Vehicle speed prediction method, apparatus and computer program product

Abstract

The present application relates to the technical field of vehicle speed prediction, and in particular to a vehicle speed prediction method, device and computer program product. The vehicle speed prediction method comprises: collecting a first driving parameter of a vehicle in a current period; determining a working condition classification corresponding to the first driving parameter based on the first driving parameter; inputting the first driving parameter, a second driving parameter of a previous period and a vehicle speed error obtained according to the previous period into a vehicle speed prediction model corresponding to the working condition classification to obtain a predicted vehicle speed output by the vehicle speed prediction model. The present application selects and adjusts the vehicle speed prediction model in a targeted manner through working condition classification, effectively improves the accuracy and real-time performance of prediction, and provides more reliable vehicle speed information support for electronic control systems such as an anti-lock braking system (ABS) and an electronic stability control system (ESC).

Description

Technical Field

[0001] This invention relates to the field of vehicle speed prediction technology, and in particular to a vehicle speed prediction method, apparatus and computer program product. Background Technology

[0002] The accuracy and effectiveness of vehicle speed prediction are crucial factors affecting the performance of anti-lock braking systems (ABS) and electronic stability control systems (ESC). Therefore, improving the timeliness and accuracy of vehicle speed prediction is of great significance for optimizing the strategies of the vehicle's electronic control system. Commonly used vehicle speed prediction methods are divided into two types: data-driven and model-driven.

[0003] Data-driven vehicle speed prediction methods suffer from poor filtering and nonlinear mapping capabilities, making it difficult to achieve accurate and real-time vehicle speed prediction. Model-driven methods, on the other hand, often employ highly complex high-precision models, potentially impacting real-time prediction performance. Simplifying models, however, may sacrifice prediction accuracy, compromising both accuracy and timeliness. Therefore, improving the accuracy and real-time performance of vehicle speed prediction is a pressing technical challenge. Summary of the Invention

[0004] To address the aforementioned technical issues, this invention provides a vehicle speed prediction method, apparatus, and computer storage medium.

[0005] According to a first aspect of the present invention, a vehicle speed prediction method is provided, comprising: collecting a first driving parameter of a vehicle in the current cycle; determining a working condition classification corresponding to the first driving parameter based on the first driving parameter; inputting the first driving parameter, a second driving parameter of the previous cycle, and a vehicle speed error obtained according to the previous cycle into a vehicle speed prediction model corresponding to the working condition classification, and obtaining the predicted vehicle speed output by the vehicle speed prediction model.

[0006] Furthermore, according to the vehicle speed prediction method of the first aspect of the present invention, the first driving parameter includes a first vehicle speed and a first acceleration; determining the operating condition classification corresponding to the first driving parameter based on the first driving parameter includes: determining the operating condition classification corresponding to the first driving parameter based on the first acceleration and the acceleration threshold value.

[0007] Furthermore, according to the vehicle speed prediction method of the first aspect of the present invention, the training method of the vehicle speed prediction model includes: acquiring historical driving parameters of the vehicle; dividing the historical driving parameters into several working condition blocks based on working condition classification; preprocessing the first historical driving parameters of each working condition block to obtain the target historical driving parameters corresponding to each working condition block; for each working condition block: initializing the weights and bias parameters of the initial vehicle speed prediction model based on a Long Short-Term Memory (LSTM) network; iteratively training the initialized initial vehicle speed prediction model based on the second vehicle speed, the second acceleration, and the third vehicle speed in the (t-1)th period and the vehicle speed prediction deviation in the (t-1)th period of the target historical driving parameters; updating the weights based on the backpropagation algorithm during training; and obtaining the trained vehicle speed prediction model corresponding to the working condition classification when the initial vehicle speed prediction model meets the training termination condition.

[0008] Furthermore, according to the vehicle speed prediction method of the first aspect of the present invention, the first historical driving parameters of each working condition block are preprocessed to obtain the target historical driving parameters corresponding to each working condition block, including: performing wavelet decomposition and reconstruction on the first historical driving parameters of each working condition block to obtain the target historical driving parameters corresponding to each working condition block.

[0009] Furthermore, according to the vehicle speed prediction method of the first aspect of the present invention, initializing the weights and bias parameters of the initial vehicle speed prediction model based on the Long Short-Term Memory Network (LSTM) includes: initializing the weights and bias parameters of the initial vehicle speed prediction model based on the Long Short-Term Memory Network (LSTM) based on the driving condition classification.

[0010] Furthermore, according to the vehicle speed prediction method of the first aspect of the present invention, during training, updating the weights based on the backpropagation algorithm includes: defining a standard deviation function during training, calculating the gradient of the standard deviation function with respect to the weights using the backpropagation algorithm and the Adam optimization algorithm, and updating the weights of the initial vehicle speed prediction model according to the gradient.

[0011] Furthermore, according to the vehicle speed prediction method of the first aspect of the present invention, when the initial vehicle speed prediction model meets the training termination condition, a vehicle speed prediction model corresponding to the completed working condition classification is obtained, including: determining a model evaluation function based on the predicted vehicle speed output by the initial vehicle speed prediction model and its corresponding actual vehicle speed; when the model evaluation function meets a preset condition, the initial vehicle speed prediction model meets the training termination condition, and a vehicle speed prediction model corresponding to the completed working condition classification is obtained.

