Shock absorber force value prediction method, device, equipment and medium

Abstract

The invention provides a shock absorber force value prediction method, device, equipment and medium, a multi-round k-fold cross validation mode is utilized, a training sample set of small samples is utilized to train an initial random forest regression model, target vehicle model parameters of a vehicle are input into the trained random forest regression model, and the force value of the shock absorber is predicted. According to the method, the corresponding target shock absorber force value can be estimated, and the data of the currently developed vehicle model parameters and shock absorber force values are limited, and the number of training sample sets is small, so that k-fold cross validation can be repeatedly carried out for multiple times, the training sample sets are fully utilized, the generalization ability and accuracy of the model are improved, and in addition, the method has the advantages of being high in practicability and easy to popularize. The optimal hyper-parameter is determined by utilizing a Bayesian optimization mode, so that the operation rate of model optimization can be improved. In a word, the force value of the shock absorber is predicted through the random forest regression model, manual debugging is not needed, a large amount of manpower and time are saved, the prediction accuracy is greatly improved, usability is high, the calculation speed is high, and good robustness is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of vehicle shock absorbers, and in particular to a shock absorber force value prediction method, device, equipment and medium. BACKGROUND

[0002] As an important component on a vehicle, a shock absorber can reduce the impact and vibration on the vehicle caused by uneven road surfaces, thereby improving the driving stability and comfort of the vehicle. In order to meet the driving needs of users in different working conditions, the force value of the shock absorber needs to be accurately adjusted to achieve relatively balanced vehicle control.

[0003] Currently, the force value of the shock absorber is usually manually adjusted by the adjustment personnel according to their own experience, which is highly subjective and can lead to inaccurate adjustment results. Manual adjustment also wastes a lot of manpower and time. In addition, due to the large number of vehicle models and the large differences between vehicles, it is difficult to accurately adjust the force value of the shock absorber for a vehicle model. Therefore, it is an urgent technical problem to provide a suitable shock absorber force value prediction method. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a shock absorber force value prediction method, device, equipment and medium, which predicts the force value of the shock absorber through a random forest regression model, without the need for manual adjustment, saving a lot of manpower and time, greatly improving the prediction accuracy, and being easy to use, fast in calculation and having good robustness. The specific scheme is as follows:

[0005] In one aspect, the present application provides a shock absorber force value prediction method, which comprises:

[0006] Step 101, dividing the training sample set by k-fold cross-validation method to obtain k subsets; the k subsets include k-1 sub-training sets and the remaining 1 sub-validation set; k is an integer greater than 1; the training sample set includes vehicle model parameters and the corresponding shock absorber force values of the vehicle model parameters;

[0007] Step 102, training an initial random forest regression model with first hyperparameters through the sub-training set to obtain a random forest regression model, testing the random forest regression model with the sub-validation set to determine the first determination coefficient of the random forest regression model; the first hyperparameters include the optimal decision tree number, the maximum depth of the decision tree, the maximum feature number of the decision tree and the minimum impurity number of the node split;

[0008] Step 103, calculating the average value of the k first determination coefficients obtained in the k-fold cross-validation to obtain a second determination coefficient;

[0009] Step 104, repeating steps 101 to 103 until a first preset number of times is reached, obtaining a plurality of second decision coefficients and calculating an average value to obtain a third decision coefficient;

[0010] Step 105, repeating steps 101 to 104 within the hyperparameter space until a second preset number of times is reached, obtaining a plurality of third decision coefficients; the plurality of third decision coefficients correspond one-to-one to the plurality of first hyperparameters;

[0011] Step 106, performing Bayesian optimization on the hyperparameters according to the plurality of third decision coefficients and the plurality of first hyperparameters to determine an optimal hyperparameter and a target random forest regression model corresponding to the optimal hyperparameter; the target random forest regression model is used to output a target shock absorber force value according to an input target vehicle parameter.

[0012] In another aspect, the embodiments of the present application also provide a shock absorber force value prediction device, the device comprising:

[0013] The division unit is configured to perform step 101, divide the training sample set using a k-fold cross-validation method to obtain k subsets; the k subsets include k-1 sub-training sets and the remaining 1 sub-validation set; k is an integer greater than 1; the training sample set includes vehicle parameters and shock absorber force values corresponding to the vehicle parameters;

[0014] The training unit is configured to perform step 102, train an initial random forest regression model with first hyperparameters through the sub-training set to obtain a random forest regression model, test the random forest regression model using the sub-validation set, and determine a first decision coefficient of the random forest regression model; the first hyperparameters include the number of optimal decision trees, the maximum depth of the decision tree, the maximum number of features of the decision tree, and the minimum number of impurities for node splitting;

