Method, platform and storage medium for optimizing vehicle passive safety model, and electronic device

Abstract

The application discloses a kind of optimization method of vehicle passive safety model, platform and storage medium, electronic equipment, method includes: constructing knowledge base, wherein, knowledge base includes the vehicle type parameter of sample vehicle, collision parameter, corresponding to passive safety measure of collision parameter, and passive safety model;According to the vehicle type parameter and collision parameter of target vehicle, from knowledge base, based on similarity determine passive safety model to be optimized;Based on the passive safety measure constraint condition of target vehicle, iteration optimization is carried out to passive safety model to be optimized, and target passive safety model and target passive safety measure are obtained, to be used for the passive safety configuration of target vehicle.This method quickly establishes the passive safety model to be optimized of target vehicle using the constructed knowledge base, and the target passive safety model and target passive safety measure of target vehicle can be quickly obtained by iteration optimization to passive safety model to be optimized.

Description

Technical Field

[0001] This invention relates to the field of vehicle constraint system optimization, and more particularly to an optimization method, platform, storage medium, and electronic device for a vehicle passive safety model. Background Technology

[0002] With the diversification of intelligent driving technology and intelligent cockpit scenarios, occupant protection also needs to provide multi-scenario, multi-dimensional protection. As intelligent cockpit technology upgrades and L2+ level autonomous driving becomes more widespread, user demands for occupant restraint systems have shifted from "passive collision protection" to "active prediction, personalized adaptation, and full-scenario coverage."

[0003] In the development of automotive constraint systems, optimizing safety performance is a complex system problem involving multiple variables, strong coupling, and high computational costs. Current optimization paradigms rely heavily on engineer experience and suffer from fragmented toolchains, resulting in significant bottlenecks. Summary of the Invention

[0004] This invention aims to at least partially address one of the technical problems in related technologies. Therefore, one objective of this invention is to propose an optimization method for a vehicle passive safety model, enabling the optimization of the vehicle passive safety model to reuse optimization experience and improving the optimization efficiency of the vehicle passive safety model.

[0005] The second objective of this invention is to propose an optimization platform for a vehicle passive safety model.

[0006] A third objective of this invention is to provide a computer-readable storage medium.

[0007] The fourth objective of this invention is to provide an electronic device.

[0008] To achieve the above objectives, a first aspect of the present invention proposes an optimization method for a vehicle passive safety model. The method includes: constructing a knowledge base, wherein the knowledge base includes vehicle model parameters, collision parameters, passive safety measures corresponding to the collision parameters, and a passive safety model for the sample vehicle. Based on the vehicle model parameters and collision parameters of the target vehicle, a passive safety model to be optimized is determined from the knowledge base based on similarity; based on the passive safety measure constraints of the target vehicle, the passive safety model to be optimized is iteratively optimized to obtain the target passive safety model and target passive safety measures, which are used for the passive safety configuration of the target vehicle.

[0009] According to the vehicle passive safety model optimization method of the present invention, the passive safety model to be optimized of the target vehicle is quickly established by utilizing the constructed knowledge base. By iteratively optimizing the passive safety model to be optimized, the target passive safety model and target passive safety measures of the target vehicle are obtained, so that the optimization experience of the vehicle passive safety model can be reused and the optimization efficiency of the vehicle passive safety model is improved.

[0010] In addition, the optimization method for the vehicle passive safety model proposed in the above embodiments of the present invention may also have the following additional technical features: According to one embodiment of the present invention, the iterative optimization of the passive safety model to be optimized based on the passive safety measures constraints of the target vehicle includes: searching for the optimal passive safety model within the parameter range defined by the passive safety measures constraints; during the search process, evaluating the damage prediction uncertainty of the passive safety model to be optimized in the searched parameter region and determining the parameter uncertainty region; determining sampling points within the parameter uncertainty region and performing high-fidelity simulation on the sampling points to generate corresponding simulated damage values; and using the simulated damage values, the sampling points, and the vehicle model parameters and collision parameters of the target vehicle to incrementally train the passive safety model to be optimized.

[0011] According to one embodiment of the present invention, the damage prediction uncertainty of the passive safety model to be optimized in the searched parameter region is evaluated by a meta-model.

[0012] According to an embodiment of the present invention, the step of evaluating the damage prediction uncertainty of the passive safety model to be optimized in the searched parameter region and determining the parameter uncertainty region includes: obtaining the uncertainty quantification index corresponding to each of the searched parameter points through the meta-model as an evaluation value; and determining the parameter uncertainty region based on the searched parameter points whose evaluation values ​​are greater than a preset evaluation threshold.

[0013] According to an embodiment of the present invention, determining the sampling point within the parameter uncertainty region includes: determining the sampling point within the parameter uncertainty region using a preset sampling decision criterion, wherein the preset sampling decision criterion includes at least one of the following: maximizing prediction variance criterion, expected improvement criterion, and upper confidence boundary criterion.

[0014] According to an embodiment of the present invention, the iterative optimization of the passive safety model to be optimized based on the passive safety measures constraints of the target vehicle further includes: after the optimization converges, performing high-fidelity simulation verification on the candidate optimal solutions in the candidate optimal solution set, wherein the candidate optimal solution set is obtained by optimization with the objective of minimizing the damage prediction value output by the passive safety model to be optimized; if the verification is successful, the candidate optimal solution that has been verified is taken as the target passive safety measure; if the verification fails, the candidate optimal solution that has failed the verification is taken as new training data, and incremental training and optimization are performed again.

