Knowledge and data combined driving vehicle load estimation method

By designing a bilayer neural network architecture, combining vehicle dynamics model and real-time sensor data, the problem of insufficient sensor dependence and real-time performance of load estimation methods in the existing technology is solved, and accurate and stable real-time estimation of vehicle loads is achieved, reducing the Cost and noise impacts improve robustness and applicability.

CN118940626APending Publication Date: 2024-11-12TONGJI UNIV

Patent Information

Application Number
CN202410993106.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Existing vehicle load estimation methods generally require additional sensors, which are difficult to deploy and cost in real time, are limited in scope of application, and have high requirements for data volume and characteristics, resulting in slow solution speed and poor real-time performance, which cannot meet certain real-time applications. The need for insufficient consideration of the joint driving of knowledge and data leads to poor robustness and applicability.

Method used

Design a bilayer neural network architecture, extract vehicle dynamic model features by solving differential equations and combine real-time sensor data, train neural networks using forward propagation and backpropagation algorithms to reduce loss functions, realize load estimation, and automatically update vehicle CAN data. Fusion of vehicle dynamics models and real-time sensor data to reduce the impact of data noise using deep learning technology.

Benefits of technology

It realizes accurate and stable real-time estimation of vehicle loads, reduces dependence on additional sensors, reduces costs, improves estimation accuracy and robustness, and is suitable for complex nonlinear systems to meet real-time application needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940626A_ABST
    Figure CN118940626A_ABST
Patent Text Reader

Abstract

The invention discloses a knowledge and data combined driving vehicle load estimation method. The method comprises the following steps: designing a double-layer neural network architecture, extracting vehicle dynamics model features, obtaining and processing real-time data of each sensor, training a double-layer neural network, solidifying a neural network structure and parameters, and obtaining real-time vehicle load estimation. According to the method, the vehicle-mounted CAN data are automatically updated and read in real time, the proposed method is jointly driven by knowledge and data, only the CAN data of the vehicle need to be acquired, and the real-time monitoring and precision of the vehicle load are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a vehicle load estimation method, and in particular to a vehicle load estimation method jointly driven by knowledge and data. Background Art

[0002] Vehicle load is an important parameter that affects vehicle control performance and driver experience. Accurate estimation of vehicle load is crucial for road traffic management, safety, fuel economy and maintenance. Traditional load estimation methods usually rely on dedicated sensors or equipment for direct measurement, which is not only costly and complex to operate, but also difficult to apply in real time. Knowledge model-driven methods usually use vehicle dynamics models and road slope models to estimate loads. Although they improve the theoretical accuracy of the system, they are often limited by model assumptions and affected by system complexity. When faced with complex working conditions in reality, large estimation errors are prone to occur. Data-driven methods usually collect a large amount of sensor data and use big data mining methods to extract key features. Although they have advantages in processing complex nonlinear systems, their reliance on a large amount of high-quality data and high requirements for computing resources make it difficult to ensure the accuracy and stability of the estimation results when faced with sensor data noise and real-time requirements.

[0003] The Chinese patent application numbered 201911392989.4, "A method for dynamic real-time calculation of vehicle load", establishes a vehicle dynamics model, introduces a deep neural network model and uses big data to reduce noise interference. After training on a cloud GPU cluster and deploying it on the edge, the vehicle-mounted module can calculate the vehicle load data in real time. Disadvantages: It requires a lot of cloud computing power, cannot accurately reduce the nonlinear noise between sensors, the model cannot be fine-tuned according to the data, and the accuracy of untrained models is poor.

[0004] The Chinese patent application "Vehicle Load Estimation Method, Device and Storage Medium" with application number 202210867195.4 proposes a combination of static and dynamic load estimation models to achieve accurate estimation of vehicle load without the need for additional pressure weighing sensors. Disadvantages: Tire pressure sensors and body posture sensors are still required, and a static model is required to estimate the load in advance, which limits the scope of application.

[0005] The Chinese patent application numbered 202111421199.1, "Method and device for identifying vehicles and their load distribution", obtains the actual vehicle speed and longitudinal acceleration of the vehicle to determine the angle between the longitudinal acceleration sensor assembly and the horizontal plane, thereby eliminating the need for other sensors to identify the vehicle and its load distribution. Disadvantages: Wheel speed sensors and body posture sensors are still required, and the patent is used to estimate the load distribution based on the vehicle condition when the load is known.

