Vehicle control method and device and vehicle

Abstract

The invention relates to the field of vehicle control, and provides a vehicle control method and device and a vehicle, an echo signal corresponding to an obstacle detection signal is received, and the moving speed of an obstacle is determined according to the echo signal; performing principal component analysis processing on the echo signal to obtain a target principal component corresponding to the echo signal; inputting the target principal component into an obstacle classification model obtained through pre-training, and outputting an obstacle type through processing of the obstacle classification model; and controlling the vehicle to run according to the moving speed of the obstacle and the type of the obstacle to avoid the obstacle. According to the determined moving speed of the obstacle and the type of the obstacle, accurate sensing and recognition of the obstacle are achieved, then the vehicle is controlled to run, the obstacle is avoided, and the running safety of the vehicle is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of vehicle control, and in particular to a vehicle control method, device and vehicle. Background Art

[0002] With the rapid development of vehicle technology, vehicles have become an important means of transportation in people's daily lives. In order to facilitate users' travel and improve their experience, driverless technology has emerged and gradually matured.

[0003] For autonomous driving technology, accurate perception of the surrounding environment has become the key to ensuring the safety and reliability of autonomous driving. Due to the influence of factors such as ambient light and weather, the recognition accuracy of the environment is low, which in turn poses a safety hazard to autonomous driving. Summary of the invention

[0004] In view of this, the purpose of the present disclosure is to propose a vehicle control method, device and vehicle to solve the current problem that the recognition accuracy of the environment is low due to the influence of factors such as ambient light and weather, thereby causing safety hazards in unmanned driving.

[0005] Based on the above purpose, a first aspect of the present disclosure provides a vehicle control method, the method comprising:

[0006] receiving an echo signal corresponding to the obstacle detection signal, and determining a moving speed of the obstacle according to the echo signal;

[0007] Performing principal component analysis on the echo signal to obtain a target principal component corresponding to the echo signal;

[0008] Inputting the target principal component into a pre-trained obstacle classification model, processing the obstacle classification model, and outputting the obstacle type;

[0009] The vehicle is controlled to travel according to the moving speed of the obstacle and the type of the obstacle to avoid the obstacle.

[0010] Specifically, performing principal component analysis on the echo signal to obtain a target principal component corresponding to the echo signal includes:

[0011] Performing analog-to-digital conversion on the echo signal to obtain a digital signal corresponding to the echo signal;

[0012] Determine a correlation coefficient matrix corresponding to the digital signal, and determine a projection matrix according to the correlation coefficient matrix;

[0013] The projection matrix is ​​used to perform dimensionality reduction processing on the digital signal to obtain the target principal component corresponding to the echo signal.

[0014] Specifically, determining the correlation coefficient matrix corresponding to the digital signal includes:

[0015] Determine an initial signal data matrix corresponding to the digital signal;

[0016] Performing data preprocessing on the initial signal data matrix to obtain a signal data matrix;

[0017] A correlation calculation is performed on the signal data matrix to obtain a correlation coefficient matrix corresponding to the digital signal.

[0018] Specifically, determining the projection matrix according to the correlation coefficient matrix includes:

[0019] Determine a plurality of eigenvalues ​​corresponding to the correlation coefficient matrix and an eigenvector corresponding to each eigenvalue;

[0020] Sort all eigenvalues ​​in descending order;

[0021] Obtain a preset dimension, select the eigenvalue of the preset dimension in sequence as the target eigenvalue, and use the eigenvector corresponding to each target eigenvalue as the target eigenvector;

[0022] Construct a projection matrix based on all target eigenvectors.

[0023] Through the above scheme, the echo signal is converted into digital form to ensure accurate conversion and transmission of the signal. At the same time, by determining the projection matrix of the digital signal corresponding to the echo signal, the projection matrix is ​​used to reduce the dimension of the digital signal, thereby reducing the dimension of the digital signal and obtaining the key feature information of the obstacle, namely the target principal component. The data subsequently input into the obstacle classification model is more concise and clear, reducing the interference of irrelevant features and improving the classification accuracy of the subsequent obstacle classification model.

[0024] Specifically, determining the moving speed of the obstacle according to the echo signal includes:

[0025] Performing analog-to-digital conversion on the echo signal to obtain a digital signal corresponding to the echo signal;

[0026] Performing windowing processing on the digital signal to obtain a signal to be processed;

[0027] Performing Fourier transform on the signal to be processed to extract Doppler frequency shift data;

[0028] The wavelength and center frequency of the obstacle detection signal are obtained, and the moving speed of the obstacle is determined according to the wavelength, the center frequency and the Doppler frequency shift data.

[0029] Through the above scheme, the spectrum leakage is reduced by windowing the digital signal. At the same time, according to the parallel Fourier transform algorithm, compared with the sequential Fourier transform calculation, the extraction accuracy and calculation efficiency of Doppler frequency shift data are improved, providing a reliable data basis for the subsequent calculation of the moving speed of obstacles.

[0030] Specifically, controlling the vehicle to travel according to the moving speed of the obstacle and the type of the obstacle to avoid the obstacle includes:

[0031] Acquire a time delay corresponding to the echo signal, and determine a relative distance between the obstacle and the vehicle according to the time delay;

[0032] Determining the angle between the obstacle detection signal and the obstacle movement direction according to the Doppler frequency shift data;

[0033] The target driving state of the vehicle is determined according to the moving speed of the obstacle, the type of obstacle, the relative distance and the angle, and the vehicle is controlled to avoid the obstacle according to the target driving state.

[0034] Specifically, the training process of the obstacle classification model includes:

[0035] Acquire a preset principal component data set and an initial obstacle classification model, wherein the preset principal component data set includes a training principal component data set, and the training principal component data set includes training principal component data and a first actual obstacle type;

[0036] Using the training principal component data set to train the initial obstacle classification model to obtain training obstacle types;

[0037] Determining a regularization objective function according to the training obstacle type and the first actual obstacle type;

[0038] Until the regularized objective function is minimized, it is determined that the training of the initial obstacle classification model is completed, and an obstacle classification model is obtained.

[0039] Through the above scheme, when determining the regularization objective function according to the type of training obstacles and the first actual obstacle type, a regularization term is introduced. By introducing the regularization term, the complexity of the obstacle classification model is constrained, the obstacle classification model is prevented from overfitting, and the generalization ability of the model is improved.

[0040] Specifically, the preset principal component data set also includes a verification principal component data set, and the verification principal component data set includes verification principal component data and a second actual obstacle type; after obtaining the obstacle classification model, it also includes:

[0041] Input the verification principal component data set into the obstacle classification model, process it through the obstacle classification model, and output the verification obstacle type;

[0042] Determining the accuracy of an obstacle classification model according to the verified obstacle type and the second actual obstacle type;

[0043] In response to the accuracy being less than a preset accuracy threshold, model parameters of the obstacle classification model and / or dimensions corresponding to the principal component analysis process are adjusted, and the obstacle classification model is retrained until the accuracy is greater than or equal to the preset accuracy threshold.

[0044] Through the above scheme, after obtaining the obstacle classification model, the accuracy of the obstacle classification model can be calculated by verifying the principal component data set. When the accuracy is lower than the preset accuracy threshold, the obtained obstacle classification model can be further trained, so that the obstacle type obtained by the subsequent application of the obstacle classification model is more consistent with the actual obstacle type, so that when the vehicle is subsequently controlled according to the obstacle type and the moving speed of the obstacle, accurate avoidance of the obstacle can be achieved, thereby improving driving safety.