[0015] According to one embodiment of the present invention, performing high-fidelity simulation verification on a candidate optimal solution includes: performing high-fidelity simulation on the candidate optimal solution to obtain the corresponding simulated damage value, and comparing the simulated damage value with the predicted damage value predicted by the passive safety model to be optimized; if the comparison result meets the preset requirements, the verification is passed; otherwise, the verification fails.

[0016] To achieve the above objectives, a second aspect of the present invention proposes an optimization platform for a vehicle passive safety model. The platform includes: a construction module for constructing a knowledge base, wherein the knowledge base includes model parameters, collision parameters, passive safety measures corresponding to the collision parameters, and a passive safety model for sample vehicles; and an optimization module for determining a passive safety model to be optimized from the knowledge base based on similarity, according to the model parameters and collision parameters of a target vehicle; and iteratively optimizing the passive safety model to be optimized based on the constraints of the passive safety measures of the target vehicle to obtain a target passive safety model and target passive safety measures for use in the passive safety configuration of the target vehicle.

[0017] To achieve the above objectives, a third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the optimization method for the vehicle passive safety model as described above.

[0018] To achieve the above objectives, a fourth aspect of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the optimization method for the vehicle passive safety model as described above.

[0019] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] Figure 1 This is a flowchart of an optimization method for a vehicle passive safety model according to an embodiment of the present invention; Figure 2 This is an embodiment of the iterative optimization process of the passive security model to be optimized according to the present invention. Figure 1 ; Figure 3 This is a flowchart illustrating the determination of an uncertain region of parameters according to an embodiment of the present invention; Figure 4 This is an embodiment of the iterative optimization process of the passive security model to be optimized according to the present invention. Figure 2 ; Figure 5 This is a flowchart of a high-fidelity simulation verification of a candidate optimal solution according to an embodiment of the present invention; Figure 6 This is a schematic diagram of an optimization platform for a vehicle passive safety model according to an embodiment of the present invention; Figure 7 This is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0021] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0022] It's important to note that the traditional serial iterative development process (currently the mainstream in engineering time) is the most prevalent but least efficient model in the industry. Its process is strictly sequential: first, engineers set a set of initial design parameters based on experience; second, these are submitted to the CAE (Computer Aided Engineering) department for time-consuming high-fidelity collision simulations (lasting from hours to several days per simulation); third, safety engineers analyze the damage results and propose modifications; finally, this cycle is repeated. Essentially, this process is a human-centered, trial-and-error "open-loop" system. Its fundamental flaws are: first, the process is severely fragmented, with inefficient and distorted data and knowledge transfer between different departments and tools; second, its exploratory capabilities are limited, constrained by human experience and simulation costs, allowing exploration of only a very small portion of the design space; and third, it heavily relies on experts, involves much repetitive work, and makes it difficult to accumulate and reuse design knowledge.

[0023] To improve optimization efficiency, related technologies attempt to introduce automation into local processes, providing auxiliary optimization methods based on isolated surrogate models. For example, a typical approach involves training an independent surrogate model using a small amount of CAE data to predict the relationship between design parameters and a certain damage index, and then combining this with optimization algorithms for optimization. However, this method has a fundamental flaw: it's a static "point-to-point" tool replacement, where the data required for training the surrogate model is pre-fixed and generated all at once. The model and data generation processes are completely disconnected. When optimization enters a new region or the existing model's accuracy is found to be insufficient, it cannot automatically trigger supplementary data generation and online model updates, potentially causing the optimization process to fail due to model inaccuracies or get stuck in local optima.

[0024] Some commercial software frameworks integrate simulation domains and optimization, and general simulation optimization platforms provide universal interfaces for connecting simulation scripts and optimization algorithms. However, they suffer from fundamental flaws; these platforms are loosely integrated, like glue. They only solve the connection problem between tools but do not address the inherent contradictions between data generation, model accuracy, and optimization objectives in constraint system optimization. The platforms themselves lack the intelligence required for safety optimization tasks; they cannot understand why sampling occurs at a particular point, when the model needs updating, or how to balance multiple damage targets. All strategies still require manual pre-configuration, resulting in a low level of intelligence.

[0025] To address the aforementioned problems, embodiments of the present invention provide a method, platform, storage medium, and electronic device for optimizing a vehicle passive safety model. The optimization method, platform, storage medium, and electronic device of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] Figure 1 This is a flowchart of an optimization method for a vehicle passive safety model according to an embodiment of the present invention. Figure 1 As shown, the optimization methods for vehicle passive safety models may include: S101, Construct a knowledge base, which may include the model parameters, collision parameters, passive safety measures corresponding to the collision parameters, and passive safety models of the sample vehicles; S102, Based on the vehicle model parameters and collision parameters of the target vehicle, determine the passive safety model to be optimized from the knowledge base based on similarity; S103, based on the passive safety measures constraints of the target vehicle, iteratively optimize the passive safety model to be optimized to obtain the target passive safety model and target passive safety measures, which are then used for the passive safety configuration of the target vehicle.

[0027] The vehicle passive safety model optimization method in this embodiment of the invention is used in a vehicle passive safety model optimization platform.

[0028] Specifically, a large number of sample vehicles, such as existing vehicle model parameters, collision parameters, corresponding passive safety measures, and passive safety models, are imported into a knowledge base. When optimizing the target passive safety model and target passive safety measures for the target vehicle, the vehicle model parameters and collision parameters of the target vehicle are obtained. Based on these parameters, the knowledge base is queried for sample vehicles in the knowledge base that are most similar to the target vehicle in terms of model parameters and collision parameters. The passive safety model corresponding to the sample vehicle with the highest similarity is recorded as the passive safety model to be optimized for the target vehicle. The constraints of the passive safety measures for the target vehicle are obtained. Based on these constraints, the passive safety model to be optimized is iteratively optimized to obtain the target passive safety model and target passive safety measures, which are then used for the passive safety configuration of the target vehicle.