[0012] Furthermore, according to the vehicle speed prediction method of the first aspect of the present invention, the model evaluation function formula is as follows:

[0013]

[0014] in, The model outputs the predicted vehicle speed for the next moment. This represents the actual vehicle speed at the next moment. J When the value is greater than 90%, the evaluation function meets the preset conditions.

[0015] According to another aspect of the present invention, a vehicle speed prediction device is provided, comprising: a data acquisition module configured to acquire a first driving parameter of a vehicle in the current period; a classification module configured to determine a working condition classification corresponding to the first driving parameter based on the first driving parameter; and a predicted vehicle speed output module configured to input the first driving parameter, a second driving parameter of the previous period, and a vehicle speed error obtained according to the previous period into a vehicle speed prediction model corresponding to the working condition classification, and obtain a predicted vehicle speed output by the vehicle speed prediction model.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements any of the vehicle speed prediction methods described above.

[0017] The vehicle speed prediction method, device, and computer program product provided by this invention collect the vehicle's driving parameters in real time for the current cycle and intelligently determine the operating condition classification based on these parameters. Then, the driving parameters of the current cycle, the driving parameters of the previous cycle, and the vehicle speed error calculated based on the previous cycle are input into a vehicle speed prediction model that matches the operating condition classification, thereby achieving accurate prediction of vehicle speed. This method effectively improves the accuracy and real-time performance of prediction by selectively selecting and adjusting the vehicle speed prediction model according to the operating condition classification, and can provide more reliable vehicle speed information support for electronic control systems such as anti-lock braking systems (ABS) and electronic stability control systems (ESC). Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is one of the flowcharts illustrating the vehicle speed prediction method provided by the present invention.

[0020] Figure 2 This is the second flowchart of the vehicle speed prediction method provided by the present invention.

[0021] Figure 3 This is a schematic diagram of wavelet decomposition provided by the present invention.

[0022] Figure 4This is a structural diagram of the Long Short-Term Memory (LSTM) network provided by this invention.

[0023] Figure 5 This is the third flowchart of the vehicle speed prediction method provided by the present invention.

[0024] Figure 6 This is a functional block diagram of the vehicle speed prediction device provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0026] The following is combined Figures 1-5 The vehicle speed prediction method of the present invention is described.

[0027] Figure 1 This is one of the flowcharts illustrating the vehicle speed prediction method provided by the present invention, such as... Figure 1 As shown, the method includes the following:

[0028] Step 101: Collect the first driving parameters of the vehicle in the current cycle.

[0029] In one embodiment of the present invention, the first driving parameter refers to the relevant driving information of the vehicle within the current driving time period (i.e., the "current period"). This first driving parameter may include, but is not limited to, vehicle speed, acceleration, driving time, steering angle, braking status, and engine status. By collecting the first driving parameter, the vehicle control system can monitor the vehicle's driving status in real time and adjust the control strategy as needed to ensure the vehicle's safety and stability. Simultaneously, this data can also be used for subsequent vehicle performance analysis and fault diagnosis.

[0030] Step 102: Based on the first driving parameters, determine the operating condition classification corresponding to the first driving parameters.

[0031] In one embodiment of the present invention, the operating condition classification divides the driving state into different categories based on the vehicle's driving parameters. A first driving parameter is compared and matched with the standard for operating condition classification, and based on the matching result, the first driving parameter can be classified into the corresponding operating condition category.

[0032] In one embodiment of the present invention, the first driving parameter includes a first vehicle speed and a first acceleration;

[0033] Based on the first driving parameters, determine the operating condition classification corresponding to the first driving parameters, including:

[0034] Based on the first acceleration and the acceleration threshold, the operating condition classification corresponding to the first driving parameter is determined.

[0035] Specifically, the first vehicle speed represents the vehicle speed in the current cycle, and the first acceleration represents the acceleration in the current cycle. The acceleration threshold can be one, a set, or multiple preset acceleration thresholds used to divide acceleration into different intervals. The determination of the acceleration threshold is usually based on empirical data, industry standards, or specific application requirements. The first acceleration is compared with the acceleration threshold, and the operating condition category corresponding to the first acceleration is determined based on which interval it falls within. For example, the operating condition category may include acceleration, constant speed, deceleration, etc. (or other classifications defined according to actual needs).

[0036] For example, the first acceleration a (m / s²) of the current period is collected. 2 The vehicle state is classified into acceleration, deceleration, and constant speed based on the magnitude of acceleration. The acceleration threshold varies depending on the vehicle; this scheme selects an acceleration threshold of 0.5 m / s². 2 The details are as follows:

[0037] ① If -0.5m / s 2 <a(m / s 2 <0.5m / s 2 If , it means the vehicle is moving at a constant speed;

[0038] ②If a(m / s 2 >0.5m / s 2 This indicates that the vehicle is accelerating;

[0039] ③If a(m / s 2 < -0.5m / s 2 This indicates that the vehicle is decelerating.

