Commercial vehicle mass estimation method, device and equipment and storage medium

Abstract

The invention discloses a commercial vehicle quality estimation method, device and equipment and a storage medium, and relates to the technical field of automobiles, and the method comprises the steps: obtaining a real-time training data set, an initial quality estimation model and a longitudinal dynamics model; estimating the quality of the real-time training data set according to the initial quality estimation model to obtain a quality estimation value; obtaining an acceleration estimation value through the longitudinal dynamic model and the mass estimation value; obtaining a total loss function according to the mass estimation value and the acceleration estimation value; optimizing the initial quality estimation model according to the total loss function to obtain a commercial vehicle quality estimation model, and estimating the quality of the commercial vehicle according to the commercial vehicle quality estimation model; according to the method, a total loss function is constructed through a mass estimation value of an initial mass estimation model and an acceleration estimation value of a longitudinal dynamical model, a physical law constraint neural network is utilized to learn and output a reasonable result, finally, the model is optimized through the total loss function, and a more accurate and reliable commercial vehicle mass estimation model is obtained.

Description

Technical Field

[0001] The present application relates to the field of automotive technology, and in particular to a commercial vehicle mass estimation method, device, equipment and storage medium. Background Art

[0002] Commercial vehicles occupy an important position in modern logistics and transportation. The accuracy of their mass estimation is of vital importance to ensuring road safety, optimizing vehicle performance, improving transportation efficiency and reducing operating costs. Accurate mass estimation helps to reasonably plan driving routes and ensure that vehicles operate within a safe load range, thereby reducing the risk of traffic accidents and extending the service life of roads. In the context of the continuous development of intelligent transportation systems, commercial vehicle mass estimation is the key foundation for realizing functions such as precise scheduling and intelligent driving assistance.

[0003] The commercial vehicle mass estimation task aims to obtain the actual mass of the vehicle in real time and accurately. However, since the mass of a commercial vehicle will change due to loading conditions during driving and is affected by a variety of complex working conditions and environmental factors, mass estimation becomes a challenging task.

[0004] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of this application is to provide a commercial vehicle mass estimation method, device, equipment and storage medium, aiming to solve the technical problem that the current model for commercial vehicle mass estimation is not accurate enough in estimating the mass of the vehicle during driving.

[0006] To achieve the above objectives, the present application proposes a commercial vehicle mass estimation method, which includes:

[0007] Obtain real-time training data sets, initial mass estimation models, and longitudinal dynamics models;

[0008] Performing quality estimation based on the real-time training data set according to the initial quality estimation model to obtain a quality estimation value;

[0009] Obtaining an acceleration estimate using the longitudinal dynamics model and the mass estimate;

[0010] Obtaining a total loss function according to the mass estimation value and the acceleration estimation value;

[0011] The initial mass estimation model is optimized according to the total loss function to obtain a commercial vehicle mass estimation model, and the commercial vehicle mass estimation is performed according to the commercial vehicle mass estimation model.

[0012] In one embodiment, the step of obtaining the longitudinal dynamics model includes:

[0013] A vehicle driving force function is constructed according to the engine torque, transmission ratio, final reducer ratio, transmission efficiency, master cylinder brake pressure, front brake efficiency factor, rear brake efficiency factor and wheel rolling radius of the target vehicle;

[0014] Construct an air resistance function based on the air resistance coefficient, air density, the frontal area of ​​the target vehicle, and the driving speed;

[0015] Constructing a rolling resistance function according to the rolling resistance coefficient, the preset vehicle speed influencing parameter and the road slope angle;

[0016] A longitudinal dynamics model is constructed according to the vehicle driving force function, the air resistance function, the rolling resistance function, the slope resistance and the acceleration resistance.

[0017] In one embodiment, the step of obtaining a real-time training data set includes:

[0018] Acquiring initial real-time data, the initial real-time data including one-to-one correspondence of resistance, acceleration, longitudinal vehicle speed, engine torque, and brake master cylinder pressure;

[0019] Obtaining a reference estimated mass based on the resistance, the acceleration, the longitudinal vehicle speed, the engine torque, and the master cylinder pressure based on a recursive least square method;

[0020] Obtaining a mass variance according to the resistance, the acceleration, the longitudinal vehicle speed, the engine torque, the brake master cylinder pressure, and the reference estimated mass through a preset neural network;

[0021] Obtaining a quality estimate confidence level according to the quality variance, and obtaining a quality label based on the quality estimate confidence level;

[0022] A real-time training data set is obtained according to the quality label and the initial real-time data.

[0023] In one embodiment, obtaining a quality estimate confidence according to the quality variance, and obtaining a quality label based on the confidence, further includes:

[0024] When the confidence of the quality estimation value is greater than or equal to a preset confidence threshold, using the reference estimated quality as a quality label;

[0025] When the confidence of the quality estimation value is less than a preset confidence threshold, the quality estimation value at the last moment is used as the quality label.

[0026] In one embodiment, obtaining a real-time training data set according to the quality label and the initial real-time data includes:

[0027] Obtaining a combined training data set according to the quality label and the initial real-time data;

[0028] Dividing the combined training data set by a sliding window to obtain a data set to be cleaned;

[0029] The data set to be cleaned is preprocessed to obtain a real-time training data set, wherein the preprocessing includes outlier removal and data normalization.

[0030] In one embodiment, obtaining a total loss function according to the mass estimation value and the acceleration estimation value includes:

[0031] Obtaining a mass loss term according to the mass label and the mass estimation value, and obtaining a physical loss term according to the acceleration label and the acceleration estimation value;

[0032] Obtaining a physical loss function and a quality loss function according to a preset loss function, the quality loss term and the physical loss term;

[0033] A total loss function is obtained according to the physical loss function and the quality loss function.

[0034] In one embodiment, obtaining a total loss function according to the physical loss function and the quality loss function includes:

[0035] Get model training time and loss hyperparameters;

[0036] Constructing a time-varying weight function according to the model training duration and the loss hyperparameter;

[0037] Obtaining physical weight and quality weight according to the time-varying weight function;

[0038] A total loss function is constructed according to the physical weight, the quality weight, the physical loss function and the quality loss function.