[0029] In one embodiment of the present invention, passive safety measures corresponding to collision parameters may include seat belt tightening measures, airbag deployment measures, etc.

[0030] This invention constructs a knowledge base based on existing vehicle model parameters, collision parameters, corresponding passive safety measures, and passive safety models. Based on the vehicle model parameters and collision parameters of the new vehicle (target vehicle), the knowledge base seeks the passive safety model with the highest similarity, denoted as the passive safety model to be optimized. Based on the passive safety measure constraints of the target vehicle, the passive safety model to be optimized is iteratively optimized to obtain a completely new model. This new model ensures that it can fully adapt to the collision scenario of the new vehicle model, ensuring that safety measures are in place and achieving the highest safety score under this collision condition.

[0031] The vehicle passive safety model optimization method of this invention utilizes a constructed knowledge base to quickly build a passive safety model of the target vehicle to be optimized. By iteratively optimizing the passive safety model to be optimized, the target passive safety model and target passive safety measures of the target vehicle can be obtained. This allows the optimization experience of the vehicle passive safety model to be reused, thereby improving the optimization efficiency of the vehicle passive safety model.

[0032] In one embodiment of the present invention, such as Figure 2 As shown, based on the constraints of the passive safety measures for the target vehicle, iterative optimization of the passive safety model to be optimized may include: S201, within the parameter range limited by the constraints of passive safety measures, optimize the passive safety model to be optimized; S202, During the optimization process, assess the damage prediction uncertainty of the passive safety model to be optimized in the searched parameter region, and determine the parameter uncertainty region; S203, determine the sampling points in the region of parameter uncertainty, and perform high-fidelity simulation on the sampling points to generate the corresponding simulated damage values; S204 uses simulated damage values, sampling points, and the vehicle model parameters and collision parameters of the target vehicle to perform incremental training on the passive safety model to be optimized.

[0033] In one specific embodiment, the damage prediction uncertainty of the passive safety model to be optimized in the region of searched parameters can be evaluated through a meta-model.

[0034] Specifically, a target optimization algorithm can be used to drive the passive safety model to be optimized, seeking the optimal solution within the parameter range defined by the constraints of passive safety measures. Specifically, the target optimization algorithm drives the passive safety model to be optimized to continuously sample and predict losses within the parameter range defined by the constraints of passive safety measures, in order to find the optimal solution. Here, the target optimization algorithm is the optimization algorithm used for the passive safety model corresponding to the sample vehicle in the knowledge base that is most similar to the target vehicle in terms of vehicle type parameters and collision parameters.

[0035] During the optimization process, a meta-model is used to evaluate the damage prediction uncertainty of the passive safety model to be optimized within the searched parameter region. The existence of parameter uncertainty regions is determined based on the evaluation values ​​output by the meta-model. If parameter uncertainty regions exist, sampling points are identified within these regions. A high-fidelity simulation verification algorithm is used to perform high-fidelity simulations on the sampling points, generating corresponding simulated damage values. Incremental training of the passive safety model to be optimized is then performed using the simulated damage values, sampling points, and the target vehicle's model parameters and collision parameters.

[0036] When incrementally training a passive security model to be optimized, only some network weights of the passive security model to be optimized can be updated, thereby quickly and cost-effectively improving the prediction accuracy of the passive security model to be optimized in the region of parameter uncertainty.

[0037] In one specific embodiment of the present invention, such as Figure 3 As shown, evaluating the damage prediction uncertainty of the passive safety model to be optimized within the searched parameter region and identifying the parameter uncertainty region may include: S301, the uncertainty quantification index corresponding to each searched parameter point is obtained through the meta-model and used as the evaluation value; S302, determine the parameter uncertainty region based on the searched parameter points whose evaluation values ​​are greater than the preset evaluation threshold.

[0038] Specifically, the searched parameter points are input into the meta-model, which outputs an uncertainty quantification index corresponding to each searched parameter point. The uncertainty quantification index output by the meta-model is used as the evaluation value of the searched parameter points. The evaluation value corresponding to each searched parameter point is compared with a preset evaluation threshold, and searched parameter points whose evaluation values ​​are greater than the preset evaluation threshold are identified. Based on the searched parameter points whose evaluation values ​​are greater than the preset evaluation threshold, the parameter uncertainty region is determined.

[0039] It should be noted that the larger the evaluation value, the greater the deviation of the damage prediction value output by the passive safety model to be optimized, and the less accurate the evaluation of the passive safety model to be optimized; the smaller the evaluation value, the smaller the deviation of the damage prediction value output by the passive safety model to be optimized, and the more accurate the evaluation of the passive safety model to be optimized.

[0040] In practice, a neighborhood clustering method can be used to determine the parameter uncertainty region based on the searched parameter points whose evaluation values ​​are greater than a preset evaluation threshold. Specifically, when the meta-model evaluates the searched parameter points, if it detects that the evaluation value of one or more consecutive searched parameter points is greater than the preset evaluation threshold, it marks the neighborhood space around the searched parameter points whose evaluation values ​​are greater than the preset evaluation threshold (e.g., a hypersphere with a radius of a certain distance centered on the searched parameter points whose evaluation values ​​are greater than the preset evaluation threshold, or a continuous subspace in the target parameter space) as the parameter uncertainty region.

