: Through dual-modal judgment of temporal matching and spatial ray distance threshold, it can accurately distinguish between sequence error, position error and critical step error, providing a reliable classification basis for differentiated retraining, and solving the defect of the existing system that can only prompt operation error, which makes it impossible for students to locate specific problems.

[0009] 2. Differentiated rework mechanism reduces resource consumption: For non-essential errors that account for a high proportion (sequence errors, position errors), a single-step rollback rework is adopted, that is, only the transformation parameters and script variables of a single object are reset, and the object is returned to its original position through smooth animation. This avoids the huge CPU and memory consumption caused by unloading the entire scene, reloading model textures, and reinitializing all objects in the traditional solution, and realizes a local repair strategy from scene-level reset to object-level reset.

[0010] 3. Organic collaboration between multiple users and local judgment: The professional division of the three types of computing resources on the server side ensures that signaling processing, complex calculations (such as flame spread and AI scoring) and status synchronization do not interfere with each other. Combined with the local operation error judgment and re-execution on the client side, it avoids training lag caused by network latency or server overload, and provides scalable underlying support for multiple users to train at the same time.

[0011] Furthermore, the state synchronization cluster adopts a command mode and an optimistic locking mechanism: each user operation is encapsulated as an instruction, the final state is determined according to the timestamp and priority, and the final state is broadcast to all clients; after the state synchronization cluster rejects the instruction corresponding to a non-final state, the rejected client smoothly pulls the operated object back to the position corresponding to the final state through an interpolation algorithm and plays a visual conflict prompt.

[0012] Existing technologies in multi-user collaborative virtual training typically employ pessimistic locking (such as global mutex locks) or full state broadcasting. The former leads to long user wait times, while the latter results in wasted bandwidth and inconsistent states. This invention adopts a command pattern and optimistic locking mechanism: each user's operation is encapsulated as an independent instruction, and the server determines the final state based on timestamps and priorities. Unselected clients smoothly roll back to the authoritative position through an interpolation algorithm, and a visual conflict warning is provided. This mechanism avoids operation waiting, allows users to operate in parallel, and performs non-blocking corrections only after a conflict occurs, ensuring eventual state consistency while maximizing the smoothness of multi-user interaction. Furthermore, compared to direct jumps, rollback and interpolation methods reduce user dizziness and improve the training experience.

[0013] Furthermore, the client-side training terminal and the server-side use dual-protocol communication: key data is exchanged through a reliable transmission protocol, while high-frequency, low-importance data is exchanged through an unreliable, low-latency transmission protocol; the key data includes operation instructions, scoring results, and uploaded score data, while the high-frequency, low-importance data includes the real-time position, rotation angle, and particle emitter status of virtual objects; the client-side uses an interpolation algorithm to smoothly compensate for state transitions caused by packet loss.

[0014] Existing virtual training systems often use a single TCP protocol. While this ensures data reliability, the transmission of high-frequency data (such as real-time position and rotation angle) introduces significant delays due to TCP's congestion control and retransmission mechanisms, causing lag in virtual object movement in multi-user scenarios. This invention divides data into critical data (operation instructions, scoring results, and score uploads) and high-frequency, low-importance data (position, rotation, and particle state), using reliable and unreliable low-latency transmission protocols respectively. Critical data, requiring absolute integrity and order, is transmitted using TCP; high-frequency data, allowing for minor packet loss, is transmitted using UDP or KCP, with client-side interpolation algorithms compensating for state transitions caused by packet loss. This design reduces perceived latency in multi-user collaborative scenarios while ensuring the reliability of core business logic.

[0015] Furthermore, the lightweight client requests include: login requests, training task acquisition requests, and grade upload requests; the complex computational tasks include: disaster evolution simulation based on cellular automata, physical models, or data-driven models, and operational standardization scoring based on artificial intelligence models.

[0016] Lightweight requests (login, task retrieval, and score upload) are characterized by short connections, low computational load, and high concurrency. They are processed by the signaling service cluster of the first-level computing resources, which avoids consuming high-performance computing resources. Complex computing tasks include disaster evolution simulations based on cellular automata / physical models / data-driven models (such as fire spread and food poisoning spread) and operation standardization scoring based on AI models. These tasks are computationally intensive and have high real-time requirements. They are processed centrally by the second-level computing resources. The client only needs to upload scene parameters or operation sequences and receive the calculation results.

[0017] Furthermore, the visual highlighting effect is achieved by adding an outline stroke post-processing effect to the object rendering pipeline; the loading key value of the prompt text includes a scene identifier and a current step identifier.

