A risk scenario library construction method, system, medium and device for occupant protection

Abstract

The application provides a risk scenario library construction method, system, medium and equipment for occupant protection, acquires multi-source data, extracts features from the multi-source data to obtain feature data, classifies risk scenarios corresponding to the feature data based on the feature data to obtain corresponding risk levels, generalizes risk scenarios with extremely high risk and high risk risk levels to obtain multi-attitude high-risk scenarios, verifies physical constraints of the multi-attitude high-risk scenarios, calculates confidence of the high-risk scenarios that pass the physical constraint verification, takes high-risk scenarios with confidence greater than a preset confidence threshold as target risk scenarios, and obtains high-risk scenarios directly used for intelligent cockpit adaptive restraint system development and multi-attitude occupant protection evaluation through multi-source data fusion and generalization, thereby providing data support for improvement of vehicle collision safety performance under the premise of ensuring accuracy and efficiency of the high-risk scenarios.

Description

Technical Field

[0001] This application relates to the field of vehicle collision safety technology, specifically to a method, system, medium, and device for constructing a risk scenario library for occupant protection. Background Technology

[0002] With the advancement of autonomous driving levels, the form of intelligent cockpits is rapidly iterating, and new cockpit technologies such as zero-gravity seats, rotating seats, and adaptive intelligent restraint systems are gradually being mass-produced and applied. Occupant seating positions during driving are no longer limited to the traditional standard driving posture, but rather exhibit a wide range of non-standard, high-degree-of-freedom postures, including leaning forward, lying down, tilting to the side, and reclining. These postures differ significantly from the traditional standard driving posture, leading to a fundamental change in the injury mechanisms of occupants during collisions.

[0003] Existing collision safety protection systems and testing procedures are primarily designed based on the protection of occupants in standard seating positions. Although non-standard seating position occupant protection, such as test conditions related to zero-gravity seats, has been gradually included in the assessment scope in recent years, the coverage of test conditions remains limited and cannot meet the needs of multi-posture occupant safety protection in intelligent cockpits.

[0004] Currently, the industry generally adopts the following technical paths to obtain verification scenarios: First, directly using standard test conditions from regulations, standards, and evaluation procedures. While these conditions are highly targeted, the data source is singular, relying only on limited conditions extracted from regulations, lacking multi-source real data support, and unable to adapt to the rapidly iterating multi-attitude occupant protection needs. Second, enterprises customize typical attitude test cases based on experience. This approach is limited to specific enterprises, highly subjective, lacks systematic data support, and is prone to problems such as incomplete scenario coverage and omission of high-risk conditions. It also fails to form a reusable and scalable scenario system. Whether using regulatory conditions or custom test cases, there is a general lack of dynamic expansion and iteration capabilities, making it impossible to adapt to the rapid updates of intelligent cockpit forms and the emergence of new attitude scenarios. Moreover, most are generalized scenarios, resulting in poor timeliness and applicability of the scenario library, making it difficult to directly meet the engineering development needs of multi-attitude occupant protection.

[0005] Therefore, a method is needed that can integrate multi-source data and systematically construct a multi-posture occupant high-risk scenario library to provide comprehensive, high-fidelity, and scalable scenario support for intelligent cockpit occupant protection. Summary of the Invention

[0006] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a method, system, medium, and device for constructing a risk scenario library for occupant protection.

[0007] According to one aspect of this application, a method for constructing a risk scenario library for occupant protection is provided, comprising: acquiring multi-source data; wherein the multi-source data includes collision accident data, enterprise backend data, collision test data, collision simulation data, and supplementary data, wherein the supplementary data is predicted and generated based on the collision test data and the collision simulation data; extracting features from the multi-source data to obtain feature data; wherein the feature data includes occupant posture features, vehicle collision condition features, equivalent damage condition features, damage risk features, and occupant identity features, wherein the equivalent damage condition features represent the relative orientation of the occupant with respect to the vehicle collision direction; classifying the risk scenarios corresponding to the feature data into risk levels based on the feature data to obtain corresponding risk levels; wherein the risk levels include extremely high risk, high risk, medium risk, and low risk; generalizing the risk scenarios with extremely high risk and high risk levels to obtain multi-posture high-risk scenarios; performing physical constraint verification on the multi-posture high-risk scenarios; calculating the confidence level of the high-risk scenarios that pass the physical constraint verification; and selecting the high-risk scenarios with a confidence level greater than a preset confidence threshold as target risk scenarios.

[0008] In one embodiment, the step of classifying the risk scenarios corresponding to the feature data into risk levels based on the feature data includes: calculating the maximum single damage value of the corresponding risk scenario based on the measured values ​​of damage indicators and the corresponding limits in the feature data; calculating the weighted comprehensive risk value of the corresponding risk scenario based on the measured values ​​of damage indicators and the corresponding limits in the feature data; and obtaining the risk level of the risk scenario based on the maximum single damage value and the weighted comprehensive risk value.

[0009] In one embodiment, calculating the maximum single damage value for a corresponding risk scenario based on the measured values ​​and corresponding limits of the damage indicators in the feature data includes: the formula for calculating the maximum single damage value is: ;in, For the first i The maximum single damage value in a risk scenario. For the first i The first risk scenario k Measured values ​​of each damage index For the first k Limits for each damage index.

[0010] In one embodiment, calculating the weighted comprehensive risk value of the corresponding risk scenario based on the measured values ​​of the damage indicators and the corresponding limits in the feature data includes: the formula for calculating the weighted comprehensive risk value is: ;in, For the first i The weighted composite risk value for each risk scenario. For the first k The weight of each damage indicator For the first i The first risk scenario k Measured values ​​of each damage index For the first k Limits for each damage index.

