Virtual reality steering control method, system and device and storage medium

Abstract

The invention relates to a virtual reality steering control method, system and device and a storage medium, and relates to the technical field of PID control, and the method comprises the steps: responding to a directional motion type selected by a user, and carrying out the mapping to a predefined classification topological structure; receiving a user physiological index data set, inputting the user physiological index data set into the classification topological structure to execute clustering operation, and outputting a user category label; for the users of the potential risk category, real-time monitoring is carried out while directional motion is authorized to be started, data are collected, and a motion state vector is generated through simulation; according to the motion state vector and the physiological index data set, a historical sample library is matched through a feature retrieval algorithm, and a risk assessment coefficient is calculated; and if the risk assessment coefficient does not exceed a preset risk critical value, regulating and controlling an equipment PID control loop according to the motion state vector to carry out virtual reality control. Self-adaptive control is achieved according to the physiological indexes and the motion state of the user, and the safety and accuracy of virtual reality steering control are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of PID control, and particularly to a steering control method, system, device and storage medium for virtual reality. Background Art

[0002] In recent years, virtual reality technology has developed rapidly, and its applications have been widely extended to many fields such as entertainment, education, training and medical treatment. In a virtual reality environment, steering control is one of the key technologies to achieve an immersive experience. The current mainstream virtual reality steering control determines the steering direction by sensing the changes in the user's body center of gravity and posture information, and performs the corresponding virtual reality scene steering simulation.

[0003] However, there are significant individual differences in the physical fitness of users, which may cause inappropriate physical stimulation to some users, trigger physiological reactions such as dizziness, nausea, and abnormal heart rate, and even lead to more serious health risks, even in simulated movements in a virtual environment. The current solution is to pre-check the user's physical condition before the user experiences virtual reality steering control to screen out those who are suitable for participating in virtual reality activities. However, this method can only identify those who are obviously not suitable for participation, and cannot effectively screen out those users who are in normal surface condition but actually have potential risks. In addition, the existing technology lacks a real-time monitoring and dynamic adjustment mechanism during the virtual reality experience after the user passes the preliminary screening. Especially when it comes to complex steering actions, it is impossible to adjust the control parameters in a timely manner according to the changes in the user's physiological state during use, so as to predict and prevent possible risk situations. Summary of the Invention

[0004] Aiming at the technical problem that the virtual reality steering control system in the prior art cannot perform accurate risk assessment and safety guarantee according to the individual differences of users, resulting in potential risk users may face health and safety hazards during use, the present invention provides a steering control method, system, device and storage medium for virtual reality to solve this problem.

[0005] The technical solution of the present invention to solve the above technical problems is as follows: In a first aspect, the present invention provides a steering control method for virtual reality, comprising: in response to a selected directional movement type of a user, mapping it to a predefined classification topology; receiving a user physiological index data set, inputting it into the classification topology to perform clustering operations, and outputting a user category label; when the user category label belongs to a potential risk category, authorizing the user to initiate the directional movement type and starting real-time monitoring to collect centroid coordinates, attitude quaternions, and operation instruction streams; based on the centroid coordinates, the attitude quaternions, and the operation instruction streams, executing a kinematic simulation engine to generate a motion state vector; according to the motion state vector and the physiological index data set, matching a historical sample library through a feature retrieval algorithm to calculate a risk assessment coefficient; if the risk assessment coefficient ≤ the preset risk critical value, regulating a device PID control loop according to the motion state vector for virtual reality control.

[0006] Optionally, the method further comprises: when the user category label belongs to an explicit risk category, refusing to authorize the user to initiate the directional movement type; when the user category label belongs to a risk-free category, authorizing the user to initiate the directional movement type.

[0007] Optionally, in response to a selected directional movement type of a user, mapping it to a predefined classification topology, comprising: constructing a physiological index feature space based on user physiological index dimensions, wherein the physiological index feature space has a first deterministic risk partition, a first probabilistic risk partition, and a first deterministic safety partition preset by a management terminal; performing a central tendency analysis on a risk transaction data set to obtain a risk clustering center threshold, and performing a boundary union truncation on the first deterministic risk partition to obtain a second deterministic risk partition; performing a central tendency analysis on a safety transaction data set to obtain a safety clustering center threshold, and performing a boundary intersection truncation on the first deterministic safety partition to obtain a second deterministic safety partition; according to the geometric relationship between the second deterministic risk partition and the second deterministic safety partition, performing a convex hull reconstruction on the first probabilistic risk partition to generate the second probabilistic risk partition; constructing the classification topology based on the second deterministic risk partition, the second probabilistic risk partition, and the second deterministic safety partition.

[0008] Optionally, performing a central tendency analysis on a risk transaction data set to obtain a risk clustering center threshold and performing a boundary update on the first deterministic risk partition to obtain a second deterministic risk partition, comprising: performing a Delphi method weight distribution on the user physiological index attribute set to obtain a physiological index weight distribution vector; based on the physiological index weight distribution vector, performing a pairwise weighted Euclidean norm calculation on the risk transaction data set to generate a risk sample distance matrix; applying a local outlier factor algorithm to the risk sample distance matrix and determining it through minimum value screening to obtain the risk clustering center threshold.

[0009] Optionally, according to the motion state vector and the physiological index data set, match the historical sample library through a feature retrieval algorithm, and calculate a risk assessment coefficient, including: configuring a first feature retrieval fault tolerance condition based on the motion state deviation threshold according to the motion state vector; configuring a second feature retrieval fault tolerance condition based on the physiological index deviation threshold according to the physiological index data set; retrieving a set of motion samples in the historical sample library that simultaneously meet the first feature retrieval fault tolerance condition and the second feature retrieval fault tolerance condition; counting the proportion of the distribution of the number of physical sign abnormality identifiers in the set of motion samples, and setting it as the risk assessment coefficient.

[0010] Optionally, based on the centroid coordinates, the attitude quaternion, and the operation instruction stream, execute a kinematic simulation engine to generate a motion state vector, including: configuring centroid coordinate record data, attitude quaternion record data, operation instruction stream record data, and motion state vector true value data with the target motion device model as a constraint; configuring three branch networks based on a long short-term memory neural network, configuring a backbone network based on a fully connected neural network, and merging the output layer of the three branch networks and the input layer of the backbone network to obtain a kinematic simulation engine network architecture; supervising and training the kinematic simulation engine network architecture through the motion state vector true value data, the centroid coordinate record data, the attitude quaternion record data, and the operation instruction stream record data to obtain the kinematic simulation engine, including: configuring a set of motion physical law formulas through a management terminal; configuring a set of physical law deviation loss terms based on the set of motion physical law formulas; configuring a mean square error function and setting it as a data loss term; supervising and training the kinematic simulation engine network architecture through the motion state vector true value data, the centroid coordinate record data, the attitude quaternion record data, and the operation instruction stream record data, and generating the kinematic simulation engine when the losses of the set of physical law deviation loss terms and the data loss term are less than or equal to the corresponding loss thresholds for more than a preset proportion in a continuous preset number of times.

