Vehicle side slip angle determination method, device and equipment and storage medium

Abstract

The invention provides a vehicle side slip angle determination method, device and equipment and a storage medium, and belongs to the field of vehicle control. The method comprises the steps that a training sample is determined according to historical driving state parameters; performing hyper-parameter optimization on the LSTM neural network by using a particle swarm algorithm, and constructing a PSOLSTM model according to the optimized hyper-parameters; training the PSOLSTM model through the training sample to obtain a side slip angle prediction model; obtaining a predicted side slip angle of the target vehicle according to a side slip angle prediction model and the current driving state parameters; constructing sampling points based on the current driving state parameters of the target vehicle, and constructing an unscented Kalman filtering module according to the sampling points; and filtering the predicted side slip angle through an unscented Kalman filtering module, and determining the side slip angle of the target vehicle. Therefore, the prediction precision can be improved through the optimized hyper-parameters and the PSOLSTM, filtering is carried out in combination with the unscented Kalman filtering module, errors can be reduced, and the safety and stability of vehicle control are improved.

Description

Technical Field

[0001] The present invention belongs to the field of vehicle control, and in particular relates to a method, device, equipment and storage medium for determining a vehicle center of mass sideslip angle. Background Art

[0002] In order to improve the active safety and handling stability of vehicles, a large number of active safety control technologies have been developed and applied. However, the fundamental of active safety control of vehicles is the accurate grasp of vehicle status information, such as the vehicle's center of mass sideslip angle, which is the core element of vehicle stability evaluation.

[0003] In actual engineering applications, direct measurement of the center of mass slip angle requires expensive special test equipment and a specific installation and fixing method. It is currently not suitable for configuration on mass-produced vehicles and is only suitable for the development and testing stage. Therefore, from the perspective of cost saving and practical application, the center of mass slip angle must be estimated based on the sensor configuration on mass-produced vehicles.

[0004] The commonly used estimation method at present is to estimate the center of mass slip angle based on the dynamic method, such as using the single track model as a state parameter observation model. Under steady-state steering conditions, the center of mass slip angle of the vehicle is a linear function of the front wheel turning angle, while the single track model focuses on the influence of the tire lateral force on the lateral motion of the vehicle. The neglect of the longitudinal force of the tire makes the estimation of this method no longer accurate under conditions of obvious side slip or vehicle acceleration. In addition, the prior art also uses a three-degree-of-freedom vehicle model including longitudinal, lateral and yaw motion to estimate the center of mass slip angle. This method has high requirements on the accuracy of the tire lateral force, and for this purpose, it is necessary to establish a high-precision vehicle model, which affects the real-time performance of the estimation result. In addition, the prior art uses a high-order vehicle dynamics model to obtain a more accurate estimate of the center of mass slip angle, but in practical applications, the real-time performance and stability of such a high-order dynamics model need to be verified.

[0005] Therefore, there are large errors in the estimation of the sideslip angle of the center of mass in the prior art, which in turn affects the safety and stability of the vehicle. Summary of the invention

[0006] In order to solve the problem of potential safety hazards in the above-mentioned vehicle control, the present invention provides a method, device, equipment and storage medium for determining the sideslip angle of the center of mass of a vehicle.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] First, a method for determining a vehicle center of mass sideslip angle is provided, the method comprising:

[0009] Acquire historical driving state parameters of the vehicle under different working conditions, and determine training samples according to the historical driving state parameters; the historical driving state parameters include a true center of mass sideslip angle and other state parameters corresponding to the true center of mass sideslip angle;

[0010] The particle swarm algorithm is used to optimize the hyperparameters of the LSTM neural network, and the PSOLSTM model is constructed according to the optimized hyperparameters; the PSOLSTM model is trained by the training samples to obtain a center of mass sideslip angle prediction model;

[0011] Obtaining current driving state parameters of the target vehicle except for the center of mass sideslip angle, inputting the current driving state parameters into the center of mass sideslip angle prediction model for prediction, and obtaining a predicted center of mass sideslip angle of the target vehicle;

[0012] Constructing sampling points based on the predicted centroid sideslip angle, and constructing an unscented Kalman filter module according to the sampling points;

[0013] The predicted center-of-mass sideslip angle is filtered by the unscented Kalman filter module to determine the center-of-mass sideslip angle of the target vehicle.

