Bayesian optimization-based virtual acceleration calibration method for parameters of chassis electric control system

Through Bayesian optimization method and importance sampling algorithm, loop iterative optimization is solved, and the local optimization problem of the high-dimensional parameter virtual calibration process of the chassis electronic control system is achieved, and the effect of accelerating virtual calibration and resource conservation is achieved.

CN119045453AActive Publication Date: 2024-11-29JILIN UNIVERSITY

Patent Information

Application Number
CN202411147630.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-11-29
Estimated Expiration
2044-08-21

AI Technical Summary

Technical Problem

The existing technology is difficult to effectively accelerate the virtual calibration process of high-dimensional parameters of the chassis electronic control system. Traditional optimization algorithms are prone to falling into local optimality or unable to find global optimal solutions in high-dimensional space, resulting in wasted experimental resources.

Method used

Based on Bayesian optimization method, the optimization objective function of the chassis electronic control system is established, a high-precision simulation model is built, and the Bayesian agent model is pre-trained and the importance sampling algorithm is used to loop iterate the optimization process to obtain the optimal parameter combination.

Benefits of technology

It effectively solves the local optimal problem, significantly accelerates the virtual calibration process of the chassis electronic control system, reduces the waste of test resources, and improves the accuracy of parameter combinations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119045453A_ABST
    Figure CN119045453A_ABST
Patent Text Reader

Abstract

The invention relates to a Bayesian optimization-based virtual acceleration calibration method for parameters of a chassis electronic control system. Comprising the following steps: 1, establishing a chassis electric control system optimization objective function; 2, completing pre-training of the Bayesian agent model; 3, building a current chassis electric control system high-precision simulation model in a simulation environment, and testing a working condition scene model; 4, sampling test condition types and test condition specific parameters by using a Bayesian probability improvement method; 5, testing by using the tested working condition type and parameters obtained by sampling in a simulation environment, and obtaining an objective function result of the to-be-calibrated system under the working condition; 6, optimizing the pre-trained proxy model based on the obtained new data; and 7, carrying out loop iteration until the target condition of the virtual acceleration calibration process is reached, and obtaining the chassis electric control system corresponding to the current optimal target function. The method can effectively solve the problem of resource waste of the multi-dimensional parameter traversal combination test of the chassis electronic control system, and accelerates the virtual calibration process of the chassis electronic control system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of chassis electronic control system virtual calibration, and specifically is a chassis electronic control system parameter virtual acceleration calibration method based on Bayesian optimization. Background Art

[0002] At present, the actual vehicle calibration process of the chassis electronic control system in "two winters and one summer" is increasingly unable to meet the increasingly shortened development cycle requirements of the chassis electronic control system. Virtual calibration of chassis electronic control system parameters has become a necessary part of chassis system development. However, due to the large number of chassis electronic control system parameters, using the method of traversing parameter combinations to find the best performance of the chassis electronic control system, that is, the optimal parameter combination, will waste huge experimental resources. How to use a certain optimization process to accelerate the virtual calibration process is becoming increasingly important. The prior art has carried out research on similar content, such as a patent for a virtual calibration method for multi-strategy parameters of hybrid vehicles (CN202410142902.2), a virtual calibration method for key parameters of an automotive electronic anti-lock braking system (CN202410486349.4), etc. However, existing technologies mostly use traditional optimization algorithms to accelerate the calibration process. These algorithms can only solve the optimal combination of a small number of variables and are difficult to adapt to the high-dimensional parameter calibration requirements of the chassis electronic control system. It is very important to build a set of accelerated calibration methods that can adapt to the high-dimensional space parameter calibration requirements of the chassis electronic control system. Summary of the invention

[0003] To solve the above problems, the present invention provides a virtual acceleration calibration method for chassis electronic control system parameters based on Bayesian optimization, establishes a chassis electronic control system optimization objective function according to the vehicle motion state and the chassis system working characteristics, completes Bayesian proxy model pre-training based on other chassis electronic control system parameter calibration process data or field test data of the system to be calibrated, builds a high-precision simulation model of the current chassis electronic control system and a test condition scenario model in a simulation environment, uses a Bayesian probability improvement method to sample the test condition type and the specific parameters of the test condition, and conducts experiments in the simulation environment to obtain the objective function result under the condition, optimizes the pre-trained proxy model based on the new data obtained, and completes the acquisition of the chassis electronic control system parameter combination with the optimal objective function through cyclic iteration.