[0045] Based on the same inventive concept, the second aspect of the present disclosure provides a vehicle control device, comprising:

[0046] A signal receiving module is configured to receive an echo signal corresponding to the obstacle detection signal, and determine the moving speed of the obstacle according to the echo signal;

[0047] a principal component determination module, configured to perform principal component analysis on the echo signal to obtain a target principal component corresponding to the echo signal;

[0048] A type determination module is configured to input the target principal component into a pre-trained obstacle classification model, and output the obstacle type after being processed by the obstacle classification model;

[0049] The obstacle avoidance control module is configured to control the vehicle to avoid the obstacle according to the moving speed of the obstacle and the type of the obstacle.

[0050] Based on the same inventive concept, the third aspect of the present disclosure proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor implements the vehicle control method as described above when executing the computer program.

[0051] Based on the same inventive concept, a fourth aspect of the present disclosure proposes a non-transitory computer-readable storage medium, which stores computer instructions for causing a computer to execute the vehicle control method as described above.

[0052] Based on the same inventive concept, the fifth aspect of the present disclosure provides a vehicle, comprising the vehicle control device described in the second aspect or the electronic device described in the third aspect or the storage medium described in the fourth aspect.

[0053] As can be seen from the above, the present disclosure proposes a vehicle control method, device and vehicle, which receive an echo signal corresponding to an obstacle detection signal, and determine the moving speed of the obstacle according to the echo signal. The echo signal is subjected to principal component analysis to obtain the target principal component corresponding to the echo signal. Since the principal component analysis can convert high-dimensional data into low-dimensional data, the data dimension is reduced through principal component analysis, and the data in the echo signal is simplified, and the key feature information in the echo signal, namely the target principal component, is obtained. For subsequent classification processing according to the target principal component, the obstacle type of the obstacle is determined, and the influence of irrelevant features on the subsequent determination of the obstacle type is reduced, so that the determined obstacle type is more accurate. The target principal component is input into a pre-trained obstacle classification model, and the obstacle type is output after being processed by the obstacle classification model. Since the obstacle classification model is a model obtained by pre-training with a large amount of data, compared with manually determining the obstacle type by analyzing the target principal component, the target principal component is input into the obstacle classification model, and the target principal component is identified and classified by the trained obstacle classification model, and the obstacle type obtained is more accurate. The obstacle can be accurately sensed and identified based on the determined moving speed and type of the obstacle, and the vehicle can be controlled to avoid the obstacle, thereby improving the safety of vehicle driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the present disclosure or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 is a flow chart of a vehicle control method according to an embodiment of the present disclosure;

[0056] Figure 2 is a structural block diagram of a vehicle control device according to an embodiment of the present disclosure;

[0057] Figure 3 It is a structural block diagram of a vehicle obstacle detection and avoidance system according to another embodiment of the present disclosure;

[0058] Figure 4 It is a schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0059] In order to make the objectives, technical solutions and advantages of the present disclosure more clearly understood, the present disclosure is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0060] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present disclosure should be understood by people with ordinary skills in the field to which the present disclosure belongs. The "first", "second" and similar words used in the embodiments of the present disclosure do not indicate any order, quantity or importance, but are only used to distinguish different components. "Including" or "comprising" and similar words mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0061] The terms used in this disclosure are explained as follows:

[0062] Continuous wave (CW): Continuous wave (CW) is a wave output by a laser in a continuous manner rather than in a pulsed manner.

[0063] Frequency Modulated Continuous Wave (FMCW): Frequency Modulated Continuous Wave (FMCW) is a method that uses frequency modulation to achieve distance resolution between signals.

[0064] Analog-to-digital converter (ADC): Analog-to-digital converter, also known as A / D converter, or ADC for short, usually refers to an electronic component that converts analog signals into digital signals.

[0065] PCA: Principal components analysis (PCA), also known as principal component analysis technology, aims to use the idea of ​​dimensionality reduction to transform multiple indicators into a few comprehensive indicators.

[0066] SVM: Support Vector Machine (SVM) is a powerful supervised learning algorithm that is widely used in classification and regression problems.

[0067] Hanning Window: Hanning Window is one of the window functions, a special case of the raised cosine window, and the sum of the spectra of three rectangular time windows.

[0068] Hamming Window: Hamming Window is a commonly used window function in digital signal processing. The Hamming Window is widely used because of its large sidelobe attenuation in the frequency domain, especially in situations where sidelobe suppression is required.

[0069] With the rapid development of vehicle technology, vehicles have become an important means of transportation in people's daily lives. In order to facilitate users' travel and improve their experience, driverless technology has emerged and gradually matured.

[0070] For unmanned driving technology, accurate perception of the surrounding environment has become the key to ensuring the safety and reliability of unmanned driving. In terms of obstacle detection, traditional visual sensors, lidar, millimeter-wave radar and other sensors have their own advantages and disadvantages. Although visual sensors can provide rich image information, they are greatly affected by lighting conditions and have poor adaptability to hidden obstacles and bad weather. Lidar is known for its high precision and high resolution, but it still faces challenges in cost, data processing complexity and performance under bad weather conditions. Millimeter-wave radar has been widely used in the field of unmanned driving due to its advantages such as low cost, small size, strong penetration and all-weather working ability. However, traditional millimeter-wave radar has limitations in identifying obstacle types, sizes and precise locations, and it is difficult to meet the needs of high-precision detection in complex environments.

[0071] In other words, due to the influence of factors such as ambient lighting and weather, the accuracy of environmental recognition is low, which in turn poses a safety hazard to unmanned driving.

[0072] Based on the above description, this embodiment proposes a vehicle control method, such as Figure 1 As shown, the method includes:

[0073] Step 101: Receive an echo signal corresponding to an obstacle detection signal, and determine a moving speed of the obstacle according to the echo signal.

[0074] In specific implementation, the vehicle includes a radar signal transmitting unit and a signal receiving unit. The radar signal transmitting unit can generate and transmit an obstacle detection signal for detecting obstacles. The obstacle detection signal is an electromagnetic wave signal with a specific frequency and waveform. According to user needs, the frequency, power, beam shape and other parameters of the obstacle detection signal can be flexibly adjusted to achieve effective detection of obstacles within different distances and angles.

[0075] Specifically, the radar signal transmitting unit transmits an obstacle detection signal according to preset parameters, and the type of the obstacle detection signal may be a continuous wave (CW) or a frequency modulated continuous wave (FMCW).

[0076] The obstacle detection signal propagates in space and is reflected after encountering an obstacle, forming an echo signal. The signal receiving unit captures the echo signal reflected by the obstacle, which contains information such as the distance and speed of the obstacle. The echo signal is identified and extracted to determine the moving speed of the obstacle.

[0077] Step 102: Perform principal component analysis on the echo signal to obtain a target principal component corresponding to the echo signal.

[0078] In specific implementation, the echo signal is subjected to principal component analysis to obtain the target principal component corresponding to the echo signal. The principal component analysis is used to reduce the dimension of the data while retaining the main information in the signal, which is the most critical feature for obstacle recognition. Through principal component analysis, the original high-dimensional radar signal data is converted into a series of low-dimensional principal components, which represent the main change patterns in the signal.

[0079] Step 103: input the target principal component into a pre-trained obstacle classification model, and output the obstacle type after being processed by the obstacle classification model.

[0080] During the specific implementation, a pre-trained obstacle classification model is obtained. The obstacle classification model is preferably a support vector machine model. The support vector machine model is a powerful classifier, which is particularly suitable for processing high-dimensional data and small samples. It separates different categories of data by finding an optimal hyperplane in the feature space, thereby achieving classification.

[0081] The target principal component obtained by principal component analysis is input into the obstacle classification model, and after being processed by the obstacle classification model, the obstacle type corresponding to the obstacle is output. The obstacle type indicates the classification category to which the obstacle belongs, and the obstacle type includes static obstacles and dynamic obstacles. The static obstacle indicates an obstacle that remains stationary for a long time, and the dynamic obstacle indicates an obstacle that moves for a long time. The long time is a preset period of time. In order to accurately distinguish static obstacles from dynamic obstacles, the long time can be half a year, a year or even longer.