[0014] Optionally, the use of a particle swarm algorithm to optimize the hyperparameters of the LSTM neural network includes:

[0015] Initialize the position and velocity of the particle swarm, and randomly generate the position of each particle in the particle swarm;

[0016] The mean square error of the particle swarm is set as the target fitness value, and the individual extreme value (p ij ) and the global extreme value (p gi );

[0017] The position of the particle (x ij ) and speed (v ij ) is updated in real time, and then the fitness value is obtained according to the latest position of the particle, and then the individual extreme value and the global extreme value are updated; the specific update formula is:

[0018] v ij (t+1)=w·v ij (t)+c1r1[p ij (t)-x ij (t)]+c2r2[p gi (t)-x ij (t)];

[0019] x ij (t+1)=x ij (t)+v ij (t+1);

[0020] When the optimization result meets the preset conditions, the update is ended and the final optimization result is output to obtain the optimized hyperparameters.

[0021] Optionally, the other state parameters include a steering wheel angle, a heading angle, a lateral acceleration, a longitudinal velocity, a lateral velocity and a yaw angular velocity of the vehicle; and the PSOLSTM model is trained by the training samples to obtain a center of mass sideslip angle prediction model including:

[0022] The steering wheel angle, heading angle, lateral acceleration, longitudinal velocity, lateral velocity and yaw rate of the vehicle are input into the PSOLSTM model to obtain a prediction result of the center of mass sideslip angle;

[0023] With the goal of minimizing the deviation between the predicted result and the true center of mass sideslip angle, the PSOLSTM model is trained to obtain a center of mass sideslip angle prediction model.

[0024] Optionally, constructing sampling points based on the predicted centroid sideslip angle includes:

[0025] The constraint formula of the sampling point is constructed according to the predicted sideslip angle of the center of mass. The constraint formula of the sampling point is:

[0026]

[0027] The weights of the sampling points are:

[0028]

[0029] in, is the mean value of the system state quantity, that is, the predicted mean value of the center of mass sideslip angle, P is the variance of the state quantity, that is, the predicted variance of the center of mass sideslip angle, n is the dimension of the state vector to be estimated, and λ is the scaling factor.

[0030] Optionally, constructing an unscented Kalman filter module according to the sampling points includes:

[0031] Calculate the predicted value, mean and variance of the sample points: When i>1, construct 2n+1 sample points:

[0032]

[0033] in, is the predicted value of the sample point, X i is the prior estimate at time i, and k is the gain at time i;

[0034] Calculate the mean and variance of the predicted sample points:

[0035]

[0036] Use an unscented transformation to handle the nonlinear transfer of means and covariances to determine new forecasts:

[0037]

[0038] Update the mean and variance of the predicted values:

[0039]

[0040] Calculate the Kalman gain of the unscented Kalman filter:

[0041]

[0042] Update state mean and variance:

[0043]

[0044] P x (k) = P x (k)-KP zz (k-1)K T .

[0045] Secondly, a device for determining the sideslip angle of the center of mass of a vehicle is also provided, the device comprising:

[0046] An acquisition module, used for acquiring historical driving state parameters of the vehicle under different working conditions, and determining training samples according to the historical driving state parameters; the historical driving state parameters include a true center of mass sideslip angle and other state parameters corresponding to the true center of mass sideslip angle;

[0047] A construction module is used to optimize the hyperparameters of the LSTM neural network using a particle swarm algorithm, and to construct a PSOLSTM model according to the optimized hyperparameters; the PSOLSTM model is trained using the training samples to obtain a center of mass sideslip angle prediction model;

[0048] The acquisition module is also used to acquire the current driving state parameters of the target vehicle except the sideslip angle of the center of mass;

[0049] A prediction module, used for inputting the current driving state parameters into the center of mass sideslip angle prediction model for prediction, and obtaining a predicted center of mass sideslip angle of the target vehicle;

[0050] The construction module is also used to construct sampling points based on the predicted centroid sideslip angle, and to construct an unscented Kalman filter module according to the sampling points;

[0051] The filtering module is used to filter the predicted center of mass sideslip angle through the unscented Kalman filter module to determine the center of mass sideslip angle of the target vehicle.

[0052] In addition, a computer-readable storage medium is provided, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the method for determining the sideslip angle of the center of mass of a vehicle described above is implemented.