[0004] The technical solution of the present invention is described as follows in conjunction with the accompanying drawings:

[0005] A virtual acceleration calibration method for chassis electronic control system parameters based on Bayesian optimization includes the following steps:

[0006] Step 1: Establish the chassis electronic control system optimization objective function according to the vehicle motion state and chassis system working characteristics;

[0007] Step 2: Pre-training the Bayesian proxy model based on the calibration process data of other chassis electronic control systems or the field test data of the system to be calibrated;

[0008] Step 3: Build a high-precision simulation model of the current chassis electronic control system and a test condition scenario model in a simulation environment;

[0009] Step 4: Use the Bayesian probability improvement method to sample the test condition type and specific parameters of the test condition;

[0010] Step 5: Use the sampled test condition type and parameters to conduct experiments in the simulation environment and obtain the objective function result of the system to be calibrated under the condition;

[0011] Step 6: Optimize the pre-trained proxy model based on the new data obtained;

[0012] Step 7: Iterate step 4 to step 6 in a loop until the target condition of the virtual acceleration calibration process is reached, and obtain the chassis electronic control system corresponding to the current optimal objective function.

[0013] Furthermore, the specific method of step one is as follows:

[0014] 11) Construct the chassis electronic control system virtual calibration vehicle motion state optimization objective function; consider the vehicle motion to establish the objective function, including the handling stability optimization objective, ride comfort optimization objective, and driving safety optimization objective;

[0015] 12) Construct the objective function for optimizing the working characteristics of the chassis electronic control system virtual calibration system; establish the objective function considering the working characteristics of the chassis electronic control system, including the optimization objectives of control accuracy, robust stability, and work efficiency;

[0016] 13) Use the hierarchical analysis method to analyze the weights of different optimization objective functions, and obtain the weighted comprehensive optimization objective function of the chassis electronic control system virtual calibration process; establish a pairwise importance discrimination matrix of different categories of optimization objectives in steps 11) and 12), calculate the maximum eigenvalue of the matrix, and calculate the consistency between the maximum eigenvalue and the discrimination matrix. If the consistency judgment passes, normalize the maximum eigenvalue to obtain the weights of different categories of optimization objectives, otherwise reconstruct the discrimination matrix; weight the optimization objectives of all categories based on the obtained optimization objective weights to obtain a comprehensive optimization objective function.

[0017] Furthermore, the specific method of step 2 is as follows:

[0018] 21) Collect the electronic control system parameter type data of other models of chassis electronic control system to be calibrated during the on-site calibration process, the vehicle motion data corresponding to the parameter data of the system to be calibrated during the actual vehicle test process, and the chassis electronic control system working status data, and use the method in step 1 to calculate the comprehensive optimization objective function corresponding to different chassis electronic control system parameters;

[0019] 22) Perform pre-training of the Bayesian proxy model and select the Gaussian process as the Bayesian proxy model. A Gaussian process is defined by the mean function μ(x) and the kernel function k(x, x′), and the formula is as follows:

[0020] f(x)~GP(μ(x),k(x,x′))

[0021] Among them, the mean function μ(x) selects the zero mean function, that is, μ(x) = 0, and the kernel function selects the radial basis function, and the formula is as follows:

[0022]

[0023] Among them, ||xx′|| 2 is the square of the Euclidean distance between the input points x and x′; σ is the kernel width parameter, which controls the smoothness of the kernel function;

[0024] 23) Use the data in step 21) to fit the Gaussian process constructed in step 22) to complete the pre-training of the Bayesian proxy model.