[0041] Alternatively, the region growing method can be used, starting from the over-standard point (the searched parameter point whose evaluation value is greater than the preset evaluation threshold) and extending outwards to the surrounding space until the searched parameter point whose evaluation value is lower than the preset evaluation threshold is encountered. The connected region formed is the parameter uncertainty region.

[0042] It should be noted that the embodiments of the present invention do not limit the specific implementation methods for determining the uncertain region of parameters.

[0043] To improve the intelligence level of vehicle passive safety model optimization, this invention utilizes a meta-model to evaluate the parameter uncertainty region of damage prediction within the parameter range defined by passive safety measures during the optimization process of the passive safety model to be optimized. High-fidelity simulated damage values ​​are then generated by sampling within this parameter uncertainty region. Incremental training of the passive safety model to be optimized is performed using the simulated damage values, sampling points, and the vehicle model parameters and collision parameters of the target vehicle. This improves the prediction accuracy of the passive safety model, forming a data-model symbiosis and constructing a collaborative optimization environment with domain awareness and self-evolution, thereby achieving the optimization of the vehicle passive safety model.

[0044] The optimization method of this invention enables the passive safety model to be optimized to seek the best within the parameter range limited by the constraints of passive safety measures. It automatically performs sampling of the uncertain parameter region, generation of incremental training data, and incremental update of the passive safety model to be optimized.

[0045] In this embodiment of the invention, the meta-model can be a Gaussian process meta-model. It should be noted that the meta-model can also be implemented using ensemble learning methods, such as training multiple isomorphic but differently initialized proxy models (e.g., neural networks). For the searched parameter points, the evaluation results output by each isomorphic but differently initialized proxy model are obtained. The variance of each evaluation result is calculated, and the obtained variance is used as an uncertainty quantification index. Alternatively, in a deep learning proxy model, techniques such as Monte Carlo Dropout can be used to perform multiple forward propagations during inference, and the variance of its output can be used as an uncertainty quantification index.

[0046] In one specific embodiment, the meta-model adopts a Gaussian process meta-model, and the uncertainty quantification index may include at least one of the predicted mean and the predicted variance. The corresponding preset evaluation threshold may include at least one of the preset mean threshold and the preset variance threshold. The preset mean threshold and the preset variance threshold can be dynamically set according to the simulation cost and model accuracy requirements.

[0047] Implementably, a Gaussian process meta-model is constructed. The training dataset of the passive safety model to be optimized is used... As input to the Gaussian process meta-model, where, The parameter points corresponding to passive safety measures for collision parameters (such as airbag deployment time and seat belt force limit). The damage values ​​(real or simulated damage values) are obtained through high-fidelity simulation. The Gaussian process meta-model is based on the training dataset. A Gaussian process model is fitted to obtain multiple prediction distributions. The training dataset for the passive safety model to be optimized can be the training dataset used by the passive safety models of the sample vehicles in the knowledge base that are most similar to the target vehicle in terms of vehicle type and collision parameters.

[0048] For the searched parameter points, the Gaussian process meta-model not only outputs the predicted mean. It also outputs the prediction variance simultaneously. The prediction variance of the Gaussian process meta-model output. This can serve as a quantification of the uncertainty of the passive safety model to be optimized at the point where the parameters have already been searched. Variance The larger the value, the lower the prediction confidence of the passive safety model to be optimized in that region.

[0049] When the Gaussian process metamodel identifies the prediction variance of a certain searched parameter point Exceeding the preset variance threshold When this occurs, the optimization process of the passive safety model to be optimized is automatically interrupted. Based on the variance distribution of the searched parameter region output by the Gaussian process meta-model, the continuously satisfied parameters are then... The subspace is identified as a high-uncertainty region and used as the parameter uncertainty region for subsequent active sampling. Sampling of the uncertain region and generation of simulated damage values ​​are triggered. Using the simulated damage values, sampling points, and the target vehicle's model and collision parameters, incremental training is performed on the passive safety model to be optimized. After completing the incremental training of the passive safety model to be optimized, optimization of the passive safety model is automatically resumed. At this point, because the accuracy of the passive safety model to be optimized has improved, the optimization algorithm can more reliably advance towards a better region.

[0050] The embodiments of this invention adopt a closed-loop mechanism of "evaluation-interruption-sampling-update-recovery", which realizes the deep interweaving and mutual driving of the two major processes of data generation and model optimization, and is the fundamental guarantee for the platform to achieve efficient exploration.

[0051] In one embodiment of the present invention, determining sampling points within a region of parameter uncertainty includes: A preset sampling decision criterion is adopted to determine sampling points in the region of parameter uncertainty. The preset sampling decision criterion includes at least one of the following: maximizing prediction variance criterion, expected improvement criterion, and upper confidence boundary criterion.

[0052] In practice, when using the maximization of prediction variance criterion, the point with the largest uncertainty quantification index within the parameter uncertainty region, such as prediction variance, should be selected. The maximum point. The mathematical expression for the criterion of maximizing prediction variance is: ,in, As the next candidate point, Candidate points The predictive variance. This criterion aims to directly explore the region of greatest cognitive uncertainty.

[0053] In practice, when adopting a pre-defined expected improvement criterion (which is more effective in the context of optimization), the current optimal objective function value is defined as... For candidate points The desired improvement is as follows: ,in, Candidate points Expected improvement value This represents the current optimal objective function value. For the objective function in The value at that location, The distribution is given by the meta-model (e.g., a normal distribution under the Gaussian process assumption). ).choose The biggest point: , As the next candidate point, Candidate points The expected improvement value. This criterion balances "exploration" and "exploitation," tending to select candidate points that are likely to significantly improve the current optimal solution.