[0053] Finally, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for determining the sideslip angle of the center of mass of a vehicle when executing the program.

[0054] The method for determining the sideslip angle of the center of mass of a vehicle provided by the present invention has the following beneficial effects:

[0055] Firstly, the particle swarm algorithm is used to optimize the hyperparameters of the LSTM neural network, and the PSOLSTM model is constructed according to the optimized hyperparameters, and then the PSOLSTM model is trained to obtain the center of mass sideslip angle prediction model; in this way, by optimizing the hyperparameters of the LSTM model, PSOLSTM can improve the prediction accuracy and generalization ability, and provide strong support for the prediction and application of vehicle form state parameters; in addition, after obtaining the predicted center of mass sideslip angle through the center of mass sideslip angle prediction model, an unscented Kalman filter module is constructed in combination with the current driving state parameters of the target vehicle to filter the predicted center of mass sideslip angle, so as to further improve the evaluation accuracy of the center of mass sideslip angle, which is beneficial to reduce errors and improve the safety and stability of vehicle control. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiment of the present invention and its design scheme, the following briefly introduces the drawings required for this embodiment. The drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0057] Figure 1 The present invention is a schematic flow chart of a method for determining a vehicle center of mass sideslip angle according to an exemplary embodiment of the present invention.

[0058] Figure 2 A PSOLSTM neural network topology diagram is provided according to an exemplary embodiment of the present invention.

[0059] Figure 3 A simulation diagram of a steering wheel angle step condition provided according to an exemplary embodiment of the present invention.

[0060] Figure 4 The present invention is a block diagram of a device for determining a sideslip angle of a vehicle's center of mass provided according to an exemplary embodiment. DETAILED DESCRIPTION

[0061] In order to enable those skilled in the art to better understand the technical solution of the present invention and implement it, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the scope of protection of the present invention.

[0062] The center of mass slip angle is the angle between the velocity V at the center of mass of the vehicle and the x-axis. The lateral dynamics control of the vehicle, including the electronic stability program, steer-by-wire control, active front / rear steering technology, etc., all use the center of mass slip angle. Its direct measurement requires expensive special test equipment and a specific installation and fixing method. It is currently not suitable for configuration on mass-produced vehicles and is only suitable for the development and testing stage. Therefore, from the perspective of cost saving and practical application, the center of mass slip angle must be estimated based on the sensor configuration on mass-produced vehicles.

[0063] The digital twin vehicle center of mass sideslip angle estimator designed in the present invention consists of two parts, including a neural network module and an unscented Kalman filter module. The neural network module estimates the pseudo center of mass sideslip angle β through easily acquired sensor signals. PSOLSTM The steering wheel angle δ can be measured by the steering angle sensor, and the vehicle heading angle φ GPS The vehicle lateral acceleration a can be measured by GPS. y , vehicle longitudinal speed V x , vehicle lateral speed V y , vehicle yaw angular velocity ω r It can be obtained by GPS combined with IMU measurement. The state estimator established based on the characteristics of the neural network can solve the problem of strong nonlinear systems. However, the pseudo-measurement value cannot be directly used for vehicle state observation because the measurement result contains noise caused by the sensor. Therefore, the pseudo-measurement value needs to be input into the unscented Kalman filter module to correct the gain value, so as to obtain an accurate estimate of the vehicle's center of mass sideslip angle β.

[0064] The technical solutions provided by various embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0065] First, the present invention provides a vehicle control method under multiple working conditions, specifically as follows: Figure 1 As shown, the following steps are included:

[0066] S101. Obtain historical driving state parameters of a vehicle under different working conditions, and determine training samples according to the historical driving state parameters.

[0067] The historical driving state parameters include the real center of mass sideslip angle and other state parameters corresponding to the real center of mass sideslip angle.

[0068] Among them, the other state parameters include the vehicle's steering wheel angle, heading angle, lateral acceleration, longitudinal speed, lateral speed, and yaw angular velocity, that is, the training sample includes the vehicle's steering wheel angle, heading angle, lateral acceleration, longitudinal speed, lateral speed, and yaw angular velocity, as well as the corresponding true center of mass sideslip angle; the steering wheel angle δ can be measured by an angle sensor, and the vehicle heading angle φ GPS The vehicle lateral acceleration a can be measured by GPS. y , vehicle longitudinal speed V x , vehicle lateral speed V y , vehicle yaw angular velocity ω r It can be measured by GPS and IMU. The center of mass deviation angle needs to be tested with expensive special test equipment.