[0025] Furthermore, the specific method of step three is as follows:

[0026] 31) Build a high-precision model of the chassis electronic control system to be calibrated, build a basic principle model of the chassis electronic control system to be calibrated based on the working principle of the chassis electronic control system, and continuously adjust various parameters in the basic principle model according to the calibration process system data of the chassis electronic control system to be calibrated, until the output result of the basic principle model can accurately map the performance of the real chassis electronic control system, thereby obtaining a high-precision simulation model of the chassis electronic control system that can be used for virtual calibration based on the basic principle model;

[0027] 32) Build a chassis electronic control system test condition scenario model. According to the chassis electronic control system test regulations, establish a chassis electronic control system calibration condition table. The condition table records all the condition types and corresponding parameters that need to be calibrated. Build the corresponding test condition scenario model in the simulation environment according to the calibration condition table.

[0028] Furthermore, the specific method of step 4 is as follows:

[0029] 41) Select the test condition to be calibrated by using the importance sampling algorithm. The importance sampling algorithm uses a new sampling probability to accelerate the sampling of the condition type with greater importance. The test condition scenarios constructed in step 32) are numbered, and the number of different condition types in step 21) is counted to obtain the corresponding probability distribution p(x). The optimization objective function results corresponding to different conditions in step 21) are counted to obtain the importance sampling probability distribution q(x) of different test conditions. The ratio of q(x) to p(x) is used as the weight of different conditions to be calibrated, and the sampling of different test condition types is completed based on the obtained weight results;

[0030] 42) The Bayesian probability improvement method is used to sample the test parameters of the calibration test. The formula of the Bayesian probability improvement acquisition function a(x) is as follows:

[0031]

[0032] Where, mean(x) is the predicted mean of the Gaussian process model’s optimized objective function at parameter x; sigma(x) is the predicted standard deviation of the Gaussian process model’s optimized objective function at parameter x; g(x + ) is the current best optimization objective function value; l is a hyperparameter used to control the balance between exploration and utilization; Φ is the cumulative distribution function of the standard normal distribution.

[0033] Furthermore, the specific method of step five is as follows:

[0034] 51) Select the calibration test condition type and parameters obtained in step 4, select the corresponding simulation test condition built in step 32) to test the high-precision model of the chassis electronic control system to be calibrated in step 31), and obtain the corresponding comprehensive optimization objective function under the virtual calibration condition.

[0035] Furthermore, the specific method of step six is ​​as follows:

[0036] 61) inputting the virtual calibration test condition parameters and the corresponding comprehensive optimization objective function obtained in step 51) into the database in step 21);

[0037] 62) Retrain the Gaussian process proxy model using the method in step 22) to obtain a new proxy model.

[0038] Furthermore, the specific method of step seven is as follows:

[0039] 71) Set the maximum number of cycles n and the minimum value u of the improvement difference of the comprehensive optimization target, and iterate steps 4 to 6 until the virtual acceleration calibration parameter reaches the maximum number of cycles n or the improvement of the comprehensive optimization target function is less than u for 10 consecutive iterations, then stop the test;

[0040] 72) Obtaining a chassis electronic control system parameter combination corresponding to the current optimal comprehensive objective function as the current optimal chassis electronic control system parameter combination;

[0041] 73) Count the standard deviation of the optimization process of different control parameters in the virtual calibration process, use this value as the standard deviation of parameter changes in the subsequent field calibration process, use the optimal parameter in step 72) as the mean, and use the parameter range within three times the standard deviation above and below the mean as the debugging space of the chassis electronic control system parameters in the subsequent field calibration test, and further accurately adjust the chassis electronic control system parameters.