[0015] The first calculation unit is configured to perform step 103, calculate an average value of the k first decision coefficients obtained in the k-fold cross-validation to obtain a second decision coefficient;

[0016] The second calculation unit is configured to perform step 104, repeat steps 101 to 103 until a first preset number of times is reached, obtain a plurality of second decision coefficients and calculate an average value to obtain a third decision coefficient;

[0017] The third calculation unit is configured to perform step 105, repeat steps 101 to 104 within the hyperparameter space until a second preset number of times is reached, obtain a plurality of third decision coefficients; the plurality of third decision coefficients correspond one-to-one to the plurality of first hyperparameters;

[0018] The optimization unit is configured to perform step 106, perform Bayesian optimization on the hyperparameters according to the plurality of third decision coefficients and the plurality of first hyperparameters, determine an optimal hyperparameter, and a target random forest regression model corresponding to the optimal hyperparameter; and the target random forest regression model is configured to output a target shock absorber force value according to an input target vehicle parameter.

[0019] In another aspect, the embodiments of the present application provide a computer device, which comprises a processor and a memory:

[0020] The memory is configured to store program code and transmit the program code to the processor.

[0021] The processor is configured to execute the method according to the instructions in the program code.

[0022] In another aspect, the embodiments of the present application provide a computer readable storage medium, which is configured to store a computer program, and the computer program is configured to execute the method according to the above aspects.

[0023] The embodiments of the present application provide a shock absorber force value prediction method, device, equipment and medium, the method comprises the following steps: step 101, dividing the training sample set by using the k-fold cross-validation method to obtain k subsets; the k subsets include k-1 sub-training sets and the remaining 1 sub-validation set; k is an integer greater than 1; the training sample set includes vehicle parameters and shock absorber force values corresponding to the vehicle parameters; step 102, training an initial random forest regression model with first hyperparameters by using the sub-training set to obtain a random forest regression model, testing the random forest regression model by using the sub-validation set, and determining a first decision coefficient of the random forest regression model; the first hyperparameters include the number of optimal decision trees, the maximum depth of the decision tree, the maximum number of features of the decision tree, and the minimum number of impurities for node splitting; step 103, calculating the average value of the k first decision coefficients obtained in the k-fold cross-validation to obtain a second decision coefficient; step 104, repeating steps 101 to 103 until a first preset number of times is reached, obtaining a plurality of second decision coefficients and calculating the average value to obtain a third decision coefficient; step 105, repeating steps 101 to 104 within the hyperparameter space until a second preset number of times is reached, obtaining a plurality of third decision coefficients; the plurality of third decision coefficients correspond to the plurality of first hyperparameters one by one; step 106, performing Bayesian optimization on the hyperparameters according to the plurality of third decision coefficients and the plurality of first hyperparameters, determining an optimal hyperparameter, and a target random forest regression model corresponding to the optimal hyperparameter; and the target random forest regression model is configured to output a target shock absorber force value according to an input target vehicle parameter.

[0024] Therefore, by inputting the target vehicle model parameters of the vehicle into the trained random forest regression model, the corresponding target shock absorber force value can be estimated. Since the developed vehicle model parameters and shock absorber force values are limited, the training sample set is small, and the k-fold cross-validation can be repeated multiple times to fully utilize the training sample set, improve the generalization ability and accuracy of the model. In addition, by using the Bayesian optimization method to determine the optimal hyperparameters, the operation speed of the model optimization can be improved. In summary, by using the random forest regression model to predict the shock absorber force value, manual debugging is not required, a large amount of manpower and time is saved, the prediction accuracy is greatly improved, and the model is easy to use, fast to calculate and has good robustness. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0026] Figure 1 A flowchart of a shock absorber force value prediction method provided by an embodiment of the present application is shown;

[0027] Figure 2 A schematic diagram of a vehicle model parameter provided by an embodiment of the present application is shown;

[0028] Figure 3 A prediction result diagram of a shock absorber force value provided by an embodiment of the present application is shown;

[0029] Figure 4 A structural block diagram of a shock absorber force value prediction device provided by an embodiment of the present application is shown;

[0030] Figure 5 A structural diagram of a computer device provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0031] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings.

[0032] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0033] For ease of understanding, the application provides a kind of damper force value prediction method, device, equipment and medium provided by the embodiment of the application are described in detail below in conjunction with the drawings.

[0034] Reference Figure 1 As shown in the figure, the flowchart of the damper force value prediction method provided by the embodiment of the application can include the following steps.

[0035] Step 101, divide the training sample set using k-fold cross-validation method to obtain k subsets.

