CN112519765A

CN112519765A - Vehicle control method, apparatus, device, and medium

Info

Publication number: CN112519765A
Application number: CN201910828972.2A
Authority: CN
Inventors: 吕雷兵; 姚冬春; 王俊平; 于宁; 杨凡
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2019-09-03
Filing date: 2019-09-03
Publication date: 2021-03-19

Abstract

The embodiment of the application discloses a vehicle control method, a vehicle control device, vehicle control equipment and a vehicle control medium, and relates to an automatic driving technology in the technical field of computers. Wherein, this vehicle control method includes: acquiring motion trail information of a dynamic obstacle identified in the driving process of the vehicle; inputting the motion trajectory information to a trained machine learning model to predict a motion intent of the dynamic obstacle; and controlling the route planning of the vehicle or the running state of the vehicle according to the predicted movement intention. According to the embodiment of the application, the motion trail of the dynamic barrier is identified through the machine learning model so as to determine the motion intention of the dynamic barrier, and then the vehicle is controlled according to the motion intention, so that the accuracy of vehicle control is improved.

Description

Vehicle control method, apparatus, device, and medium

Technical Field

The embodiment of the application relates to a computer technology, in particular to an automatic driving technology, and specifically relates to a vehicle control method, device, equipment and medium.

Background

In the automatic driving technology, an automatic driving system of a vehicle performs path planning and driving state control of the vehicle according to the surrounding situation of the vehicle.

When a vehicle runs in a lane, avoidance is needed for a dynamic obstacle appearing in the lane. Wherein the dynamic obstacle can be a lane-switching vehicle, a passing vehicle, a jammed vehicle, a pedestrian crossing a road, or other situations. These dynamic obstacles have a small amplitude of motion and may have a tentative reciprocating motion. Therefore, it is difficult for the automatic driving system to determine the movement intention of the dynamic obstacle through the recognition of a long stable movement trajectory, which causes difficulty in path planning and driving state control of the vehicle.

Disclosure of Invention

The embodiment of the application provides a vehicle control method, a vehicle control device, vehicle control equipment and a vehicle control medium, and accuracy of vehicle control is improved.

In a first aspect, an embodiment of the present application provides a vehicle control method, including:

acquiring motion trail information of a dynamic obstacle identified in the driving process of the vehicle;

inputting the motion trajectory information to a trained machine learning model to predict a motion intent of the dynamic obstacle;

and controlling the route planning of the vehicle or the running state of the vehicle according to the predicted movement intention.

According to the embodiment of the application, the movement track information of the dynamic obstacle identified in the driving process of the vehicle is acquired and input into the trained machine learning model to predict the movement intention of the dynamic obstacle, and then the route planning of the vehicle or the driving state of the vehicle is controlled according to the predicted movement intention. The problem that an automatic driving system is difficult to determine the movement intention of a dynamic barrier through a long stable movement track and difficulty is caused to path planning and driving state control of a vehicle is solved, the movement track of the dynamic barrier is identified through a machine learning model to determine the movement intention of the dynamic barrier, the vehicle is controlled according to the movement intention, and accuracy of vehicle control is improved.

In addition, according to the vehicle control method of the above embodiment of the present application, the following additional technical features may also be provided:

optionally, the motion trajectory sample set required in the training process of the machine learning model is a sample set labeled with a motion intention, and the obtaining process of the motion trajectory sample set specifically includes:

acquiring motion trail information of a dynamic obstacle acquired in the driving process of a vehicle in a lane;

determining a movement intention from the movement track information of the dynamic barrier according to a set intention phenomenon rule;

wherein the motion trail information and the binary set of motion intentions constitute the motion intention sample.

One embodiment in the above application has the following advantages or benefits: the movement intention of the dynamic barrier is determined according to the set intention phenomenon rule, and the movement track information and the binary set of the movement intention form a movement intention sample, so that the movement track analysis speed of the dynamic barrier is increased, and the movement track sample is favorably obtained.

Optionally, according to a set intention phenomenon rule, determining a movement intention from the movement trajectory information of the dynamic obstacle includes at least one of:

determining that the movement intention is a pedestrian crossing a lane if the dynamic obstacle is identified as a pedestrian and the movement trajectory information is identified as crossing the lane;

determining that the movement intention is a pedestrian going straight along a lane if the dynamic obstacle is identified as a pedestrian and the movement trajectory information is identified as going straight along the same edge of the lane;

determining that the movement is intended to switch lanes or overtake if the dynamic barrier is identified as a vehicle and the movement trajectory information is identified as entering a current lane from an adjacent lane, being in front of the current vehicle.

One embodiment in the above application has the following advantages or benefits: the complexity of determining the movement intention of the dynamic obstacle is simplified by determining the movement intention of the dynamic obstacle according to the signs which show the movement intention and appear in the movement track of the dynamic obstacle.

