CN114162145A

CN114162145A - Automatic vehicle driving method and device and electronic equipment

Info

Publication number: CN114162145A
Application number: CN202210033315.0A
Authority: CN
Inventors: 张艺浩; 韩志华; 徐修信; 郭立群
Original assignee: Zhitu Shanghai Intelligent Technology Co ltd; Suzhou Zhitu Technology Co Ltd
Current assignee: Zhitu Shanghai Intelligent Technology Co ltd; Suzhou Zhitu Technology Co Ltd
Priority date: 2022-01-12
Filing date: 2022-01-12
Publication date: 2022-03-11

Abstract

The application provides a vehicle automatic driving method, a vehicle automatic driving device and electronic equipment, wherein the method comprises the following steps: the method comprises the steps of firstly obtaining driving information of a target vehicle, driving information of related vehicles and lane information, then predicting driving scores through a first neural network model to obtain driving scores of a plurality of different driving strategies, and finally determining the target driving strategy according to the driving scores. According to the technology, a plurality of driving strategies are determined firstly through driving decisions, then score prediction is carried out on each driving strategy by using the neural network model, so that the finally determined target driving strategy integrates effective prediction of the running performance of a target vehicle by the neural network on the basis of a traditional decision algorithm, the target driving strategy determined through the scores predicted by the first neural network model is closer to the actual complex traffic condition, and the safety of automatic driving of the vehicle is effectively improved.

Description

Automatic vehicle driving method and device and electronic equipment

Technical Field

The present disclosure relates to the field of automatic driving technologies, and in particular, to a method and an apparatus for automatically driving a vehicle, and an electronic device.

Background

With the rapid development of artificial intelligence technology and urban intelligent rail transit, the autonomous vehicles, namely, the autonomous driving system, have gradually entered the daily lives of the public, the autonomous driving system is a comprehensive system that collects numerous high and new technologies, and environmental information acquisition and intelligent decision control as key links depend on innovation and breakthrough of a series of high and new technologies such as sensor technology, image recognition technology, electronic and computer technology and control technology.

How to make reasonable and predictive driving decisions for an automatic driving vehicle under a dynamic complex environment to ensure that the vehicle runs fast and smoothly is a very challenging research subject in the field of automatic driving at present, and in the prior art, a driving strategy is usually determined according to a decision algorithm of a finite state.

However, in the conventional decision algorithm based on rules and finite states, decision making needs to be performed according to a specific driving situation, and a real traffic driving environment is a complex situation in which a plurality of factors are combined and cannot be perfectly matched with a simply set driving situation, so that a driving strategy determined by the conventional decision algorithm cannot meet an actual complex road condition, and dangerous situations such as collision can occur.

Disclosure of Invention

In view of the above, an object of the present application is to provide a method and an apparatus for automatically driving a vehicle, and an electronic device, so as to improve safety of automatic driving of the vehicle.

In a first aspect, an embodiment of the present application provides a method for automatically driving a vehicle, where the method includes: acquiring first running information of a target vehicle at the current moment, second running information corresponding to related vehicles which are far away from the target vehicle within a first specified range, and lane information of related lanes which are far away from the target vehicle within a second specified range; predicting driving scores of the target vehicle driving with a plurality of different predetermined driving strategies through the first neural network model, the first driving information, the second driving information and the lane information; wherein the driving score is used for representing the driving condition and the vehicle performance of the target vehicle in the preset driving strategy; the first neural network model is obtained by training corresponding sample data under different traffic conditions; and determining a target driving strategy from the plurality of different driving strategies according to the driving scores respectively corresponding to the plurality of different driving strategies so that the target vehicle drives according to the target driving strategy at the next moment of the current moment.

Further, the relevant vehicles within a specified range from the target vehicle include at least one of: the distance between the target vehicle and the target vehicle in the lane of the target vehicle is smaller than a first distance threshold value and the first related vehicle is located in front of the driving direction of the target vehicle; a second relevant vehicle which is in the right lane of the lane where the target vehicle is located, is away from the target vehicle by a distance smaller than a second distance threshold value and is located in front of the driving direction of the target vehicle; a third related vehicle which is located behind the driving direction of the target vehicle and has a distance with the target vehicle smaller than a third distance threshold value in a lane on the right side of the lane where the target vehicle is located; a fourth relevant vehicle which is in the left lane of the lane where the target vehicle is located, is away from the target vehicle by a distance smaller than a fourth distance threshold value and is located in front of the driving direction of the target vehicle; and a fifth relevant vehicle which is located behind the target vehicle in the driving direction and has a distance with the target vehicle smaller than a fifth distance threshold value in the right lane of the lane where the target vehicle is located.

Further, the first traveling information of the target vehicle at the current time includes a first position and a first speed of the target vehicle; the step of acquiring second travel information corresponding to a relevant vehicle within a specified range from the target vehicle includes: acquiring a second position and a second speed of the relevant vehicle at the current moment; calculating a relative position of the second position with respect to the first position, and a relative velocity of the second velocity with respect to the first velocity; and determining the relative position and the relative speed as second running information corresponding to the relevant vehicle.

Further, the method further comprises: if the lane where the target vehicle is located is the rightmost lane in the road, setting second driving information of the relevant vehicle located in the right lane of the lane where the target vehicle is located to be 0; and if the lane where the target vehicle is located is the leftmost lane in the road, setting the second driving information of the relevant vehicle located in the left lane of the lane where the target vehicle is located to be 0.

Further, the related lanes comprise a lane where the target vehicle is located, a right lane of the lane where the target vehicle is located, and a left lane of the lane where the target vehicle is located; the step of acquiring lane information of a relevant lane within a second specified range from the target vehicle includes: taking the position of the vehicle at the current moment as an origin position; calculating the transverse distance and the longitudinal distance between each position and the origin position in a first preset number of continuous positions in the related lane; wherein the distance between every two adjacent positions in a first preset number of consecutive positions is equal; and determining a set of the transverse distance and the longitudinal distance corresponding to each position in the first preset number of continuous positions as the lane information of the relevant lane within a second specified range away from the target vehicle.

Further, the plurality of driving strategies includes at least any two of: accelerating straight running, keeping speed straight running, decelerating straight running, straight emergency braking, changing lanes to the left lane at a constant speed and changing lanes to the right lane at a constant speed.