[0011] In one embodiment, the process of generalizing the risk scenarios with extremely high risk and high risk levels to obtain multi-pose high-risk scenarios includes: performing parameter interpolation on the risk scenarios with extremely high risk and high risk levels under parameter distribution constraints to obtain interpolated scenarios; performing parameter recombination on the risk scenarios with extremely high risk and high risk levels to obtain recombined scenarios; generating boundary scenarios that produce damage greater than a set damage threshold based on the parameter boundaries of the risk scenarios with extremely high risk and high risk levels; and obtaining the multi-pose high-risk scenarios based on the interpolated scenarios, the recombined scenarios, and the boundary scenarios.

[0012] In one embodiment, calculating the confidence level of a high-risk scenario verified by physical constraints includes: calculating the feature distance between a high-risk scenario verified by physical constraints and a real high-risk scenario; and calculating the confidence level of a high-risk scenario verified by physical constraints based on the feature distance.

[0013] In one embodiment, calculating the confidence level of a high-risk scenario verified by physical constraints based on the feature distance includes: the confidence level is calculated using the following formula: ;in, For confidence level, This represents the feature distance between high-risk scenarios verified through physical constraints and real-world high-risk scenarios. To preset the maximum feature distance threshold, , For high-risk scenarios verified through physical constraints, the first i One feature parameter, To verify the first high-risk scenario through physical constraints, the nearest neighbor of the high-risk scenario is a real high-risk scenario. i One feature parameter, n This represents the total number of feature parameters for the risk scenario.

[0014] According to another aspect of this application, a risk scenario library construction system for occupant protection is provided, comprising: a multi-source data acquisition module for acquiring multi-source data; wherein the multi-source data includes collision accident data, enterprise backend data, collision test data, collision simulation data, and supplementary data, wherein the supplementary data is predicted and generated based on the collision test data and the collision simulation data; and a feature data extraction module for extracting features from the multi-source data to obtain feature data; wherein the feature data includes occupant posture features, vehicle collision condition features, equivalent damage condition features, damage risk features, and occupant identity features, wherein the equivalent damage condition features represent the occupant's relative position to the vehicle collision direction. The system includes: a risk level classification module, used to classify risk scenarios corresponding to the feature data based on the feature data to obtain corresponding risk levels; wherein the risk levels include extremely high risk, high risk, medium risk, and low risk; a high-risk scenario generalization module, used to generalize the risk scenarios with risk levels of extremely high risk and high risk to obtain multi-pose high-risk scenarios; a physical constraint verification module, used to perform physical constraint verification on the multi-pose high-risk scenarios; a confidence calculation module, used to calculate the confidence level of high-risk scenarios that pass the physical constraint verification; and a target scenario determination module, used to identify high-risk scenarios with confidence levels greater than a preset confidence threshold as target risk scenarios.

[0015] According to another aspect of this application, a computer-readable storage medium is provided, the storage medium storing a computer program for performing any of the methods described above.

[0016] According to another aspect of this application, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; the processor being configured to perform any of the methods described above.

[0017] This application provides a method, system, medium, and device for constructing a risk scenario library for occupant protection, which acquires multi-source data. This multi-source data includes collision accident data, enterprise backend data, collision test data, collision simulation data, and supplementary data, with the supplementary data predicted and generated based on the collision test data and collision simulation data. Feature extraction is performed on the multi-source data to obtain feature data. This feature data includes occupant posture features, vehicle collision condition features, equivalent damage condition features, damage risk features, and occupant identity features. The equivalent damage condition features represent the relative orientation of the occupant with respect to the vehicle collision direction. Based on the feature data, the risk scenarios corresponding to the feature data are classified into levels. The process involves determining the corresponding risk levels, including extremely high risk, high risk, medium risk, and low risk. For extremely high and high risk scenarios, scenario generalization is performed to obtain multi-pose high-risk scenarios. Physical constraint verification is then performed on these multi-pose high-risk scenarios. The confidence level of high-risk scenarios that pass physical constraint verification is calculated. High-risk scenarios with confidence levels greater than a pre-set confidence threshold are selected as target risk scenarios. Through multi-source data fusion and generalization, high-risk scenarios directly applicable to the development of intelligent cockpit adaptive constraint systems and multi-pose occupant protection assessments are obtained. This ensures the accuracy and efficiency of high-risk scenarios while providing data support for improving vehicle collision safety performance. Attached Figure Description

[0018] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 This is a flowchart illustrating a method for constructing a risk scenario library for occupant protection provided in an exemplary embodiment of this application.

[0020] Figure 2 This is a schematic diagram of the structure of a risk scenario library construction system for occupant protection provided in an exemplary embodiment of this application.

[0021] Figure 3 This is a structural diagram of an electronic device provided in an exemplary embodiment of this application. Detailed Implementation

[0022] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0023] Figure 1 This is a flowchart illustrating a method for constructing a risk scenario library for occupant protection provided in an exemplary embodiment of this application. Figure 1 As shown, the method for constructing the risk scenario library for occupant protection includes the following steps: Step 110: Obtain multi-source data.

