CN111930117B

CN111930117B - Steering-based lateral control method, device, equipment and storage medium

Info

Publication number: CN111930117B
Application number: CN202010763285.XA
Authority: CN
Inventors: 孙子文; 李斌; 霍达; 韩旭
Original assignee: Guangzhou Jingqi Technology Co ltd
Current assignee: Guangzhou Jingqi Technology Co ltd
Priority date: 2020-07-31
Filing date: 2020-07-31
Publication date: 2024-04-05
Anticipated expiration: 2040-07-31
Also published as: CN111930117A

Abstract

The embodiment of the invention provides a lateral control method, a device, equipment and a storage medium based on steering, wherein the method comprises the following steps: when the vehicle turns around a curve, the type and the deviation value of the curve are calculated, the lane line of the curve is used as a reference, the first target deviation value representing emergency steering and the second target deviation value representing non-emergency steering are divided by part of the deviation values, the first target deviation value and the second target deviation value are used as classified samples, an event recognition module matched with the type of the curve is updated, a target event recognition model is obtained, the part of the deviation values are input into the target event recognition model for classification, the operation representing emergency steering is recognized, the steering of the vehicle around the curve is transversely controlled according to the operation representing emergency steering, the deviation value of the vehicle driven by a user is collected in real time, the previous event recognition model is used as a basis for continuous training, the training quantity is small, the requirement of real-time performance is met, and the event recognition model conforming to the driving style of the user is gradually learned.

Description

Steering-based lateral control method, device, equipment and storage medium

Technical Field

The embodiment of the invention relates to the technical field of automatic control, in particular to a transverse control method, a device, equipment and a storage medium based on steering.

Background

When a user drives a vehicle, steering is a common operation along with the change of a road, which is also called turning, and in some cases, the user can steer greatly, which is also called sharp turning, so that the vehicle can deviate from the controllable range of the user, drift is caused, the comfort of passengers is reduced, and safety risks can occur.

Therefore, the automatic driving system can detect the large-amplitude steering and intervene the large-amplitude steering, so that the comfort of passengers is improved, and the safety risk is reduced.

Conventionally, in order to detect a large-scale steering, it is common to detect the orientation of a vehicle and set a corresponding static threshold value, and if the change in the orientation exceeds or falls below the threshold value, the steering is regarded as a large-scale steering.

However, the threshold is an empirical value, and needs to be continuously adjusted according to the conditions of different users, so that the operation is complicated.

Disclosure of Invention

The embodiment of the invention provides a transverse control method, a device, equipment and a storage medium based on steering, which are used for solving the problem that the operation for detecting large-scale steering by a user is complicated.

In a first aspect, an embodiment of the present invention provides a steering-based lateral control method, including:

when the vehicle is detected to turn at a curve, calculating the type and the deviation value of the curve, wherein the deviation value represents the degree of deviation of the vehicle from a standard direction;

dividing a part of the deviation values into a first target deviation value representing emergency steering and a second target deviation value representing non-emergency steering by taking a lane line of the curve as a reference;

updating an event recognition module matched with the curve type by taking the first target deviation value and the second target deviation value as classified samples to obtain a target event recognition model;

inputting a portion of the deviation value into the target event recognition model for classification to recognize an operation indicative of an emergency steering;

and transversely controlling the steering of the vehicle at the curve according to the emergency steering operation.

In a second aspect, an embodiment of the present invention further provides a steering-based lateral control device, including:

the deviation value calculating module is used for calculating the type and the deviation value of the curve when the vehicle is detected to turn in the curve, and the deviation value represents the degree of the vehicle deviating from the standard direction;

The deviation value dividing module is used for dividing a part of the deviation values into a first target deviation value representing emergency steering and a second target deviation value representing non-emergency steering by taking a lane line of the curve as a reference;

the target event recognition model training module is used for updating the event recognition module matched with the curve type by taking the first target deviation value and the second target deviation value as classified samples to obtain a target event recognition model;

the deviation value classification module is used for inputting part of the deviation value into the target event identification model to classify so as to identify an operation representing emergency steering;

and the transverse control module is used for transversely controlling the steering of the vehicle at the curve according to the emergency steering operation.

In a third aspect, an embodiment of the present invention further provides a computer apparatus, including:

one or more processors;

a memory for storing one or more programs,

the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the steering-based lateral control method as described in the first aspect.

In a fourth aspect, embodiments of the present invention also provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steering-based lateral control method according to the first aspect.

In this embodiment, when it is detected that a vehicle turns around a curve, the type and the deviation value of the curve are calculated, the lane line of the curve is used as a reference, the partial deviation value is divided into a first target deviation value representing emergency turning and a second target deviation value representing non-emergency turning, the first target deviation value and the second target deviation value are used as classified samples, an event recognition module matched with the type of the curve is updated, a target event recognition model is obtained, the partial deviation value is input into the target event recognition model to be classified, an operation representing emergency turning is recognized, the steering of the vehicle around the curve is transversely controlled according to the operation of emergency turning, the type of the curve is used as a training event recognition model, and the condition of emergency acceleration and deceleration is recognized, so that the calculated amount can be reduced, the facing range of the event recognition model can be reduced, the accuracy of the event recognition model can be guaranteed, the deviation value of a user driving vehicle can be collected in real time, individuation and authenticity of the deviation value can be guaranteed, the prior event recognition model is used as a basis for continuous training, the training amount is small, the requirement of real-time is met, the steering of the user can be controlled in a step-by step, the user can recognize the user recognition mode is recognized by the type, and the user can be used to recognize the emergency driving mode, the user, and the steering operation is convenient, and the user can be controlled in a safe driving mode, and convenient mode is convenient, and convenient is provided.

Drawings

Fig. 1 is a schematic structural diagram of an unmanned vehicle according to an embodiment of the present invention;

FIG. 2 is a flow chart of a lateral steering control method according to a first embodiment of the present invention;

FIG. 3 is an exemplary diagram of an emergency steering provided in accordance with a first embodiment of the present invention;

fig. 4 is a flowchart of a lateral control method based on steering according to a second embodiment of the present invention;

FIG. 5 is a schematic diagram of an event recognition model according to a second embodiment of the present invention;

FIG. 6 is a diagram illustrating a relationship between an event recognition model according to a second embodiment of the present invention;

fig. 7 is a schematic structural diagram of a steering-based lateral control device according to a third embodiment of the present invention;

fig. 8 is a schematic structural diagram of a computer device according to a fourth embodiment of the present invention.

Detailed Description

The invention is described in further detail below with reference to the drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting thereof. It should be further noted that, for convenience of description, only some, but not all of the structures related to the present invention are shown in the drawings.