[0011] Optionally, the method further includes: if the risk assessment coefficient > the preset risk critical value, adjusting the motion state vector to obtain an updated motion state vector; calculating an updated risk assessment coefficient by matching the historical sample library through a feature retrieval algorithm according to the updated motion state vector and the physiological index data set; if the updated risk assessment coefficient ≤ the preset risk critical value, regulating the device PID control loop according to the updated motion state vector for virtual reality control.

[0012] In a second aspect, the present invention provides a steering control system for virtual reality, comprising: a type mapping module configured to map to a predefined classification topology in response to a selected orientation movement type by a user; a clustering operation module configured to receive a user physiological index data set, input same into the classification topology to perform a clustering operation, and output a user category label; a real-time monitoring startup module configured to, when the user category label belongs to a potential risk category, authorize the user to start the orientation movement type and start real-time monitoring, and collect centroid coordinates, attitude quaternions, and an operation instruction stream; a simulation execution module configured to execute a kinematic simulation engine based on the centroid coordinates, the attitude quaternions, and the operation instruction stream, and generate a motion state vector; a risk assessment module configured to, according to the motion state vector and the physiological index data set, match a historical sample library through a feature retrieval algorithm, and calculate a risk assessment coefficient; and a virtual reality control module configured to, if the risk assessment coefficient ≤ a preset risk critical value, regulate a device PID control loop according to the motion state vector for virtual reality control.

[0013] In a third aspect, the present application provides an electronic device, the device comprising: a processor; and a memory for storing executable instructions of the processor; wherein, the processor is configured to execute a steering control method for virtual reality provided by the present application.

[0014] In a fourth aspect, the present application provides a computer-readable storage medium storing a computer program, the computer program being configured to execute a steering control method for virtual reality provided by the present application.

[0015] The beneficial effects of the present invention are: In response to the selected orienteering type of the user, it is mapped to a predefined classification topology. According to different virtual reality steering motion types selected by the user, it can be mapped into the corresponding predefined topology, providing a basic framework for subsequent user classification. Receive the user's physiological index dataset, input it into the classification topology to perform clustering operations, and output the user category label, thereby determining the risk category to which the user belongs and achieving precise classification of the user. When the user category label belongs to the potential risk category, authorize the user to start the orienteering type and start real-time monitoring, collecting the centroid coordinates, attitude quaternions, and operation instruction streams. By specially managing the users identified as potential risk categories, while authorizing them to carry out virtual reality steering activities, start real-time monitoring, collect necessary motion data, and provide real-time information for risk assessment. Based on the centroid coordinates, attitude quaternions, and operation instruction streams, execute the kinematic simulation engine to generate a motion state vector, providing dynamic information for precisely evaluating the user's risk. According to the motion state vector and the physiological index dataset, match the historical sample library through the feature retrieval algorithm, calculate the risk assessment coefficient, and achieve precise quantification of the risk of the user's current state. If the risk assessment coefficient ≤ the preset risk critical value, regulate the device PID control loop according to the motion state vector for virtual reality control, realizing safe and stable virtual reality steering control.

[0016] Through the above technical solutions, the present invention solves the technical problem in the prior art that the virtual reality steering control system cannot perform precise risk assessment and safety guarantee according to the individual differences of users, resulting in potential health and safety hazards for potential risk users during use, and achieves the technical effect of effectively identifying potential risk users and providing dynamic safety guarantee, and realizing adaptive control according to the user's physiological indexes and motion states, effectively improving the safety and accuracy of virtual reality steering control. Brief Description of the Drawings

[0017] Figure 1 It is a schematic flow chart of a virtual reality steering control method provided by the present invention; Figure 2 It is a schematic structural diagram of a virtual reality steering control system provided by the present invention; Figure 3 It is a schematic structural diagram of an electronic device provided by the present invention; Figure 4 It is a schematic structural diagram of a computer-readable storage medium provided by the present invention.

[0018] In the drawings, the components represented by each reference numeral are as follows: Type mapping module 11, clustering operation module 12, real-time monitoring startup module 13, simulation execution module 14, risk assessment module 15, virtual reality control module 16, electronic device 200, memory 210, processor 220, first computer program 211, computer-readable storage medium 300, second computer program 311. Detailed implementation manners

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined.

[0021] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or description". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or having more advantages than other embodiments. The following description is provided to enable any person skilled in the art to make and use the present invention. Details are set forth for the purpose of explanation in the following description. It should be understood that those of ordinary skill in the art can recognize that the present invention can be practiced without these specific details. In other instances, well-known structures and processes are not described in detail to avoid unnecessary details from obscuring the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0022] Embodiment 1, as Figure 1 shown, the embodiment of the present invention provides a steering control method for virtual reality, including: S100: In response to the selected directional movement type of the user, map it to a predefined classification topology structure.

[0023] Specifically, when receiving the type of directional movement selected by the user through the virtual reality device interface (such as specific movement modes like horizontal turning, vertical pitching, rotation, etc.), the corresponding classified topological structure established in advance is mapped and selected according to this type of directional movement. Among them, the classified topological structure is a multi-dimensional data model of the type of directional movement constructed based on the user's physiological index dimension; this structure takes various physiological indexes (such as heart rate, blood pressure, blood oxygen saturation, etc.) as coordinate axes to form a multi-dimensional feature space. In this feature space, three types of regions are pre-divided, namely the deterministic risk partition, the probabilistic risk partition, and the deterministic safety partition. Among them, the deterministic risk partition refers to the partition where, under specific physiological index conditions, there is a clear health risk for the user to perform the movement of this type of directional movement; the probabilistic risk partition refers to the partition where, under specific physiological index conditions, there is a potential risk for the user to perform the movement of this type of directional movement and further monitoring is required; the deterministic safety partition indicates the partition where, under specific physiological index conditions, the user can safely perform the movement of this type of directional movement.

[0024] Since different types of directional movements have different physiological stimuli on the human body. For example, high-speed rotation actions have a greater stimulus on the vestibular system and may cause symptoms such as dizziness and nausea, while simple translational movements have relatively less stimulus on the human body. Therefore, each type of directional movement corresponds to a specific classified topological structure, and the risk partition boundaries and safety thresholds in it are different, reflecting the potential impact of the type of directional movement on users in different physiological states.

[0025] By mapping and selecting the corresponding classified topological structure according to the type of directional movement selected by the user, it provides a reference standard for the subsequent evaluation of the user category label, so as to judge which partition the physiological indexes of the current user belong to. In this way, the most suitable screening criteria and monitoring parameters can be adopted for different types of directional movements, so as to accurately judge the risk of the user performing the selected type of directional movement.

