Vehicle simulation model optimization method and device, storage medium and electronic equipment

Abstract

The application provides a vehicle simulation model optimization method and device, a storage medium and an electronic device. According to the vehicle simulation model optimization method, a time sequence model can be used to process first vehicle operation data sequences collected in the operation process of a preset vehicle, so as to obtain second vehicle operation data sequences collected when the preset vehicle is in a target working condition. After at least part of the data in the second vehicle operation data sequences is automatically input into an initial vehicle simulation model, the initial vehicle simulation model can be automatically optimized by combining the difference between the vehicle simulation result output by the initial vehicle simulation model and the second vehicle operation data sequences collected by the actual vehicle, so as to obtain an optimized vehicle simulation model. The method is beneficial to reducing the artificial resources consumed in the vehicle simulation model optimization process and improving the vehicle simulation model optimization efficiency and convenience.

Description

Technical Field

[0001] This invention relates to the field of simulation model technology, and in particular to a method, apparatus, storage medium and electronic device for optimizing vehicle simulation models. Background Technology

[0002] With the continuous development of simulation technology, people have gradually begun to build vehicle simulation models to simulate and analyze aspects such as vehicle dynamic performance, handling characteristics, safety, fuel efficiency, and environmental impact. Because high-precision vehicle simulation models can gradually replace some real-vehicle testing, they not only reduce time, labor, and material costs but also avoid the personal safety issues associated with real-vehicle testing, thus gaining wider adoption. Currently, vehicle simulation testing typically requires manual processing of large amounts of vehicle operating data collected from real vehicles to obtain the effective data needed for the vehicle simulation model to run; furthermore, it requires continuous adjustment of the vehicle simulation model's parameters based on human experience to improve simulation accuracy, thus consuming a significant amount of manpower and posing a challenge to the optimization process of vehicle simulation models. Summary of the Invention

[0003] Based on this, the present invention provides a vehicle simulation model optimization method, device, storage medium and electronic device. By using this vehicle simulation model optimization method, the vehicle operation data required for the vehicle simulation model to run is automatically extracted, and the vehicle simulation model is automatically optimized based on the extracted vehicle operation data. This can reduce the manual resources required for the vehicle simulation model optimization process and improve the efficiency and convenience of vehicle simulation model optimization.

[0004] On one hand, the present invention provides a method for optimizing a vehicle simulation model, the method comprising:

[0005] Acquire the first vehicle operation data sequence collected during the operation of the preset vehicle;

[0006] The first vehicle operation data sequence is processed using a time series model to obtain a second vehicle operation data sequence collected under the preset vehicle's target operating condition; wherein, the time series model is a deep learning model for processing sequence data.

[0007] Input at least a portion of the data from the second vehicle operation data sequence into the initial vehicle simulation model to obtain the vehicle simulation results output by the initial vehicle simulation model;

[0008] Based on the vehicle simulation results and the second vehicle operation data sequence, the initial vehicle simulation model is optimized to obtain an optimized vehicle simulation model.

[0009] Furthermore, in some embodiments, the step of processing the first vehicle operation data sequence using a time series model to obtain a second vehicle operation data sequence collected when the preset vehicle is under target operating conditions includes:

[0010] Obtain a third vehicle operation data sequence of a preset type from the first vehicle operation data sequence; wherein, the third vehicle operation data sequence of the preset type includes: a vehicle operation data sequence used to reflect the operating condition of the preset vehicle;

[0011] The third vehicle operation data sequence of the preset type is input into the time series model to obtain the fourth vehicle operation data sequence of the preset type that is collected when the preset vehicle is in the target working condition, as output by the time series model.

[0012] The step of inputting at least a portion of the data from the second vehicle operation data sequence into the initial vehicle simulation model includes:

[0013] By calling a preset application programming interface, the preset type of fourth vehicle operation data sequence is input into the initial vehicle simulation model.

[0014] Furthermore, in some embodiments, the step of processing the first vehicle operation data sequence using a time series model to obtain the second vehicle operation data sequence collected when the preset vehicle is under target operating conditions further includes:

[0015] Determine the target data collection time period corresponding to the fourth vehicle operation data sequence;

[0016] Obtain the fifth vehicle operation data sequence collected within the target data collection time period from the first vehicle operation data sequence;

[0017] The optimization process for the initial vehicle simulation model based on the vehicle simulation results and the second vehicle operation data sequence to obtain an optimized vehicle simulation model includes:

[0018] Based on the vehicle simulation results and the fifth vehicle operation data sequence, the initial vehicle simulation model is optimized to obtain an optimized vehicle simulation model.

[0019] Furthermore, in some embodiments, before processing the first vehicle operation data sequence using a time series model, the method further includes:

[0020] Obtain vehicle operation data sequence samples of the preset type carrying tagged data; wherein, the tagged data is used to reflect the preset operating conditions corresponding to the vehicle operation data sequence samples;

[0021] The initial time series model is trained using vehicle operation data sequence samples of the preset type carrying the labeled data to obtain the time series model.