[0069] For example, in order to make the center of mass slip angle neural network adapt to various vehicle operating conditions, the example vehicle model equipped with a stable control system in the vehicle dynamics simulation software Carsim can be used, and the joint simulation with Matlab / Simulink can be performed. On the road with different adhesion coefficients, the driver conducts 300 groups of experiments at different speeds under the angle step condition, double lane change condition, snake condition, fishhook condition and figure eight condition (three simulation experiments are conducted under the same working condition, vehicle speed, and road adhesion coefficient, and the average value is taken as the data set), and the current vehicle steering wheel angle, vehicle heading angle, vehicle lateral acceleration, vehicle longitudinal speed, vehicle lateral speed, vehicle yaw angular velocity, and center of mass slip angle of the vehicle under different working conditions are obtained. The original sample data obtained is subjected to minimum-maximum normalization processing, retaining the relationship existing in the original data, and each eigenvalue is mapped to [0,1] respectively to achieve the purpose of normalization. The entire sample data set is randomly divided into 60% training set and 40% test set. The training set is used to determine the hyperparameters in the neural network model, and the test set is used to test the generalization performance between the proposed neural network algorithm and the comparison algorithm.

[0070] S102. Construct a center of mass sideslip angle prediction model based on a neural network.

[0071] In this step, the particle swarm algorithm can be used to optimize the hyperparameters of the LSTM neural network, and the PSOLSTM model can be constructed according to the optimized hyperparameters; the PSOLSTM model is trained with the training samples to obtain the center of mass sideslip angle prediction model.

[0072] Specifically, a PSOLSTM neural network model is established to estimate the vehicle center of mass sideslip angle. The neural network model is mainly composed of an input layer, a fully connected layer, an optimization layer, and an output layer. The input layer is responsible for receiving input information. The fully connected layer is the core structure of the entire neural network, and its internal structure is mainly composed of some neural units. The number of nodes in the fully connected layer is the factor that has the greatest impact on the training results of the entire model. The optimization layer is responsible for optimizing the number of neurons in the fully connected layer (ls), the initial learning rate (lr), and the learning rate reduction factor (lrdf). The output layer is responsible for outputting the results and performing denormalization on the obtained data.

[0073] During the training process of the model, the determination of its model hyperparameters is crucial, which directly affects the accuracy of the entire network model. First, the vehicle steering wheel angle, vehicle heading angle, vehicle lateral acceleration, vehicle longitudinal speed, vehicle lateral speed, and vehicle yaw rate measured by the sensor are imported into the input layer of the neural network. After the data is processed by the input layer, the network begins to train. Finally, the corresponding parameters of the model are adjusted according to the estimated results of the model. When the prediction results of the model meet the requirements, the training is completed. The network topology of PSOLSTM is as follows: Figure 2 shown.

[0074] The learning rate is the most important hyperparameter of the LSTM neural network, followed by the network size, while the momentum gradient has little effect on the final result. In order to match the LSTM structure with the data characteristics of the vehicle driving state, a PSOLSTM prediction model was constructed. The hyperparameter optimization was performed with the number of neurons in the fully connected layer (ls), the initial learning rate (lr) and the learning rate reduction factor (lrdf) as the optimization targets.

[0075] For example, the first step is data processing. Different input data has different dimensions, and the difference in values ​​may be large, which will directly lead to a large difference in the training speed of the model. Therefore, the data must be further normalized. The collected data is classified. First, a mapping function is used to map the data to [-1, 1], and the normalized experimental data is divided into two data sets: "training" and "testing".

[0076] The second step is to optimize the hyperparameters. First, the position and speed of the particle swarm are initialized, and the position of each particle in the particle swarm is randomly generated; secondly, the mean square error of the particle swarm is set as the target fitness value, and the individual extreme value (p ij ) and the global extreme value (p gi ); then the particle position (x ij ) and speed (v ij ) is updated in real time, and then the fitness value is obtained according to the latest position of the particle, and then the individual extreme value and the global extreme value are updated; the specific update formula is:

[0077] v ij (t+1)=w·v ij (t)+c1r1[p ij (t)-x ij (t)]+c2r2[p gi (t)-x ij (t)];

[0078] x ij (t+1)=x ij (t)+v ij (t+1);

[0079] Finally, when the optimization results meet the preset conditions, the update is ended and the final optimization results are output to obtain the optimized hyperparameters.