[0026] S200: Receive the user physiological index data set, input it into the classified topological structure to perform clustering operations, and output the user category label.

[0027] Specifically, receive the user physiological index data set, which contains various real-time collected physiological parameter information, such as physiological indexes like heart rate, blood pressure, blood oxygen saturation, etc. These physiological index data are collected in real time through relevant sensing devices (such as wearable devices, heart rate monitors, etc.).

[0028] After receiving the user's physiological indicator data set, the multi-dimensional physiological indicator data is used as a feature vector and input into the mapped classification topology. In the multi-dimensional feature space of the classification topology, a clustering operation is performed on the input physiological indicator data to match and compare the user's current physiological state with the pre-defined partitions in the classification topology. The clustering operation process determines which partition the current user's physiological state falls into by calculating the distance or similarity between the current user's physiological indicator and the center point of each partition. By performing the clustering operation, the position of the current user's physiological indicator in the feature space can be accurately located, and its relationship with each risk partition can be determined. Based on the result of the clustering operation, the user category label is output, which may be an explicit risk category, a potential risk category, or a risk-free category. Among them, the explicit risk category indicates that the user's current physiological indicator is located in a deterministic risk partition and there is a clear health risk; the potential risk category indicates that the user's current physiological indicator is located in a probabilistic risk partition and there is a certain potential risk; the risk-free category indicates that the user's current physiological indicator is located in a deterministic safety partition and the selected directional movement can be safely performed.

[0029] By mapping the user's physiological indicator dataset to a classification topology and performing clustering operations, it is possible to quickly and accurately evaluate the suitability of the user's current physiological state for executing the directional movement selected by the user, providing a basis for subsequent safety control strategies, thereby achieving safe management of the virtual reality steering control process.

[0030] S300: When the user category tag belongs to the potential risk category, the authorized user starts the directional movement type and starts real-time monitoring to collect the center of mass coordinates, attitude quaternion and operation instruction stream.

[0031] Specifically, first, the result of the user category label is judged. When the user category label is determined to belong to the potential risk category, that is, the user's current physiological indicators are located in the probabilistic risk partition in the classification topology, the conditional authorization state is entered. In this state, the user is authorized to start the previously selected orienteering type, but at the same time, the safety monitoring mechanism is triggered to ensure the safety of the user during the orienteering process with potential risks.

[0032] After the authorized user activates the orienteering type, real-time monitoring is immediately initiated to continuously and uninterruptedly collect three types of data, namely the centroid coordinates, attitude quaternion, and operation instruction stream. Among them, the centroid coordinates are collected in real time through the inertial measurement unit configured on the user's body, reflecting the three-dimensional spatial position change of the user's body center of gravity; the attitude quaternion is obtained by calculating through the head-mounted display and IMU sensors, accurately describing the three-dimensional spatial attitude and rotation state of the user's head and body; the operation instruction stream is a sequence of control signals sent by the user through various input devices (such as gamepads, somatosensory controllers, etc.). The real-time collection of these three types of data constitutes a comprehensive monitoring of the user's motion state. The centroid coordinates reflect the movement trajectory of the user's body center of gravity and can detect whether the user has unstable shakes or tilts; the attitude quaternion can accurately depict the spatial attitude of the user's head and body, helping to identify potential uncoordinated movements; the operation instruction stream reflects the user's subjective intention and control will and can be used to judge the user's reaction ability and operation accuracy.

[0033] By collecting the centroid coordinates, attitude quaternion, and operation instruction stream, a comprehensive monitoring mechanism for the user's motion state in the virtual reality environment is established, providing a complete data basis for subsequent risk assessment. This real-time monitoring mechanism is a necessary safety guarantee measure for potential risk users, aiming to timely capture abnormal states that may cause discomfort or danger to the user and provide a basis for subsequent safety intervention.

[0034] S400: Based on the centroid coordinates, attitude quaternion, and operation instruction stream, execute the kinematic simulation engine to generate a motion state vector.

[0035] Specifically, the three types of data of the centroid coordinates, attitude quaternion, and operation instruction stream collected in real time are used as input parameters and sent to the kinematic simulation engine for processing. The kinematic simulation engine is a computational model implemented based on neural network technology and is used to obtain the motion state parameters of the user in the virtual reality environment. This kinematic simulation engine is trained with a large amount of historical data to establish a mapping relationship between the input parameters and the motion state.

[0036] After the kinematic simulation engine processes the centroid coordinates, attitude quaternion, and operation instruction stream, it generates a motion state vector, which specifically includes quantization parameters such as angular velocity, linear acceleration, and joint torque. Among them, the angular velocity describes the rate of the user's rotational action, the linear acceleration represents the acceleration state of the user's body in space, and the joint torque reflects the magnitude of the torque borne by the user's joints. These parameters together constitute a comprehensive quantitative description of the user's motion state.

[0037] By executing the kinematic simulation engine, the user's real-time motion data is converted into a standardized motion state vector, enabling the accurate assessment of the possible impact of the user's current motion state on their physiological condition and providing a quantitative basis for subsequent risk assessment.

[0038] S500: According to the motion state vector and the physiological index data set, match the historical sample library through the feature retrieval algorithm, and calculate the risk assessment coefficient.

[0039] Specifically, use the generated motion state vector and the received user physiological index data set to perform matching and retrieval in the historical sample library through the feature retrieval algorithm to calculate the risk assessment coefficient for the current user to perform the directional motion.

[0040] First, based on the motion state vector and the motion state deviation threshold, configure the first feature retrieval fault tolerance condition. This fault tolerance condition allows a certain degree of deviation in motion parameters such as angular velocity, linear acceleration, and joint torque during the retrieval process to ensure that sufficiently similar historical samples can be found. At the same time, based on the physiological index data set and the physiological index deviation threshold, configure the second feature retrieval fault tolerance condition, allowing a certain tolerance space for physiological parameters such as heart rate, blood pressure, and blood oxygen saturation during the retrieval and matching. After configuring these two types of fault tolerance conditions, perform a retrieval operation in the historical sample library to find a set of motion samples that simultaneously meet the first feature retrieval fault tolerance condition and the second feature retrieval fault tolerance condition. These samples reflect historical cases similar to the current user's motion state and physiological condition, including the physical sign changes and safety records of the users in these cases. Subsequently, count the proportion of the number of samples with physical sign abnormality identification in the set of these motion samples to the total number of samples, and set this proportion as the risk assessment coefficient. The value range of this risk assessment coefficient is from 0 to 1, and the larger the value, the higher the risk level.

[0041] Through the risk assessment method based on historical data, it is possible to determine the risk level that the user may face when performing a specific directional motion by using the existing empirical data, providing a quantitative basis for subsequent control decisions, and thus achieving more precise safety protection.

[0042] S600: If the risk assessment coefficient ≤ the preset risk critical value, adjust the device PID control loop according to the motion state vector for virtual reality control.