[0042] The beneficial effects of the present invention are:

[0043] The present invention solves the problem that traditional optimization algorithms fall into local optimal solutions or cannot find global optimal solutions during the adjustment process of high-dimensional parameter space. It can effectively solve the problem of waste of multi-dimensional parameter traversal combination test resources of chassis electronic control systems and accelerate the virtual calibration process of chassis electronic control systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] Embodiment 1

[0048] This embodiment provides a method for virtual acceleration calibration of chassis electronic control system parameters based on Bayesian optimization, including the following steps:

[0049] Step 1: Establish the chassis electronic control system optimization objective function according to the vehicle motion state and chassis system working characteristics, as follows:

[0050] 11) Construct the chassis electronic control system virtual calibration vehicle motion state optimization objective function;

[0051] Considering vehicle motion, establish objective functions, including optimization objectives for handling stability, ride comfort, and driving safety;

[0052] 12) Construct the operating characteristic optimization objective function of the chassis electronic control system virtual calibration system;

[0053] Considering the working characteristics of the chassis electronic control system, the objective function is established, including the optimization objectives of control accuracy, robust stability, and work efficiency;

[0054] 13) Use the hierarchical analysis method to analyze the weights of different optimization objective functions, and obtain the weighted comprehensive optimization objective function of the chassis electronic control system virtual calibration process; establish a pairwise importance discrimination matrix of different categories of optimization objectives in steps 11) and 12), calculate the maximum eigenvalue of the matrix, and calculate the consistency between the maximum eigenvalue and the discrimination matrix. If the consistency judgment passes, normalize the maximum eigenvalue to obtain the weights of different categories of optimization objectives, otherwise reconstruct the discrimination matrix; weight the optimization objectives of all categories based on the obtained optimization objective weights to obtain a comprehensive optimization objective function.

[0055] Step 2: Pre-train the Bayesian proxy model based on the calibration process data of other chassis electronic control systems or the field test data of the system to be calibrated, as follows:

[0056] 21) Collect the electronic control system parameter type data of other models of chassis electronic control system to be calibrated during the on-site calibration process, the vehicle motion data corresponding to the parameter data of the system to be calibrated during the actual vehicle test process, and the chassis electronic control system working status data, and use the method in step 1 to calculate the comprehensive optimization objective function corresponding to different chassis electronic control system parameters;

[0057] 22) Perform pre-training of the Bayesian proxy model and select the Gaussian process as the Bayesian proxy model. A Gaussian process is defined by the mean function μ(x) and the kernel function k(x, x′), and the formula is as follows:

[0058] f(x)~GP(μ(x),k(x,x′))

[0059] Among them, the mean function μ(x) selects the zero mean function, that is, μ(x) = 0, and the kernel function selects the radial basis function, and the formula is as follows:

[0060]

[0061] Among them, ||xx′|| 2 is the square of the Euclidean distance between the input points x and x′; σ is the kernel width parameter, which controls the smoothness of the kernel function;

[0062] 23) Use the data in step 21) to fit the Gaussian process constructed in step 22) to complete the pre-training of the Bayesian proxy model.

[0063] Step 3: Build a high-precision simulation model of the current chassis electronic control system and a test scenario model in the simulation environment, as follows:

[0064] 31) Build a high-precision model of the chassis electronic control system to be calibrated;

[0065] Based on the working principle of the chassis electronic control system, a basic principle model of the chassis electronic control system to be calibrated is built. According to the calibration process system data of the chassis electronic control system to be calibrated, various parameters in the basic principle model are continuously adjusted until the output result of the basic principle model can accurately map the performance of the real chassis electronic control system, thereby obtaining a high-precision simulation model of the chassis electronic control system that can be used for virtual calibration based on the basic principle model;

[0066] 32) Build a chassis electronic control system test condition scenario model. According to the chassis electronic control system test regulations, establish a chassis electronic control system calibration condition table. The condition table records all the condition types and corresponding parameters that need to be calibrated. Build the corresponding test condition scenario model in the simulation environment according to the calibration condition table.