Optionally, the obtaining process of the required exercise intention sample set in the training process of the machine learning model specifically includes:

acquiring a stable driving control instruction of the vehicle aiming at the motion track information of the dynamic obstacle as the motion intention;

One embodiment in the above application has the following advantages or benefits: the stable driving control instruction of the vehicle for the motion trail information of the dynamic obstacle is used as the motion intention, and the motion trail information and the binary set of the motion intention form the motion intention sample, so that favorable conditions are provided for obtaining the motion intention sample set.

Optionally, the obtaining of the stable driving control instruction of the vehicle for the motion trajectory information of the dynamic obstacle as the motion intention includes:

in response to the dynamic obstacle entering a driving influence range of the vehicle, acquiring at least one driving control instruction for the vehicle;

selecting a last driving control command from the at least one driving control command as the stable driving control command in response to the dynamic obstacle falling out of the driving influence range of the vehicle;

and acquiring the movement intention based on the stable driving control instruction.

One embodiment in the above application has the following advantages or benefits: the stable driving control instruction is obtained from the at least one driving control instruction, and the movement intention is obtained based on the stable driving control instruction, so that the reference meaning can be provided for the subsequent dynamic obstacle movement intention.

Optionally, inputting the motion trajectory information into a trained machine learning model to predict the motion intention of the dynamic obstacle includes:

inputting the motion trail information into the machine learning model to identify the motion intention;

and if the recognized movement intention result is uncertain according to the probability value of the movement intention recognition result, returning to execute the acquisition operation of the movement track information, and overlapping the newly acquired movement track information into the historically acquired movement track information for inputting the machine learning model until the movement intention of the dynamic obstacle is determined.

One embodiment in the above application has the following advantages or benefits: when the movement intention of the dynamic barrier cannot be accurately determined according to the collected movement track information, the movement intention recognition accuracy of the dynamic barrier is improved by acquiring new movement track information and combining the newly collected movement track information and historical movement track information to perform secondary movement intention recognition.

Optionally, the movement intention is obtained by clustering and then labeling the movement trajectory samples in the movement trajectory sample set that does not contain a movement intention.

One embodiment in the above application has the following advantages or benefits: clustering the motion track samples and labeling each clustered category to obtain the motion intention of the motion track samples.

Optionally, the data format of the motion trail information is that at least one trace point recorded sequentially represents the motion trail information, and each trace point records a binary set of a speed of the dynamic barrier moving along the lane line direction and a speed of the dynamic barrier moving perpendicular to the lane line direction.

One embodiment in the above application has the following advantages or benefits: and recording the moving speed of the dynamic barrier along the direction of the lane line and the moving speed of the dynamic barrier perpendicular to the direction of the lane line through each track point, thereby providing a reliable condition for subsequently determining the moving intention of the dynamic barrier.

In a second aspect, an embodiment of the present application further provides a vehicle control apparatus, including:

the motion track acquisition module is used for acquiring motion track information of the dynamic barrier identified in the running process of the vehicle;

a predicted movement intent module to input the movement trajectory information to a trained machine learning model to predict a movement intent of the dynamic obstacle;

and the vehicle control module is used for controlling the route planning of the vehicle or the running state of the vehicle according to the predicted movement intention.

In a third aspect, an embodiment of the present application further provides an electronic device, where the electronic device includes:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the vehicle control method of any of the embodiments.

In a fourth aspect, the present embodiments also provide a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the vehicle control method according to any one of the embodiments.

Other effects of the above-described alternatives of the present application will be described below with reference to specific embodiments.

Drawings

The drawings are included to provide a better understanding of the present solution and are not intended to limit the present application. Wherein:

FIG. 1 is a schematic flow chart illustrating a vehicle control method according to an embodiment of the present disclosure;

FIG. 2 is a schematic flow chart illustrating a training process of a machine learning model according to a second embodiment of the present disclosure;

FIG. 3 is a flow chart of another machine learning model training process provided in the second embodiment of the present application;

FIG. 4 is a flowchart illustrating a further machine learning model training process provided in the second embodiment of the present application;

FIG. 5 is a schematic flow chart illustrating another vehicle control method according to a third embodiment of the present application;

FIG. 6 is a schematic structural diagram of a vehicle control device according to a fourth embodiment of the present application;

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

Detailed Description

The following description of the exemplary embodiments of the present application, taken in conjunction with the accompanying drawings, includes various details of the embodiments of the application for the understanding of the same, which are to be considered exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present application. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

The embodiment of the application provides a vehicle control method aiming at the problem that in the related art, an automatic driving system is difficult to determine the movement intention of a dynamic obstacle through the identification of a long stable movement track, which causes difficulty in path planning and driving state control of a vehicle.