[0039] In addition, to achieve the above purpose, the present application also proposes a commercial vehicle mass estimation device, the commercial vehicle mass estimation device comprising:

[0040] A data acquisition module, used to acquire real-time training data sets, initial mass estimation models, and longitudinal dynamics models;

[0041] An estimation module, configured to perform quality estimation based on the real-time training data set according to the initial quality estimation model to obtain a quality estimation value;

[0042] The estimation module is further used to obtain an acceleration estimation value through the longitudinal dynamics model and the mass estimation value;

[0043] A loss function construction module, used for obtaining a total loss function according to the mass estimation value and the acceleration estimation value;

[0044] The estimation module is also used to optimize the initial mass estimation model according to the total loss function to obtain a commercial vehicle mass estimation model, and perform commercial vehicle mass estimation according to the commercial vehicle mass estimation model.

[0045] In addition, to achieve the above-mentioned purpose, the present application also proposes a commercial vehicle mass estimation device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the commercial vehicle mass estimation method as described above.

[0046] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the commercial vehicle mass estimation method as described above are implemented.

[0047] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the commercial vehicle mass estimation method as described above are implemented.

[0048] One or more technical solutions proposed in this application have at least the following technical effects:

[0049] The total loss function is constructed by the mass estimation value of the initial mass estimation model and the acceleration estimation value of the longitudinal dynamics model. The physical laws are used to constrain the neural network to learn and output reasonable results. Finally, the model is optimized by combining the physical laws and the neural network to obtain a more accurate model for mass estimation, thereby achieving more accurate and reliable commercial vehicle mass estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0051] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0052] Figure 1 A flow chart of the first embodiment of the commercial vehicle mass estimation method of the present application;

[0053] Figure 2 A schematic diagram of a commercial vehicle longitudinal force analysis provided in Example 1 of the commercial vehicle mass estimation method of the present application;

[0054] Figure 3 A schematic diagram of a complete mass estimation process provided for the first embodiment of the commercial vehicle mass estimation method of the present application;

[0055] Figure 4 A schematic diagram of the total loss of the physical information neural network model provided in Example 1 of the commercial vehicle mass estimation method of the present application;

[0056] Figure 5 A flow chart of the second embodiment of the commercial vehicle mass estimation method provided in this application;

[0057] Figure 6 This is a schematic diagram of the module structure of the commercial vehicle mass estimation device according to an embodiment of the present application;

[0058] Figure 7 Schematic diagram of the device structure of the hardware operating environment involved in the commercial vehicle mass estimation method in the embodiment of the present application. DETAILED DESCRIPTION

[0059] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0060] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0061] The main solution of the embodiment of the present application is: obtaining a real-time training data set, an initial mass estimation model and a longitudinal dynamics model; performing mass estimation based on the real-time training data set according to the initial mass estimation model to obtain a mass estimation value; obtaining an acceleration estimation value through the longitudinal dynamics model and the mass estimation value; obtaining a total loss function according to the mass estimation value and the acceleration estimation value; optimizing the initial mass estimation model according to the total loss function to obtain a commercial vehicle mass estimation model, and performing commercial vehicle mass estimation according to the commercial vehicle mass estimation model.

[0062] In this embodiment, for ease of description, the following description is made by taking the identification of the commercial vehicle mass estimation device as the execution subject.

[0063] As commercial vehicles occupy an important position in modern logistics and transportation, the accuracy of their mass estimation is of vital importance to ensuring road safety, optimizing vehicle performance, improving transportation efficiency and reducing operating costs. Accurate mass estimation helps to reasonably plan driving routes and ensure that vehicles operate within a safe load range, thereby reducing the risk of traffic accidents and extending the service life of roads. In the context of the continuous development of intelligent transportation systems, commercial vehicle mass estimation is the key foundation for realizing functions such as precise scheduling and intelligent driving assistance.

[0064] The commercial vehicle mass estimation task aims to obtain the actual mass of the vehicle in real time and accurately. However, since the mass of a commercial vehicle will change due to loading conditions during driving and is affected by a variety of complex working conditions and environmental factors, mass estimation becomes a challenging task.

[0065] The present application provides a solution, which constructs a total loss function through the mass estimation value of the initial mass estimation model and the acceleration estimation value of the longitudinal dynamics model, uses the physical laws to constrain the neural network learning to output reasonable results, and finally optimizes the model through the total loss function to obtain a more accurate and reliable commercial vehicle mass estimation model.

[0066] It can be seen from the above embodiments that the present application discloses a commercial vehicle mass estimation method, device, equipment and storage medium, which relates to the field of automobile technology, and discloses: obtaining a real-time training data set, an initial mass estimation model and a longitudinal dynamics model; estimating the quality of the real-time training data set according to the initial mass estimation model to obtain a mass estimation value; obtaining an acceleration estimation value through the longitudinal dynamics model and the mass estimation value; obtaining a total loss function according to the mass estimation value and the acceleration estimation value; optimizing the initial mass estimation model according to the total loss function to obtain a commercial vehicle mass estimation model, and estimating the mass of the commercial vehicle according to the commercial vehicle mass estimation model; the method constructs a total loss function through the mass estimation value of the initial mass estimation model and the acceleration estimation value of the longitudinal dynamics model, uses physical laws to constrain the neural network learning to output reasonable results, and finally optimizes the model through the total loss function to obtain a more accurate and reliable commercial vehicle mass estimation model.

[0067] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a commercial vehicle mass estimation device, etc. The following takes the commercial vehicle mass estimation device as an example to illustrate this embodiment and the following embodiments.

[0068] Based on this, the present application embodiment provides a commercial vehicle mass estimation method, referring to Figure 1 , Figure 1This is a flow chart of the first embodiment of the commercial vehicle mass estimation method of the present application.

[0069] In this embodiment, the commercial vehicle mass estimation method includes steps S10 to S50:

[0070] Step S10, obtaining a real-time training data set, an initial mass estimation model and a longitudinal dynamics model.

[0071] In a feasible implementation, step S10 may include steps A11 to A13:

[0072] Step A11, obtaining a real-time training data set.

[0073] It should be noted that the real-time training data set includes multiple sets of longitudinal vehicle speed, acceleration, engine torque, brake master cylinder pressure, transmission gear, and brake signal; each set of data has a corresponding quality label.

[0074] It should be noted that the quality label can be an estimated quality obtained by estimating each group of longitudinal vehicle speed, acceleration, engine torque, brake master cylinder pressure, transmission gear, and brake signal based on the recursive least squares method, Kalman filtering, and one of its variants, and the quality label can be obtained based on the estimated quality.

[0075] It is understood that the quality label is a quality value.

[0076] Step A12, obtaining an initial quality estimation model.

[0077] It should be noted that the initial quality estimation model can be a special recursive neural network, such as a bidirectional long short-term memory network (Bi-LSTM for short), which can process and predict long-term dependencies in time series data.