[0022] Furthermore, in some embodiments, acquiring the preset type of vehicle operation data sequence samples carrying tagged data includes:

[0023] Obtain vehicle operation data sequence samples of the preset type collected from real vehicles under the preset operating conditions; and / or,

[0024] Obtain vehicle operation data sequence samples of the preset type generated by simulating the preset working conditions using other vehicle simulation models.

[0025] Furthermore, in some embodiments, the time series model is a model for processing sequence data built based on at least one of Long Short-Term Memory Networks, Recurrent Neural Networks, Gated Recurrent Units, Transformer Models, and Temporal Convolutional Networks.

[0026] Furthermore, in some embodiments, the step of optimizing the initial vehicle simulation model based on the vehicle simulation results and the second vehicle operation data sequence to obtain an optimized vehicle simulation model includes:

[0027] The simulation loss value is determined based on the vehicle simulation results and the second vehicle operation data sequence.

[0028] Based on the simulation loss value, the optimized parameter values ​​of the preset key parameters at the initial vehicle simulation model are determined using the Bayesian optimization algorithm.

[0029] The optimized vehicle simulation model is obtained by replacing the historical parameter values ​​of the preset key parameters at the initial vehicle simulation model with the optimized parameter values.

[0030] Furthermore, in some embodiments, the target operating condition includes at least one of: acceleration operating condition, braking operating condition, and steering operating condition;

[0031] The initial vehicle simulation model includes at least one of the following: a vehicle drive system simulation model, a vehicle braking system simulation model, a vehicle steering system simulation model, and a whole vehicle simulation model.

[0032] On the other hand, the present invention also provides a vehicle simulation model optimization device, comprising:

[0033] The acquisition module is used to acquire the first vehicle operation data sequence collected during the operation of the preset vehicle;

[0034] The data processing module is used to process the first vehicle operation data sequence using a time series model to obtain the second vehicle operation data sequence collected when the preset vehicle is in the target operating condition; wherein, the time series model is a deep learning model for processing sequence data.

[0035] The simulation module is used to input at least a portion of the data from the second vehicle operation data sequence into the initial vehicle simulation model to obtain the vehicle simulation results output by the initial vehicle simulation model.

[0036] The optimization module is used to optimize the initial vehicle simulation model based on the vehicle simulation results and the second vehicle operation data sequence to obtain an optimized vehicle simulation model.

[0037] On the other hand, the present invention also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0038] On the other hand, the present invention also provides an electronic device, comprising: a processor and a memory; wherein the memory stores computer-readable instructions adapted to be loaded by the processor and to execute the steps of the above-described method.

[0039] According to the vehicle simulation model optimization method provided by the present invention, a time series model can be used to process a first vehicle operation data sequence collected during the operation of a preset vehicle to obtain a second vehicle operation data sequence collected when the preset vehicle is under a target operating condition. After automatically inputting at least a portion of the data in the second vehicle operation data sequence into the initial vehicle simulation model, the initial vehicle simulation model can be automatically optimized by combining the differences between the vehicle simulation results output by the initial vehicle simulation model and the second vehicle operation data sequence collected from the actual vehicle, thereby obtaining an optimized vehicle simulation model. Since there is no need to manually select the data required for the vehicle simulation model to run from a large amount of data collected from the actual vehicle, nor is it necessary to manually input the selected data into the vehicle simulation model, nor is it necessary to manually adjust and optimize the model parameters of the vehicle simulation model, the manual resources required for the vehicle simulation model optimization process can be reduced, and the efficiency and convenience of vehicle simulation model optimization can be improved.

[0040] It should be understood that the description in the Summary of the Invention is not intended to limit the key or essential features of the embodiments of the present invention, nor is it intended to restrict the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0041] Figure 1 A flowchart illustrating a vehicle simulation model optimization method provided in an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of a vehicle simulation model optimization scenario provided in an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram of the structure of a vehicle simulation model optimization device provided in an embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0046] In the description of one or more embodiments of the present invention, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0047] Currently, vehicle simulation models integrate knowledge from fields such as dynamics, aerodynamics, thermodynamics, and electrical engineering through multidisciplinary coupling to improve simulation accuracy and reliability. As high-precision vehicle simulation models can gradually replace some real-vehicle testing, they are increasingly being applied to control system development, driver assistance system testing, and autonomous driving system verification, rapidly advancing automotive advancements in intelligence and safety.

[0048] However, obtaining high-precision vehicle simulation models is currently quite costly. On one hand, complex experiments such as suspension K&C testing and tire dynamics testing are required to obtain the necessary suspension and tire parameters for building the simulation model, which is time-consuming and financially costly. On the other hand, it relies on expert experience to manually adjust model parameters and manually compare the simulation results with real-vehicle data for verification testing and optimization. This process is not only inefficient but also fails to guarantee the accuracy and reliability of the final simulation model. Furthermore, since real-vehicle operating data often contains various operating conditions, it is necessary to manually identify and extract valid data from this vast amount of data and manually import it into the simulation model. This process is also time-consuming, posing a challenge to the optimization of the vehicle simulation model.

[0049] Therefore, how to reduce the human resources required for vehicle simulation model optimization and improve the efficiency and convenience of vehicle simulation model optimization has become an urgent problem to be solved.