A vehicle control method, apparatus, device, and medium of the embodiments of the present application are described in detail below with reference to the accompanying drawings.

Example one

Fig. 1 is a flowchart of a vehicle control method provided in an embodiment of the present application, where the embodiment of the present application is applicable to a scenario in which a vehicle is controlled based on an intention of a motion of a dynamic obstacle identified during a vehicle traveling process, and the method may be executed by a vehicle control device to implement control of a path plan or a traveling state of the vehicle, and the device may be implemented by software and/or hardware and may be integrated inside an electronic device. In this embodiment, the electronic device may be any hardware device having a data processing function, such as: vehicle-mounted computers, intelligent drivers, and the like. The method specifically comprises the following steps:

and S101, acquiring the motion trail information of the dynamic obstacle identified in the running process of the vehicle.

In this embodiment, the dynamic obstacle may include a vehicle, a pedestrian, and others. Other animals can be cats, dogs, etc., or horse cars, etc.

Usually, the dynamic barrier is in motion, i.e. the position of the dynamic barrier is different at different times. Therefore, in the embodiment, continuous multi-frame images including the dynamic obstacle can be collected through the camera arranged on the current vehicle, and the position of the dynamic obstacle in each frame of image in the multi-frame images is analyzed to obtain the motion track information of the dynamic obstacle. Or the laser point cloud data can be collected through the laser radar to obtain the position of the dynamic obstacle, and the motion track information can be obtained through continuous multi-frame collection. The embodiment of the present application does not limit the means for acquiring the motion trajectory information.

The setting position of the camera can be adaptively set according to the actual application requirements, and the setting position is not particularly limited. For example, a camera can be respectively arranged at the head and the tail of the current vehicle; alternatively, a camera or the like may be provided on the roof of the current vehicle.

It should be noted that, if a camera is arranged on the roof of the current vehicle, the present embodiment may set the camera as a 360-degree panoramic camera, so as to obtain the motion trajectory information of the dynamic obstacle in different directions.

Optionally, the motion trajectory information of the dynamic obstacle may be formed by a large number of trajectory points, a data format of the motion trajectory information of the dynamic obstacle of this embodiment is that the motion trajectory information is represented by at least one trajectory point recorded sequentially, and each trajectory point records a binary set of a speed of the dynamic obstacle moving along a lane line direction and a speed of the dynamic obstacle moving perpendicular to the lane line direction.

The track points can be position points where the dynamic barrier is located at different moments.

Since the movement of the dynamic obstacle has the time attribute, the present embodiment may sequentially record a plurality of position points of the dynamic obstacle according to the time attribute, and form the motion trajectory information of the dynamic obstacle by the recorded plurality of position points.

Correspondingly, the speed of the dynamic barrier recorded by each track point moving along the direction of the lane line and the speed of the dynamic barrier moving perpendicular to the direction of the lane line can be obtained by calculation according to the position information of the dynamic barrier in the acquisition time period.

S102, inputting the motion trail information into a trained machine learning model to predict the motion intention of the dynamic obstacle.

In the present embodiment, the movement intention of the dynamic obstacle may be determined according to the category of the dynamic obstacle.

For example, if the dynamic obstacle is a vehicle, the movement intent of the vehicle may include: lane changing, accelerating, decelerating, overtaking, plugging, sudden stop and the like. If the dynamic obstacle is a pedestrian, the intent of movement of the pedestrian may include: straight running, cutting into a lane, crossing a lane, starting, accelerating, sudden stop and the like.

Optionally, in this embodiment, the motion trajectory information of the dynamic obstacle may be input to a trained machine learning model as input data, and the machine learning model performs intention recognition on the motion trajectory of the dynamic obstacle to predict the movement intention of the dynamic obstacle.

It should be noted that, the specific process of training the machine learning model in this embodiment will be described in detail in the following examples, which are not described in detail herein.

S103, controlling the route planning of the vehicle or the running state of the vehicle according to the predicted movement intention.

The route planning of the control vehicle refers to replanning the driving route of the vehicle; controlling the running state of the vehicle may refer to reducing the running speed of the vehicle, etc., and is not particularly limited herein.

In the embodiment, if the dynamic barrier is determined to be in the safe range of the current vehicle running track according to the movement intention of the dynamic barrier, path planning or running speed control is carried out on the current vehicle;

and if the dynamic obstacle is determined not to be in the safe range of the current vehicle passing track according to the movement intention of the dynamic obstacle, keeping the current driving state of the current vehicle.

The safety range is a range that may affect safe driving of the vehicle, and the vehicle needs to perform processing such as avoidance, detour, and monitoring of an obstacle in the safety range. The safety range may be determined according to the current vehicle running speed. For example, if the current vehicle travel speed is 80 kilometers per hour (km/h), the safe range of the current vehicle travel trajectory may be 80 meters (m); for another example, if the current vehicle running speed is 100km/h, the safety range of the current vehicle running track is 100 m. I.e. the faster the vehicle is travelling, the larger the required safety range and vice versa.