Referring to fig. 1, there is shown an unmanned vehicle 100 to which an embodiment of a steering-based lateral control device in an embodiment of the present invention may be applied.

As shown in fig. 1, the unmanned vehicle 100 may include a drive control apparatus 101, a body bus 102, an ECU (Electronic Control Unit ) 103, an ECU 104, an ECU 105, a sensor 106, a sensor 107, a sensor 108, and an actuator 109, an actuator 110, and an actuator 111.

The driving control apparatus (also referred to as an onboard brain) 101 is responsible for overall intelligent control of the entire unmanned vehicle 100. The driving control apparatus 101 may be a separately provided controller such as a Programmable logic controller (Programmable LogicController, PLC), a single-chip microcomputer, an industrial controller, or the like; the device can also be equipment consisting of other electronic devices with input/output ports and operation control functions; but also a computer device installed with a vehicle driving control type application. The driving control device may analyze and process data sent from each ECU and/or data sent from each sensor received on the body bus 102, make a corresponding decision, and send an instruction corresponding to the decision to the body bus.

The body bus 102 may be a bus for connecting the driving control apparatus 101, the ECU 103, the ECU 104, the ECU 105, the sensor 106, the sensor 107, the sensor 108, and other apparatuses not shown of the unmanned vehicle 100. Because of the wide acceptance of high performance and reliability of CAN (Controller AreaNetwork ) buses, the body bus commonly used in motor vehicles is currently the CAN bus. Of course, it is understood that the body bus may be other types of buses.

The body bus 102 may send the instruction sent by the driving control device 101 to the ECU 103, the ECU 104, the ECU 105, and the ECU 103, the ECU 104, and the ECU 105 analyze the instruction and send the instruction to the corresponding executing device for execution.

The sensors 106, 107, 108 include, but are not limited to, lidar, cameras, acceleration sensors, angle sensors, and the like.

It should be noted that, the steering-based lateral control method provided by the embodiment of the present invention may be performed by the driving control apparatus 101, and accordingly, the steering-based lateral control device is generally disposed in the driving control apparatus 101.

It should be understood that the number of unmanned vehicles, drive control devices, body buses, ECUs, actuators, and sensors in fig. 1 are merely illustrative. There may be any number of unmanned vehicles, drive control devices, body buses, ECU's, and sensors, as desired for implementation.

Example 1

Fig. 2 is a flowchart of a steering-based lateral control method according to an embodiment of the present invention, where the method may be applied to identify an emergency steering operation by a user-adaptive operation, and the method may be performed by a steering-based lateral control device, which may be implemented by software and/or hardware, may be configured in a computer device, for example, a driving control device, or the like, and specifically includes the following steps:

Step 201, when detecting that the vehicle turns in a curve, calculating the type and the deviation value of the curve.

In this embodiment, when the user drives the vehicle, an automatic driving mode may be started, and the automatic driving mode may refer to a mode in which the vehicle itself has environmental awareness, path planning, and vehicle control is autonomously implemented, that is, humanoid driving performed by controlling the vehicle with an electronic technology.

Driving modes can be classified into L0 non-Automation (No Automation), L1 driver assistance (Driver Assistance), L2 partial Automation (Partial Automation), L3 conditional Automation (Conditional Automation), L4 High Automation (High Automation), and L5 Full Automation (Full Automation) according to the degree of grasping of the vehicle manipulation task.

The automatic driving mode in this embodiment may refer to the driving modes in L1 to L3, and has an auxiliary function for the user to drive the vehicle.

In a specific implementation, the external environment and/or the internal environment of the vehicle can be detected, whether the vehicle turns around a curve or not can be judged, if yes, the acquisition angle of an angle sensor arranged in the vehicle can be continuously called in real time, the frequency for acquiring the angle is generally more than 10Hz, the angles are ordered according to time, deviation values are calculated, the deviation values represent the degree of the vehicle deviating from the standard direction, the deviation values can form a data sequence, and therefore the deviation values are used for identifying the emergency turning operation, namely the large-amplitude turning operation relative to a user, and the emergency turning generally has a certain risk.

For example, as shown in fig. 3, when the vehicle 301 is traveling on a road in the arrow direction, if the speed of the vehicle 301 is fast, the lateral movement is obvious during steering, and the driver is difficult to control and easy to drive into an adjacent lane, at this time, the vehicle 301 may detect an emergency steering event, and the steering progress of the vehicle 321 may be adjusted by assisting other measures, so as to avoid risks of driving into the adjacent lane, and of rubbing and collision with the vehicle in the adjacent lane.

It should be noted that, the condition for detecting the curve steering may be set by those skilled in the art according to the actual service requirement, and this embodiment is not limited thereto, and the curve steering is used as the condition for training the event recognition model and recognizing the operation of emergency steering, so that not only the calculation amount but also the facing range of the event recognition model may be reduced, thereby ensuring the accuracy of the event recognition model.

In one way of detecting a turn around a curve, video data may be collected to the outside of the vehicle, with a lane line as a detection target, the lane line of the lane in which the vehicle is located is detected in the video data, and the curvature of the lane line is calculated.

In one example, original image data in video data is converted into image data in an HSL (hue H, saturation S, luminance L) format, yellow and white are separated from the image data in the HSL format, and the image data in the HSL format after the yellow and white are separated is combined with the original image data to obtain target image data.

And converting the target image data into gray image data, smoothing the edge by Gaussian blur, and obtaining edge information by Canny edge detection on the smoothed gray image data.

Tracking the region of the interested edge information, removing the information of other regions, performing Hough transformation on the region of the interested edge information to obtain lane lines in the region of the interest, tracking the lane lines by using a specific color (such as red), separating a left lane from a right lane, and subsequently inserting a linear gradient to create two complete and smooth lane lines.

If the curvature is larger than the preset curvature threshold value, the curve degree of the lane is larger, and the lane can be determined to be a curve and the vehicle turns in the curve.

At this time, a curvature range in which the curvature is located is searched for in a plurality of preset curvature ranges as a target range, wherein each curvature range is associated with a type such as a small curve, a medium curve, a large curve, and the like.

The type of the target range association is set as the type of the curve in which the current vehicle is located.

The angle sensor is called to acquire the angle of the vehicle on the curve, namely the direction of the vehicle.

And calculating a difference value between the angle and the standard direction by taking the designated direction as the standard direction, and taking the difference value as a deviation value.

In addition, after the angle is acquired, the angle may be preprocessed, so that subsequent calculation of the deviation value is facilitated, for example, data that the difference between the current angle and the angle at the previous and subsequent moments exceeds the angle threshold is removed, which is not limited in this embodiment.