[0040] Based on the above description, the operating conditions are divided into three types: constant speed, acceleration, and deceleration.

[0041] Step 103: Input the first driving parameter, the second driving parameter of the previous cycle, and the vehicle speed error obtained from the previous cycle into the vehicle speed prediction model corresponding to the working condition classification, and obtain the predicted vehicle speed output by the vehicle speed prediction model.

[0042] In one embodiment of the present invention, the second driving parameter of the previous cycle includes driving parameters such as vehicle speed and acceleration of the previous cycle. This second driving parameter can be obtained through sensors on the vehicle or through the vehicle's CAN bus; the method of acquisition is not limited in the present invention. The vehicle speed error obtained in the previous cycle refers to the error between the predicted vehicle speed value and the actual value of the previous cycle. The predicted vehicle speed value of the previous cycle is the predicted vehicle speed value output by the vehicle speed prediction model in the previous cycle. This vehicle speed prediction model is pre-trained. The actual value refers to the vehicle speed of the previous cycle, which can be obtained through the vehicle's CAN bus or wheel speed sensors, vehicle speed sensors, etc. Different operating condition classifications correspond to different vehicle speed prediction models. The first driving parameter, the second driving parameter of the previous cycle, and the vehicle speed error are input into the vehicle speed prediction model corresponding to the operating condition classification. The vehicle speed prediction model calculates based on the input parameters and outputs the predicted vehicle speed value. Predicted vehicle speed reflects the vehicle's likely speed over a future period and can be used to optimize driving strategies, such as adjusting speed in advance and avoiding sudden acceleration or deceleration, thereby improving driving safety and comfort. It can also be used in the planning and scheduling of intelligent transportation systems, such as optimizing traffic signal control and predicting traffic congestion.

[0043] In summary, according to the technical solution provided by the embodiments of the present invention, by collecting the vehicle's driving parameters in real time for the current cycle and intelligently determining the operating condition classification based on these parameters, the driving parameters of the current cycle, the driving parameters of the previous cycle, and the vehicle speed error calculated based on the previous cycle are jointly input into a vehicle speed prediction model that matches the operating condition classification, thereby achieving accurate prediction of vehicle speed. This method effectively improves the accuracy and real-time performance of prediction by selectively selecting and adjusting the vehicle speed prediction model according to the operating condition classification, and can provide more reliable vehicle speed information support for electronic control systems such as anti-lock braking system (ABS) and electronic stability control system (ESC).

[0044] Figure 2 This is the second flowchart of the vehicle speed prediction method provided by the present invention.

[0045] like Figure 2 As shown, according to the vehicle speed prediction method provided by the present invention, the training method of the vehicle speed prediction model specifically includes the following:

[0046] Step 201: Obtain the vehicle's historical driving parameters.

[0047] In one embodiment of the present invention, obtaining historical driving parameters of a vehicle mainly refers to obtaining a series of driving-related data information generated during the vehicle's past driving process. These parameters can reflect the vehicle's driving status, performance, etc. For example, historical driving parameters can be obtained from the vehicle's driving records, and these historical driving parameters may include parameters such as vehicle speed, acceleration, and timestamps.

[0048] Step 202: Based on the operating condition classification, divide the historical driving parameters into several operating condition blocks.

[0049] In one embodiment of this disclosure, the driving condition classification divides the driving state into different categories based on the vehicle's driving parameters. Feature parameters reflecting the vehicle's driving state, such as vehicle speed, distance traveled, maximum speed, and acceleration, are extracted from historical driving parameters. These extracted feature parameters are compared with the driving condition classification, thereby dividing the historical driving parameters into several driving condition blocks, each corresponding to a driving condition classification. This driving condition block division helps the intelligent transportation system more accurately identify and understand the vehicle's driving state and behavior, thereby enabling better traffic signal control, traffic congestion prediction, and route planning.

[0050] Step 203: Preprocess the first historical driving parameters of each working condition block to obtain the target historical driving parameters corresponding to each working condition block.

[0051] In one embodiment of the present invention, the first historical driving parameter is the historical driving parameter divided into each driving condition block according to the driving condition classification. The data in each driving condition block is preprocessed, such as denoising and normalization, to facilitate better model learning.

[0052] In one embodiment of the present invention, the first historical driving parameters of each driving condition block are preprocessed to obtain the target historical driving parameters corresponding to each driving condition block, including:

[0053] Wavelet decomposition and reconstruction are performed on the first historical driving parameters of each working condition block to obtain the target historical driving parameters corresponding to each working condition block.

[0054] Specifically, wavelet decomposition and wavelet reconstruction are performed on the first historical driving parameters for each driving condition block to obtain the target historical driving parameters corresponding to each driving condition block. This step aims to reduce glitches and noise in the first historical driving parameters. Preprocessing the data through wavelet decomposition and reconstruction effectively reduces noise and glitches, thereby improving the model's training performance and prediction accuracy.