[0024] The multi-source data includes collision accident data, enterprise backend data, collision test data, collision simulation data, and supplementary data. The supplementary data is generated based on collision test data and collision simulation data. Specifically, collision accident data is obtained from collision accident cases of smart cockpit vehicles through a traffic accident in-depth investigation system. It focuses on collecting occupant injury data, vehicle EDR information, and accident environment information under non-standard seating postures such as semi-reclined, zero-gravity, and side-tilt positions. This type of data has the highest physical realism and can serve as the baseline truth for the scenario library. Enterprise backend data includes collision accident backend data collected by the enterprise, such as remote monitoring data, accident self-reporting data, and vehicle-to-everything (V2X) accident triggering data. It covers occupant posture, vehicle status, and damage information in actual collision accidents of smart cockpit vehicles, serving as a supplementary source of accident data. The crash test data comprises real vehicle and slide test data, specifically including: crash types such as frontal, side, pole impact, rear-end collision, and oblique angle; crash speeds set to two levels: standard and standard + 10%; seat tilt angles set to three levels, such as standard sitting, semi-reclining, and zero-gravity posture; occupant posture types covering various typical non-standard sitting postures such as upright sitting, reclining, and side sitting; and constraint system response and dummy injury data collected as calibration benchmarks for the simulation model. The crash simulation data is based on a vehicle-occupant constraint system simulation model built using simulation software. First, the model is calibrated against real vehicle and slide test data to ensure simulation accuracy. Based on the calibrated model, parameter combinations such as crash speed, seat tilt angle, occupant posture, and crash type are expanded in batches to generate a simulation dataset covering a wider parameter space. The supplementary data uses the collected collision test data and collision simulation data as training benchmarks to build a predictive model. It generates more detailed parameter combinations within the existing parameter space, fills in sparse data areas, and improves data density. At the same time, it conducts compliant targeted expansion for scarce scenarios such as extreme large-angle lying, rollover, and multiple superimposed conditions. The supplementary data is aligned with the distribution of actual accident data to ensure that the generated data conforms to the statistical distribution of real accident data.

[0025] Preferably, the collision simulation data and supplementary data need to be calibrated using collision accident data, enterprise back-end data, or collision test data to ensure the accuracy of the collision simulation data and supplementary data, while also having broad parameter space coverage.

[0026] Preferably, this application preprocesses the multi-source data by: removing abnormal data, such as data missing by more than 30% or parameters that are obviously distorted; and standardizing units, such as unifying speed to km / h, angle to °, and damage parameters to international standard units.

[0027] Step 120: Extract features from the multi-source data to obtain feature data.

[0028] The feature data includes occupant posture features, vehicle collision condition features, equivalent damage condition features, damage risk features, and occupant identity features. The equivalent damage condition features represent the occupant's relative direction to the vehicle's collision direction. Specifically, occupant posture features include: seating posture type, seat tilt angle, torso tilt angle, occupant H-point, head center of mass coordinates, knee coordinates, and ankle coordinates. Vehicle collision condition features include: collision type, such as frontal, side, pole impact, rear-end collision, oblique angle, rollover, and multiple condition superposition; collision speed, collision angle, collision acceleration, and restraint system operating status and parameters. Equivalent damage condition features: Because the occupant seating direction in a smart cockpit differs from traditional upright seating, the overall vehicle collision direction is not equal to the direction of the occupant injury response. The occupant injury mode is determined by the occupant's relative direction to the vehicle's collision direction, not the vehicle collision type itself. Therefore, this application introduces equivalent damage conditions; specifically, let the vehicle collision direction be θ. 车辆 Using the vehicle coordinate system as a reference (0° is directly forward, 90° is to the right, 180° is directly backward, and -90° is to the left), the occupant's facing direction is θ. 座椅 Then the relative impact angle of the occupants with respect to the direction of collision is: θ 相对 =min(|θ 车辆 θ 座椅 |, 360 - |θ 车辆 θ 座椅 |); The equivalent damage condition is determined by θ 相对 It is determined that: 0°±30° is the equivalent frontal angle, 90°±30° is the equivalent lateral angle, 180°±30° is the equivalent rear-end collision angle, and angles around 45° or 135° are equivalent oblique angles. Rollover scenarios are handled separately. Injury risk characteristics include, but are not limited to, head injury (HIC15, head acceleration), neck injury, chest injury (compression, chest acceleration), shoulder injury (shoulder force), abdominal injury (abdominal force, abdominal compression), lumbar spine injury (lumbar load), pelvic injury (pelvic force), thigh injury (axial force), knee injury (slippage), and lower leg injury (tibialic index, lower leg force). Occupant identification characteristics include: occupant physical characteristics, such as gender, height, weight, and body type percentile, used for adaptation limits and restraint system operating status, etc.

[0029] Step 130: Based on the feature data, classify the risk scenarios corresponding to the feature data into risk levels to obtain the corresponding risk levels.

[0030] The risk levels are categorized as extremely high risk, high risk, medium risk, and low risk. This application classifies the risk scenarios corresponding to the extracted feature data into risk levels to determine the risk level of each risk scenario.

[0031] Step 140: Generalize the risk scenarios with extremely high risk and high risk levels to obtain multi-pose high-risk scenarios.

[0032] This application selects risk scenarios with extremely high risk and high risk levels for scenario generalization in order to obtain more high-risk scenarios.

[0033] Step 150: Perform physical constraint verification for high-risk multi-pose scenarios.

[0034] This application performs physical constraint verification on the generalized high-risk multi-pose scenarios to determine whether the generalized high-risk scenarios conform to engineering practice and collision physics.

[0035] Step 160: Calculate the confidence level of the high-risk scenario verified by physical constraints.

[0036] This application determines the authenticity of a high-risk scenario by calculating the confidence level of a physically constrained high-risk scenario.

[0037] Step 170: Select high-risk scenarios with a confidence level greater than the preset confidence threshold as target risk scenarios.