Further, the first neural network model is obtained by training through the following steps: acquiring sample data; the sample data comprises speed information and position information of the target sample vehicle, speed information and position information of a related sample vehicle related to the target sample vehicle, and lane information of a lane corresponding to the target sample vehicle; calculating a loss value of the initial neural network model corresponding to each moment in a second preset number of continuous moments according to the sample data and the loss function; wherein the loss function comprises one or more of a collision parameter, an energy consumption parameter, a lane change penalty parameter and a driving efficiency parameter; and adjusting parameters of the initial neural network model according to the loss value, and determining the corresponding initial neural network model as the first neural network model when the training stopping condition is met.

Further, the method further comprises: determining prediction scores respectively corresponding to a plurality of driving strategies at the next moment in the current moment in a second preset number of continuous moments through the initial neural network model; taking the driving strategy with the maximum prediction score as the driving strategy at the next moment, and acquiring sample data of the target sample vehicle at the next moment generated by driving of the target sample vehicle under the driving strategy at the next moment; and continuing to train the initial neural network model based on the sample data at the next moment.

Further, the step of determining the target driving maneuver from the plurality of different driving maneuvers according to the driving scores corresponding to the plurality of different driving maneuvers, respectively, includes: determining the driving strategy with the driving score higher than a preset score threshold value as an alternative driving strategy set; and determining a target driving strategy from the alternative driving strategy set.

Further, the step of determining the target driving strategy from the set of candidate driving strategies includes: judging whether the driving strategy with the highest driving score in the alternative driving strategy set meets a preset safe driving condition or not; if so, determining the driving strategy as a target driving strategy; if not, the driving strategy is deleted from the set of alternative driving strategies.

Further, the method further comprises: if the target driving strategy is to change the lane to the left lane at a constant speed or to change the lane to the right lane at a constant speed, controlling the vehicle to change the lane by the following formula:

wherein the content of the first and second substances,

indicating the angle of rotation, theta, of the steering wheel_nAnd theta_fRespectively representing the included angles formed by the front first position, the front second position and the self-vehicle position on the target lane, delta t is the time step length, k_f、k_nAnd k_ITo representA constant of driving behavior, the target lane is a lane for which the target vehicle is ready to switch from the current lane in which the target vehicle is located.

In a second aspect, an embodiment of the present application further provides a vehicle automatic driving device, including: the information acquisition module is used for acquiring first running information of a target vehicle at the current moment, second running information corresponding to related vehicles within a first specified range away from the target vehicle and lane information of related lanes within a second specified range away from the target vehicle; a prediction module to predict a driving score for the target vehicle to travel with a plurality of different predetermined driving strategies via a first neural network model, the first travel information, the second travel information, and the lane information; wherein the driving score is used to characterize a driving condition and a vehicle performance of the target vehicle driving in the predetermined driving strategy; the first neural network model is obtained by training corresponding sample data under different traffic conditions; and the target driving strategy determining module is used for determining a target driving strategy from the plurality of different driving strategies according to the driving scores respectively corresponding to the plurality of different driving strategies so as to enable the target vehicle to drive according to the target driving strategy at the next moment of the current moment.

In a third aspect, an embodiment of the present application further provides an electronic device, which includes a processor and a memory, where the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the vehicle automatic driving method according to the first aspect.

In a fourth aspect, embodiments of the present application further provide a computer-readable storage medium storing computer-executable instructions that, when invoked and executed by a processor, cause the processor to implement the vehicle autopilot method of the first aspect described above.

Compared with the prior art, the method has the following beneficial effects:

according to the automatic vehicle driving method, the automatic vehicle driving device and the electronic equipment, firstly, the driving information of a target vehicle, the driving information of a relevant vehicle and lane information are obtained, then, the driving score is predicted through the first neural network model, the driving scores of a plurality of different driving strategies are obtained, and finally, the target driving strategy is determined according to the driving scores. According to the technology of the application, a plurality of driving strategies are firstly determined through driving decisions, then score prediction is carried out on each driving strategy by utilizing the neural network model, so that the finally determined target driving strategy is combined with effective prediction of the running performance of a target vehicle by the neural network on the basis of a traditional decision algorithm, and as the first neural network model is obtained through sample data training under various different traffic conditions, the target driving strategy determined through the scores predicted by the first neural network model is closer to the actual complex traffic conditions, and the safety of automatic driving of the vehicle is effectively improved.

Additional features and advantages of the disclosure will be set forth in the description which follows, or in part may be learned by the practice of the above-described techniques of the disclosure, or may be learned by practice of the disclosure.

In order to make the aforementioned objects, features and advantages of the present disclosure more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the detailed description of the present application or the technical solutions in the prior art, the drawings needed to be used in the detailed description of the present application or the prior art description will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative efforts.

Fig. 1 is a schematic structural diagram of an electronic system according to an embodiment of the present disclosure;

FIG. 2 is a flow chart of a method for automatic driving of a vehicle according to an embodiment of the present disclosure;

FIG. 3 is a flow chart of another method for automatic vehicle driving provided by an embodiment of the present application;

fig. 4 is a schematic structural diagram of a first neural network model according to an embodiment of the present disclosure;

FIG. 5 is a flowchart of a method for training a first neural network model according to an embodiment of the present disclosure;

FIG. 6 is a schematic structural diagram of an automatic driving device for a vehicle according to an embodiment of the present disclosure;

fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

To make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the present application will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

The main purposes of autonomous driving are to reduce manpower costs, increase transportation efficiency, degrade energy consumption, and avoid traffic accidents, etc. With the increasing development of the field of artificial intelligence, the automatic driving technology of the motor vehicle is mature day by day. Early autodrive mainly developed the focus for safe and smooth driving experience. In recent years, how to improve the driving efficiency and reduce the fuel consumption on the basis of safety has drawn more attention. It remains challenging how autonomous vehicles make reasonably predictive driving decisions in dynamic complex environments to ensure fast and smooth vehicle travel. Conventional decision making algorithms based on rules and finite states, although successful in some autonomous driving tasks, have one of their greatest drawbacks in that their decision making needs to be based on the particular driving situation. This makes it difficult to generalize more complex real traffic driving environments. Based on this, the embodiment of the application provides a vehicle automatic driving method, a device and an electronic device, so as to improve the safety of vehicle automatic driving.

Referring to fig. 1, a schematic diagram of an electronic system 100 is shown. The electronic system can be used for realizing the automatic vehicle driving method and device.