[0050] Based on this, the present invention proposes a vehicle simulation model optimization method. This method can use a time series model to process a first vehicle operation data sequence collected during the operation of a preset vehicle to obtain a second vehicle operation data sequence collected when the preset vehicle is under a target operating condition. After automatically inputting at least a portion of the data from the second vehicle operation data sequence into the initial vehicle simulation model, the method can automatically optimize the initial vehicle simulation model by combining the differences between the vehicle simulation results output by the initial vehicle simulation model and the second vehicle operation data sequence collected from the actual vehicle, thereby obtaining an optimized vehicle simulation model. Since there is no need to manually select the data required for the vehicle simulation model to run from a large amount of data collected from the actual vehicle, nor is it necessary to manually input the selected data into the vehicle simulation model, nor is it necessary to manually adjust and optimize the model parameters of the vehicle simulation model, the manual resources required for the vehicle simulation model optimization process can be reduced, thereby improving the efficiency and convenience of vehicle simulation model optimization.

[0051] Please see Figure 1 This is a flowchart illustrating a vehicle simulation model optimization method provided in an embodiment of the present invention. From a programming perspective, the executing entity of this process can be a program used for optimizing the vehicle simulation model. Alternatively, the executing entity can also be a device equipped with the aforementioned program; no specific limitation is made thereto.

[0052] The following is about Figure 1 The process shown will be explained in detail. The vehicle simulation model optimization method may specifically include the following steps:

[0053] Step S102: Obtain the first vehicle operation data sequence collected during the operation of the preset vehicle.

[0054] In this embodiment of the invention, in order to fully understand the behavior and state of the vehicle under different operating conditions, build a simulation scenario with good realism, and facilitate the verification of the accuracy and reliability of the vehicle simulation results generated by the vehicle simulation model, a first vehicle operation data sequence generated by the vehicle under various environments and operating conditions can be collected in advance to obtain real vehicle data.

[0055] The preset vehicle is typically a real vehicle of a preset type required for simulation using a vehicle simulation model, or it may be a real vehicle of other types or models, without specific limitations. The first vehicle operation data sequence may typically include real vehicle operation data recorded and saved using sensors at the preset vehicle or in other forms. For example, it may include, but is not limited to: the preset vehicle's own position data, motion state data, control signal data, driving task data, etc., arranged in chronological order of collection time from early to late; in addition, the first vehicle operation data sequence may also include: static environmental element data, dynamic environmental element data, traffic participant element data, meteorological element data, etc., around the preset vehicle; without specific limitations.

[0056] In practical applications, in order to ensure the comprehensiveness and practicality of the real vehicle data obtained, the preset vehicle can be driven in different environments such as urban roads, highways, and rural roads, and can alternately perform actions such as accelerating, decelerating, turning, and braking, so as to obtain the first vehicle operation data sequence generated by the preset vehicle under various operating conditions.

[0057] Step S104: Process the first vehicle operation data sequence using a time series model to obtain the second vehicle operation data sequence collected under the preset vehicle's target operating condition; wherein, the time series model is a deep learning model for processing sequence data.

[0058] In this embodiment of the invention, in order to fully understand and simulate the performance of the vehicle system under different scenarios, it is usually necessary to perform data analysis and mining on the first vehicle operation data sequence collected from the actual vehicle. By identifying the scenario in which the vehicle is located and the key behaviors it performs during the data analysis and mining process, it is possible to extract the second vehicle operation data sequence collected when the preset vehicle is in the target working condition from the massive amount of actual vehicle data, obtain valuable data, and eliminate other invalid data.

[0059] In this embodiment of the invention, since the deep learning model used to process sequence data can automatically extract and fuse features from the sequence data and capture long-term dependencies in the sequence data, it can efficiently and accurately complete tasks such as classification and prediction for the sequence data. Furthermore, since the first vehicle operation data sequence can be obtained by sorting the vehicle operation data collected at each time point according to the data collection time from early to late, a time series model can be built and trained using the deep learning model used to process sequence data. This time series model can then be used to classify the first vehicle operation data sequence, thereby efficiently and conveniently extracting the second vehicle operation data sequence generated under the preset vehicle target operating conditions from the first vehicle operation data sequence.

[0060] Step S106: Input at least a portion of the data from the second vehicle operation data sequence into the initial vehicle simulation model to obtain the vehicle simulation results output by the initial vehicle simulation model.

[0061] In this embodiment of the invention, the initial vehicle simulation model can refer to a data model used to simulate the behavior and performance of a vehicle or its various functional systems under various operating conditions. In practical applications, vehicle simulation software such as MATLAB, CarSim, TruckSim, ADAMS, and SUMO already possess powerful modeling, solving, and visualization capabilities, capable of meeting simulation requirements at different levels and with varying needs. Therefore, vehicle simulation software can be used to build the initial vehicle simulation model. Of course, other methods can also be used to build the initial vehicle simulation model; no specific limitations are imposed on this.

[0062] In practical applications, at least a portion of the second vehicle operation data sequence collected from the actual vehicle under the target operating conditions can be input into the initial vehicle simulation model as driver control information, or to build a test scenario, so that the initial vehicle simulation model can perform simulation tests on the working conditions of the vehicle or a specified type of vehicle system under the target operating conditions, and generate vehicle simulation results.