[0054] In practice, when using the upper confidence boundary criterion, select The largest point, of which, Candidate points The upper confidence boundary value, To predict the mean, For balancing parameters, This represents the predicted standard deviation.

[0055] It should be noted that, while maintaining the core concepts of task integration, data-model symbiosis, and knowledge accumulation, proactive sampling decision-making can also introduce expected improvement criteria in addition to being based on uncertainty. That is, it evaluates the potential improvement expected by each candidate sampling point to the current optimization objective function value, selects the point with the highest expectation for simulation, and makes data generation directly target performance improvement.

[0056] The optimization method in this embodiment of the invention solves the problem of disconnect between data, model and optimization goal. Data generation does not take into account model requirements and optimization progress, model updates lag behind design exploration, and optimization goals (such as the weight of multiple damage indicators) are rigid. The three cannot dynamically coordinate according to the task status.

[0057] In one embodiment of the present invention, such as Figure 4 As shown, based on the constraints of the passive safety measures for the target vehicle, iterative optimization of the passive safety model to be optimized may also include: S401, after the optimization converges, perform high-fidelity simulation verification on the candidate optimal solutions in the candidate optimal solution set. The candidate optimal solution set is obtained by optimizing the damage prediction value output by the passive safety model to be optimized. S402, If the verification passes, the candidate optimal solution that passes the verification will be used as the target passive security measure; S403 If the verification fails, the candidate optimal solution that failed the verification will be used as new training data, and incremental training and optimization will be performed.

[0058] Specifically, with the objective of minimizing the damage prediction value output by the passive safety model to be optimized, the passive safety model is optimized to obtain a set of candidate optimal solutions. The candidate optimal solutions in the set are then input into a high-fidelity simulation verification algorithm, which performs simulation verification on the input candidate optimal solutions. If the verification passes, the verified candidate optimal solution is recorded as the target passive safety measure (target optimal solution) and stored in a knowledge base. If the verification fails, the passive safety model to be optimized continues to be iteratively trained until the target optimal solution is obtained.

[0059] In one specific embodiment, such as Figure 5 As shown, high-fidelity simulation verification is performed on the candidate optimal solution, including: S501, perform high-fidelity simulation on the candidate optimal solution to obtain the corresponding simulated damage value, and compare the simulated damage value with the predicted damage value predicted by the passive safety model to be optimized; S502. If the comparison result meets the preset requirements, the verification is successful; otherwise, the verification fails.

[0060] Specifically, candidate optimal solutions from the candidate optimal solution set are input into a high-fidelity simulation verification algorithm. The algorithm performs high-fidelity simulation on the candidate optimal solutions and outputs simulated damage values. The simulated damage values ​​are then compared with the predicted damage values ​​predicted by the passive safety model to be optimized. If the difference between the simulated damage values ​​and the predicted damage values ​​is within a preset difference range, the verification is considered successful; otherwise, it is considered unsuccessful.

[0061] In this embodiment of the invention, after the optimal solution of the objective is verified with high fidelity, its complete parameters, data and Pareto front information are all stored in the knowledge base. The whole process realizes the synchronization of optimization task execution and platform experience accumulation, enabling the system to have self-evolution capability.

[0062] In one embodiment of the present invention, based on a knowledge base, an optimized task workflow can be established according to the vehicle model parameters and collision parameters of the target vehicle.

[0063] The optimization task workflow in this embodiment of the invention can be automatically generated by the optimization platform or edited by the user; this embodiment of the invention does not limit this.

[0064] For example, in the interactive interface provided by the optimization platform, the vehicle model parameters and collision parameters of the target vehicle are input. After obtaining the vehicle model parameters and collision parameters, the optimization platform automatically generates an optimization task workflow.

[0065] For example, users can edit relevant information based on the target vehicle's model parameters and collision parameters through the interactive interface provided by the optimization platform. The optimization platform can then generate an optimization task workflow based on the user-edited information. Specifically, the interactive interface provided by the optimization platform can be a graphical interface where users can drag and drop required models and data from the knowledge base based on the target vehicle's model parameters and collision parameters to form an executable optimization task workflow. For example, when minimizing driver injury in a frontal collision of the target vehicle, users can drag and drop not only the required model from the knowledge base, but also the collision parameters for minimizing a frontal collision, as well as driver injury data during a frontal collision.

[0066] In one specific embodiment, the user inputs the vehicle model parameters and collision parameters of the target vehicle through the graphical working engine provided by the optimization platform. The optimization platform initiates the optimization task workflow based on recommendations from the target vehicle model parameter and collision parameter knowledge base: it rapidly optimizes the passive safety model to be optimized, while simultaneously using a meta-model to evaluate the predictive uncertainty of the passive safety model in real time. When a high uncertainty region is detected, high-fidelity simulation is proactively triggered to generate key data and incrementally update the passive safety model to be optimized to improve its predictive accuracy, thus forming a data-model symbiosis.

[0067] The optimization method in this invention generates an optimization task workflow based on the target vehicle's model parameters and collision parameters, solving the systemic efficiency bottleneck caused by the fragmentation of toolchain, data flow, and knowledge flow in the development of passive safety configurations for the target vehicle. By updating the passive safety model to be optimized based on the optimization task workflow, the problems of process fragmentation and automation breakpoints are resolved. Previously, different tools were used at each stage from conceptual design to verification and optimization, resulting in inconsistent data formats and interfaces, making it impossible to form an end-to-end automated workflow and leading to numerous manual pre-breakpoints.