That is to say, in the embodiment of the present application, by determining the movement intention of the dynamic obstacle and comparing the movement intention with the safety range of the current vehicle passing trajectory, if the dynamic obstacle appears in the safety range of the current vehicle passing trajectory, it indicates that the current vehicle continues to travel along the current travel path according to the current travel state and may collide with the dynamic obstacle, and at this time, the travel path of the current vehicle needs to be re-planned to implement deceleration and/or detour, so as to avoid colliding with the dynamic obstacle; if the dynamic obstacle does not appear in the safe range of the current vehicle passing track, the situation that the current vehicle continuously runs according to the current running state and does not collide with the dynamic obstacle is shown, and at the moment, the running state of the current vehicle does not need to be adjusted, so that the intelligent control on the running condition of the vehicle is realized, and the situation that the vehicle snakes to avoid the dynamic obstacle is avoided.

According to the vehicle control method provided by the embodiment of the application, the movement track information of the dynamic obstacle identified in the driving process of the vehicle is acquired and input into the trained machine learning model to predict the movement intention of the dynamic obstacle, and then the route planning of the vehicle or the driving state of the vehicle is controlled according to the predicted movement intention. The problem that an automatic driving system is difficult to determine the movement intention of a dynamic barrier through a long stable movement track and difficulty is caused to path planning and driving state control of a vehicle is solved, the movement track of the dynamic barrier is identified through a machine learning model to determine the movement intention of the dynamic barrier, the vehicle is controlled according to the movement intention, and accuracy of vehicle control is improved. In an automatic driving system, it is necessary to acquire the conditions of the surrounding environment in real time and adjust the states such as the travel path and the speed of the vehicle in accordance with the acquired conditions, thereby achieving safe driving. However, the movement track of some dynamic obstacles can quickly indicate the intention, namely the movement intention of the dynamic obstacles is identified within a plurality of frames, so that the corresponding vehicle control reflection is made. However, the motion trajectory of some dynamic obstacles is unstable, and is in a tentative state for a relatively long time, so that it is difficult to recognize the movement intention. Typically other vehicles and pedestrians, are also observing the surrounding environment, trying to switch lanes or cross lanes, etc. In such cases, it is difficult for an autonomous vehicle to make a correct determination and plan its own path and speed in a short, tentative motion trajectory. If a conservative and safe processing mode, such as speed reduction, is adopted at this time, the driving safety can be ensured, but the normal driving speed of the vehicle is very disturbed, and the situation that other vehicles and pedestrians need to wait may occur frequently. Therefore, the technical scheme provided by the embodiment of the application can learn a large amount of motion track sample data with heuristics and unknown intentions by means of a machine learning model, so that motion intentions can be predicted by the aid of the motion track points as few as possible, and response control of the vehicle is performed.

Example two

The following describes in detail the training process of the machine learning model in the embodiment of the present application with reference to fig. 2 to 4.

It should be noted that, in the embodiment of the present application, the motion trajectory samples required by the training process of the machine learning model may be divided into: the sample set marked with the movement intention and the sample set marked with no movement intention.

First, referring to fig. 2 to fig. 3, when a motion trajectory sample set required in a training process of a machine learning model in an embodiment of the present application is a sample set labeled with a motion intention, a process of training the machine learning model according to the sample set labeled with the motion intention will be specifically described.

Fig. 2 is a schematic flowchart of a machine learning model training process according to a second embodiment of the present disclosure. The machine learning model training process of the embodiment of the application specifically comprises the following steps:

s201, acquiring the motion trail information of the dynamic obstacle acquired in the driving process of the vehicle in the lane.

The implementation process and principle of S201 in this embodiment are similar to those of the first embodiment in which the motion trajectory information of the dynamic obstacle is obtained, and details of the implementation process and principle are not repeated here, and for a specific process, refer to S101 in the first embodiment.

And S202, determining a movement intention from the movement track information of the dynamic obstacle according to a set intention phenomenon rule.

In the embodiment of the present application, the intended phenomenon setting rule may be set according to actual application requirements, and is not specifically limited herein.

The intention phenomenon rule is set to be that signs indicating the movement intention appear in the movement track, and when the training sample is obtained, the movement intention can be waited for to appear through the collection of the movement track for a longer time. And screening out the samples which show clear movement intentions from the mass collected movement track data.

For example, according to the present embodiment, the determining of the movement intention from the movement trajectory information of the dynamic obstacle according to the set intention phenomenon rule includes at least one of the following:

determining that the movement intention is to switch lanes, overtake or jam if the dynamic obstacle is identified as a vehicle and the movement trajectory information is identified as entering a current lane from an adjacent lane, being in front of the current vehicle.