[0082] Exemplarily, the static obstacle includes at least one of the following: a wall, a utility pole, a tree, a curbstone, etc. The dynamic obstacle includes at least one of the following: a vehicle, a pedestrian, a pet, a bird, etc.

[0083] Step 104: Control the vehicle to travel according to the moving speed of the obstacle and the type of the obstacle to avoid the obstacle.

[0084] In specific implementation, different types of obstacles have different common movement characteristics. For example, moving vehicles generally travel along lanes, and moving vehicles generally maintain a constant speed for a certain period of time. However, moving pedestrians move more freely, resulting in a more complex movement route. Therefore, the vehicle can be controlled according to the moving speed and type of the obstacle to avoid the obstacle.

[0085] Exemplarily, if the obstacle is a moving vehicle, the distance between the obstacle and the vehicle is determined based on the moving speed of the obstacle, and the speed of the vehicle is controlled based on the distance to ensure that the distance between the vehicle and the obstacle is greater than a preset safety distance.

[0086] In another example, if the obstacle is a moving pedestrian, assuming that the moving pedestrian is in front of the vehicle and is crossing the road in front of the vehicle, the pedestrian is monitored in real time through the in-vehicle camera device to obtain the distance between the pedestrian and the vehicle. The target moving speed of the vehicle is determined based on the distance between the pedestrian and the vehicle and the moving speed of the pedestrian, and the vehicle is then controlled to travel at the target moving speed to avoid collision with the pedestrian.

[0087] Through the above scheme, the echo signal corresponding to the obstacle detection signal is received, and the moving speed of the obstacle is determined according to the echo signal. The echo signal is subjected to principal component analysis to obtain the target principal component corresponding to the echo signal. Since the principal component analysis can convert high-dimensional data into low-dimensional data, the data dimension is reduced through principal component analysis, and the data in the echo signal is simplified, and the key feature information in the echo signal, namely the target principal component, is obtained. For subsequent classification processing according to the target principal component, the obstacle type of the obstacle is determined, and the influence of irrelevant features on the subsequent determination of the obstacle type is reduced, so that the determined obstacle type is more accurate. The target principal component is input into the obstacle classification model obtained by pre-training, and the obstacle type is output after being processed by the obstacle classification model. Since the obstacle classification model is a model obtained by pre-training with a large amount of data, compared with manually determining the obstacle type by analyzing the target principal component, the target principal component is input into the obstacle classification model, and the target principal component is identified and classified by the trained obstacle classification model, and the obstacle type obtained is more accurate. The obstacle can be accurately sensed and identified based on the determined moving speed and type of the obstacle, and the vehicle can be controlled to avoid the obstacle, thereby improving the safety of vehicle driving.

[0088] In some embodiments, when determining the target principal component corresponding to the echo signal, the digital signal corresponding to the echo signal may be determined first, and then the projection matrix corresponding to the digital signal may be determined, and the projection matrix may be used to implement dimensionality reduction processing. Therefore, in step 102, the principal component analysis processing is performed on the echo signal to obtain the target principal component corresponding to the echo signal, which specifically includes:

[0089] Step 1021, performing analog-to-digital conversion on the echo signal to obtain a digital signal corresponding to the echo signal;

[0090] Step 1022, determining a correlation coefficient matrix corresponding to the digital signal, and determining a projection matrix according to the correlation coefficient matrix;

[0091] Step 1023: Perform dimensionality reduction processing on the digital signal using the projection matrix to obtain the target principal component corresponding to the echo signal.

[0092] In specific implementation, in order to ensure accurate conversion and transmission of the signal, the echo signal is converted into digital form to obtain a digital signal corresponding to the echo signal. In this embodiment, the analog-to-digital conversion of the echo signal is achieved by a high-precision analog-to-digital converter.

[0093] Determine a correlation coefficient matrix corresponding to the digital signal, wherein the specific determination method of the correlation coefficient matrix includes:

[0094] Step a, determining an initial signal data matrix corresponding to the digital signal;

[0095] Step b, performing data preprocessing on the initial signal data matrix to obtain a signal data matrix;

[0096] Step c: performing correlation calculation on the signal data matrix to obtain a correlation coefficient matrix corresponding to the digital signal.

[0097] Specifically, an initial signal data matrix corresponding to the digital signal is determined, wherein the initial signal data matrix includes multiple features measured multiple times, that is, observation values ​​of multiple samples in multiple feature dimensions.

[0098] The initial signal data matrix is ​​subjected to data preprocessing to obtain a signal data matrix. The data preprocessing includes steps such as data cleaning, denoising and normalization. The signal data matrix is ​​subjected to correlation calculation to obtain a correlation coefficient matrix corresponding to the digital signal.

[0099] In this embodiment, the specific process of performing correlation calculation on the signal data matrix to obtain the correlation coefficient matrix corresponding to the digital signal includes:

[0100] First, the covariance matrix corresponding to the signal data matrix is ​​calculated, and the covariance matrix is ​​a matrix composed of covariances between random variables. After obtaining the covariance matrix, the corresponding correlation coefficient matrix is ​​calculated according to the covariance matrix. The correlation coefficient matrix is ​​used to count the linear correlation between two random variables. The correlation coefficient matrix represents the Pearson correlation coefficient between the two variables, which is defined as the covariance of the two variables divided by the product of their standard deviations.

[0101] After determining the correlation coefficient matrix, a projection matrix is ​​determined according to the correlation coefficient matrix. The specific method of determining the projection matrix includes:

[0102] Step A, determining a plurality of eigenvalues ​​corresponding to the correlation coefficient matrix and an eigenvector corresponding to each eigenvalue;

[0103] Step B, sorting all eigenvalues ​​in descending order;

[0104] Step C, obtaining a preset dimension, selecting the eigenvalues ​​in the previous preset dimension as target eigenvalues, and taking the eigenvectors corresponding to each target eigenvalue as target eigenvectors;

[0105] Step D: construct a projection matrix based on all target feature vectors.

[0106] During specific implementation, the correlation coefficient matrix is ​​subjected to eigenvalue decomposition to obtain a plurality of eigenvalues ​​corresponding to the correlation coefficient matrix and an eigenvector corresponding to each of the plurality of eigenvalues.

[0107] All eigenvalues ​​are counted and sorted in descending order. A preset dimension is obtained, which is the dimension after dimensionality reduction. The eigenvalues ​​with the first preset dimension are selected from the sorted eigenvalues ​​as the target eigenvalues, and the eigenvectors corresponding to each target eigenvalue are used as the target eigenvectors, and a projection matrix is ​​constructed based on all the target eigenvectors.

[0108] In this embodiment, the preset dimension may be determined by a set dimension, or by a dimension determined by retaining a certain proportion of variance (such as 95%) or by using the elbow rule of eigenvalues.

[0109] After constructing the projection matrix, the projection matrix is ​​used to perform dimensionality reduction processing on the digital signal, that is, according to the projection matrix, the signal data matrix is ​​projected into a low-dimensional space to obtain the number of principal components after dimensionality reduction, and the number of principal components is the target principal component corresponding to the echo signal.

[0110] Exemplarily, assuming that the initial signal data matrix is ​​X, a specific example of the principal component analysis process is:

[0111] The initial signal data matrix contains multiple features measured multiple times. The initial signal data matrix is ​​standardized so that the mean of each feature is 0 and the variance is 1, and a signal data matrix is ​​obtained. The signal data matrix is ​​expressed by the formula:

[0112]

[0113] Among them, X′ is the signal data matrix, μ is the mean of the feature, and σ is the standard deviation of the feature.