[0067] Step 4: Use the Bayesian probability improvement method to sample the test condition type and specific parameters of the test condition, as follows:

[0068] 41) Select the test condition to be calibrated by using the importance sampling algorithm. The importance sampling algorithm uses a new sampling probability to accelerate the sampling of the condition type with greater importance. The test condition scenarios constructed in step 32) are numbered, and the number of different condition types in step 21) is counted to obtain the corresponding probability distribution p(x). The optimization objective function results corresponding to different conditions in step 21) are counted to obtain the importance sampling probability distribution q(x) of different test conditions. The ratio of q(x) to p(x) is used as the weight of different conditions to be calibrated, and the sampling of different test condition types is completed based on the obtained weight results;

[0069] 42) The Bayesian probability improvement method is used to sample the test parameters of the calibration test. The formula of the Bayesian probability improvement acquisition function a(x) is as follows:

[0070]

[0071] Where, mean(x) is the predicted mean of the Gaussian process model’s optimized objective function at parameter x; ssigma(x) is the predicted standard deviation of the Gaussian process model’s optimized objective function at parameter x; g(x +) is the current best optimization objective function value; l is a hyperparameter used to control the balance between exploration and utilization; Φ is the cumulative distribution function of the standard normal distribution.

[0072] Step 5: Use the sampled test condition type and parameters in the simulation environment to conduct experiments and obtain the objective function results of the system to be calibrated under this condition, as follows:

[0073] 51) Select the calibration test condition type and parameters obtained in step 4, select the corresponding simulation test condition built in step 32) to test the high-precision model of the chassis electronic control system to be calibrated in step 31), and obtain the corresponding comprehensive optimization objective function under the virtual calibration condition.

[0074] Step 6: Optimize the pre-trained proxy model based on the new data obtained, as follows:

[0075] 61) inputting the virtual calibration test condition parameters and the corresponding comprehensive optimization objective function obtained in step 51) into the database in step 21);

[0076] 62) Retrain the Gaussian process proxy model using the method in step 22) to obtain a new proxy model.

[0077] Step 7: Iterate step 4 to step 6 until the target condition of the virtual acceleration calibration process is reached, and obtain the chassis electronic control system corresponding to the current optimal objective function, as follows:

[0078] 71) Set the maximum number of cycles n and the minimum value u of the improvement difference of the comprehensive optimization target, and iterate steps 4 to 6 until the virtual acceleration calibration parameter reaches the maximum number of cycles n or the improvement of the comprehensive optimization target function is less than u for 10 consecutive iterations, then stop the test;

[0079] 72) Obtaining a chassis electronic control system parameter combination corresponding to the current optimal comprehensive objective function as the current optimal chassis electronic control system parameter combination;

[0080] 73) Count the standard deviation of the optimization process of different control parameters in the virtual calibration process, use this value as the standard deviation of parameter changes in the subsequent field calibration process, use the optimal parameter in step 72) as the mean, and use the parameter range within three times the standard deviation above and below the mean as the debugging space of the chassis electronic control system parameters in the subsequent field calibration test, and further accurately adjust the chassis electronic control system parameters.

[0081] In summary, the present invention solves the problem that traditional optimization algorithms fall into local optimality or cannot find the global optimal solution during the adjustment process of high-dimensional parameter space. It can effectively solve the problem of waste of multi-dimensional parameter traversal combination test resources of chassis electronic control systems and accelerate the virtual calibration process of chassis electronic control systems.

[0082] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the protection scope of the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention. These simple variations all belong to the protection scope of the present invention.

[0083] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

[0084] In addition, various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.