[0068] The knowledge base of this invention stores a large number of historical case records. Each historical case record is a structured object, which includes at least: scene feature vector: including collision type encoding, vehicle speed range, occupant body size percentile and other quantitative features; optimization problem configuration: including optimization objectives (such as minimizing HIC, chest compression amount) and their weights, design variable constraint range; result data: Pareto optimal solution set, the finally adopted solution and its corresponding damage index values.

[0069] Calculate the similarity between the target vehicle's model parameters and collision parameters and each historical case (model parameters and collision parameters of each sample vehicle) in the knowledge base. Similarity calculation can use cosine similarity or weighted Euclidean distance. For categorical features such as collision type in the collision parameters, one-hot encoding can be performed before calculation; for continuous features such as speed, normalization is required. Retrieve the top K historical cases with the highest similarity and perform case reasoning. Recommended content may include: initial design parameters: taking the average of the Pareto solution set centers or optimal solutions of similar cases; optimization target weight suggestions: analyzing the engineers' final choice preferences in similar cases and statistically analyzing the distribution of their target weights; initialization of the passive safety model to be optimized: directly loading the passive safety model trained on the most similar case as the passive safety model to be optimized for fine-tuning.

[0070] It should be noted that the initial design parameters refer to the design parameters of the target vehicle (i.e., engineering design variables). When performing optimization tasks for new vehicle models or new projects, the optimization platform provides the starting point recommended by engineers. For example, it suggests starting the seat belt force limit value in Newtons or the airbag ignition time in millimeters. By statistically analyzing the distribution of target weights, a high-precision passive safety model already trained from the most similar historical cases can be directly loaded, and its model parameters and weights can be used as the initial values ​​for the passive safety model to be optimized. By reusing models trained in similar tasks, the passive safety model to be optimized does not need to learn from scratch; it only needs fine-tuning based on the existing foundation to quickly achieve high accuracy, thereby significantly reducing the dependence on initial high-fidelity simulation data and improving the optimization startup speed.

[0071] The optimization method of this invention utilizes a multi-objective trade-off knowledge base to transform historical optimization experience into domain intelligence that can guide search strategies, enabling the system to have self-evolution capabilities. This solves the problem of knowledge not being able to be accumulated and reused; simulation data, model parameters, and empirical rules accumulated from each optimization task are scattered across various locations and cannot be systematically accumulated into platform intelligence to accelerate the development of subsequent vehicle models or new scenarios.

[0072] The following is a specific embodiment of the optimization method for the vehicle passive safety model in this invention: Task definition and workflow configuration. In the graphical workflow orchestrator of the optimization platform's graphical interface, users can drag and drop the required modules from the component library according to the optimization goals, configure data generation parameters, model types, optimization algorithms, and target weights to form an executable workflow.

[0073] Initialization and First Round Optimization. The optimization platform, based on similar case recommendations from a knowledge base, initializes the passive safety model to be optimized and the optimization strategy. The optimization algorithm then begins iterative processes, driving the passive safety model to rapidly evaluate a large number of design parameters and search for preliminary candidate optimal sets of connections.

[0074] Model-data symbiotic loop activation. Throughout the optimization process, the meta-model continuously evaluates the predictive uncertainty of the passive safety model to be optimized. Once a high-uncertainty region is identified (i.e., a region where optimization might get lost due to model inaccuracy), active sampling is immediately triggered to determine the few sampling points that can best reduce the uncertainty in that region. The passive safety model to be optimized then undergoes online incremental training to improve its accuracy in real time.

[0075] Verification and Knowledge Accumulation. After the passive safety model to be optimized converges, the output Pareto optimal solution set will be automatically submitted for high-fidelity simulation verification. The verified design scheme and its corresponding complete workflow, parameters, data, model, and Pareto front relationship will be structured and stored in a multi-objective trade-off knowledge base, forming new case knowledge.

[0076] Closed-loop self-evolution. When a new optimization task is initiated, the platform retrieves similar historical cases from the knowledge base, provides intelligent recommendations to users (such as initial design points and target weights), and may directly reuse or fine-tune existing high-precision proxy models, thereby achieving the acceleration effect of solidifying previous work so that subsequent work can be directly used.

[0077] The optimization method of this invention constructs a demand-driven dynamic data flow closed loop, which differs from the traditional toolchain that simply links data generation, passive safety model modeling, and parameter optimization. This invention, based on a demand-driven optimization task workflow for the target vehicle, drives the optimization of the passive safety model to be optimized. During the optimization process, a meta-model is used to evaluate the uncertainty of the passive safety model in real time. This uncertainty is then used as a control signal to proactively and purposefully trigger the generation of high-fidelity simulation data and incrementally update the passive safety model to be optimized. This establishes a dynamically balanced symbiotic relationship between "data scarcity" and "model accuracy."

[0078] Because this invention employs a collaborative symbiotic mechanism, the optimization method described in this embodiment produces unexpected technical effects compared to traditional isolated proxy model optimization methods: Simulation costs are significantly reduced. While achieving the same optimization accuracy, the platform reduces the number of high-fidelity simulation calls by approximately 70% through active sampling, as each simulation directly targets the weak points in the model's understanding, avoiding blind sampling. The optimization process escapes local optima. When performing multi-objective optimization on the seatbelt-airbag system of a certain vehicle model, traditional methods, due to model inaccuracies, fall into local optima after 50 iterations. However, this invention's platform, within the same number of iterations, successfully escapes the local optimum through three rounds of active sampling and model update cycles, finding a better design scheme that reduces damage values ​​by approximately 15%. System performance self-evolves over time. After completing the first vehicle model project, the platform's knowledge base stores the Pareto front features of that model. When developing a second similar vehicle model, the platform's recommended initial strategy improves the optimization convergence speed by 40%, demonstrating its ability to accumulate and reuse domain knowledge.