[0036] In the embodiment of the application, the training sample set can be collected, which can include vehicle parameters and corresponding damper force values of vehicle parameters. Vehicle parameters are related parameters of a vehicle, which can uniquely identify a type of vehicle. The damper force value can be understood as the force of the damper on the vehicle. In the training sample set, there can be many corresponding relationships between vehicle parameters and damper force values.

[0037] Since there is no open source data set for vehicle parameters and damper force value data sets for different vehicle types, the actual development of vehicle data is limited, resulting in a small amount of data in the training sample set. Therefore, in order to make more full use of all the data in the training sample set, the training sample set can be divided using k-fold cross-validation method to obtain k subsets, wherein k is an integer greater than 1.

[0038] Specifically, the k subsets can include k-1 sub-training sets and the remaining 1 sub-validation set; the sub-training set is used to train the model, and the sub-validation set is used to verify the model. For example, when k is 5, 4 subsets are used as sub-training sets, and the remaining 1 subset is used as sub-validation set.

[0039] In each verification, k-1 subsets can be selected from k subsets as sub-training sets, i.e. in k rounds of cross-validation process, each subset will be used as a sub-validation set, so as to enrich the model training data, fully mine the existing training sample set, and improve the generalization ability and accuracy of the model.

[0040] Among them, k-fold cross-validation method can be five-fold cross-validation method or ten-fold cross-validation method, so as to cross-verify as many times as possible, provide more seed training set composition methods, and divide the training sample set more carefully, make more full use of the training sample set, and further improve the generalization ability of the model.

[0041] In one possible implementation, the vehicle parameters may include at least one of the following parameters: drive shaft type, front suspension structure, rear suspension structure, driving performance type, wheelbase, front track, rear track, center of gravity, front roll center height, rear roll center height, front axle load, rear axle load, tire specifications, front suspension frequency, rear suspension frequency, front suspension longitudinal flexibility, rear suspension longitudinal flexibility, roll gradient, front stabilizer bar diameter, rear stabilizer bar diameter, front spring stiffness, rear spring stiffness, vehicle three-axis inertia, front shock absorber leverage ratio, and rear shock absorber leverage ratio.

[0042] As an example, refer to Figure 2 As shown, drive shaft types can include front-wheel drive, rear-wheel drive, and four-wheel drive. Front suspension structures can be MacPherson struts and double wishbones, while rear suspension structures can be torsion beams, blade arms, H-arms, and five-link suspensions. Driving performance types (tuning styles) can include comfort, balanced, and handling. Different driving performance types can have different damper force values. For example, the damper force value for comfort can be less than that for balanced. A smaller damper force value results in smoother vehicle movement, while a larger damper force value makes the vehicle movement closer to the road's natural undulations, providing a stronger road feel. Tire specifications can include tire pressure, tire thickness, tire width, aspect ratio, and rim size. The vehicle's three-axis inertia can include CVW (Curb Vehicle Weight) pitch inertia, CVW yaw inertia, and CVW roll inertia.

[0043] Vehicles are typically equipped with front and rear shock absorbers. Each shock absorber possesses both restoring force and compressive force. Since there is no force without relative motion, the force provided by the shock absorber is a damping force, a force that opposes relative motion. Specifically, the restoring force can be understood as the downward pulling force provided by the shock absorber to the vehicle body when it moves upward relative to the wheels; that is, the force that pulls the vehicle body downward. The compressive force can be understood as the supporting force provided by the shock absorber to the vehicle body when it moves downward relative to the wheels; that is, the force that supports the vehicle body upward. Therefore, the shock absorber force value can include at least one of the following parameters: the restoring force of the front shock absorber, the compressive force of the front shock absorber, the restoring force of the rear shock absorber, and the compressive force of the rear shock absorber.

[0044] Furthermore, in order to more accurately determine the damper force value, the corresponding damper force value can be set according to different damper speeds. That is, different damper speeds correspond to different damper force values. The damper speed can be understood as the speed of the piston rod relative to the damper cylinder. The direction of movement is perpendicular to the ground, that is, up and down movement. For example, the damper force value can increase with the increase of the damper speed, thereby ensuring that there is a high damper force value when the vehicle body moves up and down violently, so as to ensure the stability of the vehicle body.

[0045] That is, for each shock absorber, there can be a shock absorber restoring force and a shock absorber compression force at different shock absorber speeds. For example, the shock absorber speed is 0.05 m / s, 0.13 m / s, 0.26 m / s, 0.39 m / s, 0.52 m / s, 1.0 m / s, 1.5 m / s, respectively, corresponding to the respective front shock absorber restoring force, front shock absorber compression force, rear shock absorber restoring force and rear shock absorber compression force.