[0043] Specifically, first, compare the obtained risk assessment coefficient with the preset risk critical value. The preset risk critical value is the upper limit of risk tolerance preset according to safety requirements, for example, 0.72, which represents the maximum allowable risk level.

[0044] When the risk assessment coefficient is less than or equal to the preset risk critical value, it is determined that the risk of the current user performing the selected orienteering is within an acceptable range, so the authorization to continue the virtual reality control process is given. In this case, according to the generated motion state vector (including parameters such as angular velocity, linear acceleration, joint torque, etc.), the PID control loop of the virtual reality device is regulated to achieve the simulation and feedback of the user's orienteering motion.

[0045] The PID control loop can accurately control parameters such as the steering force, speed, and acceleration of the virtual reality device according to the user's motion state vector. By adjusting the parameters of the PID control loop, the smoothness and responsiveness of the virtual reality steering control are ensured, providing a realistic and safe virtual reality experience for the user.

[0046] Through the risk assessment-based control strategy, on the premise of ensuring user safety, the precise adjustment of the virtual reality steering control is achieved, avoiding operations that may cause user discomfort or health risks, while maintaining a good user experience and operational responsiveness.

[0047] Furthermore, the embodiments of the present application further include: S710: When the user category label belongs to the explicit risk category, reject the authorization for the user to start the orienteering type; S720: When the user category label belongs to the risk-free category, authorize the user to start the orienteering type.

[0048] Specifically, when it is determined that the user category label belongs to the explicit risk category, that is, the user's current physiological indicators are within the deterministic risk partition, directly reject the authorization for the user to start the selected orienteering type. In this case, the user's physiological indicators clearly show that there are significant health risks in performing this orienteering, such as high heart rate, abnormal blood pressure, or insufficient blood oxygen saturation, etc. Therefore, for safety reasons, prevent the user from performing actions that may cause health problems, thereby achieving the protection of high-risk users.

[0049] When it is determined that the user category label belongs to the risk-free category, that is, the user's current physiological indicators are within the deterministic safety partition, directly authorize the user to start the selected orienteering type without starting an additional real-time monitoring mechanism. In this case, the user's physiological indicators indicate that their physical condition is fully suitable for performing the selected orienteering and there are no potential health risks, so the user can be directly allowed to operate, simplifying the process and enhancing the user experience.

[0050] Through a security control mechanism based on users' physiological indicators, a prohibition strategy is implemented for users with obvious risks, conditional authorization is implemented for users with potential risks and monitoring is strengthened, and direct authorization is implemented for users without risks. This hierarchical control mechanism can adopt differentiated security strategies for users with different risk levels, which not only ensures the health and safety of users, but also maximally improves the usability and user experience of the virtual system.

[0051] Further, in response to the selected type of directed movement of the user, it is mapped to a predefined classification topology structure, including: S110: Construct a physiological index feature space based on the dimension of users' physiological indicators, where the physiological index feature space has a first deterministic risk partition, a first probabilistic risk partition, and a first deterministic security partition preset by the management terminal; S120: Perform a central tendency analysis according to the risk transaction data set to obtain a risk clustering center threshold, and perform a boundary union truncation on the first deterministic risk partition to obtain a second deterministic risk partition; S130: Perform a central tendency analysis according to the security transaction data set to obtain a security clustering center threshold, and perform a boundary intersection truncation on the first deterministic security partition to obtain a second deterministic security partition; S140: According to the geometric relationship between the second deterministic risk partition and the second deterministic security partition, perform a convex hull reconstruction on the first probabilistic risk partition to generate a second probabilistic risk partition; S150: Construct a classification topology structure based on the second deterministic risk partition, the second probabilistic risk partition, and the second deterministic security partition.

[0052] In a preferred implementation manner, to construct a classification topology structure, first, a physiological index feature space is constructed based on the dimension of users' physiological indicators. This physiological index feature space uses physiological indicators such as heart rate, blood pressure, and blood oxygen saturation as coordinate axes to form a multi-dimensional feature space. In this feature space, three types of partitions are preset, namely a first deterministic risk partition (the physiological state area that is clearly not suitable for performing this movement), a first probabilistic risk partition (the physiological state area with a certain risk possibility), and a first deterministic security partition (the physiological state area suitable for performing this movement). These initial partitions are preliminary divisions preset based on medical knowledge and experience.

[0053] Then, perform a central tendency analysis using the collected risk transaction dataset. The risk transaction dataset contains historical record data of users having physiological abnormalities during the execution of relevant orienteering movements. By statistically analyzing the distribution characteristics of these data, determine the aggregation region of risk events in the physiological index feature space, thereby obtaining the risk clustering center threshold. This threshold reflects the combined feature of physiological indices that lead to risk events. Subsequently, perform a boundary union truncation operation on the first deterministic risk partition. Through the boundary union truncation, take the union of the boundary of the first deterministic risk partition and the boundary of the newly identified risk clustering center threshold, that is, take the smaller value when comparing the boundaries, so as to expand the coverage of the risk partition and obtain a more stringent second deterministic risk partition. For example, if the boundary condition of the original first deterministic risk partition is "heart rate > 160", and the risk clustering center threshold finds that "heart rate > 150" may lead to risks, then take the union of the two (that is, take the smaller value 150) to expand the risk partition range and form a more stringent second deterministic risk partition. This operation ensures that a wider range of potential risk situations can be captured and improves the safety guarantee level.

[0054] Meanwhile, perform a central tendency analysis based on the safety transaction dataset to obtain the safety clustering center threshold. The safety transaction dataset contains historical record data of users having no abnormalities during the execution of relevant orienteering movements. By analyzing these data, the characteristic threshold of the safe operation area can be determined. Subsequently, perform a boundary intersection truncation on the first deterministic safety partition. Through the boundary intersection truncation, take the intersection of the boundary of the original first deterministic safety partition and the boundary of the newly identified safety clustering center threshold, that is, take the smaller value when comparing the boundaries, so as to narrow the coverage of the safety partition and obtain a more stringent second deterministic safety partition. For example, if the boundary condition of the original first deterministic safety partition is "heart rate < 140", and the safety clustering analysis finds that only "heart rate < 130" can ensure absolute safety, then take the intersection of the two (that is, take the smaller value 130) to narrow the safety partition range and form a more stringent second deterministic safety partition. This operation further improves the rigor of the safety partition determination and reduces the risk of misjudgment.