As shown in fig. 1, an electronic system 100 includes one or more processing devices 102 and one or more memory devices 104. Optionally, electronic system 100 may also include input devices 106, output devices 108, and one or more data acquisition devices 110, which may be interconnected via a bus system 112 and/or other form of connection mechanism (not shown). It should be noted that the components and structure of the electronic system 100 shown in fig. 1 are exemplary only, and not limiting, and the electronic system may have some of the components in fig. 1, as well as other components and structures, as desired.

Processing device 102 may be a server, a smart terminal, or a device containing a Central Processing Unit (CPU) or other form of processing unit having data processing capabilities and/or instruction execution capabilities that may process data for other components in electronic system 100 and may control other components in electronic system 100 to perform vehicle autopilot functions.

Storage 104 may include one or more computer program products that may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. Volatile memory can include, for example, Random Access Memory (RAM), cache memory (or the like). The non-volatile memory may include, for example, Read Only Memory (ROM), a hard disk, flash memory, and the like. One or more computer program instructions may be stored on a computer-readable storage medium and executed by processing device 102 to implement the client functionality (implemented by the processing device) of the embodiments of the present application described below and/or other desired functionality. Various applications and various data, such as various data used and/or generated by the applications, may also be stored in the computer-readable storage medium.

The input device 106 may be a device used by a user to input instructions and may include one or more of a keyboard, a mouse, a microphone, a touch screen, and the like.

The output device 108 may output various information (e.g., images or sounds) to the outside (e.g., a user), and may include one or more of a display, a speaker, and the like.

The data acquisition device 110 may acquire the data to be processed and store the data to be processed in the storage 104 for use by other components.

For example, the devices used for implementing the method, apparatus and electronic device for automatic driving of a vehicle according to the embodiment of the present application may be integrally disposed, or may be disposed in a decentralized manner, such as integrally disposing the processing device 102, the storage device 104, the input device 106 and the output device 108, and disposing the data collecting device 110 at a designated position where data can be collected. When the above-described devices in the electronic system are integrally provided, the electronic system may be implemented as an intelligent terminal such as a camera, a smart phone, a tablet computer, a vehicle-mounted terminal, and the like.

Fig. 2 is a flowchart of an automatic driving method for a vehicle according to an embodiment of the present application, and as shown in fig. 2, the method specifically includes the following steps:

s202: acquiring first running information of a target vehicle at the current moment, second running information corresponding to related vehicles which are far away from the target vehicle within a first specified range, and lane information of related lanes which are far away from the target vehicle within a second specified range;

the target vehicle is a vehicle controlled by the method provided by the embodiment of the application, and the first running information may include information such as speed and position of the target vehicle. The relevant vehicle is a vehicle whose distance from the target vehicle is within a first specified range, such as a front vehicle and a rear vehicle of the same lane of the target vehicle, a front vehicle and a rear vehicle of an adjacent lane of the target vehicle, and the like. The second travel information includes speed and position information of the relevant vehicle. The lane information includes information such as a relative distance between the lane and the lane in which the target vehicle is located.

S204: predicting driving scores of the target vehicle driving with a plurality of different predetermined driving strategies through the first neural network model, the first driving information, the second driving information and the lane information; wherein the driving score is used for representing the driving condition and the vehicle performance of the target vehicle in the preset driving strategy; the first neural network model is obtained by training corresponding sample data under different traffic conditions;

the first neural network model is a driving score prediction model, and may process the input first driving information, second driving information and lane information to obtain a plurality of driving scores, each driving score corresponds to one predetermined driving strategy, for example, 5 driving strategies are preset, and the driving scores are 10, 25, 80, 45 and 96 respectively, so that which driving strategy the target vehicle adopts at the next time may be determined according to the driving scores. The training method of the first neural network model will be described in detail below, and will not be described herein again.

S206: and determining a target driving strategy from the plurality of different preset driving strategies according to the driving scores respectively corresponding to the plurality of different preset driving strategies, so that the target vehicle can drive according to the target driving strategy at the next moment of the current moment.

According to the automatic driving method for the vehicle, firstly, the driving information of the target vehicle, the driving information of the relevant vehicle and the lane information are obtained, then, the driving scores are predicted through the first neural network model, the driving scores of a plurality of different driving strategies are obtained, and finally, the target driving strategy is determined according to the driving scores. According to the technology of the application, a plurality of driving strategies are firstly determined through driving decisions, then score prediction is carried out on each driving strategy by utilizing the neural network model, so that the finally determined target driving strategy is combined with effective prediction of the running performance of a target vehicle by the neural network on the basis of a traditional decision algorithm, and as the first neural network model is obtained through sample data training under various different traffic conditions, the target driving strategy determined through the scores predicted by the first neural network model is closer to the actual complex traffic conditions, and the safety of automatic driving of the vehicle is effectively improved.

In some possible embodiments, the relevant vehicles within a specified distance from the target vehicle include at least one of:

(1) the distance between the target vehicle and the target vehicle in the lane of the target vehicle is smaller than a first distance threshold value and the first related vehicle is located in front of the driving direction of the target vehicle;

(2) a second relevant vehicle which is in the right lane of the lane where the target vehicle is located, is away from the target vehicle by a distance smaller than a second distance threshold value and is located in front of the driving direction of the target vehicle;

(3) a third related vehicle which is located behind the driving direction of the target vehicle and has a distance with the target vehicle smaller than a third distance threshold value in a lane on the right side of the lane where the target vehicle is located;

(4) a fourth relevant vehicle which is in the left lane of the lane where the target vehicle is located, is away from the target vehicle by a distance smaller than a fourth distance threshold value and is located in front of the driving direction of the target vehicle;

(5) and a fifth relevant vehicle which is located behind the target vehicle in the driving direction and has a distance with the target vehicle smaller than a fifth distance threshold value in the right lane of the lane where the target vehicle is located.

Fig. 3 is a flowchart of another method for automatically driving a vehicle according to an embodiment of the present application, which focuses on a specific implementation of how to obtain second driving information corresponding to a related vehicle, and as shown in fig. 3, the method specifically includes the following steps:

s302: acquiring first running information of a target vehicle at the current moment and lane information of a related lane within a second specified range away from the target vehicle;

the first travel information of the target vehicle at the current time includes a first position and a first speed of the target vehicle.