[0046] In this way, by enriching the parameter category of the vehicle model parameter and the type of the shock absorber force value, the initial random forest regression model can be more accurately trained, and the accuracy of the model output can be improved.

[0047] Step 102, training the initial random forest regression model with the first hyperparameter through the sub-training set to obtain a random forest regression model, testing the random forest regression model with the sub-validation set to determine the first determination coefficient of the random forest regression model.

[0048] In the embodiments of the present application, the initial random forest regression model with the first hyperparameter can be trained by using k-1 sub-training sets, and the random forest regression model is obtained after training. The first hyperparameter can include the optimal number of decision trees, the maximum depth of the decision tree, the maximum number of features of the decision tree, and the minimum number of impurities for node splitting, that is, the first hyperparameter can determine the number of decision trees in the initial random forest regression model, the maximum depth of each decision tree, the maximum number of features of each decision tree, and the minimum number of impurities for node splitting.

[0049] After training the random forest regression model, the model can be tested by using the sub-validation set, the vehicle model parameters in the sub-validation set are input into the random forest regression model to obtain the predicted shock absorber force value, which is compared with the real shock absorber force value corresponding to the vehicle model parameters to calculate the determination coefficient R 2 , which can be referred to as the first determination coefficient. For example, the predicted shock absorber force value can be the front and rear shock absorber restoring force and compression force at different shock absorber speeds.

[0050] Step 103, calculating the average value of the k first determination coefficients obtained in the k-fold cross-validation to obtain a second determination coefficient.

[0051] In the embodiments of the present application, a first determination coefficient can be determined in one training and verification process, and k first determination coefficients can be obtained after k-fold cross-validation.

[0052] For example, when performing five-fold cross-validation, in the first validation, four subsets A, B, C and D are taken as sub-training sets, and the remaining one subset E is taken as a sub-validation set to obtain a first decision coefficient r1, in the second validation, four subsets A, B, C and E are taken as sub-training sets, and the remaining one subset D is taken as a sub-validation set to obtain a first decision coefficient r2. A total of five times are performed, and all subsets are taken as a sub-validation set once, and a total of five first decision coefficients r1, r2, r3, r4, r5 are obtained, and the average of the five is calculated to obtain a second decision coefficient r'.

[0053] Step 104, repeating steps 101 to 103 until a first preset number of times is reached, obtaining a plurality of second decision coefficients and calculating the average to obtain a third decision coefficient.

[0054] In the embodiments of the present application, in order to make more full use of the data in the training data set, the subset division, model training and cross-validation process can be repeatedly performed, so that in each round of repeated process, subsets with different data can be randomly divided, for example, in the first calculation process, data 1, 2 and 3 are divided into the same subset, data 4, 5 and 6 are divided into another subset, and data 7, 8 and 9 are divided into another subset, in the second calculation, data 1, 3 and 4 can be divided into the same subset, data 2, 7 and 9 can be divided into another subset, and data 5, 6 and 8 can be divided into another subset, thereby enriching the subsets and performing multiple rounds of k-fold cross-validation, thereby determining a second decision coefficient in each round, thereby obtaining a plurality of second decision coefficients.

[0055] When the number of executions reaches the first preset number of times, the loop can be stopped, the average of the plurality of second decision coefficients is calculated to obtain the third decision coefficient, and the third decision coefficient has a corresponding relationship with the random forest regression model with the first hyperparameter.

[0056] Step 105, repeating steps 101 to 104 in the hyperparameter space until a second preset number of times is reached to obtain a plurality of third decision coefficients.

[0057] In the embodiments of the present application, the subset division, model training and multiple rounds of k-fold cross-validation process can be repeatedly performed in a certain hyperparameter space, in each repeated execution process, the randomly determined first hyperparameter will be different, thereby obtaining each third decision coefficient corresponding to each first hyperparameter. That is, in each repetition, a new first hyperparameter is determined, that is, a new random forest regression model is determined, and its corresponding third decision coefficient is obtained, that is, a plurality of third decision coefficients correspond to a plurality of first hyperparameters one by one.

[0058] It can be understood that by setting the random seed, the division mode of the training sample set is consistent when training the model corresponding to different first hyperparameters each time.

[0059] In a possible implementation, in the hyperparameter space, the number of optimal decision trees can be greater than or equal to 50 and less than or equal to 200, the maximum depth of the decision tree is greater than or equal to 1 and less than or equal to 100, the maximum number of features of the decision tree is less than or equal to the number of types of vehicle parameters, and the minimum number of impurities of the node splitting is less than or equal to 5.

