Vehicle rear wheel steering positioning method based on multi-sensor information fusion

Abstract

The invention discloses a vehicle rear wheel steering positioning method based on multi-sensor information fusion, which relates to the technical field of data encryption, and comprises the following steps: acquiring and preprocessing vehicle rear wheel data through sensors, monitoring the vehicle rear wheel data, dynamically adjusting parameter setting of the sensors according to a monitoring result, and when a vehicle generates a drift error, turning the vehicle rear wheel steering positioning method. A self-adaptive filter is used for calibration, the reading of a sensor is automatically adjusted, based on the calibrated vehicle rear wheel data, a deep learning model CNN is used for extracting a vehicle rear wheel feature vector from the calibrated vehicle rear wheel data, and based on the vehicle rear wheel feature vector, an MLP is used for constructing an information analysis fusion model. And inputting the feature vectors of the rear wheels of the vehicle into the information analysis fusion model to obtain comprehensive feature vectors, accurately positioning the positions of the rear wheels of the vehicle by using a support vector machine (SVM) based on the comprehensive feature vectors, outputting a positioning result, adjusting the steering angle of the rear wheels according to the positioning result, and feeding back the result to the control unit.

Description

Technical Field

[0001] The invention relates to the technical field of vehicle control, in particular to a vehicle rear wheel steering positioning method based on multi-sensor information fusion. Background Art

[0002] With the rapid development of intelligent transportation and autonomous driving technologies, precise control and positioning of vehicles have become an important research direction. In this context, vehicle rear-wheel steering positioning technology has attracted much attention due to its key role in vehicle handling performance and driving safety. Traditional rear-wheel steering positioning methods mainly rely on data from a single sensor, which can provide information such as vehicle acceleration, angular velocity, and speed. However, the data from a single sensor is easily affected by environmental noise and equipment drift, resulting in low positioning accuracy. In order to solve this problem, multi-sensor information fusion technology has gradually been applied to vehicle rear-wheel steering positioning. Multi-sensor information fusion technology improves positioning accuracy by combining data from multiple sensors. Common multi-sensor information fusion methods include Kalman filtering, particle filtering, and Bayesian fusion. These methods have improved positioning accuracy to a certain extent, but there are still some limitations. With the development of deep learning technology, multi-sensor information fusion methods based on deep learning have gradually become a research hotspot. Deep learning models, such as convolutional neural networks (CNNs) and multi-layer perceptrons (MLPs), can extract deep feature information from high-dimensional, nonlinear data, further improving the effect of multi-sensor information fusion.

[0003] Although multi-sensor information fusion technology has made certain progress in vehicle rear-wheel steering positioning, the existing technology still has some shortcomings. The existing data preprocessing methods are often limited to simple filtering and standardization processing, and fail to fully consider the synergistic effect between sensors. The feature extraction stage mostly uses traditional machine learning algorithms, such as support vector machines (SVMs) and decision trees, which perform poorly when processing high-dimensional and nonlinear data. The information fusion stage usually adopts a simple weighted average or voting mechanism, which fails to fully utilize the powerful learning ability of the deep learning model, resulting in unsatisfactory fusion effect. The existing technology has deficiencies in dynamically adjusting sensor parameters and cannot dynamically adjust the sensor parameter settings according to real-time monitoring results, thereby affecting the adaptability of the system. These problems limit the application effect of multi-sensor information fusion technology in vehicle rear-wheel steering positioning, and new solutions are needed to overcome these deficiencies. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a vehicle rear wheel steering positioning method based on multi-sensor information fusion to solve the problem that a simple weighted average or voting mechanism is usually adopted in the information fusion stage, which fails to fully utilize the powerful learning ability of the deep learning model, resulting in unsatisfactory fusion effect.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a vehicle rear wheel steering positioning method using multi-sensor information fusion, which comprises:

[0008] Collect vehicle rear wheel data through sensors and perform preprocessing;

[0009] Monitor the vehicle's rear wheel data and dynamically adjust the sensor's parameter settings based on the monitoring results. When the vehicle has a drift error, use an adaptive filter for calibration to automatically adjust the sensor readings.