[0080] The third step is to use the parameters obtained by particle swarm optimization to establish the PSOLSTM model and train the network.

[0081] Specifically, the vehicle's steering wheel angle, heading angle, lateral acceleration, longitudinal velocity, lateral velocity and yaw angular velocity are input into the PSOLSTM model to obtain a prediction result; with the goal of minimizing the deviation between the prediction result and the true center of mass sideslip angle, the PSOLSTM model is trained to obtain a center of mass sideslip angle prediction model.

[0082] Among them, the specific process steps of the PSOLSTM neural network model are as follows:

[0083] Step 1: Collect vehicle driving state parameters in real time under different working conditions, determine the search space of the particle swarm according to the constraints, and complete the initialization.

[0084] Step 2: After initialization, the fitness value of each data sample is solved, and after comparing the target fitness value, the difference between the sample and the target value is used to distinguish them into discoverers, explorers, and wanderers. The one with the smallest difference is the discoverer, and the rest are explorers and wanderers.

[0085] Step 3: Update the positions of the three individuals in step 2. If there are too many particles in the new position, the particle returns to the last valid position, and the remaining particles recalculate the target fitness value and reallocate roles.

[0086] Step 4: Check whether the convergence requirements are met. If not, jump to step 3. If yes, go directly to step 5.

[0087] Step 5: Output the globally optimal particle information and obtain the weights and thresholds of the optimal neural network.

[0088] The fourth step is to evaluate the network estimation results after completing the network training using the previously divided test set data. Finally, the root mean square error (RMSE), mean absolute error (MAE) and maximum absolute error (ME) are selected as the performance evaluation indicators. The specific formula is as follows:

[0089]

[0090] ME=max(|SSA′(t)-SSA(t)|);

[0091] Among them, SSA′(t) is the true value of the sideslip angle at time t; SSA(t) is the predicted value of the sideslip angle at time t; T is the total number of samples for the estimated test.

[0092] The optimization process of this model is to use PSO to optimize the key parameters such as the number of neurons, learning rate and maximum number of iterations in the LSTM model, and establish the model based on the obtained optimal parameters as the PSOLSTM neural network parameters.

[0093] S103, predicting the target vehicle according to the center of mass sideslip angle prediction model to obtain a predicted center of mass sideslip angle.

[0094] Specifically, the current driving state parameters of the target vehicle except the center of mass sideslip angle may be first obtained, and the current driving state parameters may be input into the center of mass sideslip angle prediction model for prediction to obtain the predicted center of mass sideslip angle of the target vehicle.

[0095] In this step, the current driving state parameters are input into the center of mass sideslip angle prediction model to obtain the predicted center of mass sideslip angle output by the center of mass sideslip angle prediction model.

[0096] S104: constructing sampling points based on the predicted sideslip angle of the center of mass, and building an unscented Kalman filter module according to the sampling points.

[0097] The unscented Kalman filter overcomes the shortcomings of the traditional Kalman filter and the extended Kalman filter. It no longer performs a large number of samplings, but only selects a limited number of sampling points, and uses the sampling points to approximate the state distribution. It does not need to calculate the Jacobian matrix, and the calculation accuracy is higher.

[0098] State equations and measurement equations are used to describe the nonlinear system model. For vehicle state parameter estimation, a nonlinear discrete system is established:

[0099] x(k+1)=f(x(k))+w(k);

[0100] z(k)=h(x(k))+v(k);

[0101] Among them, x(k) is the system state matrix, z(k) is the observation matrix, w(k), v(k) are the system noise and measurement noise, and they are assumed to be uncorrelated zero-mean white noise.