S304: acquiring a second position and a second speed of the relevant vehicle at the current moment;

s306: calculating a relative position of the second position with respect to the first position, and a relative velocity of the second velocity with respect to the first velocity;

s308: and determining the relative position and the relative speed as second running information corresponding to the relevant vehicle.

In some examples, if the lane in which the target vehicle is located is the rightmost lane in the road, the second driving information of the relevant vehicle located in the right lane of the lane in which the target vehicle is located is set to 0;

in other examples, if the lane in which the target vehicle is located is the leftmost lane in the road, the second travel information of the relevant vehicle located in the left lane of the lane in which the target vehicle is located is set to 0.

The right side lane of the rightmost lane described above, and the left side lane of the leftmost lane correspond to lanes that are out of bounds, meaning that the subject lane in question does not exist in the real road. For example, when we are driving on the rightmost lane, we consider the information of the own vehicle lane and the obstacles on the left and right 3 lanes. In this case, the corresponding own lane and left lane are real, and thus the information of the obstacle corresponds to the real information. But the right lane does not exist, which is a virtual concept. To maintain consistency, we assume that the right lane is full of virtual vehicles in this case to ensure that the out-of-bounds lane is inaccessible.

S310: predicting driving scores of the target vehicle driving with a plurality of different predetermined driving strategies through the first neural network model, the first driving information, the second driving information and the lane information;

s312: and determining a target driving strategy from the plurality of different preset driving strategies according to the driving scores respectively corresponding to the plurality of different preset driving strategies, so that the target vehicle can drive according to the target driving strategy at the next moment of the current moment.

By the method provided by the embodiment, the information of the target vehicle, the related vehicle and the lane can be fully considered in the process of predicting the vehicle driving strategy, so that the driving score corresponding to the driving strategy is more consistent with the actual traffic driving condition.

In some possible embodiments, the relevant lanes include a lane in which the target vehicle is located, a right-side lane of the lane in which the target vehicle is located, and a left-side lane of the lane in which the target vehicle is located; based on this, the lane information of the relevant lane may be acquired by:

(1) taking the position of the vehicle at the current moment as an origin position;

(2) calculating the transverse distance and the longitudinal distance between each position and the origin position in a first preset number of continuous positions in the related lane; wherein the distance between every two adjacent positions in a first preset number of consecutive positions is equal;

(3) and determining a set of the transverse distance and the longitudinal distance corresponding to each position in the first preset number of continuous positions as the lane information of the relevant lane within a second specified range away from the target vehicle.

In some possible implementations, the plurality of driving strategies in the above-described embodiments includes at least any two of: accelerating straight running, keeping speed straight running, decelerating straight running, straight emergency braking, changing lanes to the left lane at a constant speed and changing lanes to the right lane at a constant speed.

The structure of the first neural network model and the training method thereof provided by the embodiment of the present application are described below with reference to fig. 4. As shown in fig. 4, the input to the first neural network model consists of 3 parts: target vehicle characteristics, related vehicle characteristics, and lane structure characteristics. The target vehicle feature and the related vehicle feature output a feature vector with the size of 128 through the multilayer perceptron. Each lane structure feature outputs a feature vector with the size of 1024 through a series of one-dimensional convolutions. After the relevant vehicle feature vectors and the lane feature vectors are integrated, an output value with the size of 6 is finally obtained through the full-connection neural network, and the driving score corresponding to each driving strategy is shown in fig. 4. The driving score of the driving strategy represents the quality of the driving strategy, in practical application, the driving strategy with the maximum driving score value is selected as the optimal strategy selection at the current moment, and meanwhile, the feasibility of the optimal strategy is considered by integrating a conventional decision scheme based on rules or finite states.

In the training process of the first neural network model, in order to simulate the highway traffic flow under the actual condition, the tested and tested highway is 4 rows of one-way driving lanes. The peripheral vehicles are randomly selected from 18 preset vehicles (the length range is 2-18 m, the width range is 1.6-3 m), and the maximum number is 50.

In order to ensure safe driving behavior, a longitudinal Intelligent Driver (IDM) following model is adopted as a driving model of the surrounding vehicles, and a total braking (MOBIL) adaptive cruise controller caused by lane change minimization is adopted in the transverse direction.

Wherein, the expression of the longitudinal IDM cruise controller is as follows

Where v denotes the speed of the vehicle, a denotes the maximum desired acceleration, b denotes the desired rate of speed reduction, s₀Is the minimum separation between two vehicles, s represents the actual separation between two vehicles, Δ v represents the speed difference between two vehicles, v_setIndicating the desired speed, T_setThe desired time interval is indicated.

The expression of the transverse MOBIL lane change decision controller is

Wherein, a_c，a_nAnd a_oThe acceleration of the own vehicle, the acceleration of the vehicle following the lane change target lane, and the acceleration of the vehicle following the own vehicle are respectively indicated.

And

respectively indicating the addition of the vehicle after the lane-changing task is executedSpeed, acceleration of a vehicle following the lane change target lane and acceleration of a vehicle following the vehicle. b_safeIndicating the maximum rate of decrease in vehicle speed. Δ a_thIndicating an acceleration transition threshold. And when the (3) and the (4) are simultaneously satisfied, the vehicle executes a lane change task to the target lane.

On the basis of the above model structure, an embodiment of the present application further provides a training method for a first neural network model, where after an initial structure of the first neural network model is constructed and sample data is acquired, training of the model may be performed according to the method shown in fig. 5, where the method may specifically include the following steps:

s502: acquiring sample data; the sample data comprises speed information and position information of the target sample vehicle, speed information and position information of a related sample vehicle related to the target sample vehicle, and lane information of a lane corresponding to the target sample vehicle;

s504: calculating a loss value of the initial neural network model corresponding to each moment in a second preset number of continuous moments according to the sample data and the loss function; wherein the loss function comprises one or more of a collision parameter, an energy consumption parameter, a lane change penalty parameter and a driving efficiency parameter;

s506: and adjusting parameters of the initial neural network model according to the loss value, and determining the corresponding initial neural network model as the first neural network model when the training stopping condition is met.

In order to enable efficient and safe driving of the vehicle on the highway and to take the fuel consumption into account, the reward selected by the first neural network model comprises the following aspects:

whether collision occurs or not. R is applied if the own vehicle collides with another vehicle or exceeds the range of a driveway of the expressway_safeA penalty of-100 and the task ends with a failure. If no collision occurs r_safe＝0。

And secondly, the running efficiency. For each action performed, the speed of the vehicle directly determines how fast the vehicle is driving. Thus, the reward is selected in proportion to the current vehicle speed

Wherein v is the speed of the bicycle, v_desThe desired speed is indicated at 30 m/s.