[0060] In step 106, the hyperparameters are subjected to Bayesian optimization according to the plurality of third decision coefficients and the plurality of first hyperparameters, to determine optimal hyperparameters and a target random forest regression model corresponding to the optimal hyperparameters.

[0061] In the embodiments of the present application, the hyperparameters can be subjected to Bayesian optimization according to a plurality of corresponding relationships, i.e., the corresponding relationship between the third decision coefficient and the first hyperparameter, until the optimal hyperparameters are determined. The optimal hyperparameters can represent that the predicted value of the random forest regression model corresponding to the optimal hyperparameters is closest to the true value. Then, the random forest regression model can be recorded as a target random forest regression model.

[0062] The target random forest regression model is used to output a target shock absorber force value according to an input target vehicle parameter. That is, in the model application stage, the target random forest regression model can be input with a target vehicle parameter, and the target vehicle parameter is a vehicle parameter of a vehicle to be predicted. Then, the model can predict the corresponding shock absorber force value,

[0063] Reference Figure 3 As shown in the figure, (a) shows the front shock absorber force value, the horizontal coordinate is the shock absorber speed, and the positive and negative of the vertical coordinate can distinguish the restoring force and the compression force. The restoring force is positive, and the compression force is negative. Under different shock absorber speeds, the restoring force of the front shock absorber is 144N, 419N, 891N, 1089N, 1223N, 1677N, and 2106N, respectively. The compression force of the front shock absorber is -133N, -333N, -587N, -698N, -800N, -1206N, and -1604N, respectively. (b) shows the rear shock absorber force value. The restoring force is 161N, 502N, 1330N, 1629N, 1848N, 2654N, and 3431N, respectively. The compression force is -177N, -480N, -839N, -995N, -1108N, -1521N, and -1932N, respectively.

[0064] Therefore, by inputting the target vehicle model parameters of the vehicle into the trained random forest regression model, the corresponding target shock absorber force value can be estimated. Since the developed vehicle model parameters and shock absorber force values are limited, the training sample set is small, and the k-fold cross-validation can be repeated multiple times to fully utilize the training sample set, improve the generalization ability and accuracy of the model. In addition, by using the Bayesian optimization method to determine the optimal hyperparameters, the operation speed of model optimization can be improved. In summary, by using the random forest regression model to predict the shock absorber force value, the shock absorber force value can be effectively associated with the vehicle model parameters, without the need for manual debugging, saving a lot of manpower and time, and greatly improving the prediction accuracy. The shock absorber force value prediction of different vehicle models can be realized by small sample training, which is easy to use, fast in calculation and has good robustness.

[0065] In a possible implementation, the Bayesian optimization of the hyperparameters according to the plurality of third decision coefficients and the plurality of first hyperparameters to determine the optimal hyperparameters and the target random forest regression model corresponding to the optimal hyperparameters can specifically include the following steps 1061-1064.

[0066] Step 1061, determining a probability model according to the plurality of third decision coefficients and the plurality of first hyperparameters.

[0067] Specifically, the function estimation can be performed according to the plurality of third decision coefficients and the plurality of first hyperparameters, and the points are taken as initial observation points to determine the probability model. The probability model can be a TPE (Tree Parzen Estimator) proxy model, which can speed up the search speed of the hyperparameter space compared with the enumeration search and random search methods. The third decision coefficient can be taken as a negative value, and the minimum value f min of the probability model can be solved subsequently. The probability model can identify the correspondence between the hyperparameters and the decision coefficients, that is, determine the hyperparameters under which the maximum decision coefficient can be obtained, and the target random forest regression model corresponding thereto.

[0068] Step 1062, taking the expected increment as a collection function, determining the second hyperparameters by using the collection function, and taking the second hyperparameters as the first hyperparameters to repeatedly execute steps 102-104 to obtain the third decision coefficients corresponding to the second hyperparameters.

[0069] In the embodiments of the present application, the expected increment can be defined as the collection function to determine the next observation point to be calculated, that is, to determine the second hyperparameters. The model training and the process of multiple rounds of k-fold cross-validation are repeatedly executed until the third decision coefficients corresponding to the second hyperparameters are determined to obtain new observation points.

[0070] In step 1063, the probability model is updated according to the second hyperparameter and the third decision coefficient corresponding to the second hyperparameter, to obtain an updated probability model.

[0071] In the embodiments of the present application, the probability model can be updated according to the newly obtained observation point, i.e., the second hyperparameter and the third decision coefficient corresponding thereto, to obtain an updated probability model.

[0072] In step 1064, steps 1062 to 1063 are repeatedly executed until a preset optimization stopping condition is reached, to determine an optimal hyperparameter and a target random forest regression model corresponding to the optimal hyperparameter.