[0078] It is understandable that the input data of the initial mass estimation model mainly include various relevant parameters during the operation of the commercial vehicle, including: longitudinal vehicle speed, acceleration, engine torque, brake master cylinder pressure, transmission gear, and brake signal.

[0079] It should be further explained that the longitudinal vehicle speed and acceleration can reflect the basic driving state of the commercial vehicle, the engine torque reflects the driving force of the commercial vehicle, the brake master cylinder pressure and the brake signal reflect the braking process and its possible nonlinear effects, and the transmission gear position reflects the nonlinear effects brought about by the change of driving force during the gear shifting process.

[0080] It should be noted that the output data is the estimated mass of the commercial vehicle. By extracting and processing the features of the input commercial vehicle operation data, the mass of the commercial vehicle in a certain period of time in the future is predicted.

[0081] Step A13, obtaining the longitudinal dynamics model.

[0082] It is understandable that the construction of the commercial vehicle longitudinal dynamics model can be based on the longitudinal force of the commercial vehicle. For the schematic diagram of the commercial vehicle longitudinal force analysis, please refer to Figure 2 , α in the figure represents the slope angle of the road on which the commercial vehicle is traveling, F t Indicates the vehicle driving force, F w is the air resistance, F f is the rolling resistance, F i is the slope resistance, F j For acceleration resistance, the longitudinal force can be expressed as follows:

[0083] F t =F w +F f +F i +F j

[0084] It should be noted that a vehicle driving force function is constructed based on the target vehicle's engine torque, transmission ratio, main reducer ratio, transmission efficiency, master cylinder braking pressure, front brake efficiency factor, rear brake efficiency factor and wheel rolling radius; an air resistance function is constructed based on the air resistance coefficient, air density, frontal area of ​​the target vehicle and driving speed; a rolling resistance function is constructed based on the rolling resistance coefficient, preset vehicle speed influencing parameters and road slope angle; a longitudinal dynamic model is constructed based on the vehicle driving force function, the air resistance function, the rolling resistance function, slope resistance and acceleration resistance.

[0085] It should be noted that the vehicle driving force function can be constructed according to the target vehicle's engine torque, transmission ratio, final reducer ratio, transmission efficiency, master cylinder brake pressure, front brake efficiency factor, rear brake efficiency factor, and wheel rolling radius by referring to the following formula:

[0086]

[0087] Among them, T tq is the engine torque, i g is the transmission ratio, i o is the main reducer transmission ratio, η t is the transmission efficiency, P mc is the master cylinder brake pressure, k bf is the front brake efficiency factor, k br is the rear brake efficiency factor, and r is the wheel rolling radius.

[0088] It should be noted that the air resistance function can be constructed based on the air resistance coefficient, air density, windward area of ​​the target vehicle and driving speed by referring to the following formula:

[0089]

[0090] Among them, C D is the air resistance coefficient, ρ is the air density, A is the frontal area of ​​the target vehicle, and v is the speed of the commercial vehicle.

[0091] It should be noted that the rolling resistance function can be constructed according to the rolling resistance coefficient, the preset vehicle speed influence parameter and the road slope angle by referring to the following formula:

[0092] F f = mgf cosα

[0093] Where f represents the rolling resistance coefficient and α represents the road slope angle.

[0094] It should be emphasized that the rolling resistance coefficient of commercial vehicles is not a fixed value during driving, but changes with various factors such as vehicle speed, tire load, road conditions, etc. Nonlinearization is to establish a model that can more accurately reflect this changing relationship, rather than simply using a fixed linear relationship to describe the rolling resistance coefficient. For this purpose, the rolling resistance coefficient is nonlinearized, and the expression is as follows:

[0095] f=a+bv+cv 2

[0096] Among them, a represents the basic rolling resistance coefficient, and b and c reflect the influence of vehicle speed on the rolling resistance coefficient. According to the empirical formula, a can be taken as 0.0076 and b can be taken as 0.000056; however, commercial vehicles also have high-speed driving conditions. In order to fully model the nonlinearity of the rolling resistance coefficient of commercial vehicles at high speeds (80-120km / h), c can be simulated by a chassis dynamometer to simulate the conditions of commercial vehicles at high speeds. After obtaining several sets of "vehicle speed-rolling resistance coefficient" data, the fitting method is used to obtain the parameter c corresponding to each vehicle speed, and the c value can be obtained based on actual calculations.

[0097] Further, the slope resistance and acceleration resistance in the longitudinal dynamics model constructed according to the vehicle driving force function, the air resistance function, the rolling resistance function, the slope resistance and the acceleration resistance can refer to the following formula:

[0098] F i = mg sinα

[0099]

[0100] Among them, Fi Indicates slope resistance, F j represents the acceleration resistance, and δ represents the rotational mass conversion factor, which is usually 1.

[0101] It should be noted that based on the above vehicle driving force function, the air resistance function, the rolling resistance function, the slope resistance and the acceleration resistance, it can be known that the slope information is unknown information. In order to construct the longitudinal dynamics formula of mass and acceleration, the map data is introduced. According to the GNSS positioning of the commercial vehicle, the slope information of the current driving road is provided by the map data, and the road slope angle is a known value; the commercial vehicle longitudinal dynamics model can refer to the following formula:

[0102]

[0103] In this embodiment, a data-mechanism hybrid method is used to construct high-quality commercial vehicle driving data samples, which provides reliable input for commercial vehicle mass estimation based on physical information neural network, makes the rolling resistance in the longitudinal dynamics model of the commercial vehicle nonlinear, and more accurately characterizes the motion state of the commercial vehicle.

[0104] The above are only feasible implementations of step S10 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S10.

[0105] Step S20: performing quality estimation based on the real-time training data set according to the initial quality estimation model to obtain a quality estimation value.

[0106] It should be noted that the mass estimation based on the real-time training data set may be inputting the longitudinal vehicle speed, acceleration, engine torque, brake master cylinder pressure, transmission gear, and brake signal in the real-time training data set into an initial mass estimation model.

[0107] Furthermore, the network structure of the initial quality estimation model has 2 hidden layers, the number of nodes in each layer is set to 16, the number of neurons in the fully connected layer is set to 8, and the number of nodes in the output layer is 1, corresponding to the quality of the output.