Step 202, taking a lane line of a curve as a reference, and dividing a part of deviation values into a first target deviation value representing emergency steering and a second target deviation value representing non-emergency steering.

In a specific implementation, a user usually drives a vehicle within the capability range of the vehicle, the situation that emergency steering occurs is less, and in the emergency steering, the control degree of the vehicle is reduced, so that the relationship between the vehicle and a lane line can be represented, therefore, the lane line of a curve can be taken as a reference, a first target part deviation value with a higher number of partial deviation values acquired in advance is divided, the operation of emergency steering is represented, and a second target part deviation value with a lower number of partial deviation values acquired in advance is divided, and the operation of non-emergency steering is represented.

In one example, a target lane is determined, the direction of which is opposite to the direction in which the vehicle turns, i.e., if the vehicle turns left, the target lane is the lane on the right side of the vehicle, and if the vehicle turns right, the target lane is the lane on the left side of the vehicle.

The geometric relationship is used in the video data to measure the distance between the vehicle and the target lane line, the distance is a series of data, and the difference between every two adjacent distances can be calculated to determine the change trend of the distance.

If the change trend of the distance is monotonous decrease, the deviation value is determined as a first target deviation value representing emergency steering.

If the change trend of the distance is oscillation, namely the distance is maintained in a preset safety range, the deviation value is determined to be a second target deviation value representing non-emergency steering.

Of course, the above manner of dividing the first target deviation value and the second target deviation value is merely an example, and in implementing the embodiment of the present invention, other manners of dividing the first target deviation value and the second target deviation value may be set according to actual situations, for example, n deviation values with highest kurtosis values, skewness values, etc. are set as the first target deviation value, other deviation values are set as the second target deviation value, etc., which is not limited to this embodiment of the present invention. In addition, in addition to the above-mentioned manners of dividing the first target deviation value and the second target deviation value, those skilled in the art may also adopt other manners of dividing the first target deviation value and the second target deviation value according to actual needs, which is not limited in the embodiment of the present invention.

And 203, updating an event recognition module matched with the type of the curve by taking the first target deviation value and the second target deviation value as classified samples to obtain a target event recognition model.

In a specific implementation, the server collects deviation values occurring in different types of curves and marks urgent acceleration and non-urgent acceleration, so that the deviation values are used as classified samples, and an event recognition model which is universal in the types of curves is trained, namely the event recognition model can be used for recognizing deviation values of urgent steering and deviation values of non-urgent steering.

The event recognition model is a two-class model, and may be a mechanical learning model, such as an SVM (SupportVector Machine ), a Logistic (regression model), or a neural network, which is not limited in this embodiment.

Upon completion of the training, the server may distribute the event recognition model into the vehicle.

In this embodiment, on the basis of the initial event recognition model, the event recognition model may be continuously trained for driving styles of different users, that is, the event recognition model may be trained by taking the partial deviation value collected previously as a sample, to obtain the target event recognition model, and the target event recognition model may be stored in the vehicle as the event recognition model, and waiting for continuous training on the basis, so the event recognition model matched with the curve type may be the initial general event recognition model or the event recognition model continuously trained, which is not limited in this embodiment.

Step 204, inputting the partial deviation value into the target event recognition model for classification to recognize the operation representing the emergency steering.

In this embodiment, for the same driving operation triggered by the same user, a part of the deviation values acquired later may be input into the target event recognition model, so as to classify the deviation values, so as to recognize an operation representing emergency steering and an operation representing non-emergency steering.

In order to identify the identity of the user, the identity of the user can be identified through the information (such as a user account number) that the user logs in the vehicle directly or logs in the associated device when the vehicle is started, or the identity of the user can be determined by calling a camera in the vehicle to collect image data facing the driving position and performing face recognition on the image data, and the embodiment is not limited to the above.

After confirming the identity of the user, the driving operation of the vehicle triggered by the user between start and stop may be regarded as the same driving operation triggered by the same user.

If the identity of the user is not recognized, the driving operation of the driver's seat side door between the two opening and closing operations may be regarded as the same driving operation triggered by the same user.

Step 205, the steering of the vehicle in the curve is transversely controlled according to the emergency steering operation.

If the operation of triggering the emergency steering by the user is detected, the vehicle can be transversely controlled by taking the lane line as a reference and referring to the situation of the lane line, namely, the transverse movement is controlled.

In the specific implementation, on one hand, the braking force of the steering is increased, namely the steering angle is increased, the steering deviation value is increased, the vehicle is prevented from deviating from the current lane and entering an adjacent lane, and on the other hand, the speed (namely the vehicle speed) can be reduced, and the control degree of the user on the vehicle is increased, so that the distance between the vehicle and the lane line is kept within a preset safety range.

Example two

Fig. 4 is a flowchart of a lateral control method based on steering according to a second embodiment of the present invention, where the method is based on the foregoing embodiment, and further refines operations of finding an event recognition model, training a target event recognition model, and recognizing an emergency steering, and the method specifically includes the following steps:

step 401, when it is detected that the vehicle turns around a curve, calculating the type and deviation value of the curve.

Wherein the deviation value indicates the degree to which the vehicle deviates from the standard direction.

Step 402, taking a lane line of a curve as a reference, dividing a partial deviation value into a first target deviation value representing emergency steering and a second target deviation value representing non-emergency steering.

Step 403, searching an event recognition model trained for the type of curve.

In this embodiment, an event recognition model, which is distributed by the server and trained for the type of the current curve, is locally extracted at the current vehicle, and the event recognition model is associated with a standard deviation value, which represents a characteristic of a deviation value (i.e., a second target deviation value) used to train the event recognition model, which identifies a non-emergency steering.

Step 404, calculating a correlation between the second target deviation value and the standard deviation value.

After determining the event recognition model, the second target bias value may be compared with a standard bias value of the event recognition model, and a correlation between the two may be calculated, thereby measuring the degree of closeness between the two.

The standard deviation value has two forms, one of which is a data point representing the average value of the samples of the prior training event recognition model (second target deviation value) and the other of which is a data range representing the amplitude of the samples of the prior training event recognition model (second target deviation value) (i.e., the range of the same location data point between the maximum and minimum values).

If the standard deviation value is a data point, the similarity between the second target deviation value and the standard deviation value can be calculated as the correlation through an algorithm such as EDR, LCSS, DTW.

If the standard deviation value is the data range, determining the data points which fall into the data range in the second target deviation value as target points, and counting the proportion of the target points to the second target deviation value as the correlation.