If the dynamic obstacle is identified as a pedestrian and the motion trail information is identified to end suddenly at any position of the lane, determining that the motion intention is pedestrian sudden stop;

if the dynamic barrier is identified as a pedestrian and the motion trail information is identified to be changed suddenly after a period of time, determining that the motion intention is pedestrian starting;

determining that the movement intention is pedestrian acceleration if the dynamic obstacle is identified as a pedestrian and the time from the last position to the current position of the movement trajectory information is identified to be short;

determining that the movement intention is vehicle acceleration if the dynamic obstacle is recognized as a vehicle and a time from a last position to a current position of the movement trajectory information is recognized to be short;

determining that the movement intention is vehicle deceleration if the dynamic obstacle is recognized as a vehicle and the time from the last position to the current position of the movement trajectory information is recognized to be longer;

determining that the movement intention is a vehicle scram if the dynamic obstacle is identified as a vehicle and the movement trajectory information is identified to end abruptly at any position.

Further, after the movement intention is determined from the movement track information of the dynamic obstacle, the embodiment of the application may combine the movement track information of the dynamic obstacle and the determined binary combination of the movement intention into a movement intention sample, that is, a sample set marked with the movement intention may be obtained.

In the embodiment, the motion trajectory information is formed by a plurality of trajectory point sequences, and the motion intention can be determined according to the type of the dynamic obstacle. For example, if the dynamic obstacle is a pedestrian, the movement intent may include the pedestrian crossing the lane, the pedestrian going straight along the lane; as another example, if the dynamic obstacle is a vehicle, the movement intent may include vehicle acceleration, vehicle cut-in, etc.

For example, if the dynamic obstacle is a pedestrian, the acquired motion trajectory information is moved from point a to point B and from point B to point C, and the motion intention is to cross a lane, a pedestrian motion intention sample may be composed by a-B-C, cross a road.

And S203, training the machine learning model by using the acquired motion trail sample set marked with the motion intention.

That is to say, in the embodiment of the present application, the motion trail samples are input into the machine learning model, and the motion intentions corresponding to the motion trail samples are taken as training results, and the machine learning model is repeatedly trained until the motion trail samples are input, so that the corresponding motion intentions can be obtained. At this time, the model is set as a final machine learning model.

Fig. 3 is a schematic flow chart of another machine learning model training process provided in the second embodiment of the present application. The machine learning model training process of the embodiment of the application specifically comprises the following steps:

s301, acquiring the motion trail information of the dynamic obstacle acquired in the driving process of the vehicle in the lane.

The implementation process and principle of S301 are similar to those of the first embodiment in which the motion trajectory information of the dynamic obstacle is obtained, and details of the implementation process and principle are not described here, and for a specific process, refer to S101 in the first embodiment.

And S302, acquiring a stable driving control instruction of the vehicle aiming at the motion trail information of the dynamic obstacle as the motion intention.

In the embodiment of the present application, the stable driving control command is a command for causing a dynamic obstacle to deviate from the driving influence range of the current vehicle. The driving influence range of the vehicle may be the same as or different from the safety range. The driving influence range of the vehicle may refer to a range that may influence safe driving of the vehicle.

Optionally, in this embodiment, the obtaining of the stable driving control instruction of the vehicle for the motion trajectory information of the dynamic obstacle includes, as the motion intention:

In the present embodiment, the driving control instruction is an instruction adaptively taken based on the motion trajectory information of the dynamic obstacle, and since the motion trajectory information may be tentative, the control instruction may also be a series of instructions. That is, the control instructions can reflect the movement intention of the dynamic obstacle. For example, if the dynamic obstacle is a pedestrian, the pedestrian may approach the lane at the side of the lane for a while and look away from the lane for a while, and a tentative movement trajectory may appear, when the pedestrian wants to cross the lane when there are few vehicles. Autonomous vehicles typically respond to a series of tentative driving commands by a pedestrian by providing a deceleration control command when the pedestrian is identified as approaching the lane, but accelerating when the pedestrian is away from the lane. Until the pedestrian really crosses the lane, the vehicle makes the last control command of deceleration or stopping; alternatively, the pedestrian is not crossing the lane and is being exceeded by the vehicle, and the last control command for the vehicle is acceleration or normal speed. The last control instruction can be used as a stable driving control instruction for responding the pedestrian motion track information, has reference significance and can be used as a training sample. In subsequent predictions of movement intent, if the pedestrian has a similar movement trajectory, the corresponding steady driving control command may be employed to respond to the pedestrian's movement intent.

And S303, training the machine learning model by using the acquired motion trail sample set marked with the motion intention.

The implementation process and principle of S303 are the same as or similar to S203 in the second embodiment, and the specific process may be referred to as S203, which is not described herein in detail.