[0055] For example, the first historical driving parameters of different driving conditions are preprocessed based on the wavelet denoising decomposition algorithm. By using a "time-frequency" window that varies with frequency, the problems caused by a fixed window can be effectively avoided. The wavelet denoising process mainly includes wavelet decomposition and data reconstruction, as shown in the appendix. Figure 3 As shown. To reduce data glitches, wavelet decomposition and reconstruction are performed on the data. The wavelet decomposition level is set to 3 levels, and the process can be represented as follows:

[0056]

[0057] In the formula, i represents the decomposition level, H and L represent high-pass filtering and low-pass filtering, respectively. After decomposition, the data is reconstructed using wavelet, and the process can be expressed as:

[0058]

[0059] In the formula, L* and H* represent the sampling operators for low-pass and high-pass filtering, respectively. To prevent over-filtering from distorting the original data, a soft threshold function can be selected to filter high-frequency spikes in the data.

[0060] For example, the above method is used to preprocess the first historical driving parameters, which mainly include: vehicle speed at time t. (m / s), acceleration a at time t t (m / s 2 (Speed ​​at time t-1) (m / s), the above parameters are decomposed into three levels of wavelet decomposition, as shown below:

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067] Reconstruct the parameters after decomposition:

[0068]

[0069]

[0070]

[0071] The low-pass and high-pass filtering sampling operators can be customized based on the sample information. The parameters obtained after reconstruction are the target historical driving parameters.

[0072] Step 204: For each working condition block, initialize the weights and bias parameters of the initial vehicle speed prediction model based on the Long Short-Term Memory (LSTM) network.

[0073] In one embodiment of the present invention, an initial vehicle speed prediction model based on LSTM is created for each driving condition block, and the weights and bias parameters of the initial vehicle speed prediction model based on LSTM are initialized for each driving condition block. Initializing the weights and bias parameters of the initial vehicle speed prediction model based on LSTM for each driving condition block is an important step in the process of building the vehicle speed prediction model, which helps to improve prediction accuracy, enhance the model's generalization ability, and optimize driving strategies.

[0074] In one embodiment of the present invention, initializing the weights and bias parameters of the initial vehicle speed prediction model based on a Long Short-Term Memory (LSTM) network includes:

[0075] Based on the classification of driving conditions, the weights and bias parameters of the initial vehicle speed prediction model based on the Long Short-Term Memory (LSTM) network are initialized.

[0076] Specifically, the driving conditions can be classified into three types: acceleration, constant speed, and deceleration. The weights and bias parameters of the initial vehicle speed prediction model based on a Long Short-Term Memory (LSTM) network are initialized according to this classification. Initializing the model parameters based on the driving condition classification allows the model to better learn and predict the target variable (vehicle speed). This method helps improve the model's generalization ability and prediction accuracy.

[0077] For example, the weights and bias parameters of the LSTM model are initialized. The LSTM algorithm involves three gates: the forget gate, the input gate, and the output gate, as shown in the attached diagram. Figure 4 As shown, the initial weights of each gate are set to a random normal distribution with a standard deviation of 1. The weight parameters are mainly concentrated between 0 and 1. Depending on the different operating conditions, the input acceleration a is increased or decreased. t (m / s 2 The weight of ) under uniform acceleration a t (m / s 2 The weights of the acceleration signal are mainly concentrated between 0 and 1. In cases of deceleration or acceleration, the weight ratio of the acceleration signal is increased, and the input variable acceleration a... t (m / s 2 The weights of the gates are concentrated between 0.3 and 1, and the initial bias values ​​of each gate are uniformly set to 0.

[0078] In summary, by differentiating the initialization of the weights and bias parameters of the LSTM model according to the operating conditions (acceleration, constant speed, deceleration), especially by adjusting the weights of the input acceleration signal, the model can more accurately capture the characteristics of vehicle speed changes under different operating conditions, thereby improving the accuracy of vehicle speed prediction and the model's generalization ability.

[0079] Step 205: Based on the second vehicle speed in the t-th cycle, the second acceleration in the t-th cycle, the third vehicle speed in the (t-1)-th cycle, and the vehicle speed prediction deviation in the (t-1)-th cycle from the target historical driving parameters, iteratively train the initialized initial vehicle speed prediction model.

[0080] In one embodiment of the present invention, the second vehicle speed is the vehicle speed in the t-th cycle of the target historical driving parameters, the second acceleration is the acceleration in the t-th cycle of the target historical driving parameters, the third vehicle speed is the vehicle speed in the (t-1)-th cycle of the target historical driving parameters, and the vehicle speed prediction deviation refers to the deviation between the predicted vehicle speed and the third vehicle speed at the end of the (t-1)-th cycle of the target historical driving parameters. The target historical driving parameters (including vehicle speed, acceleration, and previous prediction deviations) are used as input data to iteratively train the initial vehicle speed prediction model.

[0081] For example, the input parameters (i.e., the second vehicle speed in the t-th cycle, the second acceleration in the t-th cycle, the third vehicle speed in the (t-1)-th cycle, and the vehicle speed prediction deviation in the (t-1)-th cycle) can be represented as a 1×4 matrix: ,in, =vehicle speed at time t (m / s) =Acceleration at time t (m / s) 2 ), =Speed ​​of vehicle at time t-1 (m / s) The vehicle speed prediction deviation (m / s) at time t-1 is calculated using the following formula:

[0082] (m / s)=

[0083] The above formula calculates the actual vehicle speed at time t-1. With predicted vehicle speed The difference, the output of the LSTM neural network prediction model (i.e., the aforementioned initial vehicle prediction model). To predict the vehicle speed at the next moment (m / s).