[0010] Based on the calibrated vehicle rear wheel data, a deep learning model CNN is used to extract the vehicle rear wheel feature vector from the calibrated vehicle rear wheel data;

[0011] Based on the feature vector of the rear wheel of the vehicle, a multi-layer perceptron (MLP) is used to build an information analysis and fusion model. The feature vector of the rear wheel of the vehicle is input into the information analysis and fusion model to obtain a comprehensive feature vector.

[0012] Based on the comprehensive feature vector, the support vector machine (SVM) is used to accurately locate the position of the rear wheel of the vehicle and output the positioning result;

[0013] The rear wheel steering angle is adjusted according to the positioning results and the results are fed back to the control unit.

[0014] As a preferred solution of the vehicle rear wheel steering positioning method of multi-sensor information fusion of the present invention, wherein: the vehicle rear wheel data is collected by the sensor and pre-processed, and the specific steps are:

[0015] The acceleration, angular velocity, speed and image data of the vehicle's rear wheels are collected through MEMS accelerometers, gyroscopes, magnetoresistive sensors and cameras;

[0016] Use a bandpass filter to remove random noise in the data, use the moving average method to reduce data fluctuations, and use the Z-score for standardization;

[0017] The preprocessed data set is set to x={x1, x2, x3, x4}, where x1 represents acceleration data, x2 represents angular velocity data, x3 represents speed data, and x4 represents rear wheel position image data.

[0018] As a preferred embodiment of the vehicle rear-wheel steering positioning method for multi-sensor information fusion according to the present invention, the steps are as follows: monitoring the data of the vehicle rear wheels, and dynamically adjusting the parameter settings of the sensors according to the monitoring results. The specific steps are as follows:

[0019] Use the Kalman Filter to detect the drift error of the real-time data. The expression is:

[0020]

[0021] where e(t) represents the drift error at time t, represents the ideal value at time t, and x(t) represents the real value at time t;

[0022] The value range of e(t) is [-1, 1]. When e(t) = 0, it means that the vehicle has no drift error, and continue to observe;

[0023] When e(t) < 0, it means that the vehicle has a left drift error, and trigger the adaptive filter to perform a dynamic adjustment to the right of the sensor reading;

[0024] When e(t) > 0, it means that the vehicle has a right drift error, and trigger the adaptive filter to perform a dynamic adjustment to the left of the sensor reading.

[0025] As a preferred embodiment of the vehicle rear-wheel steering positioning method for multi-sensor information fusion according to the present invention, the steps are as follows: when the vehicle has a drift error, use the adaptive filter for calibration and automatically adjust the sensor reading. The specific steps are as follows:

[0026] Use the adaptive filter to calibrate the sensor reading to eliminate the drift error. The expression is:

[0027] x cal (t) = x(t) - λ·e(t);

[0028] where x cal (t) represents the calibrated sensor reading at time t, x(t) represents the original sensor reading at time t, and λ is a scaling factor used to adjust the intensity of the calibration;

[0029] Set the thresholds T1 and T2. When 0 < e(t) ≤ T1, it means that the right drift error is small, and set a small calibration scaling factor λ for fine-tuning to the left;

[0030] When T1 < e(t) ≤ 1, it means that the right drift error is large, and set a large calibration scaling factor λ for strong calibration to the left to quickly correct the sensor reading;

[0031] When \(T2\leq e(t)\lt0\), it indicates that the drift error to the left is small, and a small calibration scale factor \(\lambda\) is set to fine-tune to the right;

[0032] When \(-1\leq e(t)\lt T2\), it indicates that the drift error to the left is large, and a large calibration scale factor \(\lambda\) is set to strongly calibrate to the right to quickly correct the sensor reading.