Of course, the foregoing manner of calculating the correlation is merely an example, and other manners of calculating the correlation may be set according to actual situations when implementing the embodiment of the present invention, which is not limited thereto. In addition, in addition to the above-mentioned ways of calculating the correlation, those skilled in the art may also use other ways of calculating the correlation according to actual needs, which is not limited in this embodiment of the present invention.

Step 405, selecting an original event recognition model from the event recognition models based on the correlation.

In this embodiment, the operation of non-emergency steering belongs to a relatively stable operation, and may represent the driving style of the user, that is, the second target deviation value identifying the non-emergency steering may represent the driving style of the user, so that in the event recognition model trained for the service scenario, an event recognition model suitable for processing the second target deviation value (i.e., matching the driving style of the user) may be searched for as the original event recognition model.

In general, the higher the correlation between the second target deviation value and the standard deviation value of the event recognition model, the higher the adaptation degree between the event recognition model and the driving style of the current user is, whereas the lower the correlation between the second target deviation value and the standard deviation value of the event recognition model is, the lower the adaptation degree between the event recognition model and the driving style of the current user is, so in this embodiment, the appropriate candidate event recognition model may be selected as the original event recognition model by referring to the correlation between the different second target deviation values and the standard deviation values of the event recognition model.

In one approach, an average of the correlations may be calculated and compared to a preset correlation threshold.

If the average value of the correlation is greater than or equal to a preset correlation threshold, a discrete value of the correlation is calculated, and the discrete value represents the discrete degree of the correlation, such as variance, standard deviation and the like.

And selecting the event recognition model with the minimum discrete value as the original event recognition model, so that the performance of the original event recognition model is kept stable, and the robustness of the original event recognition model is improved.

If the average value of the correlations is smaller than a preset correlation threshold, selecting the event recognition model with the smallest average value of the correlations as the original event recognition model, namely selecting the original event recognition model closest to the sample, and ensuring the accuracy of the original event recognition model.

Of course, the above manner of selecting the original event recognition model is merely taken as an example, and other manners of selecting the original event recognition model may be set according to practical situations when implementing the embodiment of the present invention, for example, calculating the sum of all correlations, taking the sum of all correlations as the total correlation, selecting the event recognition model with the highest total correlation as the original event recognition model, and so on. In addition, in addition to the above-mentioned manner of selecting the original event recognition model, those skilled in the art may also use other manners of selecting the original event recognition model according to actual needs, which is not limited in this embodiment of the present invention.

And step 406, updating the original event recognition module by taking the first target deviation value and the second target deviation value as classified samples to obtain a target event recognition model.

In this embodiment, for the first target deviation value, an emergency steering operation may be identified, for the second target deviation value, a non-emergency steering operation may be identified, and the first target deviation value and the second target deviation value are used as classified samples, and the original event recognition model is continuously trained to obtain the target event recognition model, so as to further improve the adaptation degree of the target event recognition model and the driving style of the user.

It should be noted that, the original event recognition model may ensure a certain accuracy, so, on one hand, before training the target event recognition model is completed, the operation of identifying the emergency steering from the deviation value by the original event recognition model may be invoked under the same type of curve, and when training the target event recognition model is completed, the operation of identifying the emergency acceleration and deceleration from the deviation value by the original event recognition model is switched to the target event recognition model, and on the other hand, the operation of identifying the emergency acceleration and deceleration from the deviation value by the target event recognition model is invoked under the same type of curve, and on the other hand, the iteration times are used as conditions for stopping training, that is, when the iteration training reaches the preset times, the training of the target event recognition model is considered to be completed, thereby ensuring instantaneity.

In one embodiment of the present invention, step 406 may include the steps of:

step 4061, a deviation value identifying an emergency steering is obtained as a new first target deviation value.

In this embodiment, the difference between the first target deviation value and the second target deviation value may be relatively small, and in order to prevent overfitting during training, some operations indicating typical emergency steering, that is, deviation values indicating emergency steering may be set in advance for event recognition models corresponding to different types of curves, and distributed to respective vehicles.

After determining the event recognition model, the deviation value of the emergency steering can be locally extracted from the current vehicle as a new first target deviation value and combined with the original first target deviation value.

Step 4062, extracting first sample features from all the first target bias values.

In this embodiment, for each first target deviation value (including the original first target deviation value and the new first target deviation value), features of dimensions such as association degree, waveform, statistics and the like may be extracted from the first target deviation value as first sample features, and an emergency is marked as a Tag (Tag).

In one example, the first sample feature includes at least one of a first sample residual, a first sample statistical feature, a second sample statistical feature, and a second sample residual, then in this example, a standard deviation value associated with the original event recognition model may be found, and a difference in the same position between the first target deviation value and the standard deviation value is calculated as the first sample residual.

If the standard deviation value is a data point, the difference value at the same position can be directly calculated with the first target deviation value, and if the standard deviation value is a data range, the intermediate value of the data range is calculated, so that the difference value at the same position is calculated with the first target deviation value.

And calculating data such as an average value, a maximum value, a minimum value, a variance, a skewness value, a kurtosis value and the like for the first residual as a first sample statistical characteristic.

And calculating data such as an average value, a maximum value, a minimum value, a variance, a skewness value, a kurtosis value and the like of the first target skewness value as a second sample statistical characteristic.

And calculating the difference value between the second sample statistical characteristic and the standard statistical characteristic (such as data of average value, maximum value, minimum value, variance, skewness value, kurtosis value and the like) of the standard deviation value at the same position, and taking the difference value as a second sample residual error.

Of course, the above first sample feature is merely an example, and other first sample features may be set according to actual situations when implementing the embodiment of the present invention, which is not limited thereto. In addition, in addition to the first sample feature described above, those skilled in the art may also use other first sample features according to actual needs, which are not limited by the embodiments of the present invention.

Step 4063, extracting a second sample feature from the second target bias value.

In this embodiment, for each second target deviation value, features of dimensions such as association degree, waveform, statistics, and the like may be extracted therefrom as first sample features, and emergency as a Tag (Tag) may be marked.

In one example, the second sample feature includes at least one of a third sample residual, a third sample statistical feature, a fourth sample statistical feature, and a fourth sample residual, and in this example, a standard deviation value associated with the original event recognition model may be found, and a difference value at the same position between the second target deviation value and the standard deviation value may be calculated as the third sample residual.

It should be noted that, if the standard deviation value is a data point, the difference value at the same position may be directly calculated with the second target deviation value, and if the standard deviation value is a data range, the intermediate value of the data range is calculated, so that the difference value at the same two positions is calculated with the first target deviation value.