[0084] In summary, through iterative training using historical vehicle speeds, accelerations, and prediction biases, the vehicle speed prediction model of this invention can learn and adapt to driving trends, improving prediction accuracy, enhancing adaptability and robustness, thereby providing better performance and user experience in practical applications.

[0085] Step 206: During training, update the weights based on the backpropagation algorithm.

[0086] In one embodiment of the present invention, during the training process of the initial vehicle speed prediction model, the backpropagation algorithm is used to calculate the gradient of the error between the predicted vehicle speed and the actual vehicle speed relative to the weights, and the weights and bias parameters of the model are updated according to the gradient to reduce the prediction error.

[0087] In one embodiment of the present invention, during training, updating the weights based on the backpropagation algorithm includes:

[0088] During training, a standard deviation function is defined, and the gradient of the standard deviation function with respect to the weights is calculated using the backpropagation algorithm. Based on the gradient, the weights of the initial vehicle speed prediction model are updated using the Adam algorithm.

[0089] Specifically, the standard deviation function measures the difference between the model's predicted vehicle speed and the actual vehicle speed. The Adam algorithm is an adaptive learning rate optimization algorithm that combines the advantages of momentum and RMSprop algorithms. In each iteration, the Adam algorithm adjusts the learning rate based on the first and second components of the gradient, thereby updating the weights. Here, the learning rate is a hyperparameter used to control the step size of weight updates and is adaptively adjusted in the Adam algorithm. During the initial training of the vehicle speed prediction model, firstly, a standard deviation function is defined to measure the difference between the model's predicted vehicle speed and the actual vehicle speed. Secondly, the gradient of the standard deviation function with respect to the model weights is calculated using the backpropagation algorithm, and the learning rate is dynamically updated using the Adam algorithm. During training, the gradient of the loss function (i.e., the aforementioned standard deviation function) with respect to the model weights is first calculated using the backpropagation algorithm. Then, the learning rate is dynamically adjusted using the Adam algorithm based on the first and second components of the gradient. After calculating the gradient and adjusting the learning rate, the weights are updated. This process iterates multiple times on the entire training set, updating the weights in each iteration based on the new learning rate and gradient. The gradients described above represent the degree of influence of each weight on the standard deviation function value. This process starts from the output layer and proceeds backward through each layer of the network until the input layer. This process iterates multiple times.

[0090] Through the above process, the model can gradually learn how to better extract features from the input data and make accurate vehicle speed predictions. Using the Adam algorithm can update weights more efficiently, improving the model's convergence speed and prediction performance.

[0091] For example, during vehicle operation, the weight parameters are adjusted in real time based on the deviation of the predicted vehicle speed from the previous time step t (as described by the aforementioned standard deviation function), and rolling iterative prediction is performed in real time. The sampling period is 100ms, and the calculation process is as follows:

[0092] ① Input gate to input vector: To extract effective information, the input gate is transformed using the sigmoid activation function, the expression of which is as follows:

[0093]

[0094] The output of the input gate is:

[0095]

[0096] In the formula, This represents the weights of the intermediate vectors of the input gate. This represents the weights of the input variables of the input gate. This indicates the output deviation of the input gate. This represents the input to the LSTM model (i.e., the aforementioned initial vehicle speed prediction model) at time t. , This represents the hidden state of the output layer at time t-1, i.e. Default value at the initial time The output of the input gate of the vehicle speed prediction model is:

[0097]

[0098]

[0099] ② The forget gate determines the information that the neural network model needs to discard. It automatically transfers the knowledge acquired at time t to the next new time. Choosing the sigmoid activation function, the forget gate can be represented as:

[0100]

[0101] In the formula, This represents the weights of the intermediate vector in the forget gate. The weights of the input variables for the forget gate are represented. This indicates the output deviation of the forget gate. This represents the input of the LSTM model at time t, i.e. , This represents the hidden state of the output layer at time t-1, i.e., the intermediate vector. The output of the forget gate in the initial vehicle speed prediction model is as follows:

[0102]

[0103] ③ Intermediate state update: The state of intermediate variables in the model is updated by forgetting old information and adding new information. The update process is represented as follows:

[0104]

[0105] In the formula, C t-1This represents the intermediate variable state at time t-1, with an initial default value of C0 = 0. t A represents the intermediate variable state calculated at time t. t This represents the input information at time t. This represents the dot product. The intermediate state values ​​of the vehicle speed prediction model can be represented as:

[0106] C t =[C t,1 C t,2 C t,3 C t,4 ] T

[0107] =[C t-1,1 *f t,1 C t-1,2 *f t,2 C t-1,3 *f t,3 C t-1,4 *f t,4 ] T

[0108] +[i t,1 *x1,i t,2 *x2,i t,3 *x3,i t,4 *x4] T

[0109] ④ The output gate determines how the neuron's state at time t affects the hidden state at the next time step. The output information of the forget gate and the input gate are combined and the intermediate variable hidden state is calculated. The sigmoid activation function is selected. The output and hidden states of the output layer are shown below:

[0110]

[0111]

[0112] In the formula, This represents the weights of the intermediate variables of the output gate. This represents the input variable weights of the output gate. This indicates the output deviation of the output gate. This represents the input of the LSTM model at time t, i.e. , This represents the hidden state of the output layer at time t-1, i.e. The output of the vehicle speed prediction model The hidden state can be represented as:

[0113]

[0114]

[0115] ④ The model iteratively updates the weights, combining the adaptive backpropagation algorithm and the Adam algorithm to update the weight parameters. The standard deviation function of the initial vehicle speed prediction model is:

[0116]

[0117] in, = This represents the predicted vehicle speed at the next moment, as output by the LSTM prediction model. = This indicates the actual vehicle speed value read via the CAN bus at the next moment.

[0118] The backpropagation output gate weight iteration process can be represented as:

[0119]

[0120]

[0121] The weight iteration process of the backpropagation forget gate can be represented as:

[0122]

[0123]

[0124] The weight iteration process of the backpropagation input gate can be represented as:

[0125]

[0126]

[0127] In the Adam algorithm, the initial learning rate is set to... =0.5, to avoid local optima during gradient optimization, the learning rate of the parameters is updated in real time. If the learning rate at time t is... Then the learning rate at time t+1 can be expressed as:

[0128]

[0129] In the formula, and Let these represent the first and second components of the gradient learning rate at time t, respectively. This represents the hyperparameter learning rate, which is typically set to 10. -3 -10 -5 between, This represents the correction term constant, typically taken as 10. -8 The learning rate is updated in real time according to the above process.

[0130] Step 207: When the initial vehicle speed prediction model meets the training termination condition, the vehicle speed prediction model corresponding to the completed working condition classification is obtained.

[0131] In one embodiment of the present invention, the training termination condition refers to setting certain conditions during the training process of the machine learning model to determine when to stop training. These conditions may include: reaching a preset number of iterations (epochs), meaning the model has completed a certain number of learning iterations on the entire training set; or, the loss function value on the validation set does not decrease significantly after several consecutive iterations, indicating that the model may be close to its optimal state; or, reaching preset performance indicators, such as accuracy or recall, satisfying a specific threshold. When the above training termination conditions are met, the model training process terminates. At this point, the model's parameters (weights and biases) are fixed, forming a trained vehicle speed prediction model for each work condition classification. Obtaining a trained model means that the model is ready for practical applications and can be deployed on vehicles for real-time vehicle speed prediction.

[0132] In summary, the technical solution provided by the embodiments of the present invention improves the accuracy of vehicle speed prediction by training the model specifically for different operating conditions. The model can adapt to different driving conditions and provide more comprehensive prediction capabilities. The real-time prediction capability of the LSTM model enables the system to respond quickly to changes in driving conditions. It can accurately predict vehicle speed changes under different driving states, providing strong support for intelligent vehicle control.

[0133] When the initial vehicle speed prediction model meets the training termination condition, the trained vehicle speed prediction model corresponding to the working condition classification is obtained, including:

[0134] Based on the predicted vehicle speed output by the initial vehicle speed prediction model and its corresponding actual vehicle speed, the model evaluation function is determined.

[0135] When the model evaluation function meets the preset conditions, the initial vehicle speed prediction model meets the training termination condition, and the vehicle speed prediction model corresponding to the completed working condition classification is obtained.

[0136] In one embodiment of the present invention, in machine learning, a model evaluation function is a metric used to measure model performance. This function can be calculated based on the model's predicted output (predicted vehicle speed) and the actual label (actual vehicle speed). When the result of the model evaluation function meets preset conditions, the model is considered to have reached a sufficient performance level, and training can be stopped. This means that the difference between the model's prediction and the actual value is small enough, or the model's prediction accuracy is high enough. Once the model evaluation function meets the preset conditions, the initial vehicle speed prediction model is considered to have completed training. At this point, the model's parameters are fixed, forming a vehicle speed prediction model that can be used in practical applications.

[0137] In summary, according to the technical solution provided by the embodiments of the present invention, during the model training process, the model performance is evaluated by comparing the vehicle speed predicted by the model with the actual vehicle speed. When this evaluation result reaches a preset performance standard, the model training ends, resulting in a vehicle speed prediction model that can be used for actual prediction. This process ensures that the model can provide accurate and reliable vehicle speed predictions in practical applications.

[0138] Furthermore, the model evaluation function formula is as follows:

[0139]

[0140] in, The model outputs the predicted vehicle speed for the next moment. This represents the actual vehicle speed at the next moment. J When the value is greater than 90%, the evaluation function meets the preset conditions, which means that the predicted vehicle speed has reached the accuracy requirement.

[0141] To explain the technical solution of this invention in detail, please refer to... Figure 5 ,like Figure 5 This is a flowchart illustrating the vehicle speed prediction method provided by the present invention.

[0142] As shown in Figure 5, the vehicle speed prediction method includes the following steps.

[0143] start:

[0144] Step 501: Collect real vehicle data.