[0038] This application selects scenarios with high confidence and high risk level as target risk scenarios obtained through generalization in order to construct a risk scenario library.

[0039] This application provides a method for constructing a risk scenario library for occupant protection, which involves acquiring multi-source data. This multi-source data includes collision accident data, enterprise backend data, collision test data, collision simulation data, and supplementary data. The supplementary data is generated based on collision test data and collision simulation data. Feature extraction is performed on the multi-source data to obtain feature data. This feature data includes occupant posture features, vehicle collision condition features, equivalent damage condition features, damage risk features, and occupant identity features. The equivalent damage condition features represent the relative orientation of the occupant with respect to the vehicle's collision direction. Based on the feature data, the risk scenarios corresponding to the feature data are classified into levels to obtain corresponding risk scenarios. Risk levels are categorized into extremely high risk, high risk, medium risk, and low risk. The process involves generalizing high-risk scenarios (extremely high and high risk) to obtain multi-pose high-risk scenarios. Physical constraints are then verified on these multi-pose high-risk scenarios. The confidence level of high-risk scenarios that pass physical constraint verification is calculated. High-risk scenarios with confidence levels greater than a pre-set confidence threshold are selected as target risk scenarios. Through multi-source data fusion and generalization, high-risk scenarios directly applicable to the development of intelligent cockpit adaptive constraint systems and multi-pose occupant protection assessments are obtained. This approach provides data support for improving vehicle collision safety performance while ensuring the accuracy and efficiency of high-risk scenarios.

[0040] In one embodiment, step 130 can be implemented as follows: based on the measured values ​​of damage indicators and the corresponding limits in the feature data, calculate the maximum single damage value of the corresponding risk scenario; based on the measured values ​​of damage indicators and the corresponding limits in the feature data, calculate the weighted comprehensive risk value of the corresponding risk scenario; and based on the maximum single damage value and the weighted comprehensive risk value, obtain the risk level of the risk scenario.

[0041] This application calculates the maximum single damage value and the weighted comprehensive risk value of the corresponding risk scenario, and combines the maximum single damage value and the weighted comprehensive risk value to obtain the risk level of the risk scenario.

[0042] In one embodiment, step 130 can be specifically implemented as follows: the formula for calculating the maximum value of a single damage is: ;in, For the first i The maximum single damage value in a risk scenario. For the first i The first risk scenario k Measured values ​​of each damage index For the first k Limits for each damage index.

[0043] The measured values ​​of the injury indicators include, but are not limited to, head HIC15, head acceleration, neck Nij, neck Fz, chest compression, chest acceleration, shoulder force, abdominal compression, abdominal force, lumbar load, pelvic force, thigh force, knee slippage, tibialic index TI, and lower leg force. The limits for the injury indicators are set based on the high-risk limits for each body part, taking into account different occupant characteristics and genders, and different injury limits are set for different percentile dummies.

[0044] In one embodiment, step 130 can be specifically implemented as follows: the formula for calculating the weighted composite risk value is: ;in, For the first i The weighted composite risk value for each risk scenario. For the first k The weight of each damage indicator For the first i The first risk scenario k Measured values ​​of each damage index For the first k Limits for each damage index.

[0045] The weights of the damage indicators are adjusted and allocated based on the equivalent damage condition, riding posture, and collision condition, and satisfy the following: =1. Specifically, the basic weights are first determined based on the equivalent injury conditions: Based on five types of conditions—equivalent frontal, equivalent side, equivalent rear-end collision, equivalent oblique angle, and equivalent rollover—the basic weight distribution of each occupant injury index is preset. For example, in the frontal equivalent condition: the head and chest have balanced weights and are greater than the neck and lower limb weights; in the side equivalent condition: the head, chest, abdomen, and pelvis have equal weights and are greater than the lower limb weights; in the rear-end collision equivalent condition: the neck has a greater weight than the head and chest, and the head and chest have a greater weight than the lower limb weights; in the oblique angle equivalent condition: the head and chest have balanced weights and are greater than the abdomen and pelvis, and the abdomen and pelvis have a greater weight than the neck and lower limb weights; in the rollover equivalent condition: the head has a greater weight than the neck and lumbar spine. Then, the basic weights are adjusted, specifically through posture correction and fine-tuning of operating conditions. Posture correction is determined by the seat back and torso tilt angles. Based on these angles, adjustments are made to semi-reclining, zero-gravity, and large-angle reclining postures, with a focus on adjusting the weights of vulnerable areas such as the lumbar spine, neck, and head, ensuring that the risk grading aligns with the injury patterns of non-standard sitting postures. As the seat back angle increases, the risk of occupant descent increases, and the weights for abdominal and lumbar spine injuries are correspondingly increased. The specific method involves adjusting the seat back angle θ... 座椅 Divided into three intervals: 20° ≤ θ in normal sitting posture 座椅 <30°, semi-reclining posture 30°≤θ 座椅 <45°, zero-gravity attitude 45°≤θ 座椅≤70°; Normal sitting posture: Correction coefficient for each part is 1.0 (i.e., no adjustment); Semi-reclining posture: Abdominal and lumbar spine weights are multiplied by a coefficient of 1.2, while the weights for other parts remain unchanged; Zero gravity posture: Abdominal and lumbar spine weights are multiplied by a coefficient of 1.4, while the weights for other parts remain unchanged; This application corrects the basic weights based on posture, and then re-normalizes them so that the sum of the weights is 1. The fine-tuning of the working conditions is based on the actual vehicle collision working condition type as the final fine-tuning item, taking into account the engineering rationality of the vehicle collision form, so that the risk score takes into account both the actual force logic of the occupants and the actual collision form of the vehicle; For single working conditions, such as frontal, side, rear-end, oblique angle, rollover, etc., the basic weights remain unchanged; For multi-working-condition superimposed collisions, that is, the vehicle is subjected to composite loads of impacts from multiple directions in the same accident, fine-tuning is performed according to the following rules: Determination of primary and secondary impact directions: Based on the vehicle velocity vector before the collision and the collision direction, the equivalent collision velocity in each direction is calculated, and the velocity component is taken. The direction with the greatest impact is the primary impact direction; the affected areas of each impact direction are determined based on the area with the highest weight under a single working condition in that direction: frontal impact affects the head and chest, side impact affects the chest, abdomen and pelvis, rear-end collision impact affects the neck, rollover impact affects the head and neck, and oblique impact affects the head, chest and torso; the weight of the corresponding area is multiplied by 1.2 to 1.3 according to the primary impact direction, and the weight of the corresponding area according to the secondary impact direction is multiplied by 1.0 to 1.1; if multiple impact directions act on the same area, the weight of that area is multiplied by 1.3, and after fine adjustment, it is normalized again.