And channel change punishment. Although lane changes are a prerequisite for overtaking, lane changes that are too frequent are not good driving behavior. To limit the lane change frequency, each lane change decision is given by r_changeA penalty of-1.

And fourthly, fuel consumption. After each macro action is executed, the actual force curve of the speed is recorded, and the fuel consumption p (L/km) of the current action during execution can be obtained by combining an engine fuel consumption model of a truck of the company. For the purpose of reducing fuel consumption, the corresponding report is r_oil＝-p/100。

In conjunction with all the above return definitions, the total return expression is

R＝r_safe+r_v+r_change+r_oil (6)

The first neural network model in the embodiment of the application adopts a DQN algorithm, and updates the action value at the current moment by adopting the maximum value of the action at the next moment. The loss function is defined as:

L＝(R+γ*argmax_a′Q(s′,a′)-Q(s,a))² (7)

wherein R represents the return, γ is the depreciation factor, and Q (s, a) is the value of the selected action a under the state s. Optimizing the training parameters by adopting a gradient descent algorithm, wherein the updating of the training parameters w satisfies

Where α is the learning rate, s represents a state quantity and is the observed state corresponding to table 1, and a represents an action and is 6 discrete macro decision behaviors.

Further, in some examples, after determining the predicted score for the next time, the manner of travel of the sample vehicle at the next time may also be determined according to the following method:

(1) determining prediction scores respectively corresponding to a plurality of driving strategies at the next moment in the current moment in a second preset number of continuous moments through the initial neural network model;

(2) taking the driving strategy with the maximum prediction score as the driving strategy at the next moment, and acquiring sample data of the target sample vehicle at the next moment generated by driving of the target sample vehicle under the driving strategy at the next moment;

(3) and continuing to train the initial neural network model based on the sample data at the next moment.

Each driving strategy is equivalent to a state at each moment in the DQN algorithm and is represented by s, different action values Q (s, a) under each state s are approximately expressed by using a first neural network model, so that for a given 'next moment' in a training sample, Q values corresponding to 6 actions can be obtained, and then the maximum value is selected as an updated iteration target under the 'current moment'. Although these values are given randomly at the beginning, they eventually converge through successive iterations.

Only a certain fixed action value is updated per iteration, depending on the information recorded in the training samples. For example, during training, a sample records that action a1 is selected in state s, and the action value corresponding to this a1 may not be the maximum. The different actions may converge gradually through iteration. Different actions are selected with a certain probability in the training process and are called exploration, because the magnitude of the true value corresponding to each action is not known at first, and the actual value can be evaluated only through certain exploration.

After training is finished, the driving strategy is to select 6 action values at the current moment when in use, and then to select the action corresponding to the maximum value as the driving strategy.

When the deep reinforcement learning intelligent agent is used after training is finished, corresponding characteristic information is extracted through the driving characteristics of the related vehicles obtained by the sensing modules of the related vehicles and the road structure information given by the map positioning module. And then obtaining each action value in the current time state through the neural network. The selection of the optimal action follows the following rules: firstly, sorting is carried out according to the magnitude of the action values, and the action corresponding to the maximum value is preferentially selected. The state machine then checks whether the action can be performed successfully. If yes, the action is executed, and if the conversion condition of the state machine is not met, the action corresponding to the left maximum action value is selected. The check of the state machine is repeated until the condition is satisfied. If all actions are not satisfied, a return instruction is triggered.

In some possible embodiments, after determining the driving scores for a plurality of different driving strategies, a target driving strategy may be selected from among the following methods: determining the driving strategy with the driving score higher than a preset score threshold value as an alternative driving strategy set; and determining a target driving strategy from the alternative driving strategy set.

Specifically, it may be determined whether the driving strategy with the highest driving score in the candidate driving strategy set meets a preset safe driving condition; if so, determining the driving strategy as a target driving strategy; if not, the driving strategy is deleted from the set of alternative driving strategies.

It should be noted that the safe driving condition in the embodiment of the present application may be determined by using a state machine, and the state machine may be implemented by using an existing decision determination algorithm. For example, when our decision model gives an indication of a lane change, detection of a trigger state machine will be detected. The state machine judges whether the distance between the current vehicle and the rear vehicle meets a series of conditions such as safe distance length, whether the relative speed between the current vehicle and the front vehicle can collide, and the like. The decision instructions are considered executable when all conditions are satisfied. The state machine adopted by the invention is simpler at present, and is mainly used for verifying the safety of the whole decision as an early-stage test.

In some possible embodiments, if it is determined that the target driving strategy is to change lane to the left lane at a constant speed or to change lane to the right lane at a constant speed, the vehicle is controlled to change lane by the following formula:

wherein the content of the first and second substances,

indicating the angle of rotation, theta, of the steering wheel_nAnd theta_fRespectively representing the included angles formed by the front first position, the front second position and the self-vehicle position on the target lane, delta t is the time step length, k_f、k_nAnd k_IA constant representing driving behavior, the target lane being a lane for which the target vehicle is ready to switch from a current lane in which the target vehicle is located.

In order to facilitate understanding, an embodiment of the present application further provides a vehicle automatic driving method in a practical application scenario, where the method specifically includes the following steps:

step 1: acquiring a first speed v1 and a first position d1 of a target vehicle at the current moment;

step 2: determining a relevant vehicle;

the related vehicle includes: within the sensing range 100m of the target vehicle, 10 vehicles are counted, namely 2 vehicles with the shortest distance in front of the current lane where the target vehicle is located, 2 vehicles with the shortest distance in front/back of the left lane and 2 vehicles with the shortest distance in front/back of the right lane.

And step 3: determining a relevant lane;

the relevant lane includes: the lane of the target vehicle, the right lane of the target vehicle and the left lane of the target vehicle;

and 4, step 4: determining the characteristics 1 of the target vehicle and the characteristics 2 of the related vehicle;

the characteristic 1 of the target vehicle is the speed at the present moment and the distance from the lane center line.

The characteristic 2 of each vehicle in the relevant vehicles is that each vehicle obtains the position and the speed of the current moment through radar and positioning information and converts the position and the speed into the relative position and the relative speed relative to the vehicle.