In another implementation scenario of the present application, the embodiment may further directly obtain, from the database, a sample set obtained by manually performing a movement intention labeling on a movement trajectory of the dynamic obstacle. And then, training the machine learning model according to the motion trail sample set.

Further, with reference to fig. 4, a process of training a machine learning model according to a sample set without a movement intention when a movement trajectory sample set required in a training process of the machine learning model in the embodiment of the present application is a sample set without a movement intention is specifically described.

Fig. 4 is a schematic flow chart of a training process of another machine learning model provided in the second embodiment of the present application. The training process of the machine learning model of the embodiment of the application specifically comprises the following steps:

s401, acquiring motion trail information of the dynamic obstacle acquired in the driving process of the vehicle in the lane.

The implementation process and principle of S401 are similar to those of the first embodiment in which the motion trajectory information of the dynamic obstacle is obtained, and details of the implementation process and principle are not described here, and for a specific process, refer to S101 in the first embodiment.

And S402, clustering the motion trail information of the dynamic barrier, and then marking intentions to obtain a motion trail sample set marked with motion intentions.

The embodiment can obtain different categories by clustering the motion tracks of the dynamic obstacles. And then, carrying out movement intention labeling on different types obtained by clustering in a manual labeling mode to obtain a movement track sample set.

And S403, training the machine learning model by using the acquired motion trail sample set marked with the motion intention.

The implementation process and principle of S403 are the same as or similar to S203 in the second embodiment, and the specific process may be referred to as S203, which is not described herein in detail.

In other words, according to the embodiment of the application, the intention of movement is labeled on different movement track categories obtained by clustering in a manual labeling mode, so that a movement track sample set is obtained, and the machine learning model is trained according to the movement track sample set.

EXAMPLE III

As can be seen from the above analysis, the embodiment of the present application inputs the motion trajectory information of the dynamic obstacle into the trained machine learning model to predict the motion intention of the dynamic obstacle, and controls the route planning of the vehicle or the driving state of the vehicle according to the predicted motion intention.

In another implementation scenario of the present application, when the motion trail information is input into the trained machine learning model for motion intention recognition, if the motion intention result cannot be recognized according to the probability value of the motion intention recognition result, the embodiment may acquire new motion trail information and superimpose the newly acquired motion trail information onto the historically acquired motion trail information, so as to input the machine learning model until the motion intention of the dynamic obstacle is determined. The following describes the above-described situation of the vehicle control method according to the embodiment of the present application with reference to fig. 5.

Fig. 5 is a schematic flowchart of another vehicle control method according to a third embodiment of the present application. As shown in fig. 5, the vehicle control method includes the steps of:

s501, acquiring the motion trail information of the dynamic obstacle identified in the running process of the vehicle.

S502, inputting the motion trail information into the machine learning model to identify the motion intention.

And S503, if the movement intention result is not determined according to the probability value of the movement intention identification result, returning to execute the acquisition operation of the movement track information, and overlapping the newly acquired movement track information into the historically acquired movement track information for inputting the machine learning model until the movement intention of the dynamic barrier is determined.

In this embodiment, in the training process, various exercise intentions may be marked on each piece of exercise trajectory information serving as a sample, and the exercise trajectory sample set marked with the exercise intentions is used to train the machine learning model, so that when the machine learning model identifies the exercise trajectory information, the probability values of the exercise trajectory information corresponding to the various exercise intentions are correspondingly obtained. If any probability value of the multiple types of movement intentions exceeds a preset threshold value, determining the movement intentions exceeding the preset threshold value as the movement intentions of the movement track information; if the probability values of the various movement intentions do not exceed the preset threshold, it is indicated that the movement intention error determined according to the movement track information of the dynamic obstacle is large, and the movement intention result is uncertain.

For example, identifying the movement intent of the dynamic obstacle by the machine learning model includes: the method comprises the steps of crossing a lane and running straight along the lane, wherein the probability value of crossing the lane is 30%, the probability value of running straight along the lane is 50%, if the preset threshold value is 70%, the probability value of neither intention exceeds 70%, and the movement intention of the dynamic obstacle cannot be determined.

Further, in this embodiment, when it is determined that the trajectory point in the motion trajectory information is less than the preset quantity value, or the speed of the trajectory point is lower than the preset speed value, it is described that the data for predicting the motion intention is not representative, and it may also be recognized that the motion intention result is uncertain.

The preset quantity value can be adaptively set according to the actual application requirement, and is not specifically limited herein. For example, the preset number value is set to 1 or 2.

The preset speed threshold is set according to the type of the dynamic obstacle. That is, when the dynamic obstacles are different, the corresponding preset speed thresholds are also different.