[0033] As a preferred solution of the vehicle rear-wheel steering positioning method for multi-sensor information fusion according to the present invention, wherein: based on the calibrated vehicle rear-wheel data, the deep learning model CNN is used to extract the vehicle rear-wheel feature vector from the calibrated vehicle rear-wheel data. The specific steps are as follows:

[0034] First, determine a composite function \(f\) for feature extraction that includes convolution, pooling, and activation operations CNN , and the expression is:

[0035] f CNN (x) = \(\sigma(W\) conv *x + b conv );

[0036] Wherein, * represents the convolution operation, \(W\) conv represents the convolution kernel weight, \(b\) conv represents the bias term, and \(\sigma\) represents the activation function;

[0037] Using the deep learning model CNN to extract the vehicle rear-wheel feature vector from the calibrated vehicle rear-wheel data, the expression is:

[0038] F i = f CNN (x cal (t));

[0039] Wherein, \(F\) i represents the \(i\)th vehicle rear-wheel feature vector, and the value range of \(i\) is \(\{1, 2, 3, 4\}\).

[0040] As a preferred solution of the vehicle rear-wheel steering positioning method for multi-sensor information fusion according to the present invention, wherein: based on the vehicle rear-wheel feature vector, a multi-layer perceptron MLP is used to construct an information analysis and fusion model, and the vehicle rear-wheel feature vector is input into the information analysis and fusion model to obtain a comprehensive feature vector. The specific steps are as follows:

[0041] Set the linear transformation of the first layer to perform feature fusion on the vehicle rear-wheel feature vector, and the expression is:

[0042] z1 = \(W1\cdot F\) i + b1;

[0043] Among them, W1 represents the weight matrix, b1 represents the bias term, and z1 represents the output after linear transformation of the first fully connected layer;

[0044] Select ReLU as the activation function and apply the activation function to the linearly transformed output z1. The expression is:

[0045] a1=ReLU(z1);

[0046] Among them, a1 represents the output after the nonlinear transformation of the first fully connected layer;

[0047] After the nonlinear transformation of the first fully connected layer, a1 is used as the input of the second layer and continues to be linearly transformed. The expression is:

[0048] z2=W2·a1+b2;

[0049] Among them, W2 represents the weight matrix, b2 represents the bias term, and z2 represents the output of the second layer;

[0050] Apply the ReLU activation function to the output of the second layer:

[0051] a2=ReLU(z2);

[0052] Set the number of iterations. After multiple layers of feature extraction, the output of the last layer is used as the comprehensive feature vector F, which is expressed as:

[0053] F=σ(W F [F1,F2,F3,F4]+b F );

[0054] Among them, [F1, F2, F3, F4] represents the comprehensive feature vector F, F i The value range of i in is {1,2,3,4}, corresponding to [F1,F2,F3,F4] respectively, and σ represents the activation function ReLU.

[0055] As a preferred solution of the vehicle rear wheel steering positioning method of multi-sensor information fusion of the present invention, wherein: based on the comprehensive feature vector, the support vector machine SVM is used to accurately locate the position of the vehicle rear wheel, and the specific steps are:

[0056] Select RBF as the kernel function K(F a ,F b ), the expression is:

[0057] K(F a ,F b )=exp(―γ∥F i ―F∥ 2 );

[0058] Among them, exp represents the exponential function, and γ represents a positive constant;

[0059] Based on the comprehensive feature vector F, the comprehensive feature vector F is input into the information analysis fusion model, and the support vector machine SVM is used to accurately locate the position of the rear wheel of the vehicle. The expression is:

[0060]

[0061] in, represents the predicted rear wheel angle of the vehicle, α i represents the Lagrange multiplier, m represents the number of training samples, y i represents the label of the training sample, and b represents the bias term.

[0062] As a preferred solution of the vehicle rear wheel steering positioning method of multi-sensor information fusion of the present invention, wherein: the output positioning result is fed back to the control unit to adjust the rear wheel steering angle, and the specific steps are:

[0063] First, calculate and adjust the rear wheel steering angle θ, the expression is:

[0064]

[0065] Among them, k p represents the proportional gain, y desired Indicates the desired angle;

[0066] Then, the calculated rear wheel steering angle θ is sent to the vehicle control unit to perform steering adjustment;

[0067] When θ>0, it means that the rear wheel of the vehicle is offset to the left and adjusted to the right;

[0068] When θ = 0, it means that the rear wheels of the vehicle have not deviated and no adjustment is required;

[0069] When θ<0, it means that the rear wheel of the vehicle is offset to the right and is adjusted to the left.