And 5: determining a characteristic 3 corresponding to the lane;

and the characteristic 3 is that every lane takes 1m as a distance interval, and the track points of 80 central lines which are 80 meters ahead are selected to be in the relative positions with the transverse direction and the longitudinal direction of the target vehicle. The dimensions of the above features are shown in table 1.

TABLE 1

It is noted that if there is no vehicle at a position within 100m in a certain direction, the corresponding relative position and speed information is filled with a constant of 0. For a lane that is beyond the boundary, the lane is considered full of vehicles and inaccessible. Therefore, the relative position of the vehicle in the lane in the transverse direction is the lane width, the relative position in the longitudinal direction is 0, and the relative speed is 0.

Step 6: determining a set of driving strategies

The driving strategy in the embodiment of the application comprises the following steps: accelerating straight running, keeping speed straight running, decelerating straight running, straight emergency braking, changing lanes to the left lane at a constant speed and changing lanes to the right lane at a constant speed. The bottom layer control of linear acceleration and deceleration is realized by changing the control parameters of the IDM adaptive cruise algorithm in the formulas (1) and (2), and the control parameters are shown in the table 2.

TABLE 2

	Longitudinal acceleration	Longitudinal deceleration
			a	2.0m/s2	0.6m/s2
b	3.0m/s2	1.0m/s2
			s₀	0m/s	5m/s
v_set	30m/s	19m/s
			T_set	1s	2s

The embodiment of the application adopts a simplified 2-point control model to complete lane change control. The following formula is shown in detail:

and on the basis of keeping constant-speed running, the lane changing operation is realized by changing the rotating angle of the steering wheel. Wherein the content of the first and second substances,

indicating the angle of rotation, theta, of the steering wheel_nAnd theta_fRespectively representing the included angles formed by the reference points of 50 meters and 100 meters in front and the self-vehicle position on the target lane, wherein delta t is 0.05 time step, k_f＝20，k_n＝10，k_IA constant of 6 indicates driving behavior.

And 7: determining the driving strategy with the driving score higher than a preset score threshold value as an alternative driving strategy set;

and 8: judging whether the driving strategy with the highest driving score in the alternative driving strategy set meets a preset safe driving condition or not;

and step 9: if so, determining the driving strategy as a target driving strategy;

step 10: the target driving strategy is to change the lane to the left lane at a constant speed or to change the lane to the right lane at a constant speed, and the target vehicle runs according to the target driving strategy at the next moment.

Specifically, the vehicle is controlled to make a lane change by the following formula:

wherein the content of the first and second substances,

Based on the above method embodiment, the present application embodiment further provides an automatic vehicle driving device, as shown in fig. 6, the device includes:

the information acquisition module 602 is used for acquiring first running information of the target vehicle at the current moment, second running information corresponding to related vehicles within a first specified range away from the target vehicle, and lane information of related lanes within a second specified range away from the target vehicle;

a prediction module 604 for predicting a driving score for the target vehicle to travel with a plurality of different predetermined driving strategies via the first neural network model, the first travel information, the second travel information, and the lane information; wherein the driving score is used for representing the driving condition and the vehicle performance of the target vehicle in the preset driving strategy; the first neural network model is obtained by training corresponding sample data under different traffic conditions;

and a target driving strategy determining module 606, configured to determine a target driving strategy from the multiple different driving strategies according to the driving scores corresponding to the multiple different driving strategies, so that the target vehicle drives according to the target driving strategy at a next time of the current time.

The automatic vehicle driving device provided by the embodiment of the application firstly acquires the running information of the target vehicle, the running information of the relevant vehicle and the lane information, then predicts the running score through the first neural network model to obtain the running scores of a plurality of different driving strategies, and finally determines the target driving strategy according to the running scores. According to the technology of the application, a plurality of driving strategies are firstly determined through driving decisions, then score prediction is carried out on each driving strategy by utilizing the neural network model, so that the finally determined target driving strategy is combined with effective prediction of the running performance of a target vehicle by the neural network on the basis of a traditional decision algorithm, and as the first neural network model is obtained through sample data training under various different traffic conditions, the target driving strategy determined through the scores predicted by the first neural network model is closer to the actual complex traffic conditions, and the safety of automatic driving of the vehicle is effectively improved.

The relevant vehicles within a specified range from the target vehicle include at least one of: the distance between the target vehicle and the target vehicle in the lane of the target vehicle is smaller than a first distance threshold value and the first related vehicle is located in front of the driving direction of the target vehicle; a second relevant vehicle which is in the right lane of the lane where the target vehicle is located, is away from the target vehicle by a distance smaller than a second distance threshold value and is located in front of the driving direction of the target vehicle; a third related vehicle which is located behind the driving direction of the target vehicle and has a distance with the target vehicle smaller than a third distance threshold value in a lane on the right side of the lane where the target vehicle is located; a fourth relevant vehicle which is in the left lane of the lane where the target vehicle is located, is away from the target vehicle by a distance smaller than a fourth distance threshold value and is located in front of the driving direction of the target vehicle; and a fifth relevant vehicle which is located behind the target vehicle in the driving direction and has a distance with the target vehicle smaller than a fifth distance threshold value in the right lane of the lane where the target vehicle is located.

The first running information of the target vehicle at the current moment comprises a first position and a first speed of the target vehicle; the information obtaining module 602 is further configured to: acquiring a second position and a second speed of the relevant vehicle at the current moment; calculating a relative position of the second position with respect to the first position, and a relative velocity of the second velocity with respect to the first velocity; and determining the relative position and the relative speed as second running information corresponding to the relevant vehicle.

The above apparatus is also for: if the lane where the target vehicle is located is the rightmost lane in the road, setting second driving information of the relevant vehicle located in the right lane of the lane where the target vehicle is located to be 0; and if the lane where the target vehicle is located is the leftmost lane in the road, setting the second driving information of the relevant vehicle located in the left lane of the lane where the target vehicle is located to be 0.

The related lanes comprise a lane where the target vehicle is located, a right lane of the lane where the target vehicle is located and a left lane of the lane where the target vehicle is located; the information obtaining module 602 is further configured to: taking the position of the vehicle at the current moment as an origin position; calculating the transverse distance and the longitudinal distance between each position and the origin position in a first preset number of continuous positions in the related lane; wherein the distance between every two adjacent positions in a first preset number of consecutive positions is equal; and determining a set of the transverse distance and the longitudinal distance corresponding to each position in the first preset number of continuous positions as the lane information of the relevant lane within a second specified range away from the target vehicle.