[0102] For the construction of the unscented Kalman filter module, we must first construct the sampling points, which are constructed as follows:

[0103] The basic principle of sample point selection is to include the three characteristics of normal distribution as much as possible, namely expectation, variance and symmetry. Therefore, the constraint formula of sampling points can be constructed according to the predicted centroid sideslip angle. The constraint formula of sampling points is:

[0104]

[0105] The weights of the adopted points are:

[0106]

[0107] in, is the mean of the system state quantity, that is, the predicted mean of the center of mass sideslip angle, P is the variance of the state quantity, that is, the predicted variance of the center of mass sideslip angle, n is the dimension of the state vector to be estimated, and λ is the scaling factor;

[0108] Secondly, the predicted value, mean and variance of the sampling points are calculated: when i>1, 2n+1 sample points are constructed:

[0109]

[0110] Calculate the mean and variance of the predicted sample points:

[0111]

[0112] Use an unscented transformation to handle the nonlinear transfer of the mean and covariance to ensure that new predictions are obtained:

[0113]

[0114] Update the mean and variance of the predicted values:

[0115]

[0116] Calculate the Kalman gain of the unscented Kalman filter:

[0117]

[0118] Update state mean and variance:

[0119]

[0120] P x (k) = Px (k)-KP zz (k-1)K T .

[0121] The state parameter estimation system optimizes the state parameter estimation value through each iterative update, thereby reducing the system error.

[0122] S105, filtering the predicted center-of-mass sideslip angle through the unscented Kalman filter module to determine the center-of-mass sideslip angle of the target vehicle.

[0123] Using the above method, the particle swarm algorithm is first used to optimize the hyperparameters of the LSTM neural network, and the PSOLSTM model is constructed according to the optimized hyperparameters, and then the PSOLSTM model is trained to obtain the center of mass sideslip angle prediction model; in this way, by optimizing the hyperparameters of the LSTM model, PSOLSTM can improve the prediction accuracy and generalization ability, and provide strong support for the prediction and application of vehicle form state parameters; in addition, after obtaining the predicted center of mass sideslip angle through the center of mass sideslip angle prediction model, an unscented Kalman filter module is constructed in combination with the current driving state parameters of the target vehicle to predict the predicted center of mass sideslip angle, which further improves the evaluation accuracy of the center of mass sideslip angle, is conducive to reducing errors and improving the safety and stability of vehicle control.

[0124] In addition, in order to verify the effectiveness of the method for determining the sideslip angle of the center of mass of a vehicle proposed in the present invention, a Carsim / Simulink joint simulation platform is built, and two working modes, open-loop simulation and closed-loop experiment, are set. In the open-loop simulation, the simulations of steering wheel angle step input conditions, serpentine conditions, double lane change conditions and sinusoidal conditions are completed respectively, and the estimation results of the sideslip angle estimator of the center of mass in the digital twin virtual vehicle model system are simulated and analyzed.

[0125] The transient response and steady-state response test of the steering wheel angle step input is an indispensable experimental project in the vehicle handling experiment. The steering wheel angle step simulation experiment was carried out in accordance with the provisions of the national standard "Automobile Handling Stability Test Method GB / T6323-2014". The vehicle speed during the simulation was 70% of the maximum speed of the tested vehicle of 120km / h and rounded to an integer multiple of 10, 80km / h. The road adhesion coefficients were 0.2 and 0.8 respectively. The simulation results are shown in the figure below. Figure 3 shown. Figure 3The estimation results of the PSOLSTM+UKF estimator model designed in this paper and the comparison of the estimation results of the UKF estimator model with the actual center of mass sideslip angle of the Cars im vehicle under the steering wheel angle step condition. When the road adhesion coefficient is 0.2, the average estimation error of the designed PSOLSTM+UKF estimator model is 0.062deg, and the average estimation error of the UKF estimator model is 0.125deg. When the road adhesion coefficient is 0.8, the average estimation error of the designed PSOLSTM+UKF estimator model is 0.181deg, and the average estimation error of the UKF estimator model is 0.264deg. The results show that the estimation accuracy of the PSOLSTM+UKF center of mass sideslip angle estimator model designed in this invention has been effectively improved, verifying the effectiveness and accuracy of the designed center of mass sideslip angle estimator model.

[0126] Secondly, the present invention also provides a device for determining the sideslip angle of the center of mass of a vehicle, such as Figure 4 As shown, including:

[0127] The acquisition module 401 is used to acquire historical driving state parameters of the vehicle under different working conditions and determine training samples according to the historical driving state parameters; the historical driving state parameters include the true center of mass sideslip angle and other state parameters corresponding to the true center of mass sideslip angle.