[0006] The Chinese patent application numbered 202210648101.4, "Vehicle Load Calculation Method, System and Aerial Work Platform", measures the gravity and support force of the tire assembly, calculates the wheel support reaction force according to the constraint conditions, and achieves accurate acquisition of vehicle load data and effective assessment of the force conditions. Disadvantages: Additional tire assemblies and force sensors are required, the vehicle load is not systematically modeled, and the algorithm is slightly less real-time.

[0007] The Chinese patent application numbered 202310024350.0, "A pure electric vehicle load estimation method based on intelligent optimization algorithm", determines the specific driving state through the dynamic balance equation of the adjacent state of the vehicle and the data analysis method, and realizes efficient load prediction by combining the vehicle information parameter adjustment, solving the complexity and convergence speed problems of the traditional method. Disadvantages: It requires accelerated or uniform speed data segments, and the real-time performance of the method is slightly weak due to the need for continuous time data. The accuracy of the method is poor due to the neglect of other resistance changes in dynamics.

[0008] The Chinese patent application "A Vehicle Load Identification System" with application number 202310677216.0 proposes a vehicle load identification system including a state monitor, a verification system and a state memory. The state monitor interacts with the vehicle's BCM, ESP, and ABS through a communication bus to monitor and analyze the static variable load, longitudinal dynamic load, lateral dynamic load, door logic, and machine learning dynamic load of the passenger car. The verification system is calibrated and proofread to finally obtain the load judgment value Load. The system can be used for driving safety and tire pressure monitoring. Disadvantages: More additional sensors are required, and the real-time performance of the regression algorithm is slightly weak.

[0009] The Chinese patent application with application number 202211179286.5, "Method and system for real-time measurement of vehicle load based on AI visual intelligent vehicle", installs an AI intelligent camera on the top of the vehicle, installs a circular or cylindrical sensor between the bottom of the spring steel plate and the axle, and installs a data integration box in the vehicle cab. The management platform pushes the data to each user terminal. Disadvantages: Additional hardware equipment such as AI visual intelligent equipment, circular or cylindrical sensors, and data integration boxes need to be installed, and the load conversion based only on the degree of spring deformation is quite limited.

[0010] In summary,

[0011] 1. Existing methods generally require the addition of additional sensors, which are difficult to deploy and have high real-time costs, and their scope of application is limited;

[0012] 2. Some methods have high requirements on data volume and data characteristics, resulting in slow solution speed and poor real-time performance, which cannot meet the needs of some real-time applications;

[0013] 3. Existing methods do not fully consider the joint driving of knowledge and data, and some methods have not considered the influence of external factors, so the robustness and applicability of the methods are poor. Summary of the invention

[0014] Purpose of the invention: The purpose of the present invention is to provide a vehicle load estimation method jointly driven by knowledge and data with high accuracy and stability, so as to realize real-time estimation of vehicle load.

[0015] Technical solution: On the one hand, the present invention provides a vehicle load estimation method driven jointly by knowledge and data.

[0016] Step 1: Design a two-layer neural network architecture;

[0017] Step 2: Construct a vehicle longitudinal dynamics model, extract the vehicle dynamics model features by solving differential equations, and use them as the input X of the first layer of the neural network. dynamics ;

[0018] Step 3: Obtain and process the CAN bus data of each sensor as the input X of the second layer of neural network sensor ;

[0019] Step 4: Calculate the output m of each sample through forward propagation i ;

[0020] Step 5: Calculate the error between the estimated load and the actual load according to the loss function;

[0021] Step 6: Calculate the gradient of the loss function with respect to each weight and bias through back propagation;

[0022] Step 7: Use the gradient descent algorithm to update the weights and biases according to the calculated gradients.

[0023] Step 8. Repeat steps 4 to 7 until the loss function When it drops to 10%, the training is terminated and the neural network structure and parameters are solidified;

[0024] Step 9: Input the real-time vehicle data into the neural network solidified in step 8, and the calculated output result is the vehicle load.