Claims

1. A virtual acceleration calibration method for chassis electronic control system parameters based on Bayesian optimization, characterized in that: The following steps are involved: Step 1: Establish the chassis electronic control system optimization objective function according to the vehicle motion state and chassis system working characteristics; Step 2: Pre-training the Bayesian proxy model based on the calibration process data of other chassis electronic control systems or the field test data of the system to be calibrated; Step 3: Build a high-precision simulation model of the current chassis electronic control system and a test condition scenario model in a simulation environment; Step 4: Use the Bayesian probability improvement method to sample the test condition type and specific parameters of the test condition; Step 5: Use the sampled test condition type and parameters to conduct experiments in the simulation environment and obtain the objective function result of the system to be calibrated under the condition; Step 6: Optimize the pre-trained proxy model based on the new data obtained; Step 7: Iterate step 4 to step 6 in a loop until the target condition of the virtual acceleration calibration process is reached, and obtain the chassis electronic control system corresponding to the current optimal objective function.

2. The method for virtual acceleration calibration of chassis electronic control system parameters based on Bayesian optimization according to claim 1 is characterized in that: The specific method of step one is as follows: 11) Construct the chassis electronic control system virtual calibration vehicle motion state optimization objective function; consider the vehicle motion to establish the objective function, including the handling stability optimization objective, ride comfort optimization objective, and driving safety optimization objective; 12) Construct the operating characteristic optimization objective function of the chassis electronic control system virtual calibration system; Considering the working characteristics of the chassis electronic control system, the objective function is established, including the optimization objectives of control accuracy, robust stability, and work efficiency; 13) Use the analytic hierarchy process to analyze the weights of different optimization objective functions, and obtain the weighted comprehensive optimization objective function of the chassis electronic control system virtual calibration process; establish a pairwise importance discrimination matrix of different categories of optimization objectives in steps 11) and 12), calculate the maximum eigenvalue of the matrix, and calculate the consistency between the maximum eigenvalue and the discrimination matrix. If the consistency judgment is passed, the maximum eigenvalue is normalized to obtain the weights of different categories of optimization objectives, otherwise the discrimination matrix is ​​reconstructed; Based on the obtained optimization target weights, the optimization targets of all classes are weighted to obtain a comprehensive optimization objective function.

3. The method for virtual acceleration calibration of chassis electronic control system parameters based on Bayesian optimization according to claim 1 is characterized in that: The specific method of step 2 is as follows: 21) Collect the electronic control system parameter type data of other models of chassis electronic control system to be calibrated during the on-site calibration process, the vehicle motion data corresponding to the parameter data of the system to be calibrated during the actual vehicle test process, and the chassis electronic control system working status data, and use the method in step 1 to calculate the comprehensive optimization objective function corresponding to different chassis electronic control system parameters; 22) Perform pre-training of the Bayesian proxy model and select the Gaussian process as the Bayesian proxy model. A Gaussian process is defined by the mean function μ(x) and the kernel function k(x, x′), and the formula is as follows: f(x)~GP(μ(x),k(x,x′)) Among them, the mean function μ(x) selects the zero mean function, that is, μ(x) = 0, and the kernel function selects the radial basis function, and the formula is as follows: Among them, ||xx′|| 2 is the square of the Euclidean distance between the input points x and x′; σ is the kernel width parameter, which controls the smoothness of the kernel function; 23) Use the data in step 21) to fit the Gaussian process constructed in step 22) to complete the pre-training of the Bayesian proxy model.

4. The method for virtual acceleration calibration of chassis electronic control system parameters based on Bayesian optimization according to claim 3 is characterized in that: The specific method of step three is as follows: 31) Build a high-precision model of the chassis electronic control system to be calibrated, build a basic principle model of the chassis electronic control system to be calibrated based on the working principle of the chassis electronic control system, and continuously adjust various parameters in the basic principle model according to the calibration process system data of the chassis electronic control system to be calibrated, until the output result of the basic principle model can accurately map the performance of the real chassis electronic control system, thereby obtaining a high-precision simulation model of the chassis electronic control system that can be used for virtual calibration based on the basic principle model; 32) Build a chassis electronic control system test condition scenario model. According to the chassis electronic control system test regulations, establish a chassis electronic control system calibration condition table. The condition table records all the condition types and corresponding parameters that need to be calibrated. Build the corresponding test condition scenario model in the simulation environment according to the calibration condition table.