[0018] Existing systems typically pre-set highlighting prompts to fixed steps, making it impossible to adjust based on dynamic user behavior. This invention employs a contour outlining post-processing effect to achieve visual highlighting, without altering the object's original material, resulting in low performance overhead and compatibility with multiple rendering pipelines. The loading key-value pair of the prompt text includes a scene identifier and a current step identifier (e.g., "hint_kitchen_step3_position"), enabling dynamic loading on demand and facilitating multi-language expansion and independent configuration for different scenarios.

[0019] Furthermore, the distance threshold is a preset distance value corresponding to the screen pixel deviation, and the time threshold is a preset maximum idle time.

[0020] The distance threshold corresponds to a preset distance value for screen pixel deviation (e.g., mapping 50 pixels on the screen to 0.2 meters in world coordinates), used to determine whether the user's click position deviates from the center of the target object, and is a quantitative expression of spatial error tolerance; the time threshold is a preset maximum idle time (e.g., 30 seconds), used to determine whether the user has timed out in the current step. Attached Figure Description

[0021] Figure 1 This is a system block diagram of an embodiment of a virtual simulation training system for emergency management of rural tourism according to the present invention.

[0022] Figure 2 This is a schematic diagram of dual-protocol communication between the client training terminal and the server in an embodiment of the present invention. Detailed Implementation

[0023] The following detailed description illustrates the specific implementation method: The basic implementation examples are as follows: Figure 1 As shown: A virtual simulation training system for emergency management in rural tourism, including a client training terminal and a server.

[0024] I. System Overall Architecture The client-side training terminal runs on a general-purpose computing device that supports 3D rendering (such as a PC, VR all-in-one machine, or mobile terminal). It is used to load and render 3D virtual scenes (such as kitchen fire scenes or food poisoning scenes), collect user interaction data such as clicks and drags, and perform local operation judgment, replay control, and highlight prompt functions.

[0025] The server is deployed in a cloud environment, employing three different sizes of computing resources for specialized tasks: The first-level computing resources (such as an 8-core, 8GB memory instance) are used to deploy a signaling service cluster to handle lightweight client requests. These lightweight requests include login requests, training task retrieval requests, and grade upload requests. These requests are characterized by short connections, low computational load, and high concurrency. Processing them by the signaling service cluster avoids consuming high-performance computing resources. The signaling service cluster refers to a group of servers responsible for processing control messages (such as login, task retrieval, and grade upload) between the client and the server.

[0026] The second-tier computing resources (such as a 16-core, 32GB memory instance) are deployed as physical computing and AI inference clusters to execute complex computational tasks that cannot be efficiently completed locally on the client side. These complex computational tasks include: disaster evolution simulations based on cellular automata, physical models, or data-driven models (such as fire spread simulations and food poisoning spread simulations), and operational compliance scoring based on artificial intelligence models. The client only needs to upload scene parameters (such as wind speed and fuel quantity) or operation sequences, and the server calculates and returns the results.

[0027] The third-level computing resources (such as an 8-core, 16GB memory instance) are deployed as a state synchronization cluster. This cluster is used to synchronize the state of virtual objects (such as position, rotation, and operation results) across multiple users operating the same virtual scene simultaneously, based on a room isolation strategy. The state synchronization cluster refers to a group of servers responsible for maintaining the consistency of the virtual object states (such as position, rotation, and operation results) across clients when multiple users operate the same virtual scene simultaneously. Resource allocation uses a room isolation strategy. Specifically, when a user enters a specific scene (such as the "kitchen fire" scene), the client requests to create or join a virtual room. The server then schedules all state synchronization logic for that room to the same server instance.

[0028] It should be noted that the above three CPU / memory configurations are only examples, and can be dynamically adjusted according to the complexity of the scenario in actual deployment.

[0029] II. Specific Implementation of Each Module in the Client-Side Training Terminal 1. Operation Error Detection Module This module is responsible for converting user actions into quantifiable error type labels, specifically configured as follows: (1) Sequence error judgment: The system presets the correct step sequence for each training scenario (e.g., the standard steps for using a fire extinguisher in a kitchen fire scenario are: lift the fire extinguisher, pull out the pin, aim at the base of the fire, and squeeze the handle). The operation error judgment module records the time sequence of the user's operation in real time and matches the current user operation with the current expected step in the preset step sequence. If the user skips "pulling out the pin" and directly performs the "aiming" operation, the module detects that the current operation is "aiming" while the expected step is "pulling out the pin", and then judges it as a sequence error.