[0046] By using the aforementioned dynamic weighting correction mechanism, the actual injury risk of occupants under non-standard postures can be accurately reflected, solving the problem of mismatch between traditional risk classification methods and the injury patterns of occupants in multiple postures, and achieving more accurate risk scoring.

[0047] Based on the maximum value of a single injury and the weighted comprehensive risk value, risk levels are classified according to priority from high to low: Extremely high risk: ≥1.5 or ≥1.5; High risk: Does not meet the criteria for extremely high risk, and ≥1.0 or ≥1.0; Medium risk: Does not meet the criteria for high risk or above, and ≥0.5 or ≥0.5; Low risk: <0.5 and <0.5.

[0048] High-risk and extremely high-risk scenarios are selected as the basis for subsequent exploration and generalization.

[0049] In one embodiment, step 140 can be implemented as follows: performing parameter interpolation on risk scenarios with extremely high and high risk levels under parameter distribution constraints to obtain interpolated scenarios; performing parameter recombination on risk scenarios with extremely high and high risk levels to obtain recombined scenarios; generating boundary scenarios that produce damage greater than a set damage threshold based on the parameter boundaries of risk scenarios with extremely high and high risk levels; and obtaining multi-pose high-risk scenarios based on interpolated scenarios, recombined scenarios, and boundary scenarios.

[0050] For high-risk areas, especially the parameter ranges and boundaries within high-risk scenarios, the generation of refined parameters remains insufficient. This application, based on selected high-risk and extremely high-risk scenarios, combines occupant posture parameters, seat tilt angle, seating orientation, and posture type to perform sparse region generalization and boundary exploration for multi-posture high-risk scenarios. This involves in-depth mining and secondary targeted generation to expand the coverage depth and breadth of the scenario library. Specifically, for sparse intervals within high-risk areas in the multi-posture parameter space, such as the 40°~65° torso tilt angle interval in a semi-reclining posture, under the parameter distribution constraints of existing high-risk scenarios, physically plausible interpolated scenarios are generated to fill the continuous posture gaps within high-risk areas, resulting in interpolated scenarios. Furthermore, using parameters such as collision conditions, equivalent damage conditions, collision speed, seat tilt angle, seating orientation, and torso tilt angle as features, an unsupervised clustering algorithm is employed to identify typical high-risk posture combinations not covered by traditional experience, such as "large tilt angle semi-reclined posture + medium-high collision speed" and "frontal collision + occupant recumbent (equivalent to rear-end collision)". This results in reconstructed scenarios to reveal the damage pattern characteristics under different postures and collision conditions. Simultaneously, with occupant injury as the optimization objective, reinforcement learning is used to actively search for extreme posture parameter combinations that may produce high damage in the boundary regions of multi-posture parameters, such as the seat tilt angle limit and special angle combinations between seating orientation and collision direction, thus discovering novel multi-posture high-risk scenarios. The specific implementation steps are as follows: First, combining the mechanical limits of the intelligent cockpit hardware and the ergonomic constraints of the occupants, the value ranges of key parameters such as seat tilt angle, seating orientation, collision speed, and collision direction are determined to construct a continuous action space. Boundary regions are explored by incrementally adjusting these parameters. Then, based on the current policy network, random sampling is performed within the boundary interval to generate multiple sets of posture collision parameter combinations close to the boundary extremes. Invalid samples exceeding mechanical limits or exhibiting posture deformities are eliminated. The filtered boundary conditions are input into a lightweight agent model to quickly predict injury indicators for the head, neck, chest, lumbar spine, and other parts of the body, calculating the risk value of a single severity indicator and the weighted comprehensive risk value. Furthermore, a reward function is constructed to evaluate the results of each exploration, guiding the agent to search for high-risk boundary regions. The reward function is defined as follows: ; in, Basic risk reward The higher the risk, the greater the reward; therefore, priority should be given to dangerous scenarios. For boundary excitation terms, The distance from the current state to the boundary of the parameter value. σ As a scale parameter, the excitation term increases as the state approaches the boundary, guiding the active exploration of boundary regions such as the seat tilt limit and the special angle between the seating orientation and the collision direction. α、β These are the weighting coefficients, α+β =1, which can improve performance in the early stages of the search. β Encourage boundary exploration; improvements can be made in the later stages of the search. α Focusing on high-risk areas, the specific weight values ​​can be dynamically adjusted according to the exploration phase.