And calculating data such as an average value, a maximum value, a minimum value, a variance, a skewness value, a kurtosis value and the like of the second residual as a third sample statistical characteristic.

And calculating data such as an average value, a maximum value, a minimum value, a variance, a skewness value, a kurtosis value and the like of the second target skewness value as a fourth sample statistical characteristic.

And calculating the difference value between the second sample statistical characteristic and the standard statistical characteristic (such as data of average value, maximum value, minimum value, variance, skewness value, kurtosis value and the like) of the standard deviation value at the same position, and taking the difference value as a fourth sample residual error.

Of course, the above second sample feature is merely an example, and other second sample features may be set according to actual situations when implementing the embodiment of the present invention, which is not limited thereto. In addition, in addition to the second sample feature, those skilled in the art may also use other second sample features according to actual needs, which is not limited by the embodiment of the present invention.

And 4064, performing migration learning on the original event recognition model by taking the first sample characteristic and the second sample characteristic as samples and taking emergency steering and non-emergency steering as classification targets to obtain a target event recognition model.

In this embodiment, the first sample feature and the second sample feature may be used as classified samples, the emergency steering and the non-emergency steering may be used as classification targets, and the first sample feature and the second sample feature may be used as samples to perform migration learning on the original event recognition model, so as to obtain the target event recognition model.

The transfer learning refers to transferring the parameters of the trained original event recognition model to a new target event recognition model to help the training of the target event recognition model, and considering that most data or tasks have correlation, the learned parameters can be shared to the new target event recognition model in a certain way through the transfer learning, so that the learning efficiency of the target event recognition model is quickened and optimized, and the instantaneity is ensured.

In a specific implementation, the migration learning can be performed on the original event recognition model by one of the following ways:

(1) Transfer Learning: all convolution layers of the pre-training model (original event recognition model) are frozen and only the custom full join layer is trained.

(2) Extract Feature Vector: the feature vectors (first sample features and second sample features) of the convolutional layer of the pre-training model (original event recognition model) on all training and testing data are calculated, then the pre-training model (original event recognition model) is thrown away, and only the customized full-connection network of the simplified configuration version is trained.

(3) Fine-tune: the partial convolution layers (typically the most convolution layers near the input) of the pre-training model (the original event recognition model) are frozen, and the remaining convolution layers (typically the partial convolution layers near the output) and the fully connected layers are trained.

In the process of transfer learning, the classification (urgent, non-urgent) of the sample prediction can be compared with the actual classification (urgent, non-urgent), so that a loss value in each round of iterative training is calculated, and parameters in the original event recognition model are updated in a manner of being based on the loss value, using gradient descent, random gradient descent and the like.

In addition, when training of the target event recognition model is completed, a standard deviation value is generated based on the second target deviation value, so that an association relationship between the target event recognition model and the second target deviation value is established and stored in the local of the current vehicle.

In one example, an average of the data points at the same position in the second target deviation value may be calculated as the data points of the standard deviation value.

In another example, the magnitudes (i.e., the range between the maximum and minimum values) of the data points at the same position in the second target deviation value may be counted as the data range of the standard deviation value.

Of course, the above-mentioned manner of calculating the standard deviation value is merely an example, and other manners of calculating the standard deviation value may be set according to actual situations when implementing the embodiment of the present invention, which is not limited thereto. In addition, in addition to the above-mentioned mode of calculating the standard deviation value, those skilled in the art may also use other modes of calculating the standard deviation value according to actual needs, which is not limited in this embodiment of the present invention.

Step 407, extracting the target feature from the partial deviation value.

In this embodiment, the deviation value may be collected in the same type of curve, and the characteristics of dimensions such as association degree, waveform, statistics, etc. may be extracted therefrom as the target characteristics.

In one example, the target feature includes at least one of a first target residual, a first target statistical feature, a second target statistical feature, and a second target residual, and then in this example, a standard deviation value associated with the target event identification model may be found, and a difference between the deviation value and the standard deviation value may be calculated as the first target residual.

Calculating data such as an average value, a maximum value, a minimum value, a variance, a skewness value, a kurtosis value and the like of the first target residual error to serve as a first target statistical feature;

calculating data such as an average value, a maximum value, a minimum value, a variance, a skewness value, a kurtosis value and the like of the deviation value as a second target statistical feature;

and calculating the difference between the second target statistical feature and the standard statistical feature (such as data of average value, maximum value, minimum value, variance, skewness value, kurtosis value and the like) of the standard deviation value as a second target residual.

Of course, the above target features are merely examples, and other target features may be set according to actual situations when implementing the embodiment of the present invention, which is not limited thereto. In addition, other target features besides those described above may be adopted by those skilled in the art according to actual needs, and the embodiments of the present invention are not limited thereto.

Step 408, in the convolutional neural network of the target event recognition model, performing convolutional processing on the target feature to output a candidate feature.

Step 409, calculating residual features for the candidate features in the residual network of the target event identification model.

Step 410, in the long-term and short-term memory network of the target event recognition model, the residual characteristics are mapped to output the category of the deviation value.

Step 411, if the category is emergency steering, determining that the deviation value indicates an emergency steering operation.

In order to ensure real-time performance, the structure of the event recognition model (including the current target event recognition model) is designed to be simpler, the target event recognition model belongs to a model under a specified curve type, the oriented scenes are more concentrated, and the simple structure can still keep higher accuracy.

In this embodiment, as shown in fig. 5, the event recognition model has a three-layer structure, which is respectively:

1. convolutional neural network (Convolutional Neural Network, CNN) 510

CNN is a type of feedforward neural network (Feedforward Neural Networks) that includes convolution computation and has a deep structure, and is one of algorithms of deep learning (deep learning). CNNs have the ability to characterize learning (representation learning), and can perform a Shift-invariant classification (Shift-invariant classification) on input information in its hierarchical structure, and are therefore also referred to as "Shift-invariant artificial neural networks (sia)".

The visual perception (visual perception) mechanism of the CNN imitation biology is constructed, and can perform supervised learning and unsupervised learning, and the convolution kernel parameter sharing and the sparsity of interlayer connection in the hidden layer enable the convolution neural network to perform grid-like feature with small calculation amount.

2. Residual network 520

In general, each layer of the network corresponds to extracting feature information of different layers, including a lower layer, a middle layer and a higher layer, and when the network is deeper, the more information of different layers is extracted, the more the combination of layer information among different layers is, the higher the "level" of the feature becomes with the deepening of the depth of the network, and the depth of the network is an important factor for realizing good effect, however, gradient dispersion/explosion becomes an obstacle for training the deep layer network, and cannot converge.