[0055] Then, based on the geometric relationship between the obtained second deterministic risk partition and the second deterministic safety partition, perform convex hull reconstruction on the first probabilistic risk partition. Specifically, analyze the spatial region between the second deterministic risk partition and the second deterministic safety partition in the multi-dimensional feature space, determine the geometric shape of the transition region between these two partitions, and generate a second probabilistic risk partition with a regular geometric shape through the convex hull algorithm. The convex hull reconstruction ensures that the second probabilistic risk partition has a clear mathematical boundary while maintaining an accurate description of the risk transition region. This processing method avoids the fragmentation and irregularity of the partition boundary, improving the execution efficiency and stability of subsequent classification algorithms. After that, integrate the obtained second deterministic risk partition, second probabilistic risk partition, and second deterministic safety partition to construct a complete classification topology structure. This structure not only includes the boundary definitions of each partition but also the topological relationships between the partitions, forming a continuous and complete classification space for subsequent user physiological index evaluation and classification.

[0056] By constructing the classification topology structure, each partition in the classification topology structure is based on both medical knowledge and experience and combines the statistical analysis of historical data, thereby improving the accuracy and reliability of user category labels.

[0057] Furthermore, perform central tendency analysis on the risk transaction data set to obtain the risk clustering center threshold, and update the boundary of the first deterministic risk partition to obtain the second deterministic risk partition, including: S121: Perform Delphi method weight distribution on the set of user physiological index attributes to obtain the physiological index weight distribution vector; S122: Based on the physiological index weight distribution vector, perform pairwise weighted Euclidean norm calculation on the risk transaction data set to generate a risk sample distance matrix; S123: Apply the local outlier factor algorithm to the risk sample distance matrix and determine it through minimum value screening to obtain the risk clustering center threshold.

[0058] In a preferred embodiment, first, perform Delphi method weight distribution on the set of user physiological index attributes. Among them, the Delphi method obtains a consensus on complex problems through multiple rounds of expert consultations. In this embodiment, medical experts are solicited through the Delphi method to evaluate the importance of different physiological indicators (such as heart rate, blood pressure, blood oxygen saturation, etc.) in judging specific directional movement risks. After multiple rounds of iteration and statistical analysis, the weight coefficients of each physiological indicator are formed. These weight coefficients form the physiological index weight distribution vector, reflecting the contribution degree of different physiological indicators to risk judgment. For example, for rotational directional movements, physiological indicators related to vestibular function may obtain higher weights, while for tilt directional movements, indicators related to blood pressure fluctuations may obtain higher weights.

[0059] Then, based on the obtained physiological index weight distribution vector, pairwise weighted Euclidean norm calculation is performed on the risk transaction dataset to generate a risk sample distance matrix. Specifically, physiological index data of all risk events are extracted from the risk transaction dataset, and each piece of data is regarded as a point in a multi-dimensional feature space. Then, the weighted Euclidean distance between any two risk sample points is calculated, where the weights of each dimension are provided by the physiological index weight distribution vector. Through this calculation, a risk sample distance matrix is generated, and each element in this matrix represents the distance between the corresponding two risk samples in the weighted feature space. This weighted calculation method takes into account the importance differences of different physiological indexes, making the differences in important indexes have a greater impact on the distance calculation.

[0060] After that, the local outlier factor algorithm is applied to the risk sample distance matrix, and the risk clustering center threshold is obtained through minimum value screening. The local outlier factor algorithm is a density-based outlier detection method that can evaluate the local density deviation of data points. The local outlier factor value of each risk sample is calculated by applying the local outlier factor algorithm, and this value reflects the density situation of the sample point relative to its neighborhood. The lower the local outlier factor value, the higher the sample density around the point, that is, the point is in the aggregation area of risk samples. By screening the minimum values of the local outlier factor values, the aggregation centers of risk samples are determined, and the physiological index values of these center points are extracted as the risk clustering center threshold. These thresholds precisely define the combined characteristic of physiological indexes with high incidence of risk events, providing data support for the boundary update of the first deterministic risk partition.

[0061] By means of intelligent analysis based on historical risk data, the risk clustering center threshold is determined, providing a reliable basis for the subsequent optimization of the risk partition boundary, thereby improving the safety assessment accuracy in virtual reality steering control.

[0062] Furthermore, according to the motion state vector and the physiological index dataset, the historical sample library is matched through the feature retrieval algorithm, and the risk assessment coefficient is calculated, including: S510: According to the motion state vector, based on the motion state deviation threshold, configure the first feature retrieval fault tolerance condition; S520: According to the physiological index dataset, based on the physiological index deviation threshold, configure the second feature retrieval fault tolerance condition; S530: Retrieve the set of motion samples in the historical sample library that simultaneously meet the first feature retrieval fault tolerance condition and the second feature retrieval fault tolerance condition; S540: Statistically calculate the proportion of the distribution quantity with the physical sign abnormality identifier in the set of motion samples, and set it as the risk assessment coefficient.

[0063] In a preferred embodiment, first, according to the generated motion state vector, based on a preset motion state deviation threshold, a first feature retrieval fault tolerance condition is configured. The motion state vector includes multiple parameters such as angular velocity, linear acceleration, and joint torque, and a corresponding deviation tolerance range is set for each parameter. For example, the fault tolerance range of the angular velocity is set to ±5° / s, the fault tolerance range of the linear acceleration is set to ±0.2 m / s², and the fault tolerance range of the joint torque is set to ±0.5 N·m, thereby obtaining the first feature retrieval fault tolerance condition. The first feature retrieval fault tolerance condition allows for certain parameter deviations during the historical sample retrieval process, thereby increasing the number of effectively matched samples and improving the statistical reliability of the retrieval results. At the same time, according to the user's current physiological index data set, based on a preset physiological index deviation threshold, a second feature retrieval fault tolerance condition is configured. The physiological index data set includes multiple indexes such as heart rate, blood pressure, and blood oxygen saturation, and a corresponding deviation tolerance range is also set for each index. For example, the fault tolerance range of the heart rate is set to ±5 beats per minute, the fault tolerance range of the systolic blood pressure is set to ±5 mmHg, and the fault tolerance range of the blood oxygen saturation is set to ±1%, thereby obtaining the second feature retrieval fault tolerance condition. The second feature retrieval fault tolerance condition ensures that slight fluctuations in physiological indexes can be tolerated when retrieving similar historical samples, improving the flexibility and applicability of the retrieval.

[0064] Then, a retrieval operation is performed in the historical sample library to find a set of motion samples that simultaneously meet the first feature retrieval fault tolerance condition and the second feature retrieval fault tolerance condition. The historical sample library is the user usage data accumulated over a long time, which contains detailed records of a large number of different users performing various orientation movements under different physiological states, as well as the corresponding physical sign changes. By performing a dual-condition retrieval in this database, historical cases similar to the current user's motion state and physiological state are found to form a set of motion samples. These samples represent the actual usage situations and health responses of other users under similar conditions. Then, the proportion of the number of samples with physical sign abnormality marks in the set of motion samples to the total number of samples is statistically calculated, and this proportion is set as the risk assessment coefficient. The physical sign abnormality mark refers to the marks of adverse reactions such as abnormal acceleration of the heart rate, drastic fluctuations in blood pressure, dizziness, and nausea in the historical records. By calculating the frequency of physical sign abnormalities in similar situations, a risk assessment coefficient between 0 and 1 is obtained. For example, if there are 100 similar cases retrieved in the set of motion samples, and 75 of them have physical sign abnormalities, then the risk assessment coefficient is 0.75, indicating that there is a 75% probability that the current user will have physical sign abnormalities when performing this orientation movement.