[0025] Further, the method of step 1 is as follows:

[0026] 1.1 Construct the first layer of neural network to process the output data of the vehicle dynamics model and extract higher-dimensional features, as follows:

[0027]

[0028] Where, X dynamicsis the output feature vector of the vehicle dynamics model, which contains 3 features; v is the vehicle speed; T is the engine output torque; α is the road slope angle; H1 is the output of the first layer of neural network, with a dimension of 5; f1 is the activation function, which takes the ReLU function; W1 is the weight matrix of the first layer of neural network, with a dimension of 5×3; b1 is the bias vector of the first layer of neural network, with a dimension of 5

[0029] 1.2 Construct the second layer of neural network to process real-time sensor data and extract real-time features, as follows:

[0030]

[0031] Where, X sensor is a real-time sensor data vector containing n features: x1, x2, …, x n ; H2 is the output of the second layer neural network, with a dimension of 5; f2 is the activation function, which is the ReLU function; W2 is the weight matrix of the second layer neural network, with a dimension of 5×n; b2 is the bias vector of the second layer neural network, with a dimension of 5

[0032] 1.3 Construct the output layer and estimate the load by combining the outputs H1 and H2 of the first and second layers of the neural network, as follows:

[0033]

[0034] In the formula, H concat is a 10-dimensional vector; m is the vehicle load estimation result; f out is a linear activation function; W out is the weight matrix of the output layer, with a dimension of 1×10; b out is the bias scalar of the output layer.

[0035] Furthermore, the method of step 2 is as follows:

[0036] Where δ is the rotation mass conversion factor; t is time; F traction F is the vehicle traction force; air is air resistance; F μ is the rolling resistance; F g is the gravity; i g is the main reduction ratio; i t is the transmission gear ratio; η is the transmission efficiency; r is the tire rolling radius; ρ is the air density; C is the air resistance coefficient; A is the vehicle's frontal area; μ is the rolling resistance coefficient; and g is the acceleration of gravity.

[0037] Furthermore, the method of step 3 is as follows:

[0038] 3.1 Obtain basic parameters of vehicles carrying CAN devices;

[0039] 3.2 Obtain vehicle-mounted CAN data in real time;

[0040] 3.3 Preprocess the vehicle CAN data in step 3.2 according to the time window granularity and clean the abnormal data;

[0041] 3.4 Select a single sensor and normalize the feature of this dimension, as follows:

[0042]

[0043] In the formula, x n is the normalized sensor data; n is the sensor at X sensor Corresponding dimension; x is the sensor data before normalization; x min is the minimum value of the sensor data; x max is the maximum value of the sensor data.

[0044] 3.5 Repeat step 3.4 until all sensor data are normalized.

[0045] Furthermore, the method of step 4 is as follows:

[0046] 4.1 Obtain input feature X through step 2 dynamics ;

[0047] 4.2 Calculate the output H1 of the first layer of the neural network through step 1.1;

[0048] 4.3 Obtain input feature X through step 3 sensor ;

[0049] 4.4 Calculate the output H2 of the second layer neural network through step 1.2;

[0050] 4.5 Based on the output H1 obtained in step 4.2 and the output H2 obtained in step 4.4, calculate the final load estimation result m through step 1.3 i .

[0051] Furthermore, the method of step 5 is as follows:

[0052] 5.1 Construct the loss function as follows:

[0053]

[0054] In the formula, is the mean square error MSE; N is the number of training data samples; m i is the estimated load of the ith sample; m true,i is the actual load of the ith sample;

[0055] 5.2 Calculate m using the loss function of step 5.1 i and m true,i The error between .

[0056] Furthermore, the method of step 6 is as follows:

[0057] 6.1 Calculate the gradient of the loss function relative to the output layer as follows:

[0058]

[0059] 6.2 Calculate the gradient of the loss function relative to the second layer of the neural network as follows:

[0060]

[0061] 6.3 Calculate the gradient of the loss function relative to the first layer of the neural network as follows:

[0062]

[0063] Furthermore, the method of step 7 is as follows:

[0064] 7.1 Update the output layer parameters as follows:

[0065]

[0066] In the formula, Ω is the learning rate,

[0067] 7.2 Update the parameters of the second layer neural network as follows:

[0068]

[0069] 7.3 Update the parameters of the first layer of neural network as follows:

[0070]

[0071] Compared with the prior art, the present invention has the following beneficial effects:

[0072] 1. The present invention is implemented through the following technical solutions: designing a two-layer neural network architecture, extracting vehicle dynamics model characteristics, acquiring and processing real-time data from each sensor, training a two-layer neural network, solidifying the neural network structure and parameters, and obtaining real-time vehicle load estimation.