[0063] The initial vehicle simulation model can be of various types, including, but not limited to, vehicle drive system simulation models, vehicle braking system simulation models, vehicle steering system simulation models, and whole vehicle simulation models. The target operating conditions can include, but are not limited to, acceleration conditions, braking conditions, and steering conditions. It is understood that when using different types of initial vehicle simulation models to conduct simulation tests for different vehicle operating conditions, the types of the second vehicle operating data sequences input to the initial vehicle simulation models can be the same or different; and the types of data included in the vehicle simulation results output by the corresponding initial vehicle simulation models can be the same or different, without specific limitations.

[0064] To facilitate understanding, an example is provided. Assuming the initial vehicle simulation model is a vehicle braking system simulation model, the types of second vehicle operating data sequences input to the initial vehicle simulation model can include, but are not limited to: braking torque, wheel speed, differential output shaft torque, steering wheel angle, steering input torque, etc. The vehicle simulation results output by the initial vehicle simulation model can include: vehicle speed, vehicle acceleration, vehicle longitudinal forces, vehicle heading angle, etc. Alternatively, assuming the initial vehicle simulation model is a whole vehicle simulation model, the types of second vehicle operating data sequences input to the initial vehicle simulation model can include, but are not limited to: driving torque, wheel speed, differential output shaft torque, transmission output shaft torque, engine speed, steering wheel angle, steering input torque, etc. The vehicle simulation results output by the initial vehicle simulation model can include: vehicle speed, vehicle acceleration, vehicle heading angle, vehicle heading angular velocity, center of gravity acceleration, vehicle travel distance, etc. No specific limitations are imposed on these.

[0065] Step S108: Based on the vehicle simulation results and the second vehicle operation data sequence, optimize the initial vehicle simulation model to obtain an optimized vehicle simulation model.

[0066] In this embodiment of the invention, a large difference between the vehicle simulation result and the second vehicle operation data sequence indicates poor accuracy and reliability of the initial vehicle simulation model; conversely, a small difference indicates good accuracy and reliability. Therefore, based on the difference between the vehicle simulation result and the second vehicle operation data sequence, and in conjunction with a parameter optimization algorithm, at least some parameters within the initial vehicle simulation model can be automatically optimized to improve the accuracy and reliability of the optimized vehicle simulation model.

[0067] In one feasible implementation, the step of processing the first vehicle operation data sequence using a time series model to obtain the second vehicle operation data sequence collected under the preset vehicle's target operating condition may include:

[0068] Obtain a third vehicle operation data sequence of a preset type from the first vehicle operation data sequence; wherein the third vehicle operation data sequence of the preset type includes: a vehicle operation data sequence used to reflect the operating conditions of the preset vehicle.

[0069] The third vehicle operation data sequence of the preset type is input into the time series model to obtain the fourth vehicle operation data sequence of the preset type, which is collected when the preset vehicle is in the target operating condition, as output by the time series model.

[0070] Correspondingly, inputting at least a portion of the data from the second vehicle operation data sequence into the initial vehicle simulation model may include:

[0071] By calling a preset application programming interface, the preset type of fourth vehicle operation data sequence is input into the initial vehicle simulation model.

[0072] In this embodiment of the invention, since the input data required for the initial vehicle simulation model to run can include vehicle operation data reflecting the vehicle's operating conditions, the initial vehicle simulation model can simulate the operation of the vehicle or vehicle system under specified operating conditions. Based on this, a third vehicle operation data sequence reflecting the preset vehicle operating conditions can be extracted from the numerous first vehicle operation data sequences collected from actual vehicles. Then, the vehicle operating conditions corresponding to each vehicle operation data in the third vehicle operation data sequence are identified, thereby dividing the third vehicle operation data sequence into fourth vehicle operation data sequences collected under each vehicle operating condition.

[0073] Subsequently, based on vehicle simulation requirements, a fourth vehicle operation data sequence collected under the target operating condition can be input into the initial vehicle simulation model by calling the preset application programming interface at the initial vehicle simulation model. This data serves as driver control signals related to the steering wheel, accelerator, and brakes, allowing the initial vehicle simulation model to respond and perform simulation processing based on the fourth vehicle operation data sequence, thereby generating vehicle simulation results. Since there is no need for manual sifting of the input data required for the vehicle simulation model to run from a large amount of data collected from real vehicles, nor for manual input of the sifted data into the vehicle simulation model, the manual resources required for the vehicle simulation model optimization process can be reduced, improving the efficiency and convenience of vehicle simulation model optimization.

[0074] In one feasible implementation, before processing the first vehicle operation data sequence using a time series model, the process may further include:

[0075] Obtain vehicle operation data sequence samples of the preset type carrying tagged data; wherein, the tagged data is used to reflect the preset operating conditions corresponding to the vehicle operation data sequence samples.

[0076] The initial time series model is trained using vehicle operation data sequence samples of the preset type carrying the labeled data to obtain the time series model.