[0108] The first layer of LSTM is used to preliminarily extract the time series features in the vehicle operation data, and the second layer of LSTM further abstracts and processes these features to obtain a more advanced feature representation. The bidirectional LSTM consists of two LSTM nodes. In the commercial vehicle mass estimation, one node inputs the positive sequence of vehicle operation data (including longitudinal vehicle speed, acceleration and other features), and the other node inputs the corresponding reverse sequence. The nodes with positive input get the sequence output, state and hidden state, and the nodes with reverse input get the corresponding sequence output, state and hidden state.

[0109] The output sequence of the bidirectional LSTM is obtained by combining the outputs of the forward and reverse order nodes. At the same time, the bidirectional LSTM can obtain two hidden state outputs. The network splices the hidden states of each LSTM node in each LSTM hidden layer to obtain high-dimensional features that characterize the operation of commercial vehicles. The above features are used as the input of the fully connected layer to estimate the quality of commercial vehicles.

[0110] Step S30: obtaining an estimated acceleration value by using the longitudinal dynamics model and the mass estimation value.

[0111] It can be understood that the longitudinal dynamics model has been obtained as follows:

[0112]

[0113] Among them, dv / dt can be directly understood as the estimated value of acceleration to be calculated, and m represents the estimated value of mass; the other parameters on the right side of the equation are all known through the acquired real-time training data set.

[0114] In a specific implementation, the acceleration estimate can be calculated by inputting the longitudinal vehicle speed, acceleration, engine torque, brake master cylinder pressure, transmission gear, brake signal and mass estimate in the acquired real-time training data set into the longitudinal dynamics model.

[0115] Step S40: Obtain a total loss function according to the mass estimation value and the acceleration estimation value.

[0116] It should be noted that the total loss function can include a quality loss function and a physical loss function.

[0117] It should be noted that the mass estimation value is calculated based on the longitudinal vehicle speed, acceleration, engine torque, brake master cylinder pressure, transmission gear, and brake signal corresponding to the mass labels in the real-time training data set.

[0118] Furthermore, a quality loss function is constructed based on the quality estimation values ​​and the corresponding quality labels in the real-time training dataset.

[0119] It should be noted that the physical loss function can be constructed by an acceleration label and an acceleration estimate, and the acceleration estimate can be calculated by substituting the mass estimate into the longitudinal dynamics model of the commercial vehicle.

[0120] In a feasible implementation, step S40 may include steps A41 to A43:

[0121] Step A41, obtaining a mass loss term according to the mass tag and the mass estimation value, and obtaining a physical loss term according to the acceleration tag and the acceleration estimation value.

[0122] It should be noted that the mass loss term can be the difference between the mass label and the mass estimate; the physical loss term can be the difference between the acceleration estimate and the acceleration label.

[0123] Step A42, obtaining a physical loss function and a quality loss function according to a preset loss function, the quality loss term and the physical loss term.

[0124] It should be noted that in order to train a LSTM network with better results, it is very important to choose a suitable loss function. In the acceleration estimation component of the commercial vehicle, the mean square error is used as the preset loss function.

[0125] It is understandable that the physical loss function and quality loss function obtained based on the preset loss function can refer to the following formula:

[0126]

[0127] Among them, MSE m represents the quality loss function, MSE a represents the physical loss function, n represents the total number of training data, m i Indicates quality label, a i represents the acceleration label, represents the quality estimate, Represents the estimated value of acceleration.

[0128] Step A43, obtaining a total loss function according to the physical loss function and the quality loss function.

[0129] It should be noted that the total loss function obtained according to the physical loss function and the quality loss function can refer to the following formula:

[0130] L total =α(t)L LSTM +(1-α(t))L phy

[0131] Among them, L LSTM represents the quality loss function, L phy represents the acceleration loss function, α(t) is a time-varying weight, represents the weight of the quality loss function, and (1-α(t)) represents the weight of the acceleration loss function.

[0132] In a specific implementation, the model training time and loss hyperparameters are obtained; a time-varying weight function is constructed according to the model training time and the loss hyperparameters; physical weight and quality weight are obtained according to the time-varying weight function; a total loss function is constructed according to the physical weight, the quality weight, the physical loss function and the quality loss function.

[0133] The model training duration may be the duration for training the LSTM model using data in a real-time training data set, and the loss hyperparameter may be a constant.

[0134] It should be noted that the time-varying weight function constructed according to the model training time and the loss hyperparameters can refer to the following formula:

[0135]

[0136] Among them, e represents a constant, β represents the loss hyperparameter, which is generally taken as 0.5.

[0137] It should be noted that when setting the weight of the physical loss term, over-reliance on the data-driven part may lead to underfitting of the model. At this time, appropriately increasing the weight of the physical loss term and using the prior knowledge of the physical model can improve the performance of the model. When the sample data is abundant and accurate, the weight can be adjusted appropriately to give the data-driven part more opportunities to learn the detailed information in the data and further optimize the quality estimation results. Therefore, in order to better guide the model to learn accurate quality estimates, it is necessary to set reasonable weights for each loss of the physical information neural network.

[0138] It should be emphasized that the weight setting method is adaptive and dynamic. At the beginning of training, since the parameters of the Bi-LSTM network have not yet fully learned effective information, a larger weight is given to the physical loss term (such as 0.7 for the physical loss term and 0.3 for the LSTM quality estimation component term) to allow the network to perform preliminary training under the strong constraints of physical laws, avoiding model divergence or falling into unreasonable parameter space due to the instability of the data-driven part. As the training progresses to the middle stage, when the quality estimation performance index of the Bi-LSTM network on the validation set begins to decline, it indicates that it gradually learns effective information in the data. At this time, the weight of the physical loss term is gradually reduced, and the weight of the quality estimation part of the Bi-LSTM network is increased. For example, according to certain attenuation rules (such as reduction and increase after each training batch), adjustments are made to make the data-driven part play a greater role in the model and further optimize the quality estimation results. In the later stage of model training, when the performance indicators of the model on the validation set tend to be stable, small-scale fine-tuning is performed based on the current weight combination and performance (such as adjusting the step size around the current weight value) to find a better weight balance and ensure that the model achieves the best mass estimation accuracy while taking into account the physical laws. At the same time, the boundary conditions for weight adjustment are set to prevent the weight value from exceeding the reasonable range, ensuring the stability and effectiveness of model training.

[0139] It should be further explained that in longitudinal dynamics, acceleration is the direct response of the vehicle under the action of various forces. The acceleration predicted by the constrained neural network is consistent with the acceleration calculated according to the physical equations, which is actually indirectly constraining the network's understanding and estimation of the vehicle mass and other forces. When the network can accurately predict acceleration, it means that it has correctly learned the relationship between the vehicle mass and various forces, so that it can get an accurate mass estimate through this learning process.