The residual network is introduced into the event recognition model, and when the event recognition model is transmitted in the forward direction, an input signal can be directly transmitted to a high layer from any low layer, so that the problem of network degradation can be solved to a certain extent due to the inclusion of an identity mapping, an error signal can be directly transmitted to the low layer without any intermediate weight matrix transformation, the problem of gradient dispersion can be relieved to a certain extent, the information can be transmitted more smoothly in the forward and backward directions, the problems of gradient disappearance and gradient explosion in the training process of the event recognition model can be effectively solved, the number of layers of the network is not increased, and an accurate training result can be obtained.

3. Long Short-Term Memory (LSTM) 530

LSTM is a time-cycled neural network specifically designed to solve the long-term dependency problem of the general RNN (cycled neural network).

LSTM is a type of neural network that contains LSTM blocks (blocks) or others, possibly described as intelligent network elements, because it can memorize values for indefinite lengths of time, with a gate in the block that can determine if input is important enough to be memorized and can not be output.

LSTM has four S-function units, the leftmost Bian Hanshu would be the input of the block, if the case may be, and the right three would go through gate to determine if input can go into the block, the second left is input gate, if the yield here is approximately zero, the value here would be blocked from going to the next layer. The third on the left is the for gate, which when it produces a value that is approximately zero, forgets the value remembered in the block. The fourth, right most input is the output gate, which determines whether the input in the block memory can be output.

In this embodiment, in the target event recognition model, target features are input into CNN, and the CNN performs convolution processing on the target features, and candidate features are output to a residual network, the residual network calculates residual features for the candidate features and outputs to LSTM, and the LSTM performs feature mapping on the residual features and outputs a class of a deviation value.

If the category of the output deviation value is non-emergency steering, it is determined that the deviation value represents an operation of non-emergency steering.

If the category of the output deviation value is emergency steering, it is determined that the deviation value represents an operation of emergency acceleration and deceleration.

By applying the embodiment of the invention, the event recognition model can be used as a node, the trained dependency relationship is used as a directed edge, a tree structure is generated, and iterative training is continuously carried out along with the accumulation of the deviation value of the driving of the user, so that the event recognition model with high adaptation degree to the driving style of the user is generated, and the personalized and high-precision recognition acceleration and deceleration operation is realized.

The tree structure comprises Root nodes and leaf nodes, paths between the Root nodes and the leaf nodes are traversed to serve as model links, the model links represent the direction of iterative training, reasonable iterative training can be screened out through effectiveness judgment of the iterative training, a final event recognition model is generated, namely a plurality of event recognition models are arranged in the model links, father-son relations exist among the event recognition models, the event recognition model serving as a child node depends on the event recognition model serving as a father node to train, namely the event recognition model serving as the father node is an original event recognition model, and the event recognition model serving as a child node is a target event recognition model.

The Root node Root is a universal event recognition model trained by a server, and a molecular node is started along the Root node Root and is a leaf node when the molecular node is continuously subdivided until no child node exists.

It should be noted that, one event recognition model may have a plurality of parent-child relationships, in a certain parent-child relationship, a certain event recognition model may be used as a child node, and in other parent-child relationships, the event recognition model may be used as a parent node, which is not limited in this embodiment.

For example, for the tree structure shown in fig. 6, the following model links may be divided:

1、Root→A1→A2→A3→A4→A5→A6

2、Root→A1→A2→A3→A4→A41

3、Root→B1→B2→B3→B4

4、Root→B1→B2→B21

5、Root→B1→B2→B3→B31

6、Root→C1→C2→C3

7、Root→C1→C21→C22

for the 1 st model link, for the parent-child relationship between A1, A2, A1 is the parent node, A2 is the child node, A2 is the parent node, A3 is the child node, and so on.

It should be noted that, for simplicity of description, the method embodiments are shown as a series of acts, but it should be understood by those skilled in the art that the embodiments are not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the embodiments. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred embodiments, and that the acts are not necessarily required by the embodiments of the invention.

Example III

Fig. 7 is a block diagram of a steering-based lateral control device according to a third embodiment of the present invention, which may specifically include the following modules:

a deviation value calculating module 701, configured to calculate, when it is detected that the vehicle turns in a curve, a type of the curve and a deviation value, the deviation value representing a degree to which the vehicle deviates from a standard direction;

the deviation value dividing module 702 is configured to divide a portion of the deviation values into a first target deviation value representing an emergency steering and a second target deviation value representing a non-emergency steering, with a lane line of the curve as a reference;

a target event recognition model training module 703, configured to update an event recognition module matched with the curve type by using the first target deviation value and the second target deviation value as classified samples, to obtain a target event recognition model;

a bias value classification module 704, configured to input a part of the bias values into the target event recognition model to classify the target event recognition model, so as to recognize an operation representing an emergency steering;

a lateral control module 705 for laterally controlling the steering of the vehicle at the curve according to the emergency steering operation.

In one embodiment of the present invention, the offset calculation module 701 includes:

The lane line detection sub-module is used for detecting lane lines of lanes where vehicles are located and calculating the curvature of the lane lines;

the curve steering determining submodule is used for determining that the lane is a curve and the vehicle is steered in the curve if the curvature is larger than a preset curvature threshold value;

the curvature range searching sub-module is used for searching a curvature range where the curvature is located in a plurality of preset curvature ranges and taking the curvature range as a target range, wherein each curvature range is associated with a type;

a curve type setting sub-module, configured to set a type associated with the target range as a type of the curve;

an angle acquisition sub-module for acquiring an angle of the vehicle on the curve;

and the angle difference value calculation sub-module is used for calculating the difference value between the angle and the standard direction by taking the designated direction as the standard direction, and taking the difference value as the deviation value.

In one embodiment of the present invention, the offset value dividing module 702 includes:

the target lane line determining sub-module is used for determining a target lane line, and the direction of the target lane line is opposite to the direction in which the vehicle turns;

the distance measurement submodule is used for measuring the distance between the vehicle and the target lane line;

The first target deviation value determining submodule is used for determining the deviation value as a first target deviation value representing emergency steering if the distance monotonically decreases;

and the second target deviation value determining submodule is used for determining the deviation value as a second target deviation value representing non-emergency steering if the distance is within a preset safety range.

In one embodiment of the present invention, the target event recognition model training module 703 includes:

the event recognition model searching sub-module is used for searching an event recognition model trained for the curve type, and the event recognition model is associated with a standard deviation value;

a correlation calculation sub-module for calculating a correlation between the second target deviation value and the standard deviation value;

an original event recognition model selection sub-module for selecting an original event recognition model from the event recognition models based on the correlation;

and the original event recognition module updating sub-module is used for updating the original event recognition module by taking the first target deviation value and the second target deviation value as classified samples to obtain a target event recognition model.