[0077] The time series model can be a model for processing sequence data built based on at least one of Long Short-Term Memory Network, Recurrent Neural Network, Gated Recurrent Unit, Transformer Model and Temporal Convolutional Network.

[0078] In this embodiment of the invention, a Recurrent Neural Network (RNN) is an artificial neural network suitable for processing and predicting temporal dependencies and time-series information in sequential data. A Long Short-Term Memory Network (LSTM) is a recurrent neural network adept at capturing long-term dependencies in time-series data, solving the gradient vanishing or exploding problems that original recurrent neural networks encounter when processing long-series data. A Gate Recurrent Unit (GRU) introduces a gating mechanism, enabling the network to better capture long-term dependencies while also improving network efficiency. The Transformer model is a neural network architecture based on self-attention, which allows the model to consider information from all positions simultaneously at each position in the sequence, thereby capturing long-distance dependencies within the sequence. A Temporal Convolutional Network (TCN) is a model that uses convolutional operations to process time-series data; compared to RNNs, it has advantages in parallelization, long-term dependency capture, and stability. As can be seen, the above deep learning models are all applicable to sequence data processing scenarios. Therefore, the above models and other deep learning models that can process sequence data can be used to build an initial time series model.

[0079] Subsequently, vehicle operation data sequence samples carrying label data reflecting the preset operating conditions of the vehicles corresponding to the sample data can be used to train the initial time series model. This enables the trained time series model to identify the vehicle operating conditions corresponding to the vehicle operation data sequence. Thus, the trained time series model can automatically identify the fourth vehicle operation data sequence generated under the target operating condition from the first vehicle operation data sequence collected from real vehicles, which helps ensure the accuracy and reliability of the extracted fourth vehicle operation data sequence.

[0080] In one feasible implementation, acquiring the preset type of vehicle operation data sequence samples carrying tagged data may include:

[0081] Obtain vehicle operation data sequence samples of the preset type collected from real vehicles under the preset operating conditions; and / or,

[0082] Obtain vehicle operation data sequence samples of the preset type generated by simulating the preset working conditions using other vehicle simulation models.

[0083] In this embodiment of the invention, vehicle operation data sequence samples generated under preset operating conditions can be extracted either from real vehicle data or from vehicle simulation results generated by other vehicle simulation models with good simulation effects. Label data reflecting the preset operating conditions corresponding to the vehicle operation data sequence samples is then set to train the initial time series model. This helps reduce the difficulty of obtaining model training samples, thereby better meeting the training sample quantity requirements of the time series model training process.

[0084] In practical applications, since time series models are typically needed to extract vehicle operation data sequences of a predetermined type collected under predetermined operating conditions and input to the initial vehicle simulation model, in order to reduce the computational load of the model training process and improve model training efficiency, the vehicle operation data sequence samples can be configured to contain only the aforementioned predetermined types of vehicle operation data sequences, without needing to include all types of vehicle operation data sequences contained in the actual vehicle data collected, thus offering good flexibility. Of course, the vehicle operation data sequence samples can also contain other types of vehicle operation data sequences besides the aforementioned predetermined types, without specific limitations.

[0085] In one feasible implementation, optimizing the initial vehicle simulation model based on the vehicle simulation results and the second vehicle operation data sequence to obtain an optimized vehicle simulation model may include:

[0086] The simulation loss value is determined based on the vehicle simulation results and the second vehicle operation data sequence.

[0087] Based on the simulation loss value, the optimized parameter values ​​of the preset key parameters at the initial vehicle simulation model are determined using a Bayesian optimization algorithm.

[0088] The optimized vehicle simulation model is obtained by replacing the historical parameter values ​​of the preset key parameters at the initial vehicle simulation model with the optimized parameter values.

[0089] In this embodiment of the invention, Bayesian optimization is a global optimization algorithm based on Bayes' theorem. It intelligently selects evaluation points by constructing a surrogate model, aiming to find the global optimum with fewer evaluation iterations. Due to its advantages such as efficiency, robustness, and flexibility, Bayesian optimization has been widely applied in hyperparameter tuning within machine learning. Based on this, the preset key parameters to be optimized in the initial vehicle simulation model can be determined in advance according to actual needs. The optimized parameter values ​​of the preset key parameters in the initial vehicle simulation model can then be automatically generated using the Bayesian optimization algorithm. These optimized parameter values ​​then replace the historical parameter values ​​of the preset key parameters in the initial vehicle simulation model, resulting in an optimized vehicle simulation model. Of course, other optimization algorithms for model parameters can also be used to determine the optimized parameter values ​​of the preset key parameters in the initial vehicle simulation model; no specific limitation is made in this regard.

[0090] In practical applications, the types of preset key parameters to be optimized for different types of initial vehicle simulation models can be the same or different. For example, for a vehicle braking system simulation model, the preset key parameters may include, but are not limited to: brake pad friction coefficient, vacuum booster curve, and brake performance curve; while for a vehicle steering system simulation model, the preset key parameters may include, but are not limited to: vehicle steering ratio, moment of inertia of the steering wheel and rotatable parts of the steering system, steering gear and its connected parts; in addition, engine characteristic parameters, clutch characteristic parameters, transmission characteristic parameters, differential characteristic parameters, etc., may also be included, without specific limitations.