[0073] In the embodiments of the present application, the acquisition function can be recalculated to determine the next observation point, until the preset optimization stopping condition is reached, for example, the number of iterations meets the requirement, so that the optimal solution, i.e., the optimal hyperparameter, can be obtained in the hyperparameter space, and the random forest regression model corresponding thereto is taken as the target random forest regression model. In this way, the next observation point can be more accurately determined, the search speed in the hyperparameter space is accelerated, and the model training speed is improved.

[0074] Based on the above shock absorber force value prediction method, the embodiments of the present application also provide a shock absorber force value prediction device, as shown in Figure 4 The device can include:

[0075] The division unit 201 is configured to perform step 101, divide the training sample set by using the k-fold cross-validation method to obtain k subsets; the k subsets include k-1 sub-training sets and the remaining 1 sub-validation set; k is an integer greater than 1; the training sample set includes vehicle type parameters and shock absorber force values corresponding to the vehicle type parameters;

[0076] The training unit 202 is configured to perform step 102, train an initial random forest regression model with a first hyperparameter by using the sub-training set to obtain a random forest regression model, test the random forest regression model by using the sub-validation set, and determine a first decision coefficient of the random forest regression model; the first hyperparameter includes the number of optimal decision trees, the maximum depth of the decision tree, the maximum number of features of the decision tree, and the minimum number of impurities for node splitting;

[0077] The first calculation unit 203 is configured to perform step 103, calculate the average value of the k first decision coefficients obtained in the k-fold cross-validation to obtain a second decision coefficient;

[0078] The second computing unit 204 is configured to execute step 104, repeatedly execute steps 101-103 until a first preset number of times is reached, obtain a plurality of second decision coefficients, and calculate an average value to obtain a third decision coefficient;

[0079] The third computing unit 205 is configured to execute step 105, repeatedly execute steps 101-104 until a second preset number of times is reached in the hyperparameter space, and obtain a plurality of third decision coefficients; the plurality of third decision coefficients correspond to the plurality of first hyperparameters one by one.

[0080] The optimization unit 206 is configured to execute step 106, perform Bayesian optimization on the hyperparameters according to the plurality of third decision coefficients and the plurality of first hyperparameters, determine an optimal hyperparameter, and determine a target random forest regression model corresponding to the optimal hyperparameter; the target random forest regression model is used to output a target shock absorber force value according to an input target vehicle parameter.

[0081] Optionally, the optimization unit is configured to execute:

[0082] Step 1061, determining a probability model according to the plurality of third decision coefficients and the plurality of first hyperparameters; the probability model identifies the correspondence between the hyperparameters and the decision coefficients;

[0083] Step 1062, taking an expected increment as a collection function, determining a second hyperparameter by using the collection function, and repeatedly executing steps 102-104 by taking the second hyperparameter as the first hyperparameter to obtain a third decision coefficient corresponding to the second hyperparameter;

[0084] Step 1063, updating the probability model according to the second hyperparameter and the third decision coefficient corresponding to the second hyperparameter to obtain an updated probability model;

[0085] Step 1064, repeatedly executing steps 1062-1063 until a preset optimization stop condition is reached, determining the optimal hyperparameter, and determining a target random forest regression model corresponding to the optimal hyperparameter.

[0086] Optionally, the vehicle parameter includes at least one of the following parameters:

[0087] Drive shaft type, front suspension structure, rear suspension structure, driving performance type, wheelbase, front wheel track, rear wheel track, center of mass, front roll center height, rear roll center height, front suspension axle load, rear suspension axle load, tire specification, front suspension bias frequency, rear suspension bias frequency, front suspension longitudinal flexibility, rear suspension longitudinal flexibility, roll gradient, front stabilizer bar diameter, rear stabilizer bar diameter, front spring stiffness, rear spring stiffness, vehicle three-axis inertia, front shock absorber lever ratio, and rear shock absorber lever ratio;

[0088] The shock absorber force value includes at least one of the following parameters:

[0089] The rebound force of the front shock absorber, the compression force of the front shock absorber, the rebound force of the rear shock absorber, and the compression force of the rear shock absorber.

[0090] Optionally, in the hyperparameter space, the number of the optimal decision trees is greater than or equal to 50 and less than or equal to 200, the maximum depth of the decision tree is greater than or equal to 1 and less than or equal to 100, the maximum number of features of the decision tree is less than or equal to the number of categories of the vehicle parameter, and the minimum number of impurities for node splitting is less than or equal to 5.

[0091] Optionally, the k-fold cross-validation method is a five-fold cross-validation method or a ten-fold cross-validation method.