[0065] Through the risk assessment based on historical data, the risk coefficient of the current user performing a specific orientation movement is calculated, providing data support for subsequent safety control decisions, thereby improving the safety and reliability of virtual reality steering control.

[0066] Furthermore, based on the centroid coordinates, attitude quaternions, and operation instruction streams, execute the kinematic simulation engine to generate a motion state vector, including: S410: Constrained by the target motion device model, configure the centroid coordinate record data, attitude quaternion record data, operation instruction stream record data, and motion state vector true value data; S420: Configure three branch networks based on the long short-term memory neural network, configure the main network based on the fully connected neural network, and merge the output layer of the three branch networks and the input layer of the main network to obtain the kinematic simulation engine network architecture; S430: Supervise and train the kinematic simulation engine network architecture through the motion state vector true value data, centroid coordinate record data, attitude quaternion record data, and operation instruction stream record data to obtain the kinematic simulation engine, including: S431: Configure the set of motion physical law formulas through the management terminal; S432: Configure the set of physical law deviation loss terms based on the set of motion physical law formulas; S433: Configure the mean square error function and set it as the data loss term; S434: Supervise and train the kinematic simulation engine network architecture through the motion state vector true value data, centroid coordinate record data, attitude quaternion record data, and operation instruction stream record data. When the losses of the set of physical law deviation loss terms and the data loss term are less than or equal to the corresponding loss thresholds for more than a preset proportion in a continuous preset number of times, generate the kinematic simulation engine.

[0067] In a preferred embodiment, first, a training data set is configured with the target motion device model as a constraint condition. Here, the target motion device model refers to the virtual reality device model actually used by the user. Since the physical characteristics and response characteristics of different devices may vary, a dedicated model needs to be constructed for a specific device model. At the same time, four types of data are configured, namely centroid coordinate record data (the three-dimensional coordinate data of the user's body center of gravity collected by the inertial measurement unit), attitude quaternion record data (the user's head and body attitude data calculated by the head-mounted display and IMU sensors), operation instruction stream record data (the control signal data sent by the user through the input device), and true value data of the motion state vector (the parameter data such as angular velocity, linear acceleration, and joint torque measured by the high-precision sensing device). These historical record data will be used as the training samples of the neural network model to construct the kinematic simulation engine. Then, a hybrid neural network architecture is constructed as the basic framework of the kinematic simulation engine. Specifically, three independent branch networks are configured based on the long short-term memory neural network to process the three types of time series data of centroid coordinates, attitude quaternions, and operation instruction streams respectively. The long short-term memory neural network is suitable for processing time series data and can capture the time-dependent characteristics of the user's movements and operations. At the same time, a backbone network is configured based on the fully connected neural network to integrate the feature information from the three branch networks and generate the final motion state vector. The output layers of the three branch networks and the input layer of the fully connected backbone network are merged to form an end-to-end kinematic simulation engine network architecture, which can simultaneously process multi-modal input data and generate accurate motion state predictions.

[0068] After that, the kinematic simulation engine network architecture is supervised and trained using the true value data of the motion state vector, the centroid coordinate record data, the attitude quaternion record data, and the operation instruction stream record data to obtain the kinematic simulation engine. Specifically, the set of motion physical law formulas is configured through the management terminal. These formulas include physical principles applicable to human motion such as Newton's laws of motion, the law of conservation of angular momentum, and the law of conservation of energy, which are used to constrain the prediction results of the neural network to conform to physical laws. Then, based on the configured set of motion physical law formulas, a set of physical law deviation loss terms is constructed. These loss terms are used to quantify the deviation degree between the prediction results of the neural network and the expected results of the physical laws, ensuring that the model output conforms to basic physical constraints. For example, a loss term based on the conservation of angular momentum is constructed, which will give a large penalty value when the prediction result violates the conservation of angular momentum. After that, the mean square error function is configured as the data loss term, which is used to quantify the difference between the predicted value of the neural network and the actual measured value, thereby effectively measuring the prediction accuracy. Subsequently, the configured training dataset (including the true value data of the motion state vector, the centroid coordinate record data, the attitude quaternion record data, and the operation instruction stream record data) is used to supervise and train the kinematic simulation engine network architecture. During the training process, both the physical law deviation loss term and the data loss term are optimized simultaneously, so that the network prediction results conform to both physical law constraints and are close to the actual measured values. When, in a continuous preset number of times (such as 50 consecutive training cycles), the loss values of both the set of physical law deviation loss terms and the data loss term are less than or equal to the corresponding preset loss thresholds, and the proportion of training cycles that meet the conditions exceeds the preset proportion (such as 90%), it is considered that the model training is completed, and the final kinematic simulation engine is generated.

[0069] By constructing the kinematic simulation engine, it is possible to accurately predict the user's motion state vector (parameters such as angular velocity, linear acceleration, joint torque, etc.) based on the real-time collected centroid coordinates, attitude quaternions, and operation instruction stream data, providing a scientific basis for subsequent risk assessment, thereby improving the safety and accuracy of virtual reality steering control.

[0070] Furthermore, the embodiment of the present application further includes: S810: If the risk assessment coefficient > the preset risk critical value, adjust the motion state vector to obtain an updated motion state vector; S820: According to the updated motion state vector and the physiological index dataset, match the historical sample library through the feature retrieval algorithm, and calculate the updated risk assessment coefficient; S830: If the updated risk assessment coefficient ≤ the preset risk critical value, regulate the device PID control loop according to the updated motion state vector for virtual reality control.

[0071] In a feasible implementation, when it is found that the risk assessment coefficient is greater than the preset risk critical value, it indicates that there is a relatively high risk for the user to perform the directional movement in the current movement state, and it is not advisable to directly execute the original virtual reality control. At this time, instead of simply rejecting the user's operation request, the movement state vector is actively adjusted and optimized to reduce the potential risk. Specifically, the angular velocity parameter can be reduced, the linear acceleration parameter can be decreased, or the joint torque parameter can be lowered, making the movement state smoother and safer, thereby obtaining an updated movement state vector. For example, if the angular velocity in the original movement state vector is 30° / s, it may be adjusted to 20° / s; if the original linear acceleration is 1.5 m / s², it may be adjusted to 1.0 m / s² to reduce the physiological stimulation to the user. Then, according to the adjusted updated movement state vector and the user's current physiological index data set, the feature retrieval algorithm in step S500 is executed again to search for similar cases in the historical sample library and calculate the updated risk assessment coefficient. By re-evaluating the risk level corresponding to the adjusted movement state, the effectiveness of the optimization measure is verified to ensure that the adjusted operation can indeed reduce the user's health risk. After that, it is judged whether the updated risk assessment coefficient is less than or equal to the preset risk critical value. If the condition is met, it indicates that the adjusted movement state has reduced the risk to an acceptable range. At this time, the device PID control loop is regulated according to the updated movement state vector to execute the virtual reality control. Through the adjusted PID parameters, a more gentle and smooth virtual reality steering experience is provided to the user compared to the original request, which not only meets the user's operation requirements but also ensures the safety of the operation process.