[0091] In practical applications, the core of the Bayesian optimization algorithm lies in constructing a surrogate model (such as a Gaussian process) to describe the objective function. It involves selecting sample points in the search space and evaluating the objective function at these points to update the parameters of the surrogate model based on the newly collected sample points and the objective function values. Subsequently, based on the updated surrogate model, the next sampling point most likely to improve the objective function value can be selected. This process of sampling, updating the surrogate model, and selecting sampling points is repeated until the optimal solution is found or a stopping condition is met (such as reaching a preset number of evaluations, a time limit, or a convergence criterion).

[0092] As can be seen, the Bayesian optimization algorithm is an iterative process of continuously optimizing parameters. Therefore, after performing a parameter update to obtain an optimized vehicle simulation model, the optimized vehicle simulation model can be used to process other second vehicle running data sequences to obtain the current vehicle simulation result. Furthermore, the Bayesian optimization algorithm can be used again to determine the current optimized parameter values ​​of the preset key parameters at the initial vehicle simulation model based on the simulation loss value between the current vehicle simulation result and the aforementioned other second vehicle running data sequences. Setting the current optimized parameters as the preset key parameter values ​​at the initial vehicle simulation model helps improve the simulation accuracy and reliability of the final generated vehicle simulation model. Simultaneously, since it eliminates the need for complex quantitative testing on test benches or qualitative testing on roads, such as tire dynamics testing or vehicle braking performance testing, to obtain optimal key parameter values ​​for the vehicle simulation model, it also helps reduce the financial and time costs required for the vehicle simulation model optimization process.

[0093] In one feasible implementation, the step of processing the first vehicle operation data sequence using a time series model to obtain the second vehicle operation data sequence collected under the preset vehicle's target operating condition may further include:

[0094] Determine the target data collection time period corresponding to the fourth vehicle operation data sequence.

[0095] Obtain the fifth vehicle operation data sequence collected within the target data collection time period from the first vehicle operation data sequence.

[0096] Correspondingly, the step of optimizing the initial vehicle simulation model based on the vehicle simulation results and the second vehicle operation data sequence to obtain an optimized vehicle simulation model may include:

[0097] Based on the vehicle simulation results and the fifth vehicle operation data sequence, the initial vehicle simulation model is optimized to obtain an optimized vehicle simulation model.

[0098] Specifically, based on the vehicle simulation results and the fifth vehicle's operating data sequence, a simulation loss value is determined. Then, based on this simulation loss value, a Bayesian optimization algorithm is used to determine the optimized parameter values ​​of the preset key parameters in the initial vehicle simulation model. Finally, the optimized parameter values ​​are used to replace the historical parameter values ​​of the preset key parameters in the initial vehicle simulation model, resulting in an optimized vehicle simulation model.

[0099] In this embodiment of the invention, since a fourth type of vehicle operation data sequence of a preset type can be extracted using a time series model when the vehicle is under a preset target operating condition, when the initial vehicle simulation model is used to perform simulation testing on the target operating condition based on the fourth vehicle operation data sequence, it is usually only necessary to calculate the simulation loss between the vehicle simulation result output by the initial vehicle simulation model and the fifth vehicle operation data sequence collected under the target operating condition, without having to calculate the simulation loss with the vehicle operation data sequences collected under other operating conditions. This is beneficial to ensuring the optimization effect and reliability of the vehicle simulation model optimization based on the calculated simulation loss value.

[0100] For ease of understanding, Figure 2 This is a schematic diagram of a vehicle simulation model optimization scenario provided in an embodiment of the present invention. Figure 2 As shown, the vehicle simulation model optimization process may include a data extraction part 21, a vehicle simulation model part 22, and a Bayesian optimization part 23.

[0101] In its implementation, the data extraction section 21 can first extract a third vehicle operation data sequence of a preset type from a large number of first vehicle operation data sequences collected from actual vehicles. Then, using the time series model 201, it can extract a fourth vehicle operation data sequence 202 of a preset type collected under the target operating condition from the third vehicle operation data sequence of the preset type. The fourth vehicle operation data sequence 202 is then input to the vehicle simulation model section 22. The vehicle simulation model section 22 can generate and send vehicle simulation data to the Bayesian optimization section 23 based on the fourth vehicle operation data sequence 202.

[0102] In addition, during implementation, the data extraction section 21 can determine the target data collection time period corresponding to the fourth vehicle operation data sequence; and input the fifth vehicle operation data sequence 203 extracted from the third vehicle operation data sequence and collected within the target data collection time period into the Bayesian optimization section 23. This allows the Bayesian optimization section 23 to determine the simulation loss value between the fifth vehicle operation data sequence 203 and the aforementioned vehicle simulation data, and based on the simulation loss value, use the Bayesian optimization algorithm to determine the optimized parameter values ​​of the preset key parameters at the vehicle simulation model. The optimized parameter values ​​are then used to replace the historical parameter values ​​of the preset key parameters at the vehicle simulation model section 22 to obtain the optimized vehicle simulation model.