[0114] Solve the covariance matrix corresponding to the signal data matrix. The covariance matrix is ​​used to describe the correlation between various features in the data. The covariance matrix is ​​expressed by the formula:

[0115]

[0116] Among them, C is the covariance matrix and N is the number of samples.

[0117] The covariance matrix is ​​subjected to eigenvalue decomposition to obtain eigenvalues ​​and eigenvectors, wherein the eigenvalue solution process is expressed by the formula:

[0118] Cv i =λ i v i

[0119] Among them, v i is the eigenvector, λ i is the eigenvalue, and i is the number of eigenvalues.

[0120] All eigenvalues ​​are sorted from large to small. According to the size of the eigenvalues, the eigenvectors corresponding to the first k largest eigenvalues ​​are selected as the target eigenvectors. A projection matrix is ​​constructed according to all the target eigenvectors. Then, the projection matrix is ​​used to project the signal data matrix into a low-dimensional space to obtain a reduced-dimensional data matrix. The reduced-dimensional data matrix contains the target principal component. The projection process is expressed as follows:

[0121] Y=X′P

[0122] Among them, Y is the data matrix after dimensionality reduction, and P is the projection matrix.

[0123] Through the above scheme, the echo signal is converted into digital form to ensure accurate conversion and transmission of the signal. At the same time, by determining the projection matrix of the digital signal corresponding to the echo signal, the projection matrix is ​​used to reduce the dimension of the digital signal, thereby reducing the dimension of the digital signal and obtaining the key feature information of the obstacle, namely the target principal component. The data subsequently input into the obstacle classification model is more concise and clear, reducing the interference of irrelevant features and improving the classification accuracy of the subsequent obstacle classification model.

[0124] In some embodiments, determining the moving speed of the obstacle according to the echo signal in step 101 specifically includes:

[0125] Step 1011, performing analog-to-digital conversion on the echo signal to obtain a digital signal corresponding to the echo signal;

[0126] Step 1012, performing windowing processing on the digital signal to obtain a signal to be processed;

[0127] Step 1013, performing Fourier transform on the signal to be processed to extract Doppler frequency shift data;

[0128] Step 1014, obtaining the wavelength and center frequency of the obstacle detection signal, and determining the moving speed of the obstacle according to the wavelength, the center frequency and the Doppler frequency shift data.

[0129] In specific implementation, in order to ensure accurate conversion and transmission of the signal, the echo signal is converted into digital form to obtain a digital signal corresponding to the echo signal. In this embodiment, the analog-to-digital conversion of the echo signal is achieved by a high-precision analog-to-digital converter.

[0130] The digital signal is windowed by a preset window function to obtain a signal to be processed to reduce spectrum leakage. In this embodiment, a window function such as a Hanning window or a Hamming window is used for windowing. The form of the window function is expressed by the formula:

[0131]

[0132] Among them, w(n) is the window function, n is the sample index, and N is the window length.

[0133] Since the Fourier transform algorithm itself is decomposable and suitable for parallel processing, parallel computing technology (such as multithreading, GPU acceleration, etc.) is used to accelerate the Fourier transform operation. The signal to be processed is subjected to Fourier transform to extract Doppler frequency shift data.

[0134] The Doppler effect causes the frequency of the radar signal to change, and this change is proportional to the relative speed of the obstacle. The wavelength and center frequency of the obstacle detection signal are obtained, and the moving speed of the obstacle is determined based on the wavelength, the center frequency and the Doppler frequency shift data. The above process is expressed by the formula:

[0135]

[0136] Among them, f d is the Doppler frequency shift data, v is the moving speed of the obstacle, λ is the wavelength of the obstacle detection signal, and f0 is the center frequency of the obstacle detection signal.

[0137] In this embodiment, in order to improve the accuracy of the obstacle's moving speed, the obstacle's moving speed obtained from the frequency spectrum data of multiple periods may be averaged, and the obtained average value is used as the obstacle's moving speed to reduce noise interference.

[0138] In this embodiment, before windowing the digital signal, the digital signal may be preprocessed, and the data preprocessing method includes filtering denoising and signal enhancement. Filtering denoising aims to eliminate random noise and interference components in the signal. Signal enhancement improves the signal-to-noise ratio and clarity of the signal by means of signal amplification, phase compensation, etc.

[0139] Through the above scheme, the spectrum leakage is reduced by windowing the digital signal. At the same time, according to the parallel Fourier transform algorithm, compared with the sequential Fourier transform calculation, the extraction accuracy and calculation efficiency of Doppler frequency shift data are improved, providing a reliable data basis for the subsequent calculation of the moving speed of obstacles.

[0140] In some embodiments, step 104 specifically includes:

[0141] Step 1041, obtaining a time delay corresponding to the echo signal, and determining a relative distance between the obstacle and the vehicle according to the time delay;

[0142] Step 1042, determining the angle between the obstacle detection signal and the obstacle movement direction according to the Doppler frequency shift data;

[0143] Step 1043, determining the target driving state of the vehicle according to the moving speed of the obstacle, the type of obstacle, the relative distance and the angle, and controlling the vehicle to avoid the obstacle according to the target driving state.

[0144] In a specific implementation, the time delay corresponding to the echo signal is obtained, and the relative distance between the obstacle and the vehicle is calculated according to the time delay. The relative distance is expressed by the formula:

[0145]

[0146] Among them, d is the relative distance between the obstacle and the vehicle, c is the speed of light, and Δt is the time delay of the echo signal.

[0147] According to the Doppler frequency shift data, the moving speed of the obstacle and the wavelength of the obstacle detection signal, the angle between the obstacle detection signal and the obstacle moving direction is calculated. The above process is expressed by the formula:

[0148]

[0149] Where v is the moving speed of the obstacle, f d is the Doppler frequency shift data, λ is the wavelength of the obstacle detection signal, and θ is the angle between the obstacle detection signal and the obstacle moving direction.

[0150] The target driving state of the vehicle is determined according to the moving speed of the obstacle, the type of obstacle, the relative distance and the angle. The target driving state includes driving parameter information of the vehicle when it is driving. The driving parameter information includes the driving speed of the vehicle, the driving direction of the vehicle, etc.

[0151] In this embodiment, a specific method of determining the target driving state of the vehicle according to the moving speed of the obstacle, the obstacle type, the relative distance and the angle includes:

[0152] According to the moving speed of the obstacle and the type of the obstacle, a database is searched to determine the target vehicle speed corresponding to the moving speed of the obstacle and the type of the obstacle, and the database pre-stores the corresponding relationship between the moving speed of the obstacle and the type of the obstacle and the target vehicle speed. The current vehicle speed is obtained, and the vehicle is controlled to adjust from the current vehicle speed to the target vehicle speed.

[0153] During the adjustment process, the relative distance and angle between the vehicle and the obstacle are constantly monitored. If the relative distance is less than a preset distance threshold and / or the angle is less than a preset angle threshold, it means that the distance between the vehicle and the obstacle is too close. At this time, the database is searched according to the current relative distance and / or the current angle to determine the target steering wheel angle value of the vehicle, and the vehicle steering wheel angle is controlled to be adjusted from the current angle value to the target steering wheel angle value.

[0154] In this embodiment, another specific method of determining the target driving state of the vehicle according to the moving speed of the obstacle, the obstacle type, the relative distance and the angle includes:

[0155] Obtain the pre-trained obstacle avoidance path planning model and the current position of the vehicle, input the moving speed of the obstacle, the obstacle type and the current position of the vehicle into the obstacle avoidance path planning model, and output the target obstacle avoidance path after being processed by the obstacle avoidance path planning model. Then the vehicle drives according to the target obstacle avoidance path to avoid the obstacle.