[0070] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the vehicle rear wheel steering positioning method based on multi-sensor information fusion as described in the first aspect of the present invention.

[0071] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the vehicle rear wheel steering positioning method using multi-sensor information fusion as described in the first aspect of the present invention.

[0072] The beneficial effects of the present invention are as follows: by collecting vehicle rear wheel data through sensors and performing preprocessing, the quality and stability of the data are improved; by monitoring the vehicle rear wheel data and dynamically adjusting the sensor parameters, the cumulative error is reduced, the response speed and robustness of the system are improved; by extracting the vehicle rear wheel feature vector through the deep learning model CNN, the feature expression ability is enhanced, and the accuracy and stability of the vehicle rear wheel steering positioning are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0074] Figure 1 This is a flow chart of the vehicle rear wheel steering positioning method using multi-sensor information fusion in Example 1.

[0075] Figure 2 This is a schematic diagram of adjusting the rear wheel steering angle in Example 1. DETAILED DESCRIPTION

[0076] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0077] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0078] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0079] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a vehicle rear wheel steering positioning method based on multi-sensor information fusion, comprising the following steps:

[0080] S1 collects the vehicle rear wheel data through sensors and performs pre-processing;

[0081] The acceleration, angular velocity, speed and image data of the vehicle's rear wheels are collected through MEMS accelerometers, gyroscopes, magnetoresistive sensors and cameras;

[0082] Use a bandpass filter to remove random noise in the data, use the moving average method to reduce data fluctuations, and use the Z-score for standardization;

[0083] The preprocessed data set is set to x={x1, x2, x3, x4}, where x1 represents acceleration data, x2 represents angular velocity data, x3 represents speed data, and x4 represents rear wheel position image data.

[0084] S2 monitors the rear wheel data of the vehicle and dynamically adjusts the sensor parameter settings based on the monitoring results. When the vehicle has drift errors, it uses an adaptive filter for calibration and automatically adjusts the sensor readings.

[0085] Compared with traditional fixed parameter filtering, the adaptive filter can be dynamically adjusted according to the actual error, which means it can more effectively deal with errors under different conditions and ensure the stability and reliability of the system;

[0086] For example, when a vehicle encounters uncertain factors such as slippery roads or tire wear during driving, the adaptive filter can quickly respond to these changes and maintain high positioning accuracy;

[0087] Use Kalman Filter to detect drift error of real-time data. The expression is:

[0088]

[0089] Where e(t) represents the drift error at time t, represents the ideal value at time t, and x(t) represents the true value at time t;

[0090] The value range of e(t) is [-1,1]. When e(t) = 0, it means that the vehicle has no drift error and continues to observe;

[0091] When e(t)<0, it means that the vehicle has a drift error to the left, triggering the adaptive filter to dynamically adjust the sensor reading to the right;

[0092] When e(t)>0, it means that the vehicle has a drift error to the right, triggering the adaptive filter to dynamically adjust the sensor reading to the left;

[0093] The sensor readings are calibrated using an adaptive filter to remove drift errors, expressed as:

[0094] x cal(t) = x(t) - λ·e(t);

[0095] where x cal (t) represents the calibrated sensor reading at time t, x(t) represents the original sensor reading at time t, and λ is a scaling factor used to adjust the strength of the calibration;

[0096] Set threshold values T1 and T2. When 0 < e(t) ≤ T1, it indicates a small rightward drift error, and a small calibration scaling factor λ is set for fine-tuning to the left;

[0097] When T1 < e(t) ≤ 1, it indicates a large rightward drift error, and a large calibration scaling factor λ is set for strong calibration to the left to quickly correct the sensor reading;

[0098] When T2 ≤ e(t) < 0, it indicates a small leftward drift error, and a small calibration scaling factor λ is set for fine-tuning to the right;

[0099] When -1 ≤ e(t) < T2, it indicates a large leftward drift error, and a large calibration scaling factor λ is set for strong calibration to the right to quickly correct the sensor reading.