[0103] Please see Figure 3 This is a schematic diagram of the structure of a vehicle simulation model optimization device provided in an embodiment of the present invention. Figure 3 As shown, the vehicle simulation model optimization device 03 can be implemented as all or part of an electronic device through software, hardware, or a combination of both. According to some embodiments, the vehicle simulation model optimization device 03 may include an acquisition module 31, a data processing module 32, a simulation module 33, and an optimization module 34, specifically:

[0104] The acquisition module 31 is used to acquire the first vehicle operation data sequence collected during the operation of the preset vehicle.

[0105] The data processing module 32 is used to process the first vehicle operation data sequence using a time series model to obtain the second vehicle operation data sequence collected when the preset vehicle is in the target operating condition; wherein, the time series model is a deep learning model for processing sequence data.

[0106] The simulation module 33 is used to input at least a portion of the data in the second vehicle operation data sequence into the initial vehicle simulation model to obtain the vehicle simulation results output by the initial vehicle simulation model.

[0107] The optimization module 34 is used to optimize the initial vehicle simulation model based on the vehicle simulation results and the second vehicle operation data sequence to obtain an optimized vehicle simulation model.

[0108] Optionally, the data processing module 32 may include:

[0109] The first acquisition unit is configured to acquire a third vehicle operation data sequence of a preset type from the first vehicle operation data sequence; wherein the third vehicle operation data sequence of the preset type includes: a vehicle operation data sequence reflecting the operating condition of the preset vehicle.

[0110] The second acquisition unit is used to input the third vehicle operation data sequence of the preset type into the time series model to obtain the fourth vehicle operation data sequence of the preset type that is collected when the preset vehicle is in the target working condition, as output by the time series model.

[0111] Correspondingly, simulation module 33 can be specifically used for:

[0112] By calling a preset application programming interface, the preset type of fourth vehicle operation data sequence is input into the initial vehicle simulation model.

[0113] Optionally, the data processing module 32 may also include:

[0114] The determining unit is used to determine the target data collection time period corresponding to the fourth vehicle operation data sequence.

[0115] The third acquisition unit is used to acquire the fifth vehicle operation data sequence collected within the target data acquisition time period from the first vehicle operation data sequence.

[0116] Correspondingly, optimization module 34 can be specifically used for:

[0117] Based on the vehicle simulation results and the fifth vehicle operation data sequence, the initial vehicle simulation model is optimized to obtain an optimized vehicle simulation model.

[0118] Optional, Figure 3 The device may further include:

[0119] The sample acquisition module is used to acquire vehicle operation data sequence samples of the preset type carrying tagged data; wherein, the tagged data is used to reflect the preset operating conditions corresponding to the vehicle operation data sequence samples.

[0120] The model training module is used to train the initial time series model using vehicle operation data sequence samples of the preset type carrying the labeled data, so as to obtain the time series model.

[0121] Optionally, the sample acquisition module can be used specifically for:

[0122] Obtain vehicle operation data sequence samples of the preset type collected from real vehicles under the preset operating conditions; and / or,

[0123] Obtain vehicle operation data sequence samples of the preset type generated by simulating the preset working conditions using other vehicle simulation models.

[0124] Optionally, the time series model is a model for processing sequence data built based on at least one of Long Short-Term Memory Network, Recurrent Neural Network, Gated Recurrent Unit, Transformer Model and Temporal Convolutional Network.

[0125] Optionally, optimization module 34 may specifically include:

[0126] The first determining unit is used to determine the simulation loss value based on the vehicle simulation results and the second vehicle operation data sequence.

[0127] The second determining unit is used to determine the optimized parameter values ​​of the preset key parameters at the initial vehicle simulation model based on the simulation loss value and using a Bayesian optimization algorithm.

[0128] The parameter replacement unit is used to replace the historical parameter values ​​of the preset key parameters in the initial vehicle simulation model with the optimized parameter values ​​to obtain the optimized vehicle simulation model.

[0129] Optionally, the target operating condition may include at least one of the following: acceleration operating condition, braking operating condition, and steering operating condition.

[0130] The initial vehicle simulation model may include at least one of the following: a vehicle drive system simulation model, a vehicle braking system simulation model, a vehicle steering system simulation model, and a whole vehicle simulation model.

[0131] The above-described apparatus embodiments correspond to the method embodiments, and detailed descriptions can be found in the description of the method embodiments section, which will not be repeated here. The apparatus embodiments are derived based on the corresponding method embodiments and have the same technical effects as the corresponding method embodiments; detailed descriptions can be found in the corresponding method embodiments.

[0132] The present invention also provides a storage medium that can store a computer program. When the computer program is executed by a processor, it can implement the vehicle simulation model optimization method as described in the above embodiments. For the specific execution process, please refer to the specific description in the above embodiments, which will not be repeated here.

[0133] In one embodiment, the present invention also provides Figure 4 The diagram shows the structure of the electronic device. Figure 4At the hardware level, the electronic device may include a processor 41 and a memory 45, and may also include an internal bus 42, a network interface 43, a memory 44, and other hardware required for the service. This electronic device can be installed in a vehicle. The processor 41 can read corresponding computer-readable instructions from the memory 45 into memory and then execute them to implement the aforementioned vehicle simulation model optimization method. For the specific execution process, please refer to the detailed descriptions in the above embodiments, which will not be repeated here.