In one embodiment of the present invention, the correlation calculation submodule includes:

A similarity calculating unit, configured to calculate, as a correlation, a similarity between the second target deviation value and the standard deviation value if the standard deviation value is a data point;

or,

a target point determining unit, configured to determine, as a target point, a data point that falls within the data range in the second target deviation value if the standard deviation value is the data range;

and the proportion statistics unit is used for counting the proportion of the target point to the second target deviation value as correlation.

In one embodiment of the present invention, the original event recognition model selection submodule includes:

a correlation average value calculation unit configured to calculate an average value of the correlations;

a discrete value calculating unit, configured to calculate a discrete value of the correlation if the average value of the correlation is greater than or equal to a preset correlation threshold;

the discrete value selection unit is used for selecting the event recognition model with the minimum discrete value as an original event recognition model;

and the correlation selecting unit is used for selecting an event recognition model with the minimum average value of the correlation as an original event recognition model if the average value of the correlation is smaller than a preset correlation threshold value.

In one embodiment of the present invention, the original event recognition module update sub-module includes:

a new deviation value acquisition unit configured to acquire a deviation value identifying an emergency steering as a new first target deviation value;

a first sample feature extraction unit configured to extract a first sample feature from the first target deviation value;

a second sample feature extraction unit configured to extract a second sample feature from the second target deviation value;

the model transfer learning unit is used for taking the first sample characteristics and the second sample characteristics as samples, taking the emergency steering and the non-emergency steering as classification targets, and performing transfer learning on the original event recognition model to obtain a target event recognition model.

In an example of an embodiment of the present invention, the first sample feature includes at least one of a first sample residual, a first sample statistical feature, a second sample statistical feature, and a second sample residual, and the first sample feature extraction unit is further configured to:

searching a standard deviation value associated with the original event recognition model;

calculating a difference value between the first target deviation value and the standard deviation value as a first sample residual;

Calculating a first sample statistical feature for the first residual;

calculating a second sample statistical feature for the first target deviation value;

and calculating a difference value between the second sample statistical characteristic and the standard statistical characteristic of the standard deviation value as a second sample residual error.

In one example of the embodiment of the present invention, the second sample feature includes at least one of a third sample residual, a third sample statistical feature, a fourth sample statistical feature, and a fourth sample residual, and the second sample feature extraction unit is further configured to:

calculating a difference value between the second target deviation value and the standard deviation value as a third sample residual;

calculating a third sample statistical feature for the second residual;

calculating a fourth sample statistical feature for the second target deviation value;

and calculating a difference value between the second sample statistical characteristic and the standard statistical characteristic of the standard deviation value as a fourth sample residual error.

In one embodiment of the present invention, the target event recognition model training module 703 further includes:

a standard deviation value generation sub-module for generating a standard deviation value based on the second target deviation value when the training of the target event recognition model is completed;

And the incidence relation establishing sub-module is used for establishing an incidence relation between the target event identification model and the second target deviation value.

In one embodiment of the present invention, the standard deviation value generation submodule includes:

the data point setting unit is used for calculating the average value of the data points at the same position in the second target deviation value and taking the average value as the data point of the standard deviation value;

or,

and the data range setting unit is used for counting the amplitude of the data points at the same position in the second target deviation value and taking the amplitude as the data range of the standard deviation value.

In one embodiment of the present invention, the bias value classification module 704 includes:

a target feature extraction sub-module for extracting target features from a part of the deviation values;

the candidate feature output sub-module is used for carrying out convolution processing on the target feature in the convolution neural network of the target event identification model so as to output candidate features;

a residual feature calculation sub-module, configured to calculate residual features for the candidate features in a residual network of the target event identification model;

the class output sub-module is used for carrying out feature mapping on the residual features in the long-term and short-term memory network of the target event identification model so as to output the class of the deviation value;

And the emergency acceleration and deceleration operation determination submodule is used for determining that the deviation value represents the operation of emergency steering if the category is emergency steering.

In one example of the embodiment of the present invention, the target feature includes at least one of a first target residual, a first target statistical feature, a second target statistical feature, and a second target residual, and the target feature extraction sub-module is further configured to:

searching a standard deviation value associated with the target event identification model;

calculating a difference between part of the deviation value and the standard deviation value as a first target residual;

calculating a first target statistical feature for the first target residual;

calculating a second target statistical feature for a portion of the deviation values;

and calculating a difference value between the second target statistical feature and the standard statistical feature of the standard deviation value as a second target residual error.

In one embodiment of the present invention, the lateral control module 705 includes:

and the range keeping sub-module is used for increasing the braking force and the speed of steering so as to keep the distance between the vehicle and the lane line within a preset safety range.

The steering-based transverse control device provided by the embodiment of the invention can execute the steering-based transverse control method provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

Example IV

Fig. 8 is a schematic structural diagram of a computer device according to a fourth embodiment of the present invention. FIG. 8 illustrates a block diagram of an exemplary computer device 12 suitable for use in implementing embodiments of the present invention. The computer device 12 shown in fig. 8 is merely an example and should not be construed as limiting the functionality and scope of use of embodiments of the present invention.

As shown in FIG. 8, the computer device 12 is in the form of a general purpose computing device. Components of computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, a bus 18 that connects the various system components, including the system memory 28 and the processing units 16.

Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, and a local bus using any of a variety of bus architectures. By way of example, and not limitation, such architectures include Industry Standard Architecture (ISA) bus, micro channel architecture (MAC) bus, enhanced ISA bus, video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

Computer device 12 typically includes a variety of computer system readable media. Such media can be any available media that is accessible by computer device 12 and includes both volatile and nonvolatile media, removable and non-removable media.

The system memory 28 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM) 30 and/or cache memory 32. The computer device 12 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34 may be used to read from or write to non-removable, nonvolatile magnetic media (not shown in FIG. 8, commonly referred to as a "hard disk drive"). Although not shown in fig. 8, a magnetic disk drive for reading from and writing to a removable non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive for reading from or writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In such cases, each drive may be coupled to bus 18 through one or more data medium interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to carry out the functions of embodiments of the invention.

A program/utility 40 having a set (at least one) of program modules 42 may be stored in, for example, memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment. Program modules 42 generally perform the functions and/or methods of the embodiments described herein.