[0100] S3 Based on the calibrated vehicle rear-wheel data, use the deep learning model CNN to extract the vehicle rear-wheel feature vector from the calibrated vehicle rear-wheel data;

[0101] By using the CNN model to extract the feature vector of the vehicle rear wheel, it can not only process image data but also combine the data provided by other sensors to achieve the fusion of multi-modal information;

[0102] First, determine a composite function f for feature extraction that includes convolution, pooling, and activation operations CNN , and the expression is:

[0103] f CNN (x) = σ(W conv * x + b conv );

[0104] where * represents the convolution operation, W conv represents the convolution kernel weight, b conv represents the bias term, and σ represents the activation function;

[0105] Use the deep learning model CNN to extract the vehicle rear-wheel feature vector from the calibrated vehicle rear-wheel data, and the expression is:

[0106] F i = f CNN (x cal (t));

[0107] where Fi represents the eigenvector of the rear wheel of the i-th vehicle, and the value range of i is {1,2,3,4};

[0108] Compared with single sensor data processing, this method can provide more comprehensive and accurate vehicle status information, which is conducive to improving the accuracy and robustness of positioning.

[0109] S4 uses a multi-layer perceptron MLP to build an information analysis fusion model based on the rear wheel feature vector of the vehicle, and inputs the rear wheel feature vector of the vehicle into the information analysis fusion model to obtain a comprehensive feature vector;

[0110] The MLP model can learn the nonlinear relationship between input data through a multi-layer neural network structure and fuse feature vectors from different sources into a comprehensive feature vector;

[0111] Set the linear transformation of the first layer and perform feature fusion on the feature vector of the rear wheel of the vehicle. The expression is:

[0112] z1=W1·F i +b1;

[0113] Among them, W1 represents the weight matrix, b1 represents the bias term, and z1 represents the output after linear transformation of the first fully connected layer;

[0114] Select ReLU as the activation function and apply the activation function to the linearly transformed output z1. The expression is:

[0115] a1=ReLU(z1);

[0116] Among them, a1 represents the output after the nonlinear transformation of the first fully connected layer;

[0117] After the nonlinear transformation of the first fully connected layer, a1 is used as the input of the second layer and continues to be linearly transformed. The expression is:

[0118] z2=W2·a1+b2;

[0119] Among them, W2 represents the weight matrix, b2 represents the bias term, and z2 represents the output of the second layer;

[0120] Apply the ReLU activation function to the output of the second layer:

[0121] a2=ReLU(z2);

[0122] Set the number of iterations. After multiple layers of feature extraction, the output of the last layer is used as the comprehensive feature vector F, which is expressed as:

[0123] F=σ(W F [F1,F2,F3,F4]+bF );

[0124] Among them, [F1, F2, F3, F4] represents the comprehensive feature vector F, F i The value range of i in is {1,2,3,4}, corresponding to [F1,F2,F3,F4] respectively, and σ represents the activation function ReLU;

[0125] This process can maximize the use of all available information, reduce information loss, and thus improve the accuracy of the final positioning.

[0126] S5 uses the support vector machine (SVM) based on the comprehensive feature vector to accurately locate the position of the rear wheel of the vehicle and output the positioning result;

[0127] SVM is particularly suitable for classification and regression tasks in small sample situations;

[0128] Select RBF as the kernel function K(F a ,F b ), the expression is:

[0129] K(F a ,F b )=exp(―γ∥F i ―F∥ 2 );

[0130] Among them, exp represents the exponential function, and γ represents a positive constant;

[0131] By selecting an appropriate kernel function, SVM can handle nonlinear problems in high-dimensional space well, so that high-precision positioning can be achieved even in complex environments;

[0132] Based on the comprehensive feature vector F, the comprehensive feature vector F is input into the information analysis fusion model, and the support vector machine SVM is used to accurately locate the position of the rear wheel of the vehicle. The expression is:

[0133]

[0134] in, represents the predicted rear wheel angle of the vehicle, α i represents the Lagrange multiplier, m represents the number of training samples, y i represents the label of the training sample, and b represents the bias term.

[0135] S6 adjusts the rear wheel steering angle according to the positioning results and feeds the results back to the control unit.