[0140] In the specific implementation, the complete quality estimation process can be divided into two parts, which can be referred to in detail. Figure 3 , Figure 3 It includes data-mechanism mixed sample acquisition and physical information neural network construction; wherein, data-mechanism mixed sample acquisition can include introducing map data to provide slope information and establishing a longitudinal nonlinear dynamic model of a commercial vehicle, modeling the nonlinear factors affecting mass estimation based on a Bayesian neural network and obtaining confidence, and judging whether to correct the model-based mass estimation samples under the current working conditions according to the confidence; wherein, the physical information neural network construction includes dividing the data set and vehicle data preprocessing based on the sliding window method, constructing commercial vehicle mass estimation components based on a Bi-LSTM network, introducing physical loss items based on a nonlinear dynamic model and setting weights, training the model and outputting the final mass estimation value of the commercial vehicle.

[0141] It should be emphasized that the data used in the second part to train the model can be the data processed by the first part.

[0142] The total loss diagram of the second part of the physical information neural network (PINN) model can be referred to Figure 4 , Figure 4 In the method, the mass estimation value is obtained through the Bi-LSTM network, and the mass loss term is obtained according to the mass estimation value and the mass label. At the same time, the acceleration estimation value is input into the longitudinal dynamic model according to the mass estimation value, and the physical loss term is obtained according to the acceleration estimation value and the acceleration label. Finally, the total loss is obtained according to the mass loss term and the physical loss term.

[0143] In this implementation, by fusing with a neural network, the neural network learns the mapping relationship between vehicle operating parameters and mass under the constraints of physical laws, so as to solve the problem that the model-based commercial vehicle mass estimation method is restricted by modeling accuracy and working condition adaptability, and the data-driven method is restricted by data quality and interpretability, thereby achieving reliable and accurate commercial vehicle mass estimation.

[0144] The above are only feasible implementations of step S40 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S40.

[0145] Step S50, optimizing the initial mass estimation model according to the total loss function to obtain a commercial vehicle mass estimation model, and performing commercial vehicle mass estimation according to the commercial vehicle mass estimation model.

[0146] It should be noted that optimizing the initial quality estimation model according to the total loss function can be performed by selecting a suitable optimization algorithm to train the model. The optimization algorithm can be any one of a basic gradient update algorithm, an adaptive learning rate algorithm, and a stochastic gradient descent algorithm.

[0147] It is understandable that the model is trained on the training set and the parameters of the model are updated through multiple iterations. In each iteration, the gradient of the total loss function with respect to the model parameters is calculated, and the optimization algorithm is used to update the parameters. The performance of the model is regularly evaluated on the validation set, and the value of the loss function and the accuracy of the quality estimation are monitored (such as by calculating indicators such as mean square error and mean absolute error). If the model is found to be overfitting on the validation set (such as the loss function value starts to rise or the quality estimation error increases), measures can be taken such as stopping training early, adjusting the model complexity, or adding regularization terms.

[0148] This embodiment provides a commercial vehicle mass estimation method, which constructs a total loss function through the mass estimation value of the initial mass estimation model and the acceleration estimation value of the longitudinal dynamics model, uses physical laws to constrain the neural network learning to output reasonable results, and finally optimizes the model through the total loss function to obtain a more accurate and reliable commercial vehicle mass estimation model.

[0149] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 5 , step A11 includes steps S111 to S115:

[0150] Step S111, acquiring initial real-time data, wherein the initial real-time data includes one-to-one correspondence of resistance, acceleration, longitudinal vehicle speed, engine torque and brake master cylinder pressure.

[0151] It is understandable that the initial real-time data may be obtained by real-time detection of the target vehicle or target vehicle model for which quality estimation is to be performed according to various instruments during actual driving.

[0152] It should be noted that resistance includes vehicle driving force minus air resistance. For details, please refer to the following formula:

[0153]

[0154] It can be understood that the longitudinal vehicle speed can be the speed component of the vehicle along its traveling direction, that is, the forward or backward speed; the engine torque can be the rotational force generated by the engine output shaft, which affects the vehicle's power performance; the brake master cylinder pressure can be the pressure generated by the compression of the fluid in the brake master cylinder when the brake pedal is pressed, which is used to transmit braking force.

[0155] It should be noted that the initial real-time data may be obtained through the OBD-II (on-board diagnostic system second generation) interface commonly equipped in modern vehicles, allowing real-time data to be read by connecting a dedicated diagnostic tool or adapter, or other acquisition methods, which are not limited in this embodiment.

[0156] Step S112, obtaining a reference estimated mass based on the recursive least squares method according to the resistance, the acceleration, the longitudinal vehicle speed, the engine torque and the brake master cylinder pressure.

[0157] It should be noted that, based on the recursive least squares method, the reference estimated mass is obtained according to the resistance, the acceleration, the longitudinal vehicle speed, the engine torque and the brake master cylinder pressure by converting the model into a format that satisfies the least squares, and calculating the model output at adjacent moments by asymptotic approximation. The model output can refer to the following formula:

[0158]

[0159] Among them, h(k) represents the input data resistance, acceleration, longitudinal vehicle speed, engine torque and brake master cylinder pressure; T represents transposition, and the output y(k) can be obtained through measurement. According to the update rule of the least squares method and the estimated error, the model is iteratively optimized to obtain the estimated value of the parameter.

[0160] It should be emphasized that because the least squares method is usually used for offline parameter identification, the calculation is large and the identification takes a long time. In order to meet the needs of online estimation, the least squares format is converted into a parameter recursive estimation form. For details, refer to the following formula:

[0161]

[0162] in, represents the current estimated value, represents the estimated value at the previous moment, △ represents the correction value, and the correction value can be a pre-set value and adjusted according to the degree of correction. This embodiment does not limit this.

[0163] Step S113, obtaining the mass variance according to the resistance, the acceleration, the longitudinal vehicle speed, the engine torque, the brake master cylinder pressure and the reference estimated mass through a preset neural network.

[0164] It should be noted that the nonlinear factors affecting mass estimation are analyzed based on the longitudinal dynamics model of commercial vehicles, the nonlinear factors are modeled, the confidence of mass estimation is obtained, and it is determined whether to correct the mass estimation based on the model to improve the reliability of the sample (in preparation for the subsequent physical information neural network input).