For example, when the acquired motion trail information of the dynamic obstacle is input into a trained machine learning model for identifying the motion intention, if the motion intention result is identified as uncertain, it indicates that the motion intention of the dynamic obstacle cannot be accurately determined according to the acquired motion trail information. For this reason, the embodiment may return to the motion trail information collection operation, accumulate the newly collected motion trail information of the next frame with the multi-frame motion trail information collected historically, and then input the accumulated motion trail information into the trained machine learning model to identify the motion intention. Since the motion trajectory is further increased, it is more likely that the intention of motion is recognized. If the movement intention of the movement track is identified, ending; and if not, continuing to return to execute the motion trail information acquisition operation and the motion trail information superposition operation, and inputting the superposed motion trail information into the trained machine learning model until the motion intention of the dynamic barrier is obtained.

S504, controlling the route planning of the vehicle or the running state of the vehicle according to the predicted movement intention.

That is to say, in the embodiment, when the movement intention of the dynamic obstacle cannot be accurately determined according to the collected movement trajectory information, the movement intention recognition accuracy of the dynamic obstacle can be improved by acquiring new movement trajectory information and performing secondary movement intention recognition by combining the newly collected movement trajectory information and the historical movement trajectory information.

Example four

In order to achieve the above object, a fourth embodiment of the present application provides a vehicle control device. Fig. 6 is a schematic structural diagram of a vehicle control device according to a fourth embodiment of the present application.

As shown in fig. 6, a vehicle control apparatus according to an embodiment of the present application includes: an acquire motion trajectory module 610, a predict motion intent module 620, and a vehicle control module 630.

The motion trajectory acquiring module 610 is configured to acquire motion trajectory information of a dynamic obstacle identified in a driving process of the vehicle;

a predicted movement intent module 620 is used to input the movement trajectory information to a trained machine learning model to predict a movement intent of the dynamic obstacle;

the vehicle control module 630 is configured to control a route planning of the vehicle or a driving state of the vehicle according to the predicted movement intention.

As an optional implementation manner of the embodiment of the present application, if the motion trajectory sample set required in the training process of the machine learning model is a sample set labeled with a motion intention, the vehicle control device further includes: the device comprises a first obtaining module and a first determining module.

The first acquisition module is used for acquiring motion trail information of a dynamic obstacle acquired in the driving process of a vehicle in a lane;

the first determination module is used for determining a movement intention from the movement track information of the dynamic barrier according to a set intention phenomenon rule;

wherein the motion track information and the binary set of motion intentions constitute the motion intention sample.

As an optional implementation manner of the embodiment of the present application, the first determining module is specifically configured to:

As an optional implementation manner of the embodiment of the present application, the vehicle control apparatus further includes: and a second obtaining module.

The first acquisition module is used for acquiring the motion trail information of the dynamic barrier acquired in the driving process of the vehicle in the lane;

the second acquisition module is used for acquiring a stable driving control instruction of the vehicle aiming at the motion trail information of the dynamic obstacle as the motion intention;

As an optional implementation manner of the embodiment of the present application, the second obtaining module is specifically configured to:

As an optional implementation manner of the embodiment of the present application, the predicted movement intention module 620 is specifically configured to:

and if the movement intention result is not determined according to the probability value of the movement intention identification result, returning to execute the acquisition operation of the movement track information, and overlapping the newly acquired movement track into the historically acquired movement track for inputting the machine learning model until the movement intention of the dynamic barrier is determined.

As an optional implementation manner of the embodiment of the application, the movement intention is obtained by clustering and then labeling the movement track samples in the movement track sample set that does not contain the movement intention.

As an optional implementation manner of the embodiment of the application, the data format of the motion trajectory information is that the motion trajectory is represented by at least one trajectory point recorded sequentially, and each trajectory point records a binary group of a speed of the dynamic barrier moving along the lane line direction and a speed of the dynamic barrier moving perpendicular to the lane line direction

It should be noted that the foregoing explanation of the embodiment of the vehicle control method is also applicable to the vehicle control device of the embodiment, and the implementation principle is similar, and is not repeated here.

According to the vehicle control device provided by the embodiment of the application, the movement track information of the dynamic obstacle identified in the driving process of the vehicle is acquired and input into the trained machine learning model to predict the movement intention of the dynamic obstacle, and then the route planning of the vehicle or the driving state of the vehicle is controlled according to the predicted movement intention. The problem that an automatic driving system is difficult to determine the movement intention of a dynamic barrier through a long stable movement track and difficulty is caused to path planning and driving state control of a vehicle is solved, the movement track of the dynamic barrier is identified through a machine learning model to determine the movement intention of the dynamic barrier, the vehicle is controlled according to the movement intention, and accuracy of vehicle control is improved.