[0145] Specifically, historical driving parameters of the vehicle are collected, which may include vehicle speed and acceleration, among other things. For example, this data can be obtained by connecting to the vehicle's CAN bus, with a typical message acquisition cycle of 100ms.

[0146] Step 502: Classification of typical working conditions.

[0147] Specifically, the vehicle state is classified according to the magnitude of the acceleration value, namely acceleration, deceleration, and constant speed (three operating condition categories), and the historical driving parameters are divided into different operating condition blocks.

[0148] Step 503, Data Denoising Processing: Wavelet Decomposition and Reconstruction.

[0149] Specifically, wavelet decomposition algorithm is used to preprocess the feature data (such as vehicle speed and acceleration) of different working condition blocks (i.e., wavelet decomposition and reconstruction) to reduce spikes and noise in the data.

[0150] Step 504: Initialize prediction model parameters: LSTM weights and biases.

[0151] Specifically, initialize the weights and bias parameters of the LSTM model (i.e., the aforementioned LSTM-based initial vehicle speed prediction model). The initial values ​​of these parameters can be set to a random normal distribution with a standard deviation of 1.

[0152] Step 505: Input data into the LSTM prediction model (i.e., the aforementioned LSTM-based initial vehicle speed prediction model).

[0153] Specifically, the data processed in step 503 above is input into the initial vehicle speed prediction model based on LSTM for iterative training.

[0154] Step 506: Calculate the predicted vehicle speed based on the LSTM input gate, forget gate, and intermediate states.

[0155] Specifically, during the training process of the model, the predicted vehicle speed is calculated based on the input gate, forget gate, and intermediate state of the initial vehicle speed prediction model based on LSTM.

[0156] Step 507: Calculate the prediction accuracy evaluation function of the model.

[0157] Specifically, a prediction accuracy evaluation function (i.e., the aforementioned model evaluation function) is set. When the prediction accuracy evaluation function meets the predetermined conditions, the model training is complete.

[0158] Step 508: Adam algorithm adaptively updates the learning rate.

[0159] Specifically, during the training process of the model, in each iteration, the gradient of the loss function (i.e., the aforementioned standard deviation function) with respect to each parameter is calculated. Using the Adam algorithm, the learning rate of each parameter is adjusted in real time through the first-order and second-order components of the gradient.

[0160] Step 509: LSTM prediction model parameter update: weights.

[0161] Specifically, the LSTM prediction model parameters (such as weights) are updated based on the gradient and learning rate obtained in step 508.

[0162] The above process enables effective prediction of commercial vehicle speed information. Based on the predicted information, the chassis braking system can anticipate the vehicle's status and implement timely and effective control during braking and starting.

[0163] The above is based on the instructions attached. Figure 1-5This invention describes the vehicle speed prediction method in detail. Wavelet decomposition is used for data preprocessing to eliminate data spikes, thereby improving data quality. This provides more accurate input for subsequent model training. Furthermore, a Long Short-Term Memory (LSTM) network is used to model the collected data. LSTM models are particularly suitable for processing and predicting time-series data, capturing the time dependence of vehicle speed changes and enabling real-time vehicle speed prediction. Further, this invention classifies vehicle states based on acceleration, such as acceleration, constant speed, and deceleration. This classification helps identify the vehicle's current driving conditions, thus allowing for customized prediction models for different operating conditions. Vehicle speed prediction models are constructed separately for each operating condition classification. This targeted modeling method improves prediction accuracy because each model is optimized for driving behavior under specific conditions. Furthermore, the LSTM model performs real-time predictions based on historical data, effectively avoiding errors and influences from physical models. Finally, the combination of adaptive backpropagation and the Adam algorithm effectively avoids gradient explosion and the problem of optimal values.

[0164] Figure 6 This is a functional block diagram of the vehicle speed prediction device provided by the present invention.

[0165] like Figure 6 As shown, the vehicle speed prediction device 600 includes: a data acquisition module 601, a classification module 602, and a predicted vehicle speed output module 603.

[0166] Furthermore, the acquisition module 601 is configured to acquire the first driving parameters of the vehicle in the current cycle.

[0167] Furthermore, the classification module 602 is configured to determine the operating condition classification corresponding to the first driving parameters based on the first driving parameters.

[0168] Furthermore, the predicted vehicle speed output module 603 is configured to input the first driving parameter, the second driving parameter of the previous cycle, and the vehicle speed error obtained according to the previous cycle into the vehicle speed prediction model corresponding to the working condition classification, and obtain the predicted vehicle speed output by the vehicle speed prediction model.

[0169] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the vehicle speed prediction methods provided by the above methods.