[0165] It should be emphasized that the nonlinear factors that affect mass estimation may be one of the following or a combination of the following: ① When the vehicle is in the braking process or the transmission shifting process, the driving force is affected; ② When the vehicle is in a non-steady-state situation, the values ​​obtained from the engine torque, vehicle speed, and acceleration are abnormal; ③ When the vehicle is in a braking slip situation, the actual braking force provided by the ground does not match the braking force obtained by the calculated formula.

[0166] It should be noted that the preset neural network can be one or a combination of a Bayesian neural network, a deep Gaussian process, a deep belief network, etc., and the various resistances, accelerations, longitudinal vehicle speeds, engine torques, and brake master cylinder pressures calculated in the longitudinal dynamics model of the commercial vehicle are used as input features. The output of the Bayesian neural network is the variance of the mass estimate.

[0167] Step S114: obtaining a quality estimate confidence level according to the quality variance, and obtaining a quality label based on the quality estimate confidence level.

[0168] It is understandable that the confidence of the quality estimation value can be understood as an evaluation value of whether the quality estimation is valid and whether it is close to the true value; the quality label can be a quality value close to the true quality value.

[0169] It should be noted that the variance of the quality estimate output by the Bayesian neural network reflects the uncertainty of the quality estimate, and this uncertainty is incorporated into the confidence.

[0170] In a specific implementation, the calculation of the confidence level of the quality estimate may refer to the following formula:

[0171]

[0172] Among them, σ 2 represents the quality variance, and C represents the confidence of the quality estimate. The confidence value range is [0,1]. When the variance is extremely small, the confidence tends to 1, indicating that the quality estimate is extremely accurate; when the variance is large, the confidence tends to 0, indicating that the uncertainty of the quality estimate is large.

[0173] It should be noted that obtaining the quality label based on the confidence of the quality estimation value may be to use the quality estimation value as the quality label when the confidence is high, and to correct the quality estimation value when the confidence is low.

[0174] It should be emphasized that the mass estimation value here can be a mass estimation value calculated by the least square method based on the initial real-time data, in order to give the initial real-time data a corresponding mass value that is closer to the actual mass value of the vehicle.

[0175] In a feasible implementation, step S114 may include steps A1141 to A1142:

[0176] Step A1141: when the confidence of the quality estimation value is greater than or equal to a preset confidence threshold, use the reference estimated quality as a quality label.

[0177] It is understandable that the preset confidence threshold may be a preset confidence level at which the estimated quality is considered to be relatively close to the actual quality, and may generally be 0.8-0.9, or other values ​​may be set and adjusted according to actual conditions.

[0178] It is understandable that if the confidence of the quality estimation value is greater than or equal to the preset confidence threshold, it can be understood that the estimated quality value is closer to the actual quality and can be directly used as a quality label.

[0179] Step A1142: When the confidence of the quality estimation value is less than a preset confidence threshold, the quality estimation value at the last moment is used as a quality label.

[0180] It is understandable that the initial real-time data is collected in real time according to time. When the confidence of the quality estimation value is less than the preset confidence threshold, it is understood that the quality estimated at the current moment is inaccurate, and the quality estimation value of the previous moment (that is, the quality label of the previous moment) is used as the quality label of the current moment.

[0181] In a specific implementation, it is determined whether to correct the model-based quality estimation, and the determination method may be one of the following: based on a preset threshold, based on a gating mechanism, etc., or a combination thereof. When the determination conditions of the selected determination method are met, the model-based quality estimation is corrected to ensure that the quality estimation sample is more consistent with the actual operating state of the vehicle. Otherwise, no correction is made, and the quality estimation value at the previous moment is used as the quality estimation value at the current moment. For example, if the preset threshold is 0.9, when the confidence level is greater than 0.9, it is considered that the estimation is accurate and there is no need to correct the model-based quality estimation. Otherwise, correction is required.

[0182] In this implementation, by calculating the confidence of the mass estimation sample and correcting the estimated quality of the mass estimation sample based on the confidence, more accurate and reliable input data is provided for subsequent mass estimation based on the physical information neural network.

[0183] The above are only feasible implementations of step S114 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S114.

[0184] Step S115, obtaining a real-time training data set according to the quality label and the initial real-time data.

[0185] It is understandable that the quality labels at each moment are associated with the initial real-time numbers at each moment to obtain combined training data, and then the combined training data is divided according to unit time intervals to obtain multiple real-time training data sets.

[0186] It is understandable that the unit time interval may be 1 minute, 2 minutes, etc., and may be adjusted according to actual conditions.

[0187] In a feasible implementation, step S115 may include steps A1151 to A1153:

[0188] Step A1151, obtaining a combined training data set according to the quality label and the initial real-time data.

[0189] It is understandable that the initial real-time data may be the resistance, acceleration, longitudinal vehicle speed, engine torque and brake master cylinder pressure collected in real time at each moment.

[0190] The quality labels calculated from the real-time data at each moment are further associated and combined with the initial real-time data at each moment to obtain a combined training data set.

[0191] Step A1152, partitioning the combined training data set through a sliding window to obtain a data set to be cleaned.

[0192] It should be noted that the data set is divided based on the sliding window method and the vehicle data contained in the data set is preprocessed. The sliding window method divides the vehicle operation data into units of 1 minute. Specifically, starting from the start time of the vehicle data, the first window covers 1 minute of data from the start time. Subsequently, the window slides at a certain step size, which is usually set to 1 minute, that is, the next window contains data from 1min+1s to 2min, and so on, until all vehicle data are processed.

[0193] It should be understood that the sliding window division method can discretize continuous vehicle operation data. Its advantage is that it can effectively capture the operating characteristics of the vehicle in different time periods. In different 1-minute windows, the vehicle may be in different operating conditions such as stable driving, acceleration, deceleration or braking, or empty or fully loaded conditions, which provides reasonable time series data for subsequent mass estimation and vehicle performance analysis.

[0194] Step A1153, preprocessing the data set to be cleaned to obtain a real-time training data set, wherein the preprocessing includes outlier removal and data normalization.

[0195] It should be noted that the preprocessing of the vehicle data contained in the data set refers to removing outliers in the data and normalizing the sequence data. The method of removing outliers can be one of the box plot method, the 3σ principle, and the algorithm based on the local anomaly factor, or a combination thereof; the method of normalizing the sequence data can be one of linear normalization, nonlinear normalization, and zero mean normalization.

[0196] In this embodiment, by dividing the data set through a sliding window, the operating characteristics of the vehicle in different time periods can be effectively captured, and outliers in the data can be cleaned and normalized, so that more effective data can be obtained and the effectiveness of subsequent model training can be improved.