[0136] First, calculate and adjust the rear wheel steering angle θ, the expression is:

[0137]

[0138] Among them, k p represents the proportional gain, y desired Indicates the desired angle;

[0139] Then, the calculated rear wheel steering angle θ is sent to the vehicle control unit to perform steering adjustment;

[0140] When θ>0, it means that the rear wheel of the vehicle is offset to the left and adjusted to the right;

[0141] When θ = 0, it means that the rear wheels of the vehicle have not deviated and no adjustment is required;

[0142] When θ<0, it means that the rear wheel of the vehicle is offset to the right and is adjusted to the left;

[0143] The rear wheel steering angle is dynamically adjusted according to the positioning results. This mechanism can significantly improve the vehicle's handling performance, especially in high-speed cornering or emergency avoidance situations, which can effectively reduce the risk of skidding and improve driving safety.

[0144] This embodiment also provides a computer device, which is suitable for the vehicle rear wheel steering positioning method of multi-sensor information fusion, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the vehicle rear wheel steering positioning method of multi-sensor information fusion as proposed in the above embodiment.

[0145] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0146] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the vehicle rear wheel steering positioning method for realizing multi-sensor information fusion as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0147] In summary, the present invention improves the quality and stability of the data by collecting the vehicle rear wheel data through sensors and performing preprocessing, reduces the cumulative error and improves the response speed and robustness of the system by monitoring the vehicle rear wheel data and dynamically adjusting the sensor parameters, extracts the vehicle rear wheel feature vector through the deep learning model CNN, enhances the feature expression ability, and significantly improves the accuracy and stability of the vehicle rear wheel steering positioning.

[0148] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of a vehicle rear wheel steering positioning method based on multi-sensor information fusion is provided.

[0149] MEMS accelerometers, gyroscopes, magnetoresistive sensors and cameras are used to collect acceleration, angular velocity, speed and image data of the rear wheels of the vehicle. Random noise in the data is removed by a bandpass filter, and the moving average method is used to reduce data fluctuations. Z-score is used for standardization to obtain the preprocessed data set.

[0150] Use Kalman filter to detect drift error in real-time data, calculate drift error, and calibrate sensor readings through adaptive filter;

[0151] A deep learning model CNN is used to extract the vehicle rear wheel feature vector from the calibrated vehicle rear wheel data. Based on the vehicle rear wheel feature vector, a multi-layer perceptron MLP is used to build an information analysis fusion model. The vehicle rear wheel feature vector is input into the information analysis fusion model to obtain a comprehensive feature vector.

[0152] The support vector machine (SVM) is used to accurately locate the position of the rear wheels of the vehicle, output the positioning results, calculate and adjust the rear wheel steering angle based on the positioning results, and feed the results back to the control unit to perform steering adjustment.

[0153] The details are shown in Table 1 below:

[0154] Table 1 Vehicle rear wheel steering positioning experiment record

[0155]

[0156] The experimental results show that the drift errors of vehicles A to F are all between 0.03 and 0.07 meters, which is much lower than the drift errors in traditional methods (usually above 0.1 meters). This shows that the combination of Kalman filter and adaptive filter can effectively detect and calibrate the drift errors of vehicles, improving the stability and reliability of the system.

[0157] The vehicle rear wheel feature vector extracted by the deep learning model CNN, combined with the information analysis fusion model of the multi-layer perceptron MLP, can effectively fuse multiple sensor data and extract more representative comprehensive feature vectors, which not only improves the accuracy of feature extraction, but also enhances the robustness of the system and can maintain high performance in complex environments;

[0158] The support vector machine (SVM) performs particularly well in positioning accuracy. Experimental data show that the positioning accuracy of all vehicles is between 0.01 and 0.03 meters, which is much higher than the positioning accuracy of the prior art (usually around 0.1 meters). This shows that the content of the present invention has significant innovation and superiority in vehicle rear wheel positioning, and can effectively improve the vehicle's handling performance and safety.