[0073] 2. The present invention automatically updates the on-board CAN data and reads it in real time. By integrating the vehicle dynamics model with the real-time sensor data and combining it with deep learning technology, it can accurately estimate the load while reducing the impact of data noise.

[0074] 3. The method proposed in the present invention is jointly driven by knowledge and data. The knowledge-driven part uses the existing vehicle dynamics and road slope models to improve the reliability and accuracy of the estimation; the data-driven part realizes real-time monitoring and prediction of vehicle load through real-time sensor data and big data analysis.

[0075] 4. The present invention only needs to obtain the vehicle's CAN data to automatically estimate the vehicle load and road slope at the same time. No additional sensors, professional monitoring equipment and other hardware are required, and there is no need for traditional video recording, observation, and recording. This not only reduces the dependence on dedicated sensors and reduces costs, but also can provide higher estimation accuracy in complex nonlinear systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 This is the specific process of the vehicle load estimation method based on the joint drive of knowledge and data of the present invention. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be further described below.

[0078] The vehicle load estimation method driven by knowledge and data in this embodiment includes the following steps:

[0079] Step 1: Design a two-layer neural network architecture as follows:

[0080] 1.1 Construct the first layer of neural network to process the output data of the vehicle dynamics model and extract higher-dimensional features, as follows:

[0081]

[0082] Where, X dynamics is the output feature vector of the vehicle dynamics model, which contains 3 features; v is the vehicle speed; T is the engine output torque; α is the road slope angle; H1 is the output of the first layer of the neural network, with a dimension of 5; f1 is the activation function, which takes the ReLU function; W1 is the weight matrix of the first layer of the neural network, with a dimension of 5×3; b1 is the bias vector of the first layer of the neural network, with a dimension of 5.

[0083] 1.2 Construct the second layer of neural network to process real-time sensor data and extract real-time features, as follows:

[0084]

[0085] Where, X sensor is a real-time sensor data vector containing n features: x1, x2, …, x n; H2 is the output of the second layer neural network, with a dimension of 5; f2 is the activation function, which is the ReLU function; W2 is the weight matrix of the second layer neural network, with a dimension of 5×n; b2 is the bias vector of the second layer neural network, with a dimension of 5.

[0086] 1.3 Construct the output layer and estimate the load by combining the outputs H1 and H2 of the first and second layers of the neural network, as follows:

[0087]

[0088] In the formula, H concat is a 10-dimensional vector; m is the vehicle load estimation result; f out is a linear activation function; W out is the weight matrix of the output layer, with a dimension of 1×10; b out is the bias scalar of the output layer.

[0089] Step 2: Construct a vehicle longitudinal dynamics model, extract the vehicle dynamics model features by solving differential equations, and use them as the input X of the first layer of the neural network. dynamics , as follows:

[0090]

[0091] Where δ is the rotation mass conversion factor; t is time; F traction F is the vehicle traction force; air is air resistance; F μ is the rolling resistance; F g is the gravity; i g is the main reduction ratio; i t is the transmission gear ratio; η is the transmission efficiency; r is the tire rolling radius; ρ is the air density; C is the air resistance coefficient; A is the vehicle's frontal area; μ is the rolling resistance coefficient; and g is the acceleration of gravity.

[0092] Step 3: Obtain and process the CAN bus data of each sensor as the input X of the second layer of neural network sensor , as follows:

[0093] 3.1 Obtain basic parameters of vehicles carrying CAN devices.

[0094] 3.2 Obtain vehicle-mounted CAN data in real time.

[0095] 3.3 Preprocess the vehicle CAN data in step 3.2 according to the time window granularity and clean the abnormal data.