[0030] (2) Position Error Detection: When a user interacts with a virtual object by clicking on the screen, the module converts the user's screen click position into a ray in the world coordinate system and calculates the spatial distance between the point where the ray hits and the center of the target object. If this distance exceeds a preset distance threshold (e.g., mapping a 50-pixel deviation on the screen to 0.2 meters in world coordinates), it is determined to be a position error. This distance threshold is used to quantify the spatial error tolerance of the user's click.

[0031] (3) Critical step error determination: When the user performs the correct operation steps (i.e., sequence matching), but the selected tool or object does not meet the requirements of the current scenario (e.g., in an oil fire scenario, a foam fire extinguisher is incorrectly selected instead of a dry powder fire extinguisher), the module determines that the critical step is incorrect.

[0032] Through the above three-fold determination, this invention elevates the traditional correct / incorrect dichotomy to a multi-dimensional classification of sequence error / position error / critical step error, providing a refined basis for subsequent differentiated rework.

[0033] 2. Re-train the control module This module selects different recovery paths based on the error type labels output by the operation error determination module: (1) When the error is determined to be a sequence error or a position error (not an essential error): a single-step rollback and replay is performed. Specifically, this includes: resetting only the transformation parameters (position, rotation angle) and script variables of the object that was erroneously manipulated to the initial state of that step, and moving the object back to its original position using a smooth animation (e.g., using linear interpolation, duration 0.5 seconds). During this process, the current scene resources are not unloaded, and only the state of a single object is reset, avoiding the reloading of the entire scene.

[0034] (2) When the error is determined to be a critical step error (essential error): perform a full scene reset and replay. Specifically, this includes: first loading the transition interface (Loading interface) to avoid the user perceiving lag; then asynchronously calling the scene manager in the background to destroy the current virtual scene and release all scene resources; and then re-instantiating the entire virtual scene and all interactive objects to restore it to the initial state.

[0035] It should be noted that traditional solutions use a full-scenario reset regardless of the type of error, resulting in high peak CPU usage and long waiting times. This invention uses a single-step rollback and retry for the high proportion of sequence errors and positional errors.

[0036] 3. Highlighting module This module is used to provide dynamic guidance when users encounter difficulties or deviations in their operations, and is specifically configured as follows: Timeout trigger: The system sets a time threshold for each operation step (a preset maximum idle time, such as 30 seconds). When the user stays in the current step for longer than the time threshold, the module automatically applies a visual highlight effect to the target object to be operated on next and dynamically loads the corresponding prompt text (such as "Please pull out the fire extinguisher pin first") from the local resource table.

[0037] Location error trigger: When the operation error judgment module determines that the location is incorrect, the highlighting prompt module is also triggered, applying a highlighting effect to the target object that the user should aim at (such as the base of a fire source), and loading targeted prompts such as "Please aim at the center area of ​​the base of the fire source".

[0038] The visual highlighting effect is achieved by adding an outline stroke post-processing effect to the object's rendering pipeline (such as using the Outline component in Unity, with a yellow stroke color). This method does not change the object's original material, has low performance consumption, and is compatible with multiple rendering pipelines. The loading key-value pair of the prompt text includes a scene identifier and a current step identifier, enabling dynamic loading on demand and facilitating multi-language expansion and independent configuration for different scenes.

[0039] III. Communication Mechanism between Client and Server As attached Figure 2 As shown, the client-side training terminal and the server use dual-protocol communication: Key data (including operation instructions, scoring results, and uploaded score data) are exchanged through reliable transmission protocols (such as TCP) to ensure data integrity and order.

[0040] High-frequency, low-importance data (including the real-time position, rotation angle, and particle emitter status of virtual objects) is exchanged via unreliable low-latency transmission protocols (such as UDP or KCP). The client side uses interpolation algorithms to smoothly compensate for state transitions caused by packet loss, such as performing linear interpolation between positions in two consecutive frames, to keep the motion trajectory of the virtual object consistent.

[0041] It should be noted that while the traditional single TCP protocol can guarantee data reliability, high-frequency data will be significantly delayed due to TCP's congestion control and retransmission mechanisms, causing lag in the movement of virtual objects in multi-user scenarios.

[0042] IV. Multi-person collaboration and concurrent conflict handling When multiple learners simultaneously operate the same interactive object in the same virtual room (e.g., two learners simultaneously grabbing the same fire extinguisher), the state synchronization cluster deployed by the third-specification computing resources uses a command mode and an optimistic locking mechanism to handle concurrency conflicts. The optimistic locking mechanism is a concurrency control strategy that allows multiple users to submit operations simultaneously. The server determines the final state based on timestamps and priorities. Clients not selected are rolled back for correction, rather than locking and waiting before the operation. Specifically: Each user's action is encapsulated into an independent instruction (including action type, target object ID, timestamp, and client ID) and sent to the state synchronization cluster.