The plurality of driving strategies includes at least any two of: accelerating straight running, keeping speed straight running, decelerating straight running, straight emergency braking, changing lanes to the left lane at a constant speed and changing lanes to the right lane at a constant speed.

The first neural network model is obtained by training through the following steps: acquiring sample data; the sample data comprises speed information and position information of the target sample vehicle, speed information and position information of a related sample vehicle related to the target sample vehicle, and lane information of a lane corresponding to the target sample vehicle; calculating a loss value of the initial neural network model corresponding to each moment in a second preset number of continuous moments according to the sample data and the loss function; wherein the loss function comprises one or more of a collision parameter, an energy consumption parameter, a lane change penalty parameter and a driving efficiency parameter; and adjusting parameters of the initial neural network model according to the loss value, and determining the corresponding initial neural network model as the first neural network model when the training stopping condition is met.

The above apparatus is also for: determining prediction scores respectively corresponding to a plurality of driving strategies at the next moment in the current moment in a second preset number of continuous moments through the initial neural network model; taking the driving strategy with the maximum prediction score as the driving strategy at the next moment, and acquiring sample data of the target sample vehicle at the next moment generated by driving of the target sample vehicle under the driving strategy at the next moment; and continuing to train the initial neural network model based on the sample data at the next moment.

The target driving strategy determination module 606 is further configured to: determining the driving strategy with the driving score higher than a preset score threshold value as an alternative driving strategy set; and determining a target driving strategy from the alternative driving strategy set.

The process of determining the target driving strategy from the set of candidate driving strategies includes: judging whether the driving strategy with the highest driving score in the alternative driving strategy set meets a preset safe driving condition or not; if so, determining the driving strategy as a target driving strategy; if not, the driving strategy is deleted from the set of alternative driving strategies.

The above apparatus is also for: if the target driving strategy is to change the lane to the left lane at a constant speed or to change the lane to the right lane at a constant speed, controlling the vehicle to change the lane by the following formula:

wherein the content of the first and second substances,

The implementation principle and the generated technical effects of the vehicle automatic driving device provided by the embodiment of the application are the same as those of the foregoing method embodiment, and for the sake of brief description, reference may be made to corresponding contents in the foregoing vehicle automatic driving method embodiment where no mention is made in part of the embodiment of the device.

In order to verify the feasibility of the first neural network model provided by the embodiment of the application, the embodiment of the application adopts a CARLA automatic driving simulator to establish a driving environment of an expressway. Wherein the initial speed distribution of the peripheral vehicles is 60 km/h-120 km/h, the initial speed of the self vehicle is 60km/h, and the maximum speed is 110 km/h. The purpose is that whether the model of the invention can run on the expressway more quickly on the basis of safety or not, and simultaneously, the fuel consumption can be reduced.

The structure of 1000 test road conditions shows that the average consumed time per kilometer is 44.76s and the oil consumption is 0.211 liter by using the traditional rule-based following model. By adopting the automatic driving method for the vehicle, the collision accident does not occur when the vehicle is driven and the lane changing overtaking time is reasonably selected, the average consumed time per kilometer is 35.24s, and the oil consumption is 0.205 liter. The average time consumption is reduced by 21.27%, and the oil consumption is reduced by 2.3%.

An embodiment of the present application further provides an electronic device, as shown in fig. 7, which is a schematic structural diagram of the electronic device, where the electronic device includes a processor 1501 and a memory 1502, the memory 1502 stores computer-executable instructions that can be executed by the processor 1501, and the processor 1501 executes the computer-executable instructions to implement the above-mentioned vehicle automatic driving method.

In the embodiment shown in fig. 7, the electronic device further comprises a bus 1503 and a communication interface 1504, wherein the processor 1501, the communication interface 1504 and the memory 1502 are connected by the bus 1503.

The Memory 1502 may include a high-speed Random Access Memory (RAM) and may also include a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. The communication connection between the network element of the system and at least one other network element is implemented through at least one communication interface 1504 (which may be wired or wireless), and the internet, a wide area network, a local network, a metropolitan area network, and the like may be used. The bus 1503 may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, an EISA (Extended Industry Standard Architecture) bus, or the like. The bus 1503 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one double-headed arrow is shown in FIG. 7, but this does not indicate only one bus or one type of bus.

Processor 1501 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware or instructions in the form of software in the processor 1501. The Processor 1501 may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the device can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in a memory, and the processor 1501 reads information in the memory, and completes the steps of the vehicle automatic driving method of the foregoing embodiment in combination with hardware thereof.

The embodiment of the present application further provides a computer-readable storage medium, where the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions cause the processor to implement the above-mentioned vehicle automatic driving method, and specific implementation may refer to the foregoing method embodiment, and is not described herein again.

The method, the apparatus, and the computer program product of the electronic device for automatically driving a vehicle provided in the embodiments of the present application include a computer-readable storage medium storing program codes, where instructions included in the program codes may be used to execute the method described in the foregoing method embodiments, and specific implementations may refer to the method embodiments and are not described herein again.

Unless specifically stated otherwise, the relative steps, numerical expressions, and values of the components and steps set forth in these embodiments do not limit the scope of the present application.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

In the description of the present application, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, and are only for convenience of description and simplicity of description, and do not indicate or imply that the device or element being referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, should not be construed as limiting the present application. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.

Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present application, and are used for illustrating the technical solutions of the present application, but not limiting the same, and the scope of the present application is not limited thereto, and although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope disclosed in the present application; such modifications, changes or substitutions do not depart from the spirit and scope of the exemplary embodiments of the present application, and are intended to be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A method of automatically driving a vehicle, the method comprising:

acquiring first running information of a target vehicle at the current moment, second running information corresponding to related vehicles which are far away from the target vehicle within a first specified range, and lane information of related lanes which are far away from the target vehicle within a second specified range;

predicting a driving score for the target vehicle to travel with a plurality of different predetermined driving strategies through a first neural network model, the first driving information, the second driving information, and the lane information; wherein the driving score is used to characterize a driving condition and a vehicle performance of the target vehicle driving in the predetermined driving strategy; the first neural network model is obtained by training corresponding sample data under different traffic conditions;

and determining a target driving strategy from the plurality of different driving strategies according to the driving scores respectively corresponding to the plurality of different driving strategies, so that the target vehicle drives according to the target driving strategy at the next moment of the current moment.