EXAMPLE five

Referring to fig. 7, an embodiment of the present application provides an electronic device 700, which includes: one or more processors 720; a memory 710 communicatively coupled to the at least one processor 720; wherein the memory 710 stores instructions executable by the at least one processor 720, the instructions being executable by the at least one processor 720 to enable the at least one processor 720 to perform a vehicle control method according to any embodiment of the present application, the method comprising:

Of course, those skilled in the art will appreciate that the processor 720 may also implement the technical solutions of the vehicle control methods provided in any of the embodiments of the present application.

The electronic device 700 shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present application.

As shown in fig. 7, electronic device 700 is embodied in the form of a general purpose computing device. The components of the electronic device 700 may include, but are not limited to: one or more processors 720, a memory 710, and a bus 750 that couples the various system components (including the memory 710 and the processors 720).

Bus 750 represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, and a processor or local bus using any of a variety of bus architectures. By way of example, such architectures include, but are not limited to, 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.

Electronic device 700 typically includes a variety of computer system readable media. Such media may be any available media that is accessible by electronic device 700 and includes both volatile and nonvolatile media, removable and non-removable media.

The memory 710 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM)711 and/or cache memory 712. The electronic device 700 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 713 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 7, commonly referred to as a "hard drive"). Although not shown in FIG. 7, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In such cases, each drive may be connected to bus 750 by one or more data media interfaces. Memory 710 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the application.

A program/utility 714 having a set (at least one) of program modules 715 may be stored, for instance, in memory 710, such program modules 715 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. The program modules 715 generally perform the functions and/or methods described in any of the embodiments herein.

The electronic device 700 may also communicate with one or more external devices 760 (e.g., keyboard, pointing device, display 770, etc.), with one or more devices that enable a user to interact with the electronic device 700, and/or with any devices (e.g., network card, modem, etc.) that enable the electronic device 700 to communicate with one or more other computing devices. Such communication may occur via input/output (I/O) interface 730. Also, the electronic device 700 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the internet) via the network adapter 740. As shown in FIG. 7, the network adapter 740 communicates with the other modules of the electronic device 700 via the bus 750. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the electronic device 700, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.

The processor 720 executes various functional applications and data processing by executing programs stored in the memory 710, for example, to implement the vehicle control method provided in the embodiments of the present application.

It should be noted that the foregoing explanation of the embodiment of the vehicle control method is also applicable to the electronic device of the embodiment, and the implementation principle thereof is similar and will not be described herein again.

According to the electronic device provided by the embodiment of the application, the movement track information of the dynamic obstacle identified in the driving process of the vehicle is acquired and input into the trained machine learning model to predict the movement intention of the dynamic obstacle, and then the route planning of the vehicle or the driving state of the vehicle is controlled according to the predicted movement intention. The problem that an automatic driving system is difficult to determine the movement intention of a dynamic barrier through a long stable movement track and difficulty is caused to path planning and driving state control of a vehicle is solved, the movement track of the dynamic barrier is identified through a machine learning model to determine the movement intention of the dynamic barrier, the vehicle is controlled according to the movement intention, and accuracy of vehicle control is improved.

EXAMPLE six

The present embodiment provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute a vehicle control method according to any one of the embodiments of the present application, the method including:

Of course, the embodiments of the present application provide a non-transitory computer-readable storage medium storing computer instructions, where the computer instructions are used to enable a computer to execute the instructions, which are not limited to the method operations described above, but also can execute the related operations in the vehicle control method provided in any embodiment of the present application.

The computer-readable storage media of the embodiments of the present application may take any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination 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 the context of 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.

A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Computer program code for carrying out operations for aspects of the present application may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + +, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).

It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present application and the technical principles employed. It will be understood by those skilled in the art that the present application 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 application. Therefore, although the present application has been described in more detail with reference to the above embodiments, the present application is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A vehicle control method characterized by comprising:

2. The method according to claim 1, wherein a motion trajectory sample set required in the training process of the machine learning model is a sample set labeled with a motion intention, and the obtaining process of the motion trajectory sample set specifically includes:

3. The method according to claim 2, wherein determining the movement intention from the movement track information of the dynamic obstacle according to a set intention phenomenon rule comprises at least one of:

4. The method according to claim 1, wherein the obtaining of the set of required motor intention samples in the training process of the machine learning model specifically comprises:

5. The method according to claim 4, wherein acquiring the stable driving control instruction of the vehicle for the motion trajectory information of the dynamic obstacle as the motion intention comprises:

6. The method of claim 1, wherein inputting the motion trajectory information to a trained machine learning model to predict the intent to move of the dynamic obstacle comprises:

and if the movement intention result is not determined according to the probability value of the movement intention identification result, returning to execute the acquisition operation of the movement track information, and overlapping the newly acquired movement track information into the historically acquired movement track information for inputting the machine learning model until the movement intention of the dynamic obstacle is determined.

7. The