[0051] A policy gradient algorithm is used to update the policy network parameters, with the optimization objective of maximizing cumulative reward. This algorithm can directly handle continuous action spaces, which matches the characteristic of parameters such as seat tilt angle and seating orientation in this application, which can be adjusted within a continuous range. Each iteration outputs a set of boundary posture parameters, which are input into the surrogate model to predict damage and calculate the reward value. Based on the reward results, the policy weights are adjusted in reverse, gradually converging towards boundary parameters with higher damage and higher risk. If the reward value fluctuation is less than a preset threshold of 3% for several consecutive generations (e.g., 15 generations), the algorithm is considered converged. At this point, all boundary iteration samples are extracted and selected to meet the criteria. ≥1.0 or For extreme operating conditions ≥1.0, redundant samples with high repetition or similarity are eliminated to obtain extreme coupling high-risk scenarios that are difficult to cover by human experience and conventional experiments, thus completing the incremental expansion of boundary scenarios.

[0052] In one embodiment, step 160 can be implemented by: calculating the feature distance between the high-risk scenario verified by physical constraints and the real high-risk scenario; and calculating the confidence level of the high-risk scenario verified by physical constraints based on the feature distance.

[0053] This application is based on a set of preset physical constraint functions to eliminate generated scenarios that do not conform to the laws of collision physics or contradict the simulation logic. The physical constraint functions cover dimensions such as the rationality of dummy damage parameters, the physicality of dummy motion response animation, and the matching of constraint system parameters with working conditions, to ensure that the generated scenarios conform to engineering reality and collision physics logic.

[0054] After physical constraint verification, this application further calculates the feature distance between high-risk scenarios that pass physical constraint verification and real high-risk scenarios (covering different seat rotation angles, large reclining postures, zero-gravity modes, differentiated occupant postures, and real compliant postures under multiple seating positions), and calculates the confidence level of high-risk scenarios that pass physical constraint verification based on the feature distance.

[0055] In one embodiment, step 160 can be implemented as follows: the confidence level is calculated using the following formula: ;in, For confidence level, This represents the feature distance between high-risk scenarios verified through physical constraints and real-world high-risk scenarios. To preset the maximum feature distance threshold, , For high-risk scenarios verified through physical constraints, the first i One feature parameter, To verify the first high-risk scenario through physical constraints, the nearest neighbor of the high-risk scenario is a real high-risk scenario. i One feature parameter, n This represents the total number of feature parameters for the risk scenario.

[0056] This application maps the multidimensional input parameters of the generated high-risk scenario to a feature space. These multidimensional input parameters include seat back tilt angle, torso tilt angle, occupant orientation, seating position, occupant posture, and collision condition parameters. The multidimensional feature distance between the high-risk scenario and the real high-risk scenario is calculated to measure the degree of similarity between the generated high-risk scenario and the actual compliant cabin conditions. A smaller feature distance indicates a higher confidence level as the generated high-risk scenario closely matches the actual physical condition distribution; conversely, a larger deviation in attitude and collision parameters indicates more pronounced unreasonable issues such as attitude distortion and constraint interference, resulting in lower physical plausibility.

[0057] This application categorizes and encapsulates the generated high-risk scenario set according to multiple dimensions, and establishes multi-dimensional classification and indexing to form a combined index system that supports rapid retrieval and filtering by any combination of dimensions, specifically including: Vehicle collision scenarios: High-risk scenarios are classified according to collision type into frontal collision, side collision, rear-end collision, oblique collision, rollover collision, and multi-condition superposition collision.

[0058] Equivalent damage condition: based on relative impact angle θ 相对 It is divided into five categories: frontal equivalent, side equivalent, rear-end collision equivalent, oblique angle equivalent, and rollover equivalent, which directly characterize the injury mode suffered by the occupants.

[0059] Risk level dimension: four risk levels: low, medium, high, and extremely high, and a risk classification index is established.

[0060] Posture dimension: Subdivided by sitting posture type and seat tilt angle range to achieve precise scene retrieval.

[0061] Speed ​​dimension: low speed (≤30km / h), medium speed, high speed, etc.

[0062] Figure 2This is a schematic diagram of the structure of a risk scenario library construction system for occupant protection provided in an exemplary embodiment of this application. For example... Figure 2 As shown, the occupant protection risk scenario database construction system 20 includes: a multi-source data acquisition module 21, used to acquire multi-source data; wherein, the multi-source data includes collision accident data, enterprise backend data, collision test data, collision simulation data, and supplementary data, and the supplementary data is generated based on collision test data and collision simulation data; a feature data extraction module 22, used to extract features from the multi-source data to obtain feature data; wherein, the feature data includes occupant posture features, vehicle collision condition features, equivalent damage condition features, damage risk features, and occupant identity features, and the equivalent damage condition features represent the relative orientation of the occupant with respect to the vehicle collision direction; a risk level classification module... Block 23 is used to classify the risk scenarios corresponding to the feature data based on the feature data to obtain the corresponding risk levels; among which, the risk levels include extremely high risk, high risk, medium risk, and low risk; the high-risk scenario generalization module 24 is used to generalize the risk scenarios with risk levels of extremely high risk and high risk to obtain multi-pose high-risk scenarios; the physical constraint verification module 25 is used to perform physical constraint verification on multi-pose high-risk scenarios; the confidence calculation module 26 is used to calculate the confidence of high-risk scenarios that have passed the physical constraint verification; the target scenario determination module 27 is used to select high-risk scenarios with confidence levels greater than a preset confidence threshold as target risk scenarios.