[0128] The construction module 402 is used to optimize the hyperparameters of the LSTM neural network using the particle swarm algorithm, and to construct the PSOLSTM model according to the optimized hyperparameters; the PSOLSTM model is trained using the training samples to obtain a center of mass sideslip angle prediction model.

[0129] The acquisition module 401 is also used to acquire the current driving state parameters of the target vehicle except the sideslip angle of the center of mass.

[0130] The prediction module 403 is used to input the current driving state parameters into the center of mass sideslip angle prediction model for prediction, so as to obtain the predicted center of mass sideslip angle of the target vehicle.

[0131] The construction module 402 is also used to construct sampling points based on the predicted center of mass sideslip angle, and to construct an unscented Kalman filter module according to the sampling points.

[0132] The filtering module 404 is used to filter the predicted center of mass sideslip angle through the unscented Kalman filter module to determine the center of mass sideslip angle of the target vehicle.

[0133] Using the above device, the particle swarm algorithm is first used to optimize the hyperparameters of the LSTM neural network, and the PSOLSTM model is constructed according to the optimized hyperparameters, and then the PSOLSTM model is trained to obtain the center of mass sideslip angle prediction model; by optimizing the hyperparameters of the LSTM model, PSOLSTM can improve the prediction accuracy and generalization ability, and provide strong support for the prediction and application of vehicle form state parameters; in addition, after the predicted center of mass sideslip angle is obtained through the center of mass sideslip angle prediction model, an unscented Kalman filter module is constructed in combination with the current driving state parameters of the target vehicle to predict the predicted center of mass sideslip angle, so as to further improve the evaluation accuracy of the center of mass sideslip angle, which is conducive to reducing errors and improving the safety and stability of vehicle control.

[0134] The present invention also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the above Figure 1 The steps of a vehicle control method under multiple operating conditions are provided.

[0135] The present invention also provides a computer device. At the hardware level, the computer device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory. Of course, it may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 The steps of a vehicle control method under multiple operating conditions are provided.

[0136] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0137] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as a combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0138] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0139] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0140] It should be noted that the above specific implementation method can enable those skilled in the art to understand the invention more comprehensively, but does not limit the invention in any way. Therefore, although the present invention has been described in detail in this specification, those skilled in the art should understand that the invention can still be modified or replaced by equivalents; and all technical solutions and improvements that do not deviate from the spirit and scope of the invention are included in the protection scope of the patent for the invention. Any figure mark in the claims should not be regarded as limiting the claims involved.

Claims

1. A method for determining the sideslip angle of the center of mass of a vehicle, characterized in that: The method comprises: Acquire historical driving state parameters of the vehicle under different working conditions, and determine training samples according to the historical driving state parameters; the historical driving state parameters include a true center of mass sideslip angle and other state parameters corresponding to the true center of mass sideslip angle; The particle swarm algorithm is used to optimize the hyperparameters of the LSTM neural network, and the PSOLSTM model is constructed according to the optimized hyperparameters; the PSOLSTM model is trained by the training samples to obtain a center of mass sideslip angle prediction model; Obtaining current driving state parameters of the target vehicle except for the center of mass sideslip angle, inputting the current driving state parameters into the center of mass sideslip angle prediction model for prediction, and obtaining a predicted center of mass sideslip angle of the target vehicle; Constructing sampling points based on the predicted centroid sideslip angle, and constructing an unscented Kalman filter module according to the sampling points; The predicted center-of-mass sideslip angle is filtered by the unscented Kalman filter module to determine the center-of-mass sideslip angle of the target vehicle.