[0134] Finally, the various embodiments in this invention are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for embodiments such as storage media and electronic devices, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments.

[0135] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for optimizing a vehicle simulation model, comprising: Acquire the first vehicle operation data sequence collected during the operation of the preset vehicle; The first vehicle operation data sequence is processed using a time series model to obtain a second vehicle operation data sequence collected under the preset vehicle's target operating condition; wherein, the time series model is a deep learning model for processing sequence data. Input at least a portion of the data from the second vehicle operation data sequence into the initial vehicle simulation model to obtain the vehicle simulation results output by the initial vehicle simulation model; Based on the vehicle simulation results and the second vehicle operation data sequence, the initial vehicle simulation model is optimized to obtain an optimized vehicle simulation model.

2. The method according to claim 1, wherein processing the first vehicle operation data sequence using a time series model to obtain the second vehicle operation data sequence collected under the preset vehicle operating condition includes: Obtain a third vehicle operation data sequence of a preset type from the first vehicle operation data sequence; wherein, the third vehicle operation data sequence of the preset type includes: a vehicle operation data sequence used to reflect the operating condition of the preset vehicle; The third vehicle operation data sequence of the preset type is input into the time series model to obtain the fourth vehicle operation data sequence of the preset type that is collected when the preset vehicle is in the target working condition, as output by the time series model. The step of inputting at least a portion of the data from the second vehicle operation data sequence into the initial vehicle simulation model includes: By calling a preset application programming interface, the preset type of fourth vehicle operation data sequence is input into the initial vehicle simulation model.

3. The method according to claim 2, wherein processing the first vehicle operation data sequence using a time series model to obtain the second vehicle operation data sequence collected under the preset vehicle operating condition further includes: Determine the target data collection time period corresponding to the fourth vehicle operation data sequence; Obtain the fifth vehicle operation data sequence collected within the target data collection time period from the first vehicle operation data sequence; The optimization process for the initial vehicle simulation model based on the vehicle simulation results and the second vehicle operation data sequence to obtain an optimized vehicle simulation model includes: Based on the vehicle simulation results and the fifth vehicle operation data sequence, the initial vehicle simulation model is optimized to obtain an optimized vehicle simulation model.

4. The method according to claim 2, further comprising, before processing the first vehicle operation data sequence using a time series model: Obtain vehicle operation data sequence samples of the preset type carrying tagged data; wherein, the tagged data is used to reflect the preset operating conditions corresponding to the vehicle operation data sequence samples; The initial time series model is trained using vehicle operation data sequence samples of the preset type carrying the labeled data to obtain the time series model.

5. The method according to claim 4, wherein obtaining the preset type of vehicle operation data sequence samples carrying tagged data comprises: Obtain vehicle operation data sequence samples of the preset type collected from real vehicles under the preset operating conditions; And / or, Obtain vehicle operation data sequence samples of the preset type generated by simulating the preset working conditions using other vehicle simulation models.

6. The method according to claim 4, wherein the time series model is a model for processing sequence data built based on at least one of Long Short-Term Memory Network, Recurrent Neural Network, Gated Recurrent Unit, Transformer Model and Temporal Convolutional Network.

7. The method according to claim 1, wherein optimizing the initial vehicle simulation model based on the vehicle simulation results and the second vehicle operation data sequence to obtain an optimized vehicle simulation model comprises: Based on the vehicle simulation results and the second vehicle operation data sequence, the simulation loss value is determined; Based on the simulation loss value, the optimized parameter values ​​of the preset key parameters at the initial vehicle simulation model are determined using the Bayesian optimization algorithm. The optimized vehicle simulation model is obtained by replacing the historical parameter values ​​of the preset key parameters at the initial vehicle simulation model with the optimized parameter values.

8. The method according to any one of claims 1-7, wherein the target operating condition includes: At least one of the following: acceleration condition, braking condition, and steering condition; The initial vehicle simulation model includes at least one of the following: a vehicle drive system simulation model, a vehicle braking system simulation model, a vehicle steering system simulation model, and a whole vehicle simulation model.

9. A vehicle simulation model optimization device, comprising: The acquisition module is used to acquire the first vehicle operation data sequence collected during the operation of the preset vehicle; The data processing module is used to process the first vehicle operation data sequence using a time series model to obtain the second vehicle operation data sequence collected when the preset vehicle is in the target operating condition; wherein, the time series model is a deep learning model for processing sequence data. The simulation module is used to input at least a portion of the data from the second vehicle operation data sequence into the initial vehicle simulation model to obtain the vehicle simulation results output by the initial vehicle simulation model. The optimization module is used to optimize the initial vehicle simulation model based on the vehicle simulation results and the second vehicle operation data sequence to obtain an optimized vehicle simulation model.

10. A storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method according to any one of claims 1 to 8.

11. An electronic device, comprising: A processor and a memory; wherein the memory stores computer-readable instructions adapted to be loaded by the processor and to perform the steps of the method as claimed in any one of claims 1 to 8.

Images