[0156] Furthermore, after obtaining the target obstacle avoidance path, the relative distance and angle between the vehicle and the obstacle when the vehicle is traveling along the target obstacle avoidance path are first determined. If the relative distance is less than a preset distance threshold, and / or the angle is less than a preset angle threshold, it means that the distance between the vehicle and the obstacle is too close. The target obstacle avoidance path can be adjusted according to optimization algorithms such as reinforcement learning to ensure driving safety.

[0157] In this embodiment, when controlling the vehicle's travel, intelligent obstacle avoidance decisions can also be made in combination with the vehicle's current motion state (such as speed, acceleration, direction, etc.) and surrounding environmental conditions (such as road type, traffic rules, etc.).

[0158] In the obstacle avoidance decision-making process, multiple aspects such as safety, efficiency and comfort can also be considered. For example, sudden braking or sharp turns that may cause discomfort to passengers can be avoided as much as possible. At the same time, an obstacle avoidance path that can reach the destination the fastest can be selected while ensuring safety.

[0159] At the same time, the vehicle control system performs corresponding obstacle avoidance actions (such as turning, slowing down or stopping, etc.). During the obstacle avoidance process, the system will continuously monitor changes in the surrounding environment and the vehicle's motion state, and make real-time adjustments as needed. At the same time, feedback information during the obstacle avoidance process (such as obstacle avoidance effect, passenger response, etc.) can also be collected for subsequent optimization and improvement.

[0160] In some embodiments, the training process of the obstacle classification model used in step 103 specifically includes:

[0161] Step 10A, obtaining a preset principal component data set and an initial obstacle classification model, wherein the preset principal component data set includes a training principal component data set, and the training principal component data set includes training principal component data and a first actual obstacle type;

[0162] Step 10B, training the initial obstacle classification model using the training principal component data set to obtain training obstacle types;

[0163] Step 10C, determining a regularization objective function according to the training obstacle type and the first actual obstacle type;

[0164] Step 10D: until the regularized objective function is minimized, it is determined that the training of the initial obstacle classification model is completed, and an obstacle classification model is obtained.

[0165] During specific implementation, a preset principal component data set and an initial obstacle classification model are obtained, wherein the preset principal component data set includes a training principal component data set and a verification principal component data set, wherein the training principal component data set includes training principal component data and a first actual obstacle type, and wherein the verification principal component data set includes verification principal component data and a second actual obstacle type.

[0166] The initial obstacle classification model is trained using the training principal component data set to obtain a training obstacle type. A regularization objective function is determined according to the type of the training obstacle and the first actual obstacle type. The regularization parameter is a penalty coefficient in the objective function, which is used to balance the classification interval and the misclassified samples. Specifically, the larger the regularization parameter, the lower the tolerance of the model to errors, which is prone to overfitting. The smaller the regularization parameter, the higher the tolerance of the model to errors, which is prone to underfitting.

[0167] Specifically, the initial obstacle classification model is a support vector machine model, which divides data of different categories by finding an optimal hyperplane in the feature space, and the optimal hyperplane is defined by a normal vector w and an intercept b. For a linear support vector machine model, its decision function is expressed by the formula:

[0168] f(x)=sign(ω·x+b)

[0169] The optimal hyperplane is expressed by the formula:

[0170] ω·x+b=0

[0171] Among them, ω is the normal vector of the hyperplane, x is the training principal component data, and b is the bias.

[0172] A regularization objective function is determined according to the type of training obstacle and the first actual obstacle type. The regularization objective function is expressed by the formula:

[0173]

[0174] Among them, ξ i is the regularization term, and C is the regularization parameter.

[0175] The initial obstacle classification model is trained using the data in the training principal component data set until the regularized objective function is minimized, which indicates that the training of the initial obstacle classification model is completed and an obstacle classification model is obtained.

[0176] Through the above scheme, when determining the regularization objective function according to the type of training obstacles and the first actual obstacle type, a regularization term is introduced. By introducing the regularization term, the complexity of the obstacle classification model is constrained, the obstacle classification model is prevented from overfitting, and the generalization ability of the model is improved.

[0177] In some embodiments, the preset principal component data set further includes a verification principal component data set, and the verification principal component data set includes verification principal component data and a second actual obstacle type. After the initial obstacle classification model is trained to obtain the obstacle classification model, the obtained obstacle classification model can be verified by the verification principal component data set to further determine whether the obstacle classification model meets the requirements. The method further includes:

[0178] Step 10a, inputting the verification principal component data set into the obstacle classification model, processing the obstacle classification model, and outputting the verification obstacle type;

[0179] Step 10b, determining the accuracy of the obstacle classification model according to the verified obstacle type and the second actual obstacle type;

[0180] Step 10c, in response to the accuracy being less than a preset accuracy threshold, adjusting the model parameters of the obstacle classification model and / or the dimensions corresponding to the principal component analysis process, and retraining the obstacle classification model until the accuracy is greater than or equal to the preset accuracy threshold.

[0181] In a specific implementation, the verification principal component data in the verification principal component data set is input into the obstacle classification model, and after being processed by the obstacle classification model, the verification obstacle type corresponding to the verification principal component data is output.

[0182] The accuracy of the obstacle classification model is determined according to the verification obstacle type and the second actual obstacle type. Specifically, a first number of verification principal component data is obtained, and a second number of verification principal component data whose verification obstacle type is the same as the second actual obstacle type is determined, wherein the verification obstacle type is the same as the second actual obstacle type, indicating that the obstacle type output by the obstacle classification model is a correct type. The second number and the first number are subjected to ratio processing, and the ratio is the accuracy of the obstacle classification model.

[0183] Exemplarily, the first number of obtained verification principal component data is 100, the second number of verified principal component data for which the verification obstacle type is determined to be the same as the second actual obstacle type is 90, and the accuracy of the obstacle classification model is determined to be 90%.

[0184] The accuracy of the determined obstacle classification model is compared with a preset accuracy threshold. If the accuracy is less than the preset accuracy threshold, it means that the obstacle classification model still needs to be trained. The model parameters of the obstacle classification model are adjusted. The model parameters are parameters in the obstacle classification model, including at least one of the following: regularization term parameters, kernel function parameters, etc.

[0185] In this embodiment, since the data input to the obstacle classification model is the principal component data obtained after the principal component analysis, when the accuracy of the obstacle classification model is lower than the preset accuracy threshold, in addition to adjusting the model parameters of the obstacle classification model, the preset dimension corresponding to the principal component analysis can also be adjusted, thereby changing the principal component data obtained after the principal component analysis, thereby achieving the adjustment of the accuracy of the obstacle type output by the obstacle classification model.

[0186] A new training principal component data set is obtained, the obstacle classification model is continuously trained using the new training principal component data set, and the accuracy of the trained obstacle classification model is recalculated until the accuracy is greater than or equal to a preset accuracy threshold.

[0187] Through the above scheme, after obtaining the obstacle classification model, the accuracy of the obstacle classification model can be calculated by verifying the principal component data set. When the accuracy is lower than the preset accuracy threshold, the obtained obstacle classification model can be further trained, so that the obstacle type obtained by the subsequent application of the obstacle classification model is more consistent with the actual obstacle type, so that when the vehicle is subsequently controlled according to the obstacle type and the moving speed of the obstacle, accurate avoidance of the obstacle can be achieved, thereby improving driving safety.

[0188] In this application, when determining the obstacle type, by combining principal component analysis and obstacle classification model, compared with directly inputting the echo signal corresponding to the obstacle detection signal into the obstacle classification model, the key feature information corresponding to the echo signal is extracted through principal component analysis, which reduces the amount of data input into the obstacle classification model, reduces the interference of irrelevant information on the obstacle classification model, and improves the accuracy of the determined obstacle type. At the same time, when controlling the vehicle, in addition to the moving speed of the obstacle, the type of obstacle is also considered, and the analysis of the movement of the obstacle is more accurate, making the control of the vehicle more precise, improving the safety of vehicle driving, and reducing safety hazards.