[0092] In another aspect, an embodiment of the present application provides a computer device, as shown in Figure 5 As shown in FIG. 10, a structural diagram of a computer device provided by an embodiment of the present application, the computer device includes a processor 310 and a memory 320:

[0093] The memory 320 is configured to store program code and transmit the program code to the processor 310.

[0094] The processor 310 is configured to execute the method provided by the above-mentioned embodiments according to instructions in the program code.

[0095] The computer device can include a terminal device or a server, and the foregoing apparatus can be configured in the computer device.

[0096] In another aspect, an embodiment of the present application further provides a storage medium, the storage medium is configured to store a computer program, and the computer program is configured to execute the method provided by the above-mentioned embodiments.

[0097] Those skilled in the art can understand that all or part of the steps of the above-mentioned method embodiments can be completed by program instructions hardware, and the foregoing program can be stored in a computer readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed. The storage medium can be at least one of the following media: read-only memory (English: Read-only Memory, abbreviation: ROM), RAM, magnetic disk or optical disk, and various media that can store program codes.

[0098] The various embodiments described in this specification are presented by way of example, and each embodiment is not necessarily composed of all features described. Each embodiment describes a different set of features, and the application can include a single embodiment or a combination of different embodiments. Each embodiment is directed to what is currently believed to be a novel and non-obvious solution to problems in the art. Each embodiment is not, however, directed to the only solution to the problems in the art. The scope of the application is defined by the appended claims and equivalents thereto as allowed by law and not by the breadth, number or quality of its claims.

[0099] The above descriptions are only the preferred embodiments of the present application, and the present application is not limited to the above. Any person skilled in the art, without departing from the scope of the present application, can make many possible changes and modifications to the above disclosed methods and technical contents, or modify equivalent embodiments. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application, without departing from the scope of the present application, are still within the scope of protection of the present application.

Claims

1. A shock absorber force value prediction method characterized by, The method comprises: Step 101, dividing the training sample set by k-fold cross-validation method to obtain k subsets; the k subsets include k-1 sub-training sets and the remaining 1 sub-validation set; k is an integer greater than 1; the training sample set includes vehicle type parameters and damper force values corresponding to the vehicle type parameters; Step 102, training an initial random forest regression model with first hyperparameters by the sub-training set to obtain a random forest regression model, testing the random forest regression model with the sub-validation set to determine the first determination coefficient of the random forest regression model; the first hyperparameters include the optimal decision tree number, the maximum depth of the decision tree, the maximum feature number of the decision tree, and the minimum impurity number of node splitting; Step 103, calculating the average value of the k first determination coefficients obtained in the k-fold cross-validation to obtain a second determination coefficient; Step 104, repeating steps 101 to 103 until a first preset number of times is reached, obtaining multiple second determination coefficients and calculating the average value to obtain a third determination coefficient; Step 105, repeating steps 101 to 104 in the hyperparameter space until a second preset number of times is reached to obtain multiple third determination coefficients; multiple third determination coefficients correspond to multiple first hyperparameters one by one; Step 106, according to multiple third determination coefficients and multiple first hyperparameters, performing Bayesian optimization on hyperparameters to determine optimal hyperparameters and a target random forest regression model corresponding to the optimal hyperparameters; the target random forest regression model is used to output a target damper force value according to an input target vehicle type parameter.

2. The method of claim 1, wherein, According to multiple third determination coefficients and multiple first hyperparameters, performing Bayesian optimization on hyperparameters to determine optimal hyperparameters and a target random forest regression model corresponding to the optimal hyperparameters, comprising: Step 1061, determining a probability model according to multiple third determination coefficients and multiple first hyperparameters; the probability model identifies the correspondence between hyperparameters and determination coefficients; Step 1062, taking the expected increment as a collection function, determining a second hyperparameter using the collection function, and repeating steps 102 to 104 with the second hyperparameter as the first hyperparameter to obtain a third determination coefficient corresponding to the second hyperparameter; Step 1063, updating the probability model according to the second hyperparameter and the third determination coefficient corresponding to the second hyperparameter to obtain an updated probability model; Step 1064, repeating steps 1062 to 1063 until a preset optimization stopping condition is reached to determine the optimal hyperparameters and the target random forest regression model corresponding to the optimal hyperparameters.