[0043] The state synchronization cluster sorts instructions by their arrival timestamps, combines them with preset priority rules (such as first-come, first-served), determines the final state (i.e. the state corresponding to a valid instruction), and broadcasts the final state to all clients.

[0044] After the state synchronization cluster rejects an instruction that does not correspond to the final state, the rejected client smoothly pulls the manipulated object back to the position corresponding to the final state using an interpolation algorithm (such as linear interpolation within 0.2 seconds). Simultaneously, a visual conflict warning is displayed on the interface (such as flashing red text "Operation invalid, item has been used by another user"). The interpolation algorithm is a mathematical method used to estimate intermediate values ​​between discrete data points; in this invention, it is used to smoothly compensate for jumps in the position or angle of virtual objects caused by packet loss.

[0045] This mechanism differs from commonly used pessimistic locking (global mutex lock) or full state broadcasting in existing technologies: this invention allows users to operate in parallel, performing non-blocking corrections only after a conflict occurs, avoiding operation waiting, ensuring eventual state consistency, and maximizing the smoothness of multi-user interaction. Furthermore, the smooth correction method of rollback and interpolation reduces user dizziness compared to direct jump changes.

[0046] The above are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A virtual simulation training system for emergency management in rural tourism, comprising a client-side training terminal and a server-side component, characterized in that: The client-side training terminal includes: The operation error judgment module is configured to: record the time sequence of user operations and match the current user operation with the preset correct step sequence; if they do not match, it is judged as a sequence error; convert the user's screen click position into a ray in the world coordinate system, calculate the spatial distance between the ray's hit point and the center of the target object, and if the distance exceeds a preset distance threshold, it is judged as a position error; when the user performs the correct operation steps but the selected tool or object does not meet the scene requirements, it is judged as a critical step error. The replay control module is configured to: when a sequence error or position error is determined, execute a single-step rollback replay, which includes: resetting only the transformation parameters and script variables of the object that was erroneously manipulated to the initial state of the step, and moving the object back to its original position through a smooth animation; when a critical step error is determined, execute a full-scene reset replay, which includes: first loading the transition interface, then asynchronously destroying the current scene in the background and re-instantiating the entire virtual scene. The highlighting module is configured to automatically apply a visual highlighting effect to the target object to be operated on in the next step when the user stays in the current step for more than a preset time threshold, or when the position is determined to be incorrect. The corresponding prompt text is also dynamically loaded. The server-side includes three types of computing resources: the first type of computing resources, which deploys a signaling service cluster to handle lightweight client requests; the second type of computing resources, which deploys a physical computing and AI inference cluster to execute complex computing tasks that cannot be efficiently completed locally on the client; and the third type of computing resources, which deploys a state synchronization cluster to perform state synchronization based on a room isolation strategy when multiple users operate the same virtual scene simultaneously.

2. The virtual simulation training system for emergency management of rural tourism according to claim 1, characterized in that, The state synchronization cluster adopts a command pattern and optimistic locking mechanism: each user operation is encapsulated as an instruction, the final state is determined according to the timestamp and priority, and the final state is broadcast to all clients; after the state synchronization cluster rejects the instruction corresponding to a non-final state, the rejected client smoothly pulls the operated object back to the position corresponding to the final state through an interpolation algorithm and plays a visual conflict prompt.

3. The virtual simulation training system for emergency management of rural tourism according to claim 2, characterized in that, The client-side training terminal and the server use dual-protocol communication: key data is exchanged through a reliable transmission protocol, while high-frequency, low-importance data is exchanged through an unreliable, low-latency transmission protocol. The key data includes operation instructions, scoring results, and uploaded score data. The high-frequency, low-importance data includes the real-time position, rotation angle, and particle emitter status of virtual objects. The client side uses an interpolation algorithm to smoothly compensate for state transitions caused by packet loss.

4. The virtual simulation training system for emergency management of rural tourism according to claim 3, characterized in that, The lightweight client requests include: login requests, training task acquisition requests, and grade upload requests; the complex computational tasks include: disaster evolution simulation based on cellular automata, physical models, or data-driven models, and operational standardization scoring based on artificial intelligence models.

5. A virtual simulation training system for emergency management of rural tourism according to claim 4, characterized in that, The visual highlighting effect is achieved by adding an outline stroke post-processing effect to the object rendering pipeline; the loading key value of the prompt text includes the scene identifier and the current step identifier.

6. A virtual simulation training system for emergency management of rural tourism according to claim 5, characterized in that, The distance threshold is a preset distance value corresponding to the screen pixel deviation, and the time threshold is a preset maximum idle time.

Images