[0189] Based on the same inventive concept, in order to solve the problem that the recognition accuracy of the environment is low due to the influence of factors such as ambient light and weather, thereby causing potential safety hazards in unmanned driving, corresponding to the above embodiment, another embodiment of the present disclosure provides a vehicle control method, the method specifically comprising:

[0190] Electromagnetic wave signals such as continuous wave (CW) or frequency modulated continuous wave (FMCW) are transmitted according to preset parameters. These signals propagate in space and are reflected after encountering obstacles, forming echo signals. The echo signals reflected by obstacles are captured and converted into digital signals. To ensure accurate conversion and transmission of signals, the receiving unit needs to have a high-precision analog-to-digital converter (ADC) and a stable clock source.

[0191] The received digital signal is preprocessed. The preprocessing stage mainly includes steps such as filtering and denoising and signal enhancement. Filtering and denoising aims to eliminate random noise and interference components in the signal; signal enhancement improves the signal-to-noise ratio and clarity of the signal through signal amplification, phase compensation and other means.

[0192] Based on the preprocessed signal, the improved Fast Fourier Transform (FFT) algorithm is used to extract the Doppler shift information in the signal. Doppler shift is a frequency change phenomenon caused by the relative motion between the obstacle and the radar. By analyzing the Doppler shift information, the speed information of the obstacle can be calculated. In order to improve the calculation accuracy and efficiency, the FFT algorithm is optimized and improved, such as using window function to reduce spectrum leakage and using parallel computing to accelerate FFT operation.

[0193] Specifically, a window function such as a Hanning Window or a Hamming Window is used to perform windowing on the signal to reduce spectrum leakage. The basic form of the window function is:

[0194]

[0195] Where n is the sample index and N is the window length.

[0196] Use parallel computing technology (such as multithreading, GPU acceleration, etc.) to speed up FFT operations. The FFT algorithm itself is decomposable and suitable for parallel processing.

[0197] Obtain the wavelength and center frequency of the obstacle detection signal, and determine the moving speed of the obstacle according to the wavelength, the center frequency and the Doppler frequency shift data. The above process is expressed by the formula:

[0198]

[0199] Among them, f d is the Doppler frequency shift data, v is the moving speed of the obstacle, λ is the wavelength of the obstacle detection signal, and f0 is the center frequency of the obstacle detection signal.

[0200] The signal strength distribution is extracted and classified by combining principal component analysis and obstacle classification model. Principal component analysis is used to reduce the dimensionality of the data while retaining the main information in the signal, that is, the features that are most critical for obstacle identification. Through principal component analysis, the original high-dimensional radar signal data is converted into a series of low-dimensional principal components, which represent the main change patterns in the signal.

[0201] The obstacle classification model is a support vector machine model, which is used to classify these principal components. The support vector machine is a powerful classifier, which is particularly suitable for processing high-dimensional data and small sample situations. It separates data of different categories by finding an optimal hyperplane in the feature space, thereby achieving classification. In the system of the present invention, the support vector machine is trained to identify different types of obstacles (such as vehicles, pedestrians, trees, etc.), and makes classification decisions based on the principal components extracted by principal component analysis.

[0202] Specifically, the original high-dimensional radar signal data matrix X (which contains the observations of multiple samples in multiple feature dimensions) is first used as the input of principal component analysis. Principal component analysis calculates the correlation coefficient matrix R, and then solves its eigenvalues ​​and eigenvectors, and selects the eigenvectors corresponding to the first k largest eigenvalues ​​to form the projection matrix P. Here, k is a key parameter that determines the dimension after dimensionality reduction, which is usually determined by retaining a certain proportion of variance (such as 95%) or according to the elbow rule of eigenvalues. Using the projection matrix P, the original data X is projected into a low-dimensional space to obtain the number of principal components after dimensionality reduction Y=XP. These principal components not only reduce the dimensionality of the data, but also retain the most important change patterns in the signal, that is, those features that are most critical to obstacle recognition.

[0203] Furthermore, the principal component analysis process is described in detail, including:

[0204] First, data preprocessing is performed to preprocess the original high-dimensional radar signal data matrix X, including data cleaning, denoising, and normalization steps to ensure the quality and consistency of the data.

[0205] Calculate the correlation coefficient matrix and calculate the correlation coefficient matrix R of the preprocessed data matrix X. This step is to understand the correlation between the features and provide a basis for subsequent feature selection.

[0206] Solving for eigenvalues ​​and eigenvectors, we identify the main directions of change in the data by solving for the eigenvalues ​​and eigenvectors of the correlation coefficient matrix R. These eigenvectors will be used to construct the projection matrix P.

[0207] Construct the projection matrix P, and select the eigenvectors corresponding to the first k largest eigenvalues ​​to form the projection matrix P. The selection of k is the key, which determines the dimension after dimensionality reduction. Usually, the value of k is determined by retaining a certain proportion of variance (such as 95%) or using the elbow rule of eigenvalues.

[0208] Finally, the data is projected, and the original data X is projected into a low-dimensional space using the projection matrix P, and the number of principal components after dimensionality reduction is obtained, Y = XP. These principal components not only reduce the dimensionality of the data, but also retain the most important change patterns in the signal, that is, those features that are most critical to obstacle recognition.

[0209] Furthermore, the process of processing the obstacle classification model to obtain the obstacle type is described in detail, including:

[0210] Input the principal component data, and use the principal component data Y extracted by principal component analysis as the input of the obstacle classification model for training.

[0211] Optimal hyperplane search, the obstacle classification model separates different categories of data in the feature space by searching for an optimal hyperplane (defined by the normal vector w and the intercept b). For linear SVM, its decision function is expressed as f(x) = sign(w·x+b)

[0212] Objective function optimization, during the training process, the training data set (including the labeled obstacle types) is used to optimize the objective function. This usually involves minimizing the regularized error term to find the optimal w and b. The introduction of the regularization term helps prevent overfitting and improves the generalization ability of the model.

[0213] Parameter adjustment: By adjusting the parameters of the obstacle classification model (such as penalty parameter C and kernel function parameters), the classification performance of the model can be further optimized. The adjustment of these parameters needs to be carried out according to the specific data set and classification task.

[0214] Model training and evaluation, after completing the parameter adjustment, use the training data set to train the obstacle classification model, and use the validation data set to evaluate the model performance. If the model performance does not meet the requirements, it is necessary to return to adjust the k value of the principal component analysis or the parameters of the obstacle classification model until a satisfactory classification effect is achieved.

[0215] Classification decision,After training is completed, the obstacle classification model can make classification decisions on new radar signal data based on the principal components extracted by principal component analysis, thereby achieving fast and accurate identification of obstacle types.

[0216] The time delay corresponding to the echo signal is obtained, and the relative distance between the obstacle and the vehicle is calculated according to the time delay. The relative distance is expressed by the formula:

[0217]

[0218] Among them, d is the relative distance between the obstacle and the vehicle, c is the speed of light, and Δt is the time delay of the echo signal.

[0219] According to the Doppler frequency shift data, the moving speed of the obstacle and the wavelength of the obstacle detection signal, the angle between the obstacle detection signal and the obstacle moving direction is calculated. The above process is expressed by the formula:

[0220]

[0221] Where v is the moving speed of the obstacle, f d is the Doppler frequency shift data, λ is the wavelength of the obstacle detection signal, and θ is the angle between the obstacle detection signal and the obstacle moving direction.