[0197] The above are only feasible implementations of step S115 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S115.

[0198] This embodiment provides a commercial vehicle quality estimation method, which provides more accurate and reliable input data for subsequent quality estimation based on a physical information neural network by judging and correcting quality estimation samples. The corrected quality estimation samples can better reflect the actual quality of the vehicle under different working conditions, reduce or avoid large deviations in the physical information neural network during training and prediction due to inaccurate sample data, and the confidence calculation and judgment mechanism can be used as an effective data screening and preprocessing method to help the physical information neural network better focus on high-quality data during the learning process, thereby improving the network's learning efficiency and the accuracy of quality estimation.

[0199] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the commercial vehicle mass estimation method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0200] This application also provides a commercial vehicle mass estimation device, please refer to Figure 6 , the commercial vehicle mass estimation device comprises:

[0201] A data acquisition module 10, for acquiring a real-time training data set, an initial mass estimation model and a longitudinal dynamics model;

[0202] An estimation module 20, configured to perform quality estimation based on the real-time training data set according to the initial quality estimation model to obtain a quality estimation value;

[0203] The estimation module 20 is further used to obtain an acceleration estimation value through the longitudinal dynamics model and the mass estimation value;

[0204] A loss function construction module 30, configured to obtain a total loss function according to the mass estimation value and the acceleration estimation value;

[0205] The estimation module 20 is further used to optimize the initial mass estimation model according to the total loss function to obtain a commercial vehicle mass estimation model, and to estimate the mass of the commercial vehicle according to the commercial vehicle mass estimation model.

[0206] The commercial vehicle mass estimation device provided by the present application adopts the commercial vehicle mass estimation method in the above-mentioned embodiment, which can solve the technical problem that the current model for commercial vehicle mass estimation is not accurate enough in estimating the mass of the vehicle during driving. Compared with the prior art, the beneficial effects of the commercial vehicle mass estimation device provided by the present application are the same as the beneficial effects of the commercial vehicle mass estimation method provided by the above-mentioned embodiment, and the other technical features in the commercial vehicle mass estimation device are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0207] In one embodiment, the data acquisition module 10 is further used to construct a vehicle driving force function according to the engine torque, transmission ratio, final reducer ratio, transmission efficiency, master cylinder brake pressure, front brake efficiency factor, rear brake efficiency factor and wheel rolling radius of the target vehicle;

[0208] Construct an air resistance function based on the air resistance coefficient, air density, the frontal area of ​​the target vehicle, and the driving speed;

[0209] Constructing a rolling resistance function according to the rolling resistance coefficient, the preset vehicle speed influencing parameter and the road slope angle;

[0210] A longitudinal dynamics model is constructed according to the vehicle driving force function, the air resistance function, the rolling resistance function, the slope resistance and the acceleration resistance.

[0211] In one embodiment, the data acquisition module 10 is further used to acquire initial real-time data, wherein the initial real-time data includes one-to-one corresponding resistance, acceleration, longitudinal vehicle speed, engine torque and brake master cylinder pressure;

[0212] Obtaining a reference estimated mass based on the resistance, the acceleration, the longitudinal vehicle speed, the engine torque, and the master cylinder pressure based on a recursive least square method;

[0213] Obtaining a mass variance according to the resistance, the acceleration, the longitudinal vehicle speed, the engine torque, the brake master cylinder pressure, and the reference estimated mass through a preset neural network;

[0214] Obtaining a quality estimate confidence level according to the quality variance, and obtaining a quality label based on the quality estimate confidence level;

[0215] A real-time training data set is obtained according to the quality label and the initial real-time data.

[0216] In one embodiment, the data acquisition module 10 is further configured to use the reference estimated quality as a quality label when the confidence of the quality estimation value is greater than or equal to a preset confidence threshold;

[0217] When the confidence of the quality estimation value is less than a preset confidence threshold, the quality estimation value at the last moment is used as the quality label.

[0218] In one embodiment, the data acquisition module 10 is further used to obtain a combined training data set according to the quality label and the initial real-time data;

[0219] Dividing the combined training data set by a sliding window to obtain a data set to be cleaned;

[0220] The data set to be cleaned is preprocessed to obtain a real-time training data set, wherein the preprocessing includes outlier removal and data normalization.

[0221] In one embodiment, the loss function construction module 30 is further used to obtain a quality loss term according to the quality label and the quality estimation value, and to obtain a physical loss term according to the acceleration label and the acceleration estimation value;

[0222] Obtaining a physical loss function and a quality loss function according to a preset loss function, the quality loss term and the physical loss term;

[0223] A total loss function is obtained according to the physical loss function and the quality loss function.

[0224] In one embodiment, the loss function construction module 30 is also used to obtain the model training time and loss hyperparameters;

[0225] Constructing a time-varying weight function according to the model training duration and the loss hyperparameter;

[0226] Obtaining physical weight and quality weight according to the time-varying weight function;

[0227] A total loss function is constructed according to the physical weight, the quality weight, the physical loss function and the quality loss function.

[0228] The present application provides a commercial vehicle mass estimation device, which includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the commercial vehicle mass estimation method in the above-mentioned embodiment 1.

[0229] Reference below Figure 7 , which shows a schematic diagram of the structure of a commercial vehicle mass estimation device suitable for implementing the embodiment of the present application. The commercial vehicle mass estimation device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 7 The commercial vehicle mass estimation device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0230] like Figure 7 As shown, the commercial vehicle mass estimation device may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. Various programs and data required for the operation of the commercial vehicle mass estimation device are also stored in RAM1004. The processing device 1001, ROM1002, and RAM1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the commercial vehicle mass estimation device to communicate wirelessly or wired with other devices to exchange data. Although the figure shows a commercial vehicle mass estimation device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have instead.

[0231] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0232] The commercial vehicle mass estimation device provided by the present application adopts the commercial vehicle mass estimation method in the above embodiment, which can solve the technical problem that the current model for commercial vehicle mass estimation is not accurate enough in estimating the mass of the vehicle during driving. Compared with the prior art, the beneficial effects of the commercial vehicle mass estimation device provided by the present application are the same as the beneficial effects of the commercial vehicle mass estimation method provided by the above embodiment, and the other technical features in the commercial vehicle mass estimation device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0233] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0234] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0235] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the commercial vehicle mass estimation method in the above-mentioned embodiment.