[0063] This application provides a risk scenario library construction system for occupant protection. A multi-source data acquisition module 21 acquires multi-source data, including collision accident data, enterprise backend data, collision test data, collision simulation data, and supplementary data. The supplementary data is generated based on collision test data and collision simulation data. A feature data extraction module 22 extracts features from the multi-source data to obtain feature data. The feature data includes occupant posture features, vehicle collision condition features, equivalent damage condition features, damage risk features, and occupant identity features. The equivalent damage condition features represent the relative orientation of the occupant to the vehicle collision direction. A risk level classification module 23 classifies the risk scenarios corresponding to the feature data based on the feature data to obtain the corresponding risk levels. The risk levels include extremely high risk, high risk, medium risk, and low risk. The high-risk scenario generalization module 24 generalizes the risk scenarios with extremely high risk and high risk levels to obtain multi-posture high-risk scenarios. The physical constraint verification module 25 performs physical constraint verification on the multi-posture high-risk scenarios. The confidence calculation module 26 calculates the confidence of the high-risk scenarios that pass the physical constraint verification. The target scenario determination module 27 takes the high-risk scenarios with a confidence level greater than the preset confidence threshold as the target risk scenarios. Through multi-source data fusion and generalization, high-risk scenarios that can be directly used for the development of intelligent cockpit adaptive constraint systems and multi-posture occupant protection evaluation are obtained. While ensuring the accuracy and efficiency of high-risk scenarios, data support is provided for improving the collision safety performance of the whole vehicle.

[0064] In one embodiment, the risk level classification module 23 can be further configured to: calculate the maximum single damage value of the corresponding risk scenario based on the measured damage index value and the corresponding limit value in the feature data; calculate the weighted comprehensive risk value of the corresponding risk scenario based on the measured damage index value and the corresponding limit value in the feature data; and obtain the risk level of the risk scenario based on the maximum single damage value and the weighted comprehensive risk value.

[0065] In one embodiment, the risk level classification module 23 can be further configured such that the formula for calculating the maximum value of a single injury is: ;in, For the first i The maximum single damage value in a risk scenario. For the first i The first risk scenario k Measured values ​​of each damage index For the first k Limits for each damage index.

[0066] In one embodiment, the risk level classification module 23 can be further configured such that the weighted comprehensive risk value is calculated using the following formula: ;in, For the firsti The weighted composite risk value for each risk scenario. For the first k The weight of each damage indicator For the first i The first risk scenario k Measured values ​​of each damage index For the first k Limits for each damage index.

[0067] In one embodiment, the high-risk scenario generalization module 24 can be further configured to: perform parameter interpolation on risk scenarios with extremely high risk and high risk levels under parameter distribution constraints to obtain interpolated scenarios; perform parameter recombination on risk scenarios with extremely high risk and high risk levels to obtain recombined scenarios; generate boundary scenarios that produce damage greater than a set damage threshold based on the parameter boundaries of risk scenarios with extremely high risk and high risk levels; and obtain multi-pose high-risk scenarios based on interpolated scenarios, recombined scenarios, and boundary scenarios.

[0068] In one embodiment, the confidence calculation module 26 can be further configured to: calculate the feature distance between the high-risk scenario verified by physical constraints and the real high-risk scenario; and calculate the confidence of the high-risk scenario verified by physical constraints based on the feature distance.

[0069] In one embodiment, the confidence calculation module 26 can be further configured such that the confidence calculation formula is: ;in, For confidence level, This represents the feature distance between high-risk scenarios verified through physical constraints and real-world high-risk scenarios. To preset the maximum feature distance threshold, , For high-risk scenarios verified through physical constraints, the first i One feature parameter, To verify the first high-risk scenario through physical constraints, the nearest neighbor of the high-risk scenario is a real high-risk scenario. i One feature parameter, n This represents the total number of feature parameters for the risk scenario.

[0070] Below, for reference Figure 3 This application describes an electronic device according to embodiments thereof. The electronic device may be either or both of a first device and a second device, or a standalone device independent of them, which may communicate with the first device and the second device to receive acquired input signals from them.

[0071] Figure 3 A block diagram of an electronic device according to an embodiment of this application is illustrated.

[0072] like Figure 3 As shown, the electronic device 10 includes one or more processors 11 and memory 12.

[0073] The processor 11 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0074] The memory 12 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may execute the program instructions to implement the methods of the various embodiments of this application described above and / or other desired functions. Various contents such as input signals, signal components, and noise components may also be stored in the computer-readable storage medium.

[0075] In one example, the electronic device 10 may also include an input device 13 and an output device 14, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0076] When the electronic device is a standalone device, the input device 13 can be a communication network connector for receiving the collected input signals from the first device and the second device.

[0077] In addition, the input device 13 may also include, for example, a keyboard, a mouse, etc.

[0078] The output device 14 can output various information to the outside, including determined distance information, direction information, etc. The output device 14 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0079] Of course, for the sake of simplicity, Figure 3 Only some of the components of the electronic device 10 relevant to this application are shown in this illustration; components such as buses, input / output interfaces, etc., are omitted. In addition, the electronic device 10 may include any other suitable components depending on the specific application.

[0080] In addition to the methods and apparatus described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of this application described in the "Exemplary Methods" section above.

[0081] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0082] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of this application described in the "Exemplary Methods" section above.