2. The method for determining the vehicle center of mass sideslip angle according to claim 1, characterized in that: The use of particle swarm algorithm to optimize the hyperparameters of the LSTM neural network includes: Initialize the position and velocity of the particle swarm, and randomly generate the position of each particle in the particle swarm; The mean square error of the particle swarm is set as the target fitness value, and the individual extreme value p of each particle is obtained according to the fitness value of each particle. ij and the global extreme value p gi ; The position x of the particle ij and speed v ij Real-time updates are performed, and then the fitness value is obtained according to the latest position of the particle, and then the individual extreme value and the global extreme value are updated; the specific update formula of speed and position is: v ij (t+1)=w·v ij (t)+c1r1[p ij (t)-x ij (t)]+c2r2[p gi (t)-x ij (t)]; x ij (t+1)=x ij (t)+v ij (t+1); Among them, v ij (t+1) is the updated value of the velocity of the jth particle in the i-th dimension at time t+1, w is the inertia weight, c1 is the individual learning factor, r1 is a random number to prevent the algorithm from falling into the local optimum, [p ij (t)-x ij (t)] is the distance and direction between the current position of the jth particle in the i-th dimension and its historical optimal position, c2 is the group learning factor, r2 is a random number, [pgi(t)-xij(t)] represents the distance and direction between the current position of the jth particle in the i-th dimension and the historical optimal position of the group, p gi (t) is the optimal position found by the group at time t; x ij (t) is the updated value of the position of the jth particle in the i-th dimension at time t; When the optimization result meets the preset conditions, the update is ended and the final optimization result is output to obtain the optimized hyperparameters.

3. The method for determining the vehicle center of mass sideslip angle according to claim 1, characterized in that: The other state parameters include the vehicle's steering wheel angle, heading angle, lateral acceleration, longitudinal velocity, lateral velocity and yaw rate; The PSOLSTM model is trained by the training samples to obtain the center of mass sideslip angle prediction model including: The steering wheel angle, heading angle, lateral acceleration, longitudinal velocity, lateral velocity and yaw rate of the vehicle are input into the PSOLSTM model to obtain a prediction result of the center of mass sideslip angle; With the goal of minimizing the deviation between the predicted result and the true center of mass sideslip angle, the PSOLSTM model is trained to obtain a center of mass sideslip angle prediction model.

4. The method for determining the sideslip angle of the center of mass of a vehicle according to claim 1, characterized in that: The constructing of sampling points based on the predicted centroid sideslip angle comprises: The constraint formula of the sampling point is constructed according to the predicted sideslip angle of the center of mass. The constraint formula of the sampling point is: The weights of the sampling points are: in, is the mean value of the state quantity of the system, that is, the mean value of the predicted center of mass sideslip angle, P is the variance of the state quantity, that is, the predicted center of mass sideslip angle variance, n is the dimension of the state vector to be estimated, and λ is the scaling factor; When i>1, construct 2n+1 sample points: in, is the predicted value of the sample point, X i is the prior estimate at time i, and k is the gain at time i.

5. The method for determining the vehicle center of mass sideslip angle according to claim 4, characterized in that: Constructing an unscented Kalman filter module according to the sampling points includes: Calculate the predicted value, mean and variance of the sample points: Calculate the mean and variance of the predicted sample points: Where Q is the process noise covariance matrix; Use an unscented transformation to handle the nonlinear transfer of means and covariances to determine new forecasts: Update the mean and variance of the predicted values: Calculate the Kalman gain of the unscented Kalman filter: Update state mean and variance: P x( k ) =P x( k ) -KP zz( k-1 ) K T ; Among them, z(k) is the observed variable, K is the Kalman gain, is the state estimation, P x( k ) Estimate for the co-party.

6. A device for determining the sideslip angle of the center of mass of a vehicle, characterized in that: The device comprises: An acquisition module, used for acquiring historical driving state parameters of the vehicle under different working conditions, and determining training samples according to the historical driving state parameters; the historical driving state parameters include a true center of mass sideslip angle and other state parameters corresponding to the true center of mass sideslip angle; A construction module is used to optimize the hyperparameters of the LSTM neural network using a particle swarm algorithm, and to construct a PSOLSTM model according to the optimized hyperparameters; the PSOLSTM model is trained using the training samples to obtain a center of mass sideslip angle prediction model; The acquisition module is also used to acquire the current driving state parameters of the target vehicle except the sideslip angle of the center of mass; A prediction module, used for inputting the current driving state parameters into the center of mass sideslip angle prediction model for prediction, and obtaining a predicted center of mass sideslip angle of the target vehicle; The construction module is also used to construct sampling points based on the predicted centroid sideslip angle, and to construct an unscented Kalman filter module according to the sampling points; The filtering module is used to filter the predicted center of mass sideslip angle through the unscented Kalman filter module to determine the center of mass sideslip angle of the target vehicle.

7. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

8. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 5 when executing the program.

Images