[0222] The target driving state of the vehicle is determined according to the moving speed of the obstacle, the type of obstacle, the relative distance and the angle. The target driving state includes driving parameter information of the vehicle when it is driving. The driving parameter information includes the driving speed of the vehicle, the driving direction of the vehicle, etc.

[0223] In this embodiment, when controlling the vehicle's travel, intelligent obstacle avoidance decisions can also be made in combination with the vehicle's current motion state (such as speed, acceleration, direction, etc.) and surrounding environmental conditions (such as road type, traffic rules, etc.).

[0224] In the obstacle avoidance decision-making process, multiple aspects such as safety, efficiency and comfort can also be considered. For example, sudden braking or sharp turns that may cause discomfort to passengers can be avoided as much as possible. At the same time, an obstacle avoidance path that can reach the destination the fastest can be selected while ensuring safety.

[0225] At the same time, the vehicle control system performs corresponding obstacle avoidance actions (such as turning, slowing down or stopping, etc.). During the obstacle avoidance process, the system will continuously monitor changes in the surrounding environment and the vehicle's motion state, and make real-time adjustments as needed. At the same time, feedback information during the obstacle avoidance process (such as obstacle avoidance effect, passenger response, etc.) can also be collected for subsequent optimization and improvement.

[0226] It should be noted that the method of the embodiment of the present disclosure can be performed by a single device, such as a computer or a server. The method of the present embodiment can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present disclosure, and the multiple devices will interact with each other to complete the described method.

[0227] It should be noted that the above describes some embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0228] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present disclosure also provides a vehicle control device.

[0229] refer to Figure 2 , Figure 2 A vehicle control device according to an embodiment of the present invention comprises:

[0230] The signal receiving module 201 is configured to receive an echo signal corresponding to the obstacle detection signal, and determine the moving speed of the obstacle according to the echo signal;

[0231] The principal component determination module 202 is configured to perform principal component analysis on the echo signal to obtain a target principal component corresponding to the echo signal;

[0232] The type determination module 203 is configured to input the target principal component into a pre-trained obstacle classification model, and output the obstacle type after being processed by the obstacle classification model;

[0233] The obstacle avoidance control module 204 is configured to control the vehicle to avoid the obstacle according to the moving speed of the obstacle and the type of the obstacle.

[0234] In some embodiments, the main component determination module 202 specifically includes:

[0235] an analog-to-digital conversion unit, configured to perform analog-to-digital conversion on the echo signal to obtain a digital signal corresponding to the echo signal;

[0236] a projection matrix determining unit, configured to determine a correlation coefficient matrix corresponding to the digital signal, and determine a projection matrix according to the correlation coefficient matrix;

[0237] The target principal component determination unit is configured to perform dimensionality reduction processing on the digital signal using the projection matrix to obtain the target principal component corresponding to the echo signal.

[0238] In some embodiments, the projection matrix determination unit is specifically configured to:

[0239] Determine an initial signal data matrix corresponding to the digital signal;

[0240] Performing data preprocessing on the initial signal data matrix to obtain a signal data matrix;

[0241] A correlation calculation is performed on the signal data matrix to obtain a correlation coefficient matrix corresponding to the digital signal.

[0242] In some embodiments, the projection matrix determination unit is further configured to:

[0243] Determine a plurality of eigenvalues ​​corresponding to the correlation coefficient matrix and an eigenvector corresponding to each eigenvalue;

[0244] Sort all eigenvalues ​​in descending order;

[0245] Obtain a preset dimension, select the eigenvalue of the preset dimension in sequence as the target eigenvalue, and use the eigenvector corresponding to each target eigenvalue as the target eigenvector;

[0246] Construct a projection matrix based on all target eigenvectors.

[0247] In some embodiments, the signal receiving module 201 is specifically configured as follows:

[0248] Performing analog-to-digital conversion on the echo signal to obtain a digital signal corresponding to the echo signal;

[0249] Performing windowing processing on the digital signal to obtain a signal to be processed;

[0250] Performing Fourier transform on the signal to be processed to extract Doppler frequency shift data;

[0251] The wavelength and center frequency of the obstacle detection signal are obtained, and the moving speed of the obstacle is determined according to the wavelength, the center frequency and the Doppler frequency shift data.

[0252] In some embodiments, the obstacle avoidance control module 204 is specifically configured to:

[0253] Acquire a time delay corresponding to the echo signal, and determine a relative distance between the obstacle and the vehicle according to the time delay;

[0254] Determining the angle between the obstacle detection signal and the obstacle movement direction according to the Doppler frequency shift data;

[0255] The target driving state of the vehicle is determined according to the moving speed of the obstacle, the type of obstacle, the relative distance and the angle, and the vehicle is controlled to avoid the obstacle according to the target driving state.

[0256] In some embodiments, the device further includes a model training module, and the model training module is specifically configured to:

[0257] Acquire a preset principal component data set and an initial obstacle classification model, wherein the preset principal component data set includes a training principal component data set, and the training principal component data set includes training principal component data and a first actual obstacle type;

[0258] Using the training principal component data set to train the initial obstacle classification model to obtain training obstacle types;

[0259] Determining a regularization objective function according to the training obstacle type and the first actual obstacle type;

[0260] Until the regularized objective function is minimized, it is determined that the training of the initial obstacle classification model is completed, and an obstacle classification model is obtained.

[0261] In some embodiments, the preset principal component data set also includes a verification principal component data set, and the verification principal component data set includes verification principal component data and a second actual obstacle type, and the model training module is specifically configured as follows:

[0262] Input the verification principal component data set into the obstacle classification model, process it through the obstacle classification model, and output the verification obstacle type;

[0263] Determining the accuracy of an obstacle classification model according to the verified obstacle type and the second actual obstacle type;

[0264] In response to the accuracy being less than a preset accuracy threshold, model parameters of the obstacle classification model and / or dimensions corresponding to the principal component analysis process are adjusted, and the obstacle classification model is retrained until the accuracy is greater than or equal to the preset accuracy threshold.

[0265] Based on the same inventive concept, another embodiment of the present disclosure provides a vehicle obstacle detection and avoidance system, such as Figure 3 As shown, the vehicle obstacle detection and obstacle avoidance system is connected to the vehicle control system, and the vehicle obstacle detection and obstacle avoidance system includes a radar signal transmitting unit, a signal receiving unit, a signal processing unit and an obstacle avoidance decision unit.

[0266] The radar signal transmitting unit is configured to generate and transmit electromagnetic wave signals of specific frequency and waveform. According to application requirements, the parameters such as the frequency, power and beam shape of the transmitted signal can be flexibly adjusted to achieve effective detection of obstacles within different distances and angles.

[0267] The signal receiving unit is configured to receive electromagnetic wave signals reflected from obstacles. The received signal contains key information such as the location, speed, shape and material of the obstacle. To ensure the integrity and accuracy of the signal, the receiving unit must have high sensitivity and low noise characteristics, and be able to effectively suppress clutter interference.

[0268] Signal processing unit: This unit is the core part of the system and is configured to preprocess, extract features and classify the received radar signals. The preprocessing stage mainly includes steps such as filtering and denoising, signal enhancement, etc. to improve signal quality; the feature extraction stage uses advanced signal processing algorithms (such as fast Fourier transform, wavelet transform, etc.) to extract key feature information from the signal; the classification and recognition stage combines machine learning algorithms (such as support vector machines, neural networks, etc.) to classify and identify the extracted features to distinguish different types of obstacles.

[0269] The obstacle avoidance decision unit is configured to calculate the optimal obstacle avoidance path in real time based on the obstacle information provided by the signal processing unit, combined with the vehicle's current motion state and environmental conditions, using an optimization algorithm, and sends instructions to the vehicle control system to execute obstacle avoidance actions. To improve the intelligence level of the system, the obstacle avoidance decision unit can also introduce a reinforcement learning algorithm to optimize the obstacle avoidance strategy through continuous trial and error and learning.