[0079] This invention optimizes the vehicle constraint system by constructing a collaborative optimization environment with domain awareness and self-evolution.

[0080] This invention constructs a task-oriented, integrated, and configurable workflow engine. Specifically, it designs a graphical workflow orchestration engine that encapsulates data generation modules, surrogate model training modules, multi-objective optimization algorithm modules, and high-fidelity verification interfaces as standardized service nodes. Users can quickly assemble customized optimization task workflows for different optimization tasks (such as driver-side frontal collision protection optimization and rear-seat child seat restraint system matching) through drag-and-drop functionality. This engine provides a unified data bus and control bus, completely eliminating manual data conversion and transfer between tools, achieving fully automated workflow from scenario definition to optimal solution output, and solving the problem of process fragmentation.

[0081] This invention employs a "data-model" symbiotic mechanism based on a meta-model and active sampling. Specifically, a meta-model is introduced, whose task is to evaluate the predictive uncertainty distribution of the current passive safety model to be optimized within the existing design space. During optimization, this meta-model monitors the predictive confidence of the passive safety model in real time. When high uncertainty is detected in a parameter region, the platform does not simply sample randomly in that region, but instead initiates an "active learning" loop: generating a small number of highly informative candidate sampling points, calling high-fidelity simulations to generate optimal data, and immediately incrementally updating the passive safety model to be optimized. This establishes a demand-driven dynamic data generation model. Data generation directly serves to improve the model's accuracy in key regions, while the continuous evolution of the model provides reliable navigation for exploring better solutions. The two transform from a sequential relationship to a symbiotic one, achieving efficient collaboration and resolving the disconnect problem.

[0082] This invention employs a multi-objective trade-off knowledge base adaptive optimization strategy. Specifically, a multi-objective trade-off knowledge base is constructed to store Pareto front relationships between various damage indicators (HIC, chest compression, etc.) under different collision scenarios in historical optimization projects, the engineers' final trade-off preferences, and the corresponding design rules. When a new task is initiated, the platform recommends initial optimization objective weights and constraints based on scenario similarity. During optimization, the platform can also adaptively adjust the search strategy of the optimization algorithm (e.g., shifting from global exploration to local refinement) based on the real-time discovered Pareto front shape. This mechanism transforms individual experience into reusable and evolvable platform collective intelligence. It not only automates optimization calculations but also automates the selection and adjustment of optimization strategies, enabling the platform to evolve and become smarter with use, accelerating the accumulation and reuse of design knowledge.

[0083] The embodiments of the present invention achieve global efficiency optimization, rather than optimal work efficiency, by optimizing the ecosystem model of the architecture and enabling real-time interaction and symbiosis between the model and optimization.

[0084] The data generation logic of this invention is on-demand, incremental, and proactive sampling, dynamically driven by the uncertainty of the model, which greatly reduces the total simulation cost of obtaining effective data and has high resource utilization.

[0085] The embodiments of the present invention have the advantages of intelligent optimization process and adaptive strategy. It can call the knowledge base and adjust the search strategy according to task characteristics and real-time progress, thereby improving the convergence speed and success rate of the optimization process and reducing the dependence on the user's proficiency in optimization algorithms.

[0086] This invention manages knowledge systematically and digitizes it, storing it in a platform knowledge base that is searchable, recommendable, and evolving. This enables the continuous accumulation and appreciation of a company's core R&D assets, creating a competitive barrier.

[0087] This invention constructs a self-evolving closed loop of "data-model-optimization," where the three elements are not in a master-slave relationship but rather form a synergistic and mutually driving organic whole. Compared to general simulation platforms, the optimization platform of this invention deeply embeds domain intelligence (such as damage target trade-offs and boundary scene perception) for vehicle constraint system optimization tasks, transforming the platform from a passive executor to an active collaborator.

[0088] This invention provides an optimization platform for vehicle passive safety models.

[0089] Figure 6 This is a schematic diagram of an optimization platform for a vehicle passive safety model according to an embodiment of the present invention. Figure 6 As shown, the vehicle passive safety model optimization platform 100 may include a construction module 10 and an optimization module 20.

[0090] The construction module 10 is used to construct a knowledge base, which includes the model parameters, collision parameters, passive safety measures corresponding to the collision parameters, and a passive safety model of the sample vehicle. The optimization module 20 is used to determine the passive safety model to be optimized from the knowledge base based on similarity according to the model parameters and collision parameters of the target vehicle. Based on the passive safety measures constraints of the target vehicle, the optimization module 20 iteratively optimizes the passive safety model to be optimized to obtain the target passive safety model and target passive safety measures for use in the passive safety configuration of the target vehicle.

[0091] It should be noted that other specific implementations of the vehicle passive safety model optimization platform provided in the embodiments of the present invention can be found in other specific implementations of the vehicle passive safety model optimization method of the above embodiments of the present invention.

[0092] The collaborative optimization platform of this invention constructs a self-evolving closed loop of "data-model-optimization," where the three elements are not in a master-slave relationship but rather form a synergistic and mutually driving organic whole. Compared with general simulation platforms, the collaborative optimization platform of this invention deeply embeds domain intelligence (such as damage target trade-offs and boundary scene perception) for vehicle constraint system optimization tasks, transforming the platform from a passive executor to an active collaborator.