[0236] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0237] The computer-readable storage medium may be included in the commercial vehicle mass estimation device; or may exist independently without being assembled into the commercial vehicle mass estimation device.

[0238] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the commercial vehicle mass estimation device, the commercial vehicle mass estimation device: obtains a real-time training data set, an initial mass estimation model and a longitudinal dynamics model; performs mass estimation based on the real-time training data set according to the initial mass estimation model to obtain a mass estimation value; obtains an acceleration estimation value through the longitudinal dynamics model and the mass estimation value; obtains a total loss function according to the mass estimation value and the acceleration estimation value; optimizes the initial mass estimation model according to the total loss function to obtain a commercial vehicle mass estimation model, and performs commercial vehicle mass estimation according to the commercial vehicle mass estimation model.

[0239] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0240] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0241] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0242] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned commercial vehicle mass estimation method, and can solve the technical problem that the current model for commercial vehicle mass estimation is not accurate enough in estimating the mass of the vehicle during driving. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the commercial vehicle mass estimation method provided by the above-mentioned embodiment, and will not be repeated here.

[0243] The present application also provides a computer program product, including a computer program, which implements the steps of the commercial vehicle mass estimation method as described above when executed by a processor.

[0244] The computer program product provided by the present application can solve the technical problem that the current model for commercial vehicle mass estimation is not accurate enough in estimating the mass of the vehicle during driving. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the commercial vehicle mass estimation method provided by the above embodiment, and will not be elaborated here.

[0245] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A commercial vehicle mass estimation method, characterized in that: The commercial vehicle mass estimation method comprises: Obtain real-time training data sets, initial mass estimation models, and longitudinal dynamics models; Performing quality estimation based on the real-time training data set according to the initial quality estimation model to obtain a quality estimation value; Obtaining an acceleration estimate using the longitudinal dynamics model and the mass estimate; Obtaining a total loss function according to the mass estimation value and the acceleration estimation value; The initial mass estimation model is optimized according to the total loss function to obtain a commercial vehicle mass estimation model, and the commercial vehicle mass estimation is performed according to the commercial vehicle mass estimation model.

2. The commercial vehicle mass estimation method according to claim 1, characterized in that: The steps to obtain the longitudinal dynamics model include: A vehicle driving force function is constructed according to the engine torque, transmission ratio, final reducer ratio, transmission efficiency, master cylinder brake pressure, front brake efficiency factor, rear brake efficiency factor and wheel rolling radius of the target vehicle; Construct an air resistance function based on the air resistance coefficient, air density, the frontal area of ​​the target vehicle, and the driving speed; Constructing a rolling resistance function according to the rolling resistance coefficient, the preset vehicle speed influencing parameter and the road slope angle; A longitudinal dynamics model is constructed according to the vehicle driving force function, the air resistance function, the rolling resistance function, the slope resistance and the acceleration resistance.

3. The commercial vehicle mass estimation method according to claim 1, characterized in that: The steps to obtain a real-time training dataset include: Acquiring initial real-time data, the initial real-time data including one-to-one correspondence of resistance, acceleration, longitudinal vehicle speed, engine torque, and brake master cylinder pressure; Obtaining a reference estimated mass based on the resistance, the acceleration, the longitudinal vehicle speed, the engine torque, and the master cylinder pressure based on a recursive least square method; Obtaining a mass variance according to the resistance, the acceleration, the longitudinal vehicle speed, the engine torque, the brake master cylinder pressure, and the reference estimated mass through a preset neural network; Obtaining a quality estimate confidence level according to the quality variance, and obtaining a quality label based on the quality estimate confidence level; A real-time training data set is obtained according to the quality label and the initial real-time data.

4. The commercial vehicle mass estimation method according to claim 3, characterized in that: The method of obtaining a quality estimate confidence level according to the quality variance, and obtaining a quality label based on the confidence level, further includes: When the confidence of the quality estimation value is greater than or equal to a preset confidence threshold, using the reference estimated quality as a quality label; When the confidence of the quality estimation value is less than a preset confidence threshold, the quality estimation value at the last moment is used as the quality label.

5. The commercial vehicle mass estimation method according to claim 3, characterized in that: The obtaining of a real-time training data set according to the quality label and the initial real-time data includes: Obtaining a combined training data set according to the quality label and the initial real-time data; Dividing the combined training data set by a sliding window to obtain a data set to be cleaned; The data set to be cleaned is preprocessed to obtain a real-time training data set, wherein the preprocessing includes outlier removal and data normalization.

6. The commercial vehicle mass estimation method according to claim 1, characterized in that: The obtaining of a total loss function according to the mass estimation value and the acceleration estimation value comprises: Obtaining a mass loss term according to the mass label and the mass estimation value, and obtaining a physical loss term according to the acceleration label and the acceleration estimation value; Obtaining a physical loss function and a quality loss function according to a preset loss function, the quality loss term and the physical loss term; A total loss function is obtained according to the physical loss function and the quality loss function.

7. The commercial vehicle mass estimation method according to claim 6, characterized in that: The obtaining of the total loss function according to the physical loss function and the quality loss function comprises: Get model training time and loss hyperparameters; Constructing a time-varying weight function according to the model training duration and the loss hyperparameter; Obtaining physical weight and quality weight according to the time-varying weight function; A total loss function is constructed according to the physical weight, the quality weight, the physical loss function and the quality loss function.

8. A commercial vehicle mass estimation device, characterized in that: The commercial vehicle mass estimation device comprises: A data acquisition module, used to acquire real-time training data sets, initial mass estimation models, and longitudinal dynamics models; An estimation module, configured to perform quality estimation based on the real-time training data set according to the initial quality estimation model to obtain a quality estimation value; The estimation module is further used to obtain an acceleration estimation value through the longitudinal dynamics model and the mass estimation value; A loss function construction module, used for obtaining a total loss function according to the mass estimation value and the acceleration estimation value; The estimation module is also used to optimize the initial mass estimation model according to the total loss function to obtain a commercial vehicle mass estimation model, and perform commercial vehicle mass estimation according to the commercial vehicle mass estimation model.

9. A commercial vehicle mass estimation device, characterized in that: The device comprises: a memory, a processor, and a commercial vehicle mass estimation program stored in the memory and executable on the processor, wherein the commercial vehicle mass estimation program is configured to implement the commercial vehicle mass estimation method according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a commercial vehicle mass estimation program, and when the commercial vehicle mass estimation program is executed by the processor, the commercial vehicle mass estimation method according to any one of claims 1 to 7 is implemented.

Images