2. The method of claim 1, wherein the relevant vehicles within a specified range from the target vehicle comprise at least one of:

a first relevant vehicle which is in the lane of the target vehicle, is away from the target vehicle by a distance smaller than a first distance threshold value and is positioned in front of the driving direction of the target vehicle;

a second relevant vehicle which is in a right lane of the lane where the target vehicle is located, is less than a second distance threshold value away from the target vehicle and is located in front of the driving direction of the target vehicle;

a third related vehicle which is in the right lane of the lane where the target vehicle is located, is away from the target vehicle by a distance smaller than a third distance threshold value and is behind the driving direction of the target vehicle;

a fourth relevant vehicle which is in a left lane of the lane where the target vehicle is located, is less than a fourth distance threshold value away from the target vehicle and is located in front of the driving direction of the target vehicle;

a fifth related vehicle which is located behind the target vehicle in the driving direction and has a distance to the target vehicle smaller than a fifth distance threshold in a lane on the right side of the lane where the target vehicle is located.

3. The method of claim 1, wherein the first travel information of the target vehicle at the current time includes a first location and a first speed of the target vehicle;

the step of acquiring second travel information corresponding to a relevant vehicle within a specified range from the target vehicle includes:

acquiring a second position and a second speed of the relevant vehicle at the current moment;

calculating a relative position of the second position with respect to the first position, and a relative velocity of the second velocity with respect to the first velocity;

and determining the relative position and the relative speed as second running information corresponding to the relevant vehicle.

4. The method of claim 3, further comprising:

if the lane where the target vehicle is located is the rightmost lane in the road, setting second driving information of the relevant vehicle located in the right lane where the target vehicle is located to be 0;

and if the lane where the target vehicle is located is the leftmost lane in the road, setting the second driving information of the relevant vehicle located in the left lane of the lane where the target vehicle is located to be 0.

5. The method of claim 1, wherein the relevant lanes include a lane in which the target vehicle is located, a right side lane of the lane in which the target vehicle is located, and a left side lane of the lane in which the target vehicle is located;

a step of acquiring lane information of a relevant lane within a second specified range from the target vehicle, including:

taking the position of the vehicle at the current moment as an origin position;

calculating the transverse distance and the longitudinal distance between each position of a first preset number of continuous positions in the related lane and the position of the origin; wherein the distance between each two adjacent positions in the first preset number of consecutive positions is equal;

determining the set of the transverse distance and the longitudinal distance corresponding to each position in the first preset number of continuous positions as the lane information of the related lane within a second specified range away from the target vehicle.

6. The method of claim 1, wherein the plurality of different driving strategies includes at least any two of:

accelerating straight running, keeping speed straight running, decelerating straight running, straight emergency braking, changing lanes to the left lane at a constant speed and changing lanes to the right lane at a constant speed.

7. The method of claim 1, wherein the first neural network model is trained by:

acquiring sample data; the sample data comprises speed information and position information of a target sample vehicle, speed information and position information of a related sample vehicle related to the target sample vehicle, and lane information of a lane corresponding to the target sample vehicle;

calculating a loss value of the initial neural network model corresponding to each moment in a second preset number of continuous moments according to the sample data and the loss function; wherein the loss function comprises one or more of a collision parameter, an energy consumption parameter, a lane change penalty parameter and a driving efficiency parameter;

and adjusting parameters of the initial neural network model according to the loss value, and determining the corresponding initial neural network model as the first neural network model when the training stopping condition is met.

8. The method of claim 7, further comprising:

determining, by the initial neural network model, prediction scores corresponding to the plurality of different driving strategies at a time next to the current time in the second preset number of consecutive times;

taking the driving strategy with the maximum predicted score as a driving strategy at the next moment, and acquiring sample data of the target sample vehicle at the next moment generated by driving with the driving strategy at the next moment;

and continuing to train the initial neural network model based on the sample data at the next moment.

9. The method of claim 1, wherein determining a target driving maneuver from the plurality of different driving maneuvers based on the respective driving scores for the plurality of different driving maneuvers comprises:

determining the driving strategy with the driving score higher than a preset score threshold value as an alternative driving strategy set;

determining a target driving strategy from the set of alternative driving strategies.

10. The method of claim 9, wherein the step of determining a target driving strategy from the set of alternative driving strategies comprises:

judging whether the driving strategy with the highest driving score in the alternative driving strategy set meets a preset safe driving condition or not;

if so, determining the driving strategy as a target driving strategy;

and if not, deleting the driving strategy from the alternative driving strategy set.

11. The method according to any one of claims 1-10, further comprising:

if the target driving strategy is changing lanes to the left lane at a constant speed or changing lanes to the right lane at a constant speed, controlling the vehicle to change lanes by the following formula:

wherein the content of the first and second substances,

indicating the angle of rotation, theta, of the steering wheel_nAnd theta_fRespectively represent the eyesAn included angle formed by the first position in the front, the second position in the front and the self-parking position on the marking lane, delta t is a time step length, k_f、k_nAnd k_IA constant representing driving behavior, the target lane being a lane in which the target vehicle is prepared to switch from the lane in which the target vehicle is located.

12. An automatic driving apparatus for a vehicle, characterized in that the apparatus comprises:

the information acquisition module is used for acquiring first running information of a target vehicle at the current moment, second running information corresponding to related vehicles within a first specified range away from the target vehicle and lane information of related lanes within a second specified range away from the target vehicle;

a prediction module to predict a driving score for the target vehicle to travel with a plurality of different predetermined driving strategies via a first neural network model, the first travel information, the second travel information, and the lane information; wherein the driving score is used to characterize a driving condition and a vehicle performance of the target vehicle driving in the predetermined driving strategy; the first neural network model is obtained by training corresponding sample data under different traffic conditions;

and the target driving strategy determining module is used for determining a target driving strategy from the plurality of different driving strategies according to the driving scores respectively corresponding to the plurality of different driving strategies so as to enable the target vehicle to drive according to the target driving strategy at the next moment of the current moment.

13. An electronic device comprising a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the method of any one of claims 1-11.

14. A computer-readable storage medium having computer-executable instructions stored thereon that, when invoked and executed by a processor, cause the processor to implement the method of any one of claims 1-11.