5. The method for virtual acceleration calibration of chassis electronic control system parameters based on Bayesian optimization according to claim 4 is characterized in that: The specific method of step 4 is as follows: 41) Select the test condition to be calibrated by using the importance sampling algorithm. The importance sampling algorithm uses a new sampling probability to accelerate the sampling of the condition type with greater importance. The test condition scenarios constructed in step 32) are numbered, and the number of different condition types in step 21) is counted to obtain the corresponding probability distribution p(x). The optimization objective function results corresponding to different conditions in step 21) are counted to obtain the importance sampling probability distribution q(x) of different test conditions. The ratio of q(x) to p(x) is used as the weight of different conditions to be calibrated, and the sampling of different test condition types is completed based on the obtained weight results; 42) The Bayesian probability improvement method is used to sample the test parameters of the calibration test. The formula of the Bayesian probability improvement acquisition function a(x) is as follows: Where, mean(x) is the predicted mean of the Gaussian process model’s optimized objective function at parameter x; sigma(x) is the predicted standard deviation of the Gaussian process model’s optimized objective function at parameter x; g(x + ) is the current best optimization objective function value; l is a hyperparameter used to control the balance between exploration and utilization; Φ is the cumulative distribution function of the standard normal distribution.

6. The method for virtual acceleration calibration of chassis electronic control system parameters based on Bayesian optimization according to claim 5 is characterized in that: The specific method of step five is as follows: 51) Select the calibration test condition type and parameters obtained in step 4, select the corresponding simulation test condition built in step 32) to test the high-precision model of the chassis electronic control system to be calibrated in step 31), and obtain the corresponding comprehensive optimization objective function under the virtual calibration condition.

7. The method for virtual acceleration calibration of chassis electronic control system parameters based on Bayesian optimization according to claim 6 is characterized in that: The specific method of step six is ​​as follows: 61) inputting the virtual calibration test condition parameters and the corresponding comprehensive optimization objective function obtained in step 51) into the database in step 21); 62) Retrain the Gaussian process proxy model using the method in step 22) to obtain a new proxy model.

8. The method for virtual acceleration calibration of chassis electronic control system parameters based on Bayesian optimization according to claim 6 is characterized in that: The specific method of step seven is as follows: 71) Set the maximum number of cycles n and the minimum value u of the improvement difference of the comprehensive optimization target, and iterate steps 4 to 6 until the virtual acceleration calibration parameter reaches the maximum number of cycles n or the improvement of the comprehensive optimization target function is less than u for 10 consecutive iterations, then stop the test; 72) Obtaining a chassis electronic control system parameter combination corresponding to the current optimal comprehensive objective function as the current optimal chassis electronic control system parameter combination; 73) Count the standard deviation of the optimization process of different control parameters in the virtual calibration process, use this value as the standard deviation of parameter changes in the subsequent field calibration process, use the optimal parameter in step 72) as the mean, and use the parameter range within three times the standard deviation above and below the mean as the debugging space of the chassis electronic control system parameters in the subsequent field calibration test, and further accurately adjust the chassis electronic control system parameters.

Citation Information

Patent Citations

  • Automatic transmission control parameter pre-calibration method based on agent model

    CN110007598A

  • Automatic driving environment perception-oriented small sample in-loop learning system and method

    CN110070139A

  • PID parameter setting optimization method based on improved Bayesian model

    CN112000001A

  • Automobile prediction cruise parameter self-tuning control system based on fusion of driving mode information

    CN113401123A

  • Method for rapidly evaluating performance of hypersonic interceptor

    CN116011524A

Cited By

  • Method for carrying out model parameter optimization on wave energy device model test error

    CN119538751A

  • Chassis electric control system parameter sensitivity analysis method based on deep proxy model

    CN120370896A

  • Chassis electric control system parameter sensitivity analysis method based on deep agent model

    CN120370896B

  • Chassis system scene self-adaptive virtual calibration method based on reinforced fine tuning

    CN122386650A