[0083] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0084] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0085] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0086] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0087] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0088] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for constructing a risk scenario library for occupant protection, characterized in that, include: Acquire multi-source data; wherein, the multi-source data includes collision accident data, enterprise back-end data, collision test data, collision simulation data, and supplementary data, and the supplementary data is generated based on the collision test data and the collision simulation data; Feature extraction is performed on the multi-source data to obtain feature data; wherein, the feature data includes occupant posture features, vehicle collision condition features, equivalent damage condition features, damage risk features, and occupant identity features, and the equivalent damage condition features represent the relative orientation of the occupant with respect to the vehicle collision direction; Based on the feature data, the risk scenarios corresponding to the feature data are classified into levels to obtain the corresponding risk levels; wherein, the risk levels include extremely high risk, high risk, medium risk, and low risk. The risk scenarios with extremely high risk and high risk levels are generalized to obtain multi-pose high-risk scenarios. Physical constraint verification was performed on the aforementioned high-risk multi-pose scenarios; Calculate the confidence level of high-risk scenarios that have passed physical constraint verification; High-risk scenarios with a confidence level greater than a preset confidence threshold are designated as target risk scenarios.

2. The method for constructing a risk scenario library for occupant protection according to claim 1, characterized in that, The step of classifying the risk scenarios corresponding to the feature data into risk levels based on the feature data to obtain the corresponding risk levels includes: Based on the measured values ​​of damage indicators and their corresponding limits in the feature data, the maximum single damage value for the corresponding risk scenario is calculated. Based on the measured values ​​of damage indicators and their corresponding limits in the feature data, the weighted comprehensive risk value of the corresponding risk scenario is calculated; The risk level of the risk scenario is obtained based on the maximum value of the single damage and the weighted comprehensive risk value.

3. The method for constructing a risk scenario library for occupant protection according to claim 2, characterized in that, The calculation of the maximum single damage value for the corresponding risk scenario based on the measured values ​​of damage indicators and corresponding limits in the feature data includes: The formula for calculating the maximum value of a single injury is: ； in, For the first i The maximum single damage value in a risk scenario. For the first i The first risk scenario k Measured values ​​of each damage index For the first k Limits for each damage index.

4. The method for constructing a risk scenario library for occupant protection according to claim 2, characterized in that, The calculation of the weighted comprehensive risk value for the corresponding risk scenario based on the measured values ​​of damage indicators and corresponding limits in the feature data includes: The formula for calculating the weighted composite risk value is as follows: ； in, For the first i The weighted composite risk value for each risk scenario. For the first k The weight of each damage indicator For the first i The first risk scenario k Measured values ​​of each damage index For the first k Limits for each damage index.

5. The method for constructing a risk scenario library for occupant protection according to claim 1, characterized in that, The scenario generalization of the risk scenarios with extremely high risk and high risk levels to obtain multi-pose high-risk scenarios includes: For the risk scenarios with risk levels of extremely high risk and high risk, parameter interpolation is performed under parameter distribution constraints to obtain the interpolated scenarios; The parameters of the risk scenarios with extremely high risk and high risk levels are recombined to obtain recombined scenarios; Based on the parameter boundaries of the risk scenarios with risk levels of extremely high risk and high risk, a boundary scenario that generates damage greater than a set damage threshold is generated. Based on the interpolation scenario, the recombination scenario, and the boundary scenario, the multi-pose high-risk scenario is obtained.

6. The method for constructing a risk scenario library for occupant protection according to claim 1, characterized in that, The confidence level of the high-risk scenario verified by the calculation through physical constraints includes: Calculate the feature distance between high-risk scenarios verified by physical constraints and real high-risk scenarios; Based on the feature distance, the confidence level of the high-risk scenario verified by physical constraints is calculated.

7. The method for constructing a risk scenario library for occupant protection according to claim 6, characterized in that, The calculation of the confidence level of high-risk scenarios verified by physical constraints based on the feature distance includes: The confidence level is calculated using the following formula: ； in, For confidence level, This represents the feature distance between high-risk scenarios verified through physical constraints and real-world high-risk scenarios. To preset the maximum feature distance threshold, , For high-risk scenarios verified through physical constraints, the first i One feature parameter, To verify the first high-risk scenario through physical constraints, the nearest neighbor of the high-risk scenario is a real high-risk scenario. i One feature parameter, n This represents the total number of feature parameters for the risk scenario.

8. A system for constructing a risk scenario library for occupant protection, characterized in that, include: A multi-source data acquisition module is used to acquire multi-source data; wherein, the multi-source data includes collision accident data, enterprise back-end data, collision test data, collision simulation data, and supplementary data, and the supplementary data is generated based on the collision test data and the collision simulation data; The feature data extraction module is used to extract features from the multi-source data to obtain feature data; wherein, the feature data includes occupant posture features, vehicle collision condition features, equivalent damage condition features, damage risk features, and occupant identity features, and the equivalent damage condition features represent the relative orientation of the occupant with respect to the vehicle collision direction. The risk level classification module is used to classify the risk scenarios corresponding to the feature data based on the feature data to obtain the corresponding risk levels; wherein, the risk levels include extremely high risk, high risk, medium risk, and low risk. The high-risk scenario generalization module is used to generalize the risk scenarios with extremely high risk and high risk levels to obtain multi-pose high-risk scenarios. The physical constraint verification module is used to perform physical constraint verification on the multi-pose high-risk scenario; The confidence calculation module is used to calculate the confidence level of high-risk scenarios that have passed the physical constraint verification. The target scenario determination module is used to identify high-risk scenarios with a confidence level greater than a preset confidence threshold as target risk scenarios.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program for performing the method described in any one of claims 1-7.

10. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is used to execute the method described in any one of claims 1-7.

Images