[0159] In the experiment, the system was able to dynamically adjust the sensor parameters according to the drift error monitored in real time, ensuring the real-time and flexibility of the system. This is particularly important for the stable driving of the vehicle under complex road conditions, and can correct the deviation in time to improve the driving experience;

[0160] In summary, the present invention realizes efficient and accurate positioning of the rear wheels of a vehicle through technical means such as multi-sensor information fusion, deep learning models, adaptive filters and support vector machines. The experimental data fully proves the significant advantages of the present invention in improving vehicle handling performance and safety, and has important application value and market prospects.

[0161] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A vehicle rear wheel steering positioning method based on multi-sensor information fusion, characterized in that: including collecting and preprocessing the data of the vehicle's rear wheels through sensors monitoring the data of the vehicle's rear wheels, dynamically adjusting the parameter settings of the sensors according to the monitoring results, when the vehicle has a drift error, using an adaptive filter for calibration and automatically adjusting the sensor readings extracting the vehicle's rear wheel feature vectors from the calibrated vehicle's rear wheel data by using the deep learning model CNN based on the calibrated vehicle's rear wheel data constructing an information analysis and fusion model by using a multi-layer perceptron MLP based on the vehicle's rear wheel feature vectors, inputting the vehicle's rear wheel feature vectors into the information analysis and fusion model to obtain a comprehensive feature vector accurately positioning the position of the vehicle's rear wheels by using a support vector machine SVM based on the comprehensive feature vector and outputting the positioning result adjusting the steering angle of the rear wheels according to the positioning result and feeding the result back to the control unit 2. The vehicle rear wheel steering positioning method based on multi-sensor information fusion as claimed in claim 1, characterized in that: The specific steps of collecting and preprocessing the data of the vehicle's rear wheels through sensors are as follows collecting the acceleration, angular velocity, speed and image data of the vehicle's rear wheels through a MEMS accelerometer, gyroscope, magnetoresistive sensor and camera using a band-pass filter to remove the random noise in the data, using a moving average method to reduce the data fluctuation, and using Z-score for standardization processing setting the preprocessed data set as x = {x1, x2, x3, x4}, where x1 represents the acceleration data, x2 represents the angular velocity data, x3 represents the speed data, and x4 represents the rear wheel position image data 3. The vehicle rear wheel steering positioning method based on multi-sensor information fusion as claimed in claim 2, characterized in that: The specific steps of monitoring the data of the vehicle's rear wheels and dynamically adjusting the parameter settings of the sensors according to the monitoring results are as follows using a Kalman Filter to detect the drift error of the real-time data, and the expression is Where e(t) represents the drift error at time t, represents the ideal value at time t, and x(t) represents the true value at time t; the value range of e(t) is [-1, 1]. When e(t) = 0, it means that the vehicle has no drift error and continues to observe when e(t) < 0, it means that the vehicle has a left drift error, and triggers the adaptive filter to perform a right dynamic adjustment on the sensor readings when e(t) > 0, it means that the vehicle has a right drift error, and triggers the adaptive filter to perform a left dynamic adjustment on the sensor readings 4. The vehicle rear wheel steering positioning method based on multi-sensor information fusion as claimed in claim 3, characterized in that: The specific steps of using an adaptive filter for calibration and automatically adjusting the sensor readings when the vehicle has a drift error are as follows using an adaptive filter to calibrate the sensor readings to eliminate the drift error, and the expression is x cal (t)=x(t)―λ·e(t); Among them, x cal (t) represents the calibrated sensor reading at time t, x(t) represents the raw sensor reading at time t, and λ is a scaling factor used to adjust the strength of the calibration; setting thresholds T1 and T2. When 0 < e(t) ≤ T1, it means that the right drift error is small, and setting a small calibration scale factor λ for fine-tuning to the left when T1 < e(t) ≤ 1, it means that the right drift error is large, and setting a large calibration scale factor λ for strong calibration to the left to quickly correct the sensor readings when T2 ≤ e(t) < 0, it means that the left drift error is small, and setting a small calibration scale factor λ for fine-tuning to the right when -1 ≤ e(t) < T2, it means that the left drift error is large, and setting a large calibration scale factor λ for strong calibration to the right to quickly correct the sensor readings 5. The vehicle rear wheel steering positioning method based on multi-sensor information fusion as claimed in claim 4, characterized in that: Based on the calibrated vehicle rear wheel data, the deep learning model CNN is used to extract the vehicle rear wheel feature vector from the calibrated vehicle rear wheel data. The specific steps are: First, determine a composite function f for feature extraction that includes convolution, pooling, and activation operations CNN , the expression is: f CNN (x)=σ(W conv *x+b conv ); Among them, * represents the convolution operation, W conv represents the convolution kernel weight, b conv represents the bias term, σ represents the activation function; The deep learning model CNN is used to extract the vehicle rear wheel feature vector from the calibrated vehicle rear wheel data, and the expression is: F i =f CNN (x cal (t)); Among them, F i Represents the eigenvector of the rear wheel of the i-th vehicle, and the value range of i is {1,2,3,4}.