[0170] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0171] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0172] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A vehicle speed prediction method, characterized in that, include: Collect the vehicle's first driving parameters in the current cycle; Based on the first driving parameters, determine the operating condition classification corresponding to the first driving parameters; The first driving parameter, the second driving parameter of the previous cycle, and the vehicle speed error obtained from the previous cycle are input into the vehicle speed prediction model corresponding to the working condition classification to obtain the predicted vehicle speed output by the vehicle speed prediction model. The first driving parameters include the first vehicle speed and the first acceleration; Based on the first driving parameters, determine the operating condition classification corresponding to the first driving parameters, including: Based on the first acceleration and the acceleration threshold, the operating condition classification corresponding to the first driving parameter is determined; Among them, the acceleration threshold value is a preset one, one or more sets of acceleration thresholds, which are used to divide the acceleration into different intervals; The training method for the vehicle speed prediction model includes: Obtain the vehicle's historical driving parameters; Based on the aforementioned operating condition classification, the historical driving parameters are divided into several operating condition blocks; The first historical driving parameters of each working condition block are preprocessed to obtain the target historical driving parameters corresponding to each working condition block; For each of the aforementioned operating conditions, initialize the weights and bias parameters of the initial vehicle speed prediction model based on the Long Short-Term Memory (LSTM) network; Based on the second vehicle speed, second acceleration, and third vehicle speed in the t-th cycle and the vehicle speed prediction deviation in the (t-1)-th cycle of the target historical driving parameters, the initialized initial vehicle speed prediction model is iteratively trained. During training, the weights are updated using the backpropagation algorithm; When the initial vehicle speed prediction model meets the training termination condition, the vehicle speed prediction model corresponding to the completed working condition classification is obtained. The step of updating the weights during training based on the backpropagation algorithm includes: During training, a standard deviation function is defined, and the gradient of the standard deviation function with respect to the weights is calculated using the backpropagation algorithm. Based on the gradient, the weights of the initial vehicle speed prediction model are updated using the Adam algorithm.

2. The vehicle speed prediction method according to claim 1, characterized in that, The step of preprocessing the first historical driving parameters of each of the driving condition blocks to obtain the target historical driving parameters corresponding to each driving condition block includes: Wavelet decomposition and reconstruction are performed on the first historical driving parameters of each driving condition block to obtain the target historical driving parameters corresponding to each driving condition block.

3. The vehicle speed prediction method according to claim 1, characterized in that, The initialization of the weights and bias parameters of the initial vehicle speed prediction model based on the Long Short-Term Memory (LSTM) network includes: Based on the aforementioned operating condition classification, the weights and bias parameters of the initial vehicle speed prediction model based on the Long Short-Term Memory (LSTM) network are initialized.

4. The vehicle speed prediction method according to claim 1, characterized in that, When the initial vehicle speed prediction model meets the training termination condition, the completed vehicle speed prediction model corresponding to the working condition classification is obtained, including: Based on the predicted vehicle speed output by the initial vehicle speed prediction model and its corresponding actual vehicle speed, the model evaluation function is determined. When the model evaluation function meets the preset conditions, the initial vehicle speed prediction model meets the training termination condition, and the trained vehicle speed prediction model corresponding to the working condition classification is obtained.

5. The vehicle speed prediction method according to claim 4, characterized in that, The formula for the model evaluation function is as follows: J=(1-(|y_(act,t)-y_(pred,t) |) / y_(act,t) )×100%; Wherein, the predicted vehicle speed value at the next moment is expressed by the model output, and the actual vehicle speed value at the next moment is expressed. When the J value is greater than 90%, the evaluation function satisfies the preset condition.

6. A vehicle speed prediction device, characterized in that, include: The data acquisition module is configured to collect the vehicle's first driving parameters in the current cycle; The classification module is configured to determine the operating condition classification corresponding to the first driving parameters based on the first driving parameters; The vehicle speed prediction output module is configured to input the first driving parameter, the second driving parameter of the previous cycle, and the vehicle speed error obtained according to the previous cycle into the vehicle speed prediction model corresponding to the working condition classification, and obtain the predicted vehicle speed output by the vehicle speed prediction model. The first driving parameters include the first vehicle speed and the first acceleration; Based on the first driving parameters, determine the operating condition classification corresponding to the first driving parameters, including: Based on the first acceleration and the acceleration threshold, the operating condition classification corresponding to the first driving parameter is determined; Among them, the acceleration threshold can be one, a group or multiple preset acceleration thresholds, used to divide the acceleration into different intervals; The training method for the vehicle speed prediction model includes: Obtain the vehicle's historical driving parameters; Based on the aforementioned operating condition classification, the historical driving parameters are divided into several operating condition blocks; The first historical driving parameters of each working condition block are preprocessed to obtain the target historical driving parameters corresponding to each working condition block; For each of the aforementioned operating conditions, initialize the weights and bias parameters of the initial vehicle speed prediction model based on the Long Short-Term Memory (LSTM) network; Based on the second vehicle speed, second acceleration, and third vehicle speed in the t-th cycle and the vehicle speed prediction deviation in the (t-1)-th cycle of the target historical driving parameters, the initialized initial vehicle speed prediction model is iteratively trained. During training, the weights are updated using the backpropagation algorithm; When the initial vehicle speed prediction model meets the training termination condition, the vehicle speed prediction model corresponding to the completed working condition classification is obtained. During training, updating the weights based on the backpropagation algorithm includes: During training, a standard deviation function is defined, and the gradient of the standard deviation function with respect to the weights is calculated using the backpropagation algorithm. Based on the gradient, the weights of the initial vehicle speed prediction model are updated using the Adam algorithm.

7. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the vehicle speed prediction method as described in any one of claims 1 to 5.

Images