[0093] This invention provides a computer-readable storage medium.

[0094] In this embodiment, a computer program is stored on a computer-readable storage medium, and when the computer program is executed by a processor, it implements the optimization method for the vehicle passive safety model as described above.

[0095] This invention provides an electronic device.

[0096] In this embodiment, the electronic device may include a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, it implements the optimization method for the vehicle passive safety model as described above.

[0097] Figure 7 This is a structural block diagram of an electronic device according to an embodiment of the present invention.

[0098] like Figure 7 As shown, the electronic device 500 includes a processor 501 and a memory 503. The processor 501 and the memory 503 are connected, for example, via a bus 502. Optionally, the electronic device 500 may also include a transceiver 504. It should be noted that in practical applications, the transceiver 504 is not limited to one type, and the structure of this electronic device 500 does not constitute a limitation on the embodiments of the present invention.

[0099] Processor 501 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 501 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0100] Bus 502 may include a pathway for transmitting information between the aforementioned components. Bus 502 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 502 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0101] The memory 503 stores a computer program corresponding to the vehicle passive safety model optimization method of the above embodiments of the present invention. This computer program is controlled and executed by the processor 501. The processor 501 executes the computer program stored in the memory 503 to implement the content shown in the foregoing method embodiments.

[0102] Among them, electronic devices 500 include, but are not limited to: mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 7 The electronic device 500 shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0103] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0104] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0105] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0106] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0107] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0108] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0109] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0110] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An optimization method for a vehicle passive safety model, characterized in that, The method includes: Construct a knowledge base, wherein the knowledge base includes the model parameters, collision parameters, passive safety measures corresponding to the collision parameters, and passive safety models of the sample vehicles; Based on the vehicle model parameters and collision parameters of the target vehicle, the passive safety model to be optimized is determined from the knowledge base based on similarity. Based on the passive safety measures constraints of the target vehicle, the passive safety model to be optimized is iteratively optimized to obtain the target passive safety model and target passive safety measures, which are then used for the passive safety configuration of the target vehicle.

2. The optimization method for the vehicle passive safety model according to claim 1, characterized in that, The iterative optimization of the passive safety model to be optimized based on the passive safety measures constraints of the target vehicle includes: Within the parameter range defined by the constraints of the passive safety measures, the passive safety model to be optimized is optimized. During the optimization process, the damage prediction uncertainty of the passive safety model to be optimized in the searched parameter region is evaluated, and the parameter uncertainty region is determined. Within the region of parameter uncertainty, sampling points are determined, and high-fidelity simulation is performed on the sampling points to generate corresponding simulated damage values. The simulated damage values, the sampling points, and the vehicle model parameters and collision parameters of the target vehicle are used to incrementally train the passive safety model to be optimized.

3. The optimization method for the passive safety model according to claim 2, characterized in that, The damage prediction uncertainty of the passive safety model to be optimized in the searched parameter region is evaluated using a meta-model.

4. The optimization method for the passive safety model according to claim 3, characterized in that, The evaluation of the damage prediction uncertainty of the passive safety model to be optimized in the searched parameter region, and the determination of the parameter uncertainty region, includes: The uncertainty quantification index corresponding to each of the searched parameter points is obtained through the meta-model and used as the evaluation value; The parameter uncertainty region is determined based on the searched parameter points whose evaluation values ​​are greater than the preset evaluation threshold.

5. The optimization method for the passive safety model according to claim 2, characterized in that, Determining the sampling point within the parameter uncertainty region includes: A preset sampling decision criterion is used to determine sampling points within the parameter uncertainty region, wherein the preset sampling decision criterion includes at least one of the following: maximizing prediction variance criterion, expected improvement criterion, and upper confidence boundary criterion.

6. The optimization method for the passive safety model according to claim 1, characterized in that, The iterative optimization of the passive safety model to be optimized based on the passive safety measures constraints of the target vehicle further includes: After the optimization converges, high-fidelity simulation verification is performed on the candidate optimal solutions in the candidate optimal solution set. The candidate optimal solution set is obtained by optimizing the damage prediction value output by the passive safety model to be optimized. If the verification passes, the candidate optimal solution that passes the verification will be used as the target passive security measure. If the validation fails, the candidate optimal solution that failed the validation will be used as new training data, and incremental training and optimization will be performed.

7. The optimization method for the passive safety model according to claim 1, characterized in that, Perform high-fidelity simulation verification on the candidate optimal solution, including: High-fidelity simulation is performed on the candidate optimal solution to obtain the corresponding simulated damage value, and the simulated damage value is compared with the predicted damage value predicted by the passive safety model to be optimized. If the comparison results meet the preset requirements, the verification is successful; otherwise, the verification fails.

8. An optimization platform for a passive safety model, characterized in that, The platform includes: A construction module is used to build a knowledge base, wherein the knowledge base includes the model parameters, collision parameters, passive safety measures corresponding to the collision parameters, and passive safety models of the sample vehicles; The optimization module is used to determine the passive safety model to be optimized based on similarity from the knowledge base according to the vehicle model parameters and collision parameters of the target vehicle; and to iteratively optimize the passive safety model to be optimized based on the passive safety measure constraints of the target vehicle to obtain the target passive safety model and target passive safety measures for use in the passive safety configuration of the target vehicle.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the optimization method of the passive security model as described in any one of claims 1-7.

10. An electronic device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the computer program is executed by the processor, it implements the optimization method of the passive security model as described in any one of claims 1-7.

Images