[0072] By adjusting the movement state vector when the risk assessment coefficient is greater than the preset risk critical value, the intelligent adjustment and optimization of high-risk operations are realized, rather than simply rejecting the execution. Thus, while ensuring the safety of the user, the coherence and satisfaction of the user experience are maximally improved. This adaptive risk control strategy can dynamically adjust the control parameters according to the actual situation of the user, realizing the intelligence and personalization of virtual reality steering control, and improving the safety while ensuring the immersion in the virtual reality experience.

[0073] Embodiment 2, as Figure 2 shown, based on the same inventive concept as the virtual reality steering control method provided in Embodiment 1, the embodiment of the present invention further provides a virtual reality steering control system, including: A type mapping module 11, configured to map to a predefined classification topology structure in response to the selected directional movement type of the user; A clustering operation module 12, configured to receive the user physiological index data set, input it into the classification topology structure to perform clustering operations, and output a user category label; The real-time monitoring startup module 13 is used to, when the user category label belongs to the potential risk category, authorize the user to start the directional motion type and start real-time monitoring, and collect the centroid coordinates, attitude quaternion, and operation instruction stream; The simulation execution module 14 is used to execute the kinematic simulation engine based on the centroid coordinates, the attitude quaternion, and the operation instruction stream to generate a motion state vector; The risk assessment module 15 is used to match the historical sample library through a feature retrieval algorithm according to the motion state vector and the physiological index data set, and calculate the risk assessment coefficient; The virtual reality control module 16 is used to, if the risk assessment coefficient ≤ the preset risk critical value, regulate the device PID control loop according to the motion state vector for virtual reality control.

[0074] Furthermore, the embodiment of the present application further includes a user authorization management module, and the user authorization management module includes the following execution steps: When the user category label belongs to the explicit risk category, refuse to authorize the user to start the directional motion type; When the user category label belongs to the risk-free category, authorize the user to start the directional motion type.

[0075] Furthermore, the type mapping module 11 includes the following execution steps: Construct a physiological index feature space based on the user physiological index dimension, where the physiological index feature space has a first deterministic risk partition, a first probabilistic risk partition, and a first deterministic safety partition preset by the management terminal; Perform a central tendency analysis according to the risk transaction data set to obtain a risk clustering center threshold, and perform a boundary union truncation on the first deterministic risk partition to obtain a second deterministic risk partition; Perform a central tendency analysis according to the safety transaction data set to obtain a safety clustering center threshold, and perform a boundary intersection truncation on the first deterministic safety partition to obtain a second deterministic safety partition; According to the geometric relationship between the second deterministic risk partition and the second deterministic safety partition, perform a convex hull reconstruction on the first probabilistic risk partition to generate the second probabilistic risk partition; Construct the classification topology structure based on the second deterministic risk partition, the second probabilistic risk partition, and the second deterministic safety partition.

[0076] Furthermore, the type mapping module 11 further includes the following execution steps: Perform a Delphi method weight distribution on the user physiological index attribute set to obtain a physiological index weight distribution vector; Perform pairwise weighted Euclidean norm calculation on the risk transaction data set based on the physiological index weight distribution vector to generate a risk sample distance matrix; Apply the local outlier factor algorithm to the risk sample distance matrix and determine it through minimum value screening to obtain the risk clustering center threshold.

[0077] Furthermore, the risk assessment module 15 includes the following execution steps: Configure the first feature retrieval fault tolerance condition based on the motion state deviation threshold according to the motion state vector; Configure the second feature retrieval fault tolerance condition based on the physiological index deviation threshold according to the physiological index data set; Retrieve the set of motion samples that simultaneously meet the first feature retrieval fault tolerance condition and the second feature retrieval fault tolerance condition in the historical sample library; Count the proportion of the distribution of the number of signs of physical abnormalities in the set of motion samples, which is set as the risk assessment coefficient.

[0078] Furthermore, the simulation execution module 14 includes the following execution steps: Configure the centroid coordinate record data, attitude quaternion record data, operation instruction stream record data, and motion state vector true value data with the target motion device model as the constraint; Configure three branch networks based on the long short-term memory neural network, configure the backbone network based on the fully connected neural network, and merge the output layer of the three branch networks and the input layer of the backbone network to obtain the kinematic simulation engine network architecture; Supervise and train the kinematic simulation engine network architecture through the motion state vector true value data, the centroid coordinate record data, the attitude quaternion record data, and the operation instruction stream record data to obtain the kinematic simulation engine, including: Configure the set of motion physical law formulas through the management terminal; Configure the set of physical law deviation loss terms based on the set of motion physical law formulas; Configure the mean square error function and set it as the data loss term; Supervise and train the kinematic simulation engine network architecture through the motion state vector true value data, the centroid coordinate record data, the attitude quaternion record data, and the operation instruction stream record data. When the losses of the set of physical law deviation loss terms and the data loss term are less than or equal to the corresponding loss thresholds for more than a preset proportion in a continuous preset number of times, generate the kinematic simulation engine.

[0079] Furthermore, the embodiment of the present application further includes a risk adjustment control module, and this module includes the following execution steps: If the risk assessment coefficient > the preset risk critical value, adjust the motion state vector to obtain an updated motion state vector; According to the updated motion state vector and the physiological index data set, match the historical sample library through a feature retrieval algorithm, and calculate an updated risk assessment coefficient; If the updated risk assessment coefficient ≤ the preset risk critical value, control the device PID control loop according to the updated motion state vector for virtual reality control.

[0080] Embodiment III, please refer to Figure 3 , Figure 3 which is a schematic diagram of an embodiment of the electronic device provided by the embodiment of the present invention. As Figure 3 shown, an electronic device 200 provided by an embodiment of the present invention includes a memory 210, a processor 220, and a first computer program 211 stored on the memory 210 and executable on the processor 220. When the processor 220 executes the first computer program 211, a virtual reality steering control method is implemented.

[0081] Embodiment IV, please refer to Figure 4 , Figure 4 which is a schematic diagram of an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. As Figure 4 shown, this embodiment provides a computer-readable storage medium 300, on which a second computer program 311 is stored. When the second computer program 311 is executed by a processor, a virtual reality steering control method is implemented.