3. The method of claim 1, wherein, The vehicle type parameters include at least one of the following parameters: Drive shaft type, front suspension type, rear suspension type, driving performance type, wheelbase, front track, rear track, center of mass, front roll center height, rear roll center height, front suspension axle load, rear suspension axle load, tire specification, front suspension roll frequency, rear suspension roll frequency, front suspension longitudinal flexibility, rear suspension longitudinal flexibility, roll gradient, front stabilizer bar diameter, rear stabilizer bar diameter, front spring stiffness, rear spring stiffness, vehicle three-axis inertia, front shock absorber lever ratio, and rear shock absorber lever ratio; The shock absorber force value includes at least one of the following parameters: Recovery force of the front shock absorber, compression force of the front shock absorber, recovery force of the rear shock absorber, and compression force of the rear shock absorber.

4. The method according to any one of claims 1 to 3, characterized in that, In the hyperparameter space, the number of optimal decision trees is greater than or equal to 50 and less than or equal to 200, the maximum depth of the decision tree is greater than or equal to 1 and less than or equal to 100, the maximum number of features of the decision tree is less than or equal to the number of types of vehicle parameters, and the minimum impurity of node splitting is less than or equal to 5.

5. The method according to any one of claims 1 to 3, characterized in that, The k-fold cross-validation method is a five-fold cross-validation method or a ten-fold cross-validation method.

6. A shock absorber force value prediction device characterized by comprising: The device comprises: A division unit is configured to perform step 101, divide the training sample set by using a k-fold cross-validation method to obtain k subsets, wherein the k subsets include k-1 sub-training sets and the remaining 1 sub-validation set, k is an integer greater than 1, the training sample set includes vehicle parameters and shock absorber force values corresponding to the vehicle parameters; A training unit is configured to perform step 102, train an initial random forest regression model with first hyperparameters by using the sub-training sets to obtain a random forest regression model, test the random forest regression model by using the sub-validation set, and determine a first determination coefficient of the random forest regression model, wherein the first hyperparameters include the number of optimal decision trees, the maximum depth of the decision tree, the maximum number of features of the decision tree, and the minimum impurity of node splitting; A first calculation unit is configured to perform step 103, calculate the average value of the k first determination coefficients obtained in the k-fold cross-validation to obtain a second determination coefficient; A second calculation unit is configured to perform step 104, repeatedly perform steps 101 to 103 until a first preset number of times are reached, obtain multiple second determination coefficients, and calculate the average value to obtain a third determination coefficient; A third calculation unit is configured to perform step 105, repeatedly perform steps 101 to 104 in the hyperparameter space until a second preset number of times are reached, and obtain multiple third determination coefficients, wherein the multiple third determination coefficients correspond to the multiple first hyperparameters one by one; An optimization unit is configured to perform step 106, perform Bayesian optimization on the hyperparameters according to the multiple third determination coefficients and the multiple first hyperparameters to determine optimal hyperparameters and a target random forest regression model corresponding to the optimal hyperparameters, and the target random forest regression model is used to output a target shock absorber force value according to an input target vehicle parameter.

7. The apparatus of claim 6, wherein, The optimization unit is configured to perform: Step 1061, determine a probability model according to the multiple third determination coefficients and the multiple first hyperparameters. The probability model identifies a correspondence between a hyperparameter and a decision coefficient; Step 1062, taking the expected increment as a collection function, determining a second hyperparameter by using the collection function, and repeating steps 102 to 104 by taking the second hyperparameter as the first hyperparameter to obtain a third decision coefficient corresponding to the second hyperparameter; Step 1063, updating the probability model according to the second hyperparameter and the third decision coefficient corresponding to the second hyperparameter to obtain an updated probability model; Step 1064, repeating steps 1062 to 1063 until a preset optimization stop condition is reached to determine the optimal hyperparameter and a target random forest regression model corresponding to the optimal hyperparameter.

8. The apparatus of claim 6, wherein, The vehicle type parameters include at least one of the following parameters: Drive shaft type, front suspension structure, rear suspension structure, driving performance type, wheelbase, front track, rear track, center of mass, front roll center height, rear roll center height, front suspension axle load, rear suspension axle load, tire specification, front suspension bias frequency, rear suspension bias frequency, front suspension longitudinal flexibility, rear suspension longitudinal flexibility, roll gradient, front stabilizer bar diameter, rear stabilizer bar diameter, front spring stiffness, rear spring stiffness, vehicle three-axis inertia, front shock absorber lever ratio, and rear shock absorber lever ratio; The shock absorber force value includes at least one of the following parameters: Front shock absorber recovery force, front shock absorber compression force, rear shock absorber recovery force, and rear shock absorber compression force.

9. A computer device, comprising: The computer device includes a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the method according to the instructions in the program code.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium is used to store a computer program for executing the method according to any one of claims 1-5. The computer readable storage medium is used to store a computer program for executing the method according to any one of claims 1-5.

Images