The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), one or more devices that enable a user to interact with the computer device 12, and/or any devices (e.g., network card, modem, etc.) that enable the computer device 12 to communicate with one or more other computing devices. Such communication may occur through an input/output (I/O) interface 22. Moreover, computer device 12 may also communicate with one or more networks such as a Local Area Network (LAN), a Wide Area Network (WAN) and/or a public network, such as the Internet, through network adapter 20. As shown, network adapter 20 communicates with other modules of computer device 12 via bus 18. It should be appreciated that although not shown, other hardware and/or software modules may be used in connection with computer device 12, including, but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, data backup storage systems, and the like.

The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, for example, implementing the steering-based lateral control method provided by the embodiment of the present invention.

Example five

The fifth embodiment of the present invention further provides a computer readable storage medium, on which a computer program is stored, where the computer program when executed by a processor implements each process of the above-mentioned steering-based lateral control method, and the same technical effects can be achieved, so that repetition is avoided, and no further description is given here.

The computer readable storage medium may include, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

Note that the above is only a preferred embodiment of the present invention and the technical principle applied. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, while the invention has been described in connection with the above embodiments, the invention is not limited to the embodiments, but may be embodied in many other equivalent forms without departing from the spirit or scope of the invention, which is set forth in the following claims.

Claims

1. A steering-based lateral control method, comprising:

transversely controlling the steering of the vehicle in the curve according to the emergency steering operation;

using curve steering as a condition for training the event recognition model and recognizing the operation of the emergency steering;

when the vehicle is detected to turn at a curve, calculating the type and the deviation value of the curve, wherein the method comprises the following steps:

detecting a lane line of a lane where a vehicle is located, and calculating the curvature of the lane line;

if the curvature is greater than a preset curvature threshold, determining that the lane is a curve and the vehicle turns in the curve;

searching a curvature range in which the curvature is located in a plurality of preset curvature ranges, and taking the curvature range as a target range, wherein each curvature range is associated with a type;

setting the type of the target range association as the type of the curve;

collecting the angle of the vehicle on the curve;

and calculating a difference value between the angle and the standard direction by taking the designated direction as the standard direction, and taking the difference value as the deviation value.

2. The method according to claim 1, wherein dividing a portion of the deviation values into a first target deviation value representing an emergency steering, a second target deviation value representing a non-emergency steering, with a lane line of the curve as a reference, comprises:

Determining a target lane line, wherein the direction of the target lane line is opposite to the direction in which the vehicle turns;

measuring a distance between the vehicle and the target lane line;

if the distance monotonically decreases, determining the deviation value as a first target deviation value representing emergency steering;

and if the distance is within the preset safety range, determining the deviation value as a second target deviation value representing non-emergency steering.

3. The method according to claim 1, wherein updating the event recognition module matching the type of the curve with the first target deviation value and the second target deviation value as classified samples to obtain a target event recognition model includes:

searching an event recognition model trained for the curve type, wherein the event recognition model is associated with a standard deviation value;

calculating a correlation between the second target deviation value and the standard deviation value;

selecting an original event recognition model from the event recognition models based on the correlation;

and updating the original event recognition model by taking the first target deviation value and the second target deviation value as classified samples to obtain a target event recognition model.

4. A method according to claim 3, wherein said calculating a correlation between said second target deviation value and said standard deviation value comprises:

if the standard deviation value is a data point, calculating the similarity between the second target deviation value and the standard deviation value as a correlation;

or,

if the standard deviation value is in a data range, determining a data point which falls into the data range in the second target deviation value as a target point;

and counting the proportion of the target point to the second target deviation value as correlation.

5. A method according to claim 3, wherein said selecting an original event recognition model from said event recognition models based on said correlation comprises:

calculating an average value of the correlations;

if the average value of the correlation is greater than or equal to a preset correlation threshold value, calculating a discrete value of the correlation;

selecting the event recognition model with the minimum discrete value as an original event recognition model;

and if the average value of the correlation is smaller than a preset correlation threshold value, selecting an event recognition model with the minimum average value of the correlation as the original event recognition model.

6. A method according to claim 3, wherein updating the original event recognition model with the first target deviation value and the second target deviation value as classified samples to obtain a target event recognition model comprises:

acquiring a deviation value for identifying emergency steering as a new first target deviation value;

extracting first sample features from all the first target deviation values;

extracting a second sample feature from the second target bias value;

taking the first sample characteristics and the second sample characteristics as samples, taking the emergency steering and the non-emergency steering as classification targets, and performing migration learning on an original event recognition model to obtain a target event recognition model.

7. The method of claim 6, wherein the step of providing the first layer comprises,

the first sample feature comprises at least one of a first sample residual, a first sample statistical feature, a second sample residual, the extracting a first sample feature from the first target bias value comprising:

searching a standard deviation value associated with an original event recognition model;

Calculating a first sample statistical feature for the first residual;

calculating a difference value between the second sample statistical feature and the standard statistical feature of the standard deviation value to be used as a second sample residual error;

the second sample feature comprises at least one of a third sample residual, a third sample statistical feature, a fourth sample residual, the extracting a second sample feature from the second target bias value comprising:

calculating a third sample statistical feature for the second residual;

8. The method of claim 3, wherein the updating the event recognition module that matches the type of the curve with the first target deviation value and the second target deviation value as classified samples to obtain a target event recognition model further comprises:

When the training of the target event recognition model is completed, generating a standard deviation value based on the second target deviation value;

and establishing an association relationship between the target event identification model and the second target deviation value.

9. The method of any of claims 1-8, wherein said inputting a portion of said deviation value into said target event identification model for classification to identify an operation indicative of an emergency turn comprises:

extracting target features from a part of the deviation values;

in the convolutional neural network of the target event recognition model, performing convolutional processing on the target features to output candidate features;

calculating residual characteristics of the candidate characteristics in a residual network of the target event identification model;

in a long-term and short-term memory network of the target event identification model, performing feature mapping on the residual features to output the category of the deviation value;

if the category is emergency steering, determining that the deviation value represents an operation of emergency steering.

10. The method according to any one of claims 1-8, wherein said operation of steering the vehicle in the curve in accordance with the emergency steering is laterally controlled, comprising:

And increasing the braking force and the speed of steering so as to keep the distance between the vehicle and the lane line within a preset safety range.

11. A steering-based lateral control device, comprising:

the transverse control module is used for transversely controlling the steering of the vehicle in the curve according to the emergency steering operation;

the deviation value calculating module comprises:

12. A computer device, the computer device comprising:

one or more processors;

a memory for storing one or more programs,

When executed by the one or more processors, causes the one or more processors to implement the steering-based lateral control method of any one of claims 1-10.

13. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steering-based lateral control method according to any of claims 1-10.