[0082] It should be noted that in the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0083] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0084] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and combinations of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded computers, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in one flow Figure 1 one flow or more flows and / or blocks Figure 1 or means for implementing the functions specified in one block or more blocks.

[0085] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in one flow Figure 1 one flow or more flows and / or blocks Figure 1 or means for implementing the functions specified in one block or more blocks.

[0086] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or more flows and / or blocks Figure 1 or means for implementing the functions specified in one block or more blocks.

[0087] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once they learn the basic inventive concept.

[0088] Obviously, those skilled in the art can make various changes and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A steering control method for virtual reality, characterized in that, Including: In response to the selected orienteering type by the user, map it to a predefined classification topology structure; Receive the user's physiological index data set, input it into the classification topology structure to perform clustering operations, and output the user category label; When the user category label belongs to the potential risk category, authorize the user to start the orienteering type and start real-time monitoring, collect the centroid coordinates, attitude quaternion and operation instruction stream; Based on the centroid coordinates, the attitude quaternion and the operation instruction stream, execute the kinematic simulation engine to generate a motion state vector; According to the motion state vector and the physiological index data set, match the historical sample library through the feature retrieval algorithm and calculate the risk assessment coefficient; If the risk assessment coefficient ≤ the preset risk critical value, regulate the device PID control loop according to the motion state vector for virtual reality control.

2. The method according to claim 1, wherein It also includes: When the user category label belongs to the obvious risk category, reject the authorization for the user to start the orienteering type; When the user category label belongs to the risk-free category, authorize the user to start the orienteering type.

3. The method according to claim 1, characterized in that In response to the selected orienteering type by the user, map it to a predefined classification topology structure, including: Construct a physiological index feature space based on the user's physiological index dimension, where the physiological index feature space has a first deterministic risk partition, a first probabilistic risk partition and a first deterministic safety partition preset by the management terminal; Perform a central tendency analysis according to the risk transaction data set to obtain the risk clustering center threshold, and perform a boundary union truncation on the first deterministic risk partition to obtain a second deterministic risk partition; Perform a central tendency analysis according to the safety transaction data set to obtain the safety clustering center threshold, and perform a boundary intersection truncation on the first deterministic safety partition to obtain a second deterministic safety partition; According to the geometric relationship between the second deterministic risk partition and the second deterministic safety partition, perform a convex hull reconstruction on the first probabilistic risk partition to generate a second probabilistic risk partition; Based on the second deterministic risk partition, the second probabilistic risk partition and the second deterministic safety partition, construct the classification topology structure.

4. The method according to claim 3, characterized in that Perform a central tendency analysis according to the risk transaction data set to obtain the risk clustering center threshold, and perform a boundary update on the first deterministic risk partition to obtain a second deterministic risk partition, including: Perform a Delphi method weight distribution on the user's physiological index attribute set to obtain a physiological index weight distribution vector; Based on the physiological index weight distribution vector, perform a pairwise weighted Euclidean norm calculation on the risk transaction data set to generate a risk sample distance matrix; Apply the local outlier factor algorithm on the risk sample distance matrix and determine it through minimum value screening to obtain the risk clustering center threshold.

5. The method according to claim 1, characterized in that, According to the motion state vector and the physiological index data set, match the historical sample library through the feature retrieval algorithm and calculate the risk assessment coefficient, including: According to the motion state vector, configure the first feature retrieval fault tolerance condition based on the motion state deviation threshold; According to the physiological index data set, configure the second feature retrieval fault tolerance condition based on the physiological index deviation threshold; Retrieve a set of motion samples in the historical sample library that simultaneously satisfy the first feature retrieval fault tolerance condition and the second feature retrieval fault tolerance condition; Statistically calculate the proportion of the distribution quantity of the physical sign abnormality identifiers in the set of motion samples, and set it as the risk assessment coefficient.

6. The method according to claim 5, characterized in that, Based on the centroid coordinates, the attitude quaternion, and the operation instruction stream, execute a kinematic simulation engine to generate a motion state vector, including: Constrained by the target motion device model, configure the centroid coordinate record data, the attitude quaternion record data, the operation instruction stream record data, and the motion state vector true value data; Configure three branch networks based on the long short-term memory neural network, configure a backbone network based on the fully connected neural network, and merge the output layers of the three branch networks and the input layer of the backbone network to obtain the kinematic simulation engine network architecture; Supervise and train the kinematic simulation engine network architecture through the motion state vector true value data, the centroid coordinate record data, the attitude quaternion record data, and the operation instruction stream record data to obtain the kinematic simulation engine, including: Configure a set of motion physical law formulas through the management terminal; Based on the set of motion physical law formulas, configure a set of physical law deviation loss terms; Configure a mean square error function and set it as the data loss term; Supervise and train the kinematic simulation engine network architecture through the motion state vector true value data, the centroid coordinate record data, the attitude quaternion record data, and the operation instruction stream record data. When the losses of the set of physical law deviation loss terms and the data loss term are less than or equal to the corresponding loss thresholds for more than a preset proportion in a continuous preset number of times, generate the kinematic simulation engine.

7. The method according to claim 1, wherein It further includes: If the risk assessment coefficient > the preset risk critical value, adjust the motion state vector to obtain an updated motion state vector; According to the updated motion state vector and the physiological index data set, match the historical sample library through a feature retrieval algorithm, and calculate an updated risk assessment coefficient; If the updated risk assessment coefficient ≤ the preset risk critical value, regulate the device PID control loop according to the updated motion state vector for virtual reality control.

8. A steering control system for virtual reality, characterized in that, For implementing a steering control method for virtual reality according to any one of claims 1-7, the system includes: A type mapping module for mapping to a predefined classification topology structure in response to the selected directional motion type by the user; A clustering operation module for receiving the user physiological index data set, inputting it into the classification topology structure to perform clustering operations, and outputting a user category label; A real-time monitoring startup module for authorizing the user to start the directional motion type and starting real-time monitoring to collect centroid coordinates, attitude quaternions, and operation instruction streams when the user category label belongs to a potential risk category; A simulation execution module for executing a kinematic simulation engine based on the centroid coordinates, the attitude quaternion, and the operation instruction stream to generate a motion state vector; A risk assessment module for calculating a risk assessment coefficient by matching the historical sample library through a feature retrieval algorithm according to the motion state vector and the physiological index data set; A virtual reality control module, configured to, if the risk assessment coefficient ≤ the preset risk critical value, regulate a device PID control loop according to the motion state vector to perform virtual reality control.

9. An electronic device, characterized in that, It includes: A memory, configured to store computer software programs; A processor, configured to read and execute the computer software programs, thereby implementing a steering control method for virtual reality according to any one of claims 1-7.

10. A non-transitory computer-readable storage medium, characterized in that, A computer software program is stored in the storage medium, and when the computer software program is executed by a processor, a steering control method for virtual reality according to any one of claims 1-7 is implemented.

Images