[0096] 3.4 Select a single sensor and normalize the feature of this dimension, as follows:

[0097]

[0098] In the formula, x n is the normalized sensor data; n is the sensor at X sensor Corresponding dimension; x is the sensor data before normalization; x min is the minimum value of the sensor data; x max is the maximum value of the sensor data.

[0099] 3.5 Repeat step 3.4 until all sensor data are normalized.

[0100] Step 4: Calculate the output m of each sample through forward propagation i , as follows:

[0101] 4.1 Obtain input feature X through step 2 dynmics .

[0102] 4.2 Calculate the output H1 of the first layer of the neural network through step 1.1.

[0103] 4.3 Obtain input feature X through step 3 sensor .

[0104] 4.4 Calculate the output H2 of the second layer neural network through step 1.2.

[0105] 4.5 Based on the output H1 obtained in step 4.2 and the output H2 obtained in step 4.4, calculate the final load estimation result m through step 1.3 i .

[0106] Step 5: Calculate the error between the estimated load and the actual load based on the loss function, as follows:

[0107] 5.1 Construct the loss function as follows:

[0108]

[0109] In the formula, is the mean square error MSE; N is the number of training data samples; m i is the estimated load of the ith sample; m true,i is the actual load of the ith sample.

[0110] 5.2 Calculate m using the loss function of step 5.1 i and m true,i The error between .

[0111] Step 6: Calculate the gradient of the loss function with respect to each weight and bias through back propagation, as follows:

[0112] 6.1 Calculate the gradient of the loss function relative to the output layer as follows:

[0113]

[0114] 6.2 Calculate the gradient of the loss function relative to the second layer of the neural network as follows:

[0115]

[0116] 6.3 Calculate the gradient of the loss function relative to the first layer of the neural network as follows:

[0117]

[0118] Step 7: Use the gradient descent algorithm to update the weights and biases according to the calculated gradients, as follows:

[0119] 7.1 Update the output layer parameters as follows:

[0120]

[0121] Where θ is the learning rate.

[0122] 7.2 Update the parameters of the second layer neural network as follows:

[0123]

[0124] 7.3 Update the parameters of the first layer of neural network as follows:

[0125]

[0126] Step 8. Repeat steps 4 to 7 until the loss function When the value drops to 10%, the training is terminated and the neural network structure and parameters are solidified.

[0127] Step 9: Input the real-time vehicle data into the neural network solidified in step 8, and the calculated output result is the vehicle load.

[0128] The above is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any technician in the relevant technical field, without departing from the scope of the technical solution of the present invention, makes any form of equivalent replacement or modification to the technical solution and technical content disclosed in the present invention, which does not depart from the content of the technical solution of the present invention and still falls within the protection scope of the present invention.

Claims

1. A vehicle load estimation method driven by knowledge and data, characterized in that: The steps include: Step 1: Design a two-layer neural network architecture; Step 2: Construct a vehicle longitudinal dynamics model, extract the vehicle dynamics model features by solving differential equations, and use them as the input X of the first layer of the neural network. dynamics ; Step 3: Obtain and process the CAN bus data of each sensor as the input X of the second layer of neural network sensor ; Step 4: Calculate the output m of each sample through forward propagation i ; Step 5: Calculate the error between the estimated load and the actual load according to the loss function; Step 6: Calculate the gradient of the loss function with respect to each weight and bias through back propagation; Step 7: Use the gradient descent algorithm to update the weights and biases according to the calculated gradients. Step 8. Repeat steps 4 to 7 until the loss function When it drops to 10%, the training is terminated and the neural network structure and parameters are solidified; Step 9: Input the real-time vehicle data into the neural network solidified in step 8, and the calculated output result is the vehicle load.

2. The vehicle load estimation method driven by knowledge and data according to claim 1 is characterized in that: The method for step 1 is as follows: 1.1 Construct the first layer of neural network to process the output data of the vehicle dynamics model and extract higher-dimensional features, as follows: Where, X dynamics is the output feature vector of the vehicle dynamics model, which contains 3 features; v is the vehicle speed; T is the engine output torque; α is the road slope angle; H1 is the output of the first layer of neural network, with a dimension of 5; f1 is the activation function, which takes the ReLU function; W1 is the weight matrix of the first layer of neural network, with a dimension of 5×3; b1 is the bias vector of the first layer of neural network, with a dimension of 5 1.2 Construct the second layer of neural network to process real-time sensor data and extract real-time features, as follows: Where, X sensor is a real-time sensor data vector containing n features: x1, x2, …, x n ; H2 is the output of the second layer neural network, with a dimension of 5; f2 is the activation function, which is the ReLU function; W2 is the weight matrix of the second layer neural network, with a dimension of 5×n; b2 is the bias vector of the second layer neural network, with a dimension of 5 1.3 Construct the output layer and estimate the load by combining the outputs H1 and H2 of the first and second layers of the neural network, as follows: In the formula, H concat is a 10-dimensional vector; m is the vehicle load estimation result; f out is a linear activation function; W out is the weight matrix of the output layer, with a dimension of 1×10; b out is the bias scalar of the output layer.

3. The vehicle load estimation method driven by knowledge and data according to claim 2 is characterized in that: The method for step 2 is as follows: Where δ is the rotation mass conversion factor; t is time; F traction F is the vehicle traction force; air is air resistance; F μ is the rolling resistance; F g is the gravity; i g is the main reduction ratio; i t is the transmission gear ratio; η is the transmission efficiency; r is the tire rolling radius; ρ is the air density; C is the air resistance coefficient; A is the vehicle's frontal area; μ is the rolling resistance coefficient; and g is the acceleration of gravity.

4. The vehicle load estimation method driven by knowledge and data according to claim 3 is characterized in that: The method for step 3 is as follows: 3.1 Obtain basic parameters of vehicles carrying CAN devices; 3.2 Obtain vehicle-mounted CAN data in real time; 3.3 Preprocess the vehicle CAN data in step 3.2 according to the time window granularity and clean the abnormal data; 3.4 Select a single sensor and normalize the feature of this dimension, as follows: In the formula, x n is the normalized sensor data; n is the sensor at X sensor The corresponding dimension; x is the sensor data before normalization; x min is the minimum value of the sensor data; x max is the maximum value of the sensor data, 3.5 Repeat step 3.4 until all sensor data are normalized.

5. The vehicle load estimation method driven by knowledge and data according to claim 4 is characterized in that: The method for step 4 is as follows: 4.1 Obtain input feature X through step 2 dynamics ; 4.2 Calculate the output H1 of the first layer of the neural network through step 1.1; 4.3 Obtain input feature X through step 3 sensor ; 4.4 Calculate the output H2 of the second layer neural network through step 1.2; 4.5 Based on the output H1 obtained in step 4.2 and the output H2 obtained in step 4.4, calculate the final load estimation result m through step 1.3 i .

6. The vehicle load estimation method driven by knowledge and data according to claim 5 is characterized in that: The method for step 5 is as follows: 5.1 Construct the loss function as follows: In the formula, is the mean square error MSE; N is the number of training data samples; m i is the estimated load of the ith sample; m true,i is the actual load of the ith sample; 5.2 Calculate m using the loss function of step 5.1 i and m true,i The error between .

7. The vehicle load estimation method driven by knowledge and data according to claim 6 is characterized in that: The method for step 6 is as follows: 6.1 Calculate the gradient of the loss function relative to the output layer as follows: 6.2 Calculate the gradient of the loss function relative to the second layer of the neural network as follows: 6.3 Calculate the gradient of the loss function relative to the first layer of the neural network as follows:

8. The vehicle load estimation method driven by knowledge and data according to claim 7 is characterized in that: The method for step 7 is as follows: 7.1 Update the output layer parameters as follows: In the formula, θ is the learning rate, 7.2 Update the parameters of the second layer neural network as follows: 7.3 Update the parameters of the first layer of neural network as follows:

Citation Information

Patent Citations

  • Vehicle load dynamic real-time calculation method

    CN111027146A

  • Vehicle load estimation method and device and storage medium

    CN114919593A

  • Vehicle load calculation method and system and aerial work platform

    CN115183851A

  • Intelligent vehicle-mounted real-time vehicle load measuring method and system based on AI vision

    CN115523990A

  • Pure electric vehicle load estimation method based on intelligent optimization algorithm

    CN115973173A

Cited By

  • Vehicle load measuring method and system based on multi-sensor data fusion technology

    CN119958676A