6. The vehicle rear wheel steering positioning method based on multi-sensor information fusion as claimed in claim 5, characterized in that: The method uses a multi-layer perceptron MLP to construct an information analysis fusion model based on the rear wheel feature vector of the vehicle, inputs the rear wheel feature vector of the vehicle into the information analysis fusion model, and obtains a comprehensive feature vector. The specific steps are: Set the linear transformation of the first layer and perform feature fusion on the feature vector of the rear wheel of the vehicle. The expression is: z1=W1·F i +b1; Among them, W1 represents the weight matrix, b1 represents the bias term, and z1 represents the output after linear transformation of the first fully connected layer; Select ReLU as the activation function and apply the activation function to the linearly transformed output z1. The expression is: a1=ReLU(z1); Among them, a1 represents the output after the nonlinear transformation of the first fully connected layer; After the nonlinear transformation of the first fully connected layer, a1 is used as the input of the second layer and continues to be linearly transformed. The expression is: z2=W2·a1+b2; Among them, W2 represents the weight matrix, b2 represents the bias term, and z2 represents the output of the second layer; Apply the ReLU activation function to the output of the second layer: a2=ReLU(z2); Set the number of iterations. After multiple layers of feature extraction, the output of the last layer is used as the comprehensive feature vector F, which is expressed as: F=σ(W F ·[F1,F2,F3,F4]+b F ); Among them, [F1, F2, F3, F4] represents the comprehensive feature vector F, F i The value range of i in is {1,2,3,4}, corresponding to [F1,F2,F3,F4] respectively, and σ represents the activation function ReLU.

7. The vehicle rear wheel steering positioning method based on multi-sensor information fusion as claimed in claim 6, characterized in that: Based on the comprehensive feature vector, the support vector machine (SVM) is used to accurately locate the position of the rear wheel of the vehicle. The specific steps are as follows: Select RBF as the kernel function K(F a ,F b ), the expression is: K(F a ,F b )=exp(―γ∥F i ―F∥ 2 ); Among them, exp represents the exponential function, and γ represents a positive constant; Based on the comprehensive feature vector F, the comprehensive feature vector F is input into the information analysis fusion model, and the support vector machine SVM is used to accurately locate the position of the rear wheel of the vehicle. The expression is: in, represents the predicted rear wheel angle of the vehicle, α i represents the Lagrange multiplier, m represents the number of training samples, y i represents the label of the training sample, and b represents the bias term.

8. The vehicle rear wheel steering positioning method based on multi-sensor information fusion as claimed in claim 7, characterized in that: The output positioning result is fed back to the control unit to adjust the rear wheel steering angle. The specific steps are: First, calculate and adjust the rear wheel steering angle θ, the expression is: Among them, k p represents the proportional gain, y desired Indicates the desired angle; Then, the calculated rear wheel steering angle θ is sent to the vehicle control unit to perform steering adjustment; When θ>0, it means that the rear wheel of the vehicle is offset to the left and adjusted to the right; When θ = 0, it means that the rear wheels of the vehicle have not deviated and no adjustment is required; When θ<0, it means that the rear wheel of the vehicle is offset to the right and is adjusted to the left.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the vehicle rear wheel steering positioning method based on multi-sensor information fusion according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the vehicle rear wheel steering positioning method based on multi-sensor information fusion according to any one of claims 1 to 8 are implemented.

Images