[0270] For the convenience of description, the above device is described by dividing it into various modules according to its functions. Of course, when implementing the present disclosure, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0271] The device of the above embodiment is used to implement the corresponding vehicle control method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0272] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the vehicle control method described in any of the above embodiments is implemented.

[0273] Figure 4 A more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment is shown, and the device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are connected to each other through the bus 1050 in the device.

[0274] The processor 1010 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0275] The memory 1020 may be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.

[0276] The input / output interface 1030 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0277] The communication interface 1040 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB, network cable, etc.) or a wireless mode (such as mobile network, WIFI, Bluetooth, etc.).

[0278] The bus 1050 includes a path that transmits information between the various components of the device (eg, the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040).

[0279] It should be noted that, although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0280] The electronic device of the above embodiment is used to implement the corresponding vehicle control method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0281] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present disclosure also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the vehicle control method described in any of the above embodiments.

[0282] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0283] The computer instructions stored in the storage medium of the above embodiments are used to enable the computer to execute the vehicle control method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0284] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present application also provides a vehicle, including the vehicle control device in the above-mentioned embodiments, the electronic device in the above-mentioned embodiments, and the computer-readable storage medium in the above-mentioned embodiments, and the vehicle equipment implements the vehicle control method described in any of the above embodiments.

[0285] The vehicle of the above-mentioned embodiment is used to implement the vehicle control method described in any of the above-mentioned embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0286] It is understandable that before using the technical solutions of each embodiment of the present disclosure, the type, scope of use, usage scenarios, etc. of the personal information involved will be informed to the user in an appropriate manner, and the user's authorization will be obtained.

[0287] For example, in response to receiving an active request from a user, a prompt message is sent to the user to clearly remind the user that the operation requested to be performed will require obtaining and using the user's personal information. Thus, the user can independently choose whether to provide personal information to software or hardware such as an electronic device, application, server, or storage medium that performs the operation of the technical solution of the present disclosure according to the prompt message.

[0288] As an optional but non-limiting implementation, in response to receiving the user's active request, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. In addition, the pop-up window may also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0289] It is understandable that the above notification and the process of obtaining user authorization are merely illustrative and do not constitute a limitation on the implementation of the present disclosure. Other methods that meet relevant laws and regulations may also be applied to the implementation of the present disclosure.

[0290] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present disclosure (including the claims) is limited to these examples. Based on the concept of the present disclosure, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present disclosure as described above, which are not provided in detail for the sake of simplicity.

[0291] In addition, to simplify the description and discussion, and in order not to make the embodiments of the present disclosure difficult to understand, the known power / ground connections to the integrated circuit (IC) chips and other components may or may not be shown in the provided figures. In addition, the device can be shown in the form of a block diagram to avoid making the embodiments of the present disclosure difficult to understand, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform on which the embodiments of the present disclosure will be implemented (that is, these details should be fully within the scope of understanding of those skilled in the art). Where specific details (e.g., circuits) are set forth to describe exemplary embodiments of the present disclosure, it is apparent to those skilled in the art that the embodiments of the present disclosure can be implemented without these specific details or with changes in these specific details. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0292] Although the present disclosure has been described in conjunction with specific embodiments of the present disclosure, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0293] The embodiments of the present disclosure are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present disclosure should be included in the scope of protection of the present disclosure.

Claims

1. A vehicle control method, characterized in that: include: receiving an echo signal corresponding to the obstacle detection signal, and determining a moving speed of the obstacle according to the echo signal; Performing principal component analysis on the echo signal to obtain a target principal component corresponding to the echo signal; Inputting the target principal component into a pre-trained obstacle classification model, processing the obstacle classification model, and outputting the obstacle type; The vehicle is controlled to travel according to the moving speed of the obstacle and the type of the obstacle to avoid the obstacle.

2. The method according to claim 1, characterized in that The performing principal component analysis on the echo signal to obtain a target principal component corresponding to the echo signal includes: Performing analog-to-digital conversion on the echo signal to obtain a digital signal corresponding to the echo signal; Determine a correlation coefficient matrix corresponding to the digital signal, and determine a projection matrix according to the correlation coefficient matrix; The projection matrix is ​​used to perform dimensionality reduction processing on the digital signal to obtain the target principal component corresponding to the echo signal.

3. The method according to claim 2, characterized in that The determining of the correlation coefficient matrix corresponding to the digital signal comprises: Determine an initial signal data matrix corresponding to the digital signal; Performing data preprocessing on the initial signal data matrix to obtain a signal data matrix; A correlation calculation is performed on the signal data matrix to obtain a correlation coefficient matrix corresponding to the digital signal.

4. The method according to claim 2, characterized in that: The step of determining a projection matrix according to the correlation coefficient matrix comprises: Determine a plurality of eigenvalues ​​corresponding to the correlation coefficient matrix and an eigenvector corresponding to each eigenvalue; Sort all eigenvalues ​​in descending order; Obtain a preset dimension, select the eigenvalue of the preset dimension in sequence as the target eigenvalue, and use the eigenvector corresponding to each target eigenvalue as the target eigenvector; Construct a projection matrix based on all target eigenvectors.

5. The method according to claim 1, characterized in that The determining the moving speed of the obstacle according to the echo signal comprises: Performing analog-to-digital conversion on the echo signal to obtain a digital signal corresponding to the echo signal; Performing windowing processing on the digital signal to obtain a signal to be processed; Performing Fourier transform on the signal to be processed to extract Doppler frequency shift data; The wavelength and center frequency of the obstacle detection signal are obtained, and the moving speed of the obstacle is determined according to the wavelength, the center frequency and the Doppler frequency shift data.

6. The method according to claim 5, characterized in that The controlling the vehicle to travel according to the moving speed of the obstacle and the type of the obstacle to avoid the obstacle includes: Acquire a time delay corresponding to the echo signal, and determine a relative distance between the obstacle and the vehicle according to the time delay; Determining the angle between the obstacle detection signal and the obstacle movement direction according to the Doppler frequency shift data; The target driving state of the vehicle is determined according to the moving speed of the obstacle, the type of obstacle, the relative distance and the angle, and the vehicle is controlled to avoid the obstacle according to the target driving state.

7. The method according to claim 1, characterized in that The training process of the obstacle classification model includes: Acquire a preset principal component data set and an initial obstacle classification model, wherein the preset principal component data set includes a training principal component data set, and the training principal component data set includes training principal component data and a first actual obstacle type; Using the training principal component data set to train the initial obstacle classification model to obtain training obstacle types; Determining a regularization objective function according to the training obstacle type and the first actual obstacle type; Until the regularized objective function is minimized, it is determined that the training of the initial obstacle classification model is completed, and an obstacle classification model is obtained.

8. The method according to claim 7, characterized in that The preset principal component data set also includes a verification principal component data set, and the verification principal component data set includes verification principal component data and a second actual obstacle type; After obtaining the obstacle classification model, it also includes: Input the verification principal component data set into the obstacle classification model, process it through the obstacle classification model, and output the verification obstacle type; Determining the accuracy of an obstacle classification model according to the verified obstacle type and the second actual obstacle type; In response to the accuracy being less than a preset accuracy threshold, model parameters of the obstacle classification model and / or dimensions corresponding to the principal component analysis process are adjusted, and the obstacle classification model is retrained until the accuracy is greater than or equal to the preset accuracy threshold.

9. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method according to any one of claims 1 to 8 is implemented.

10. A vehicle, characterized in that: The vehicle includes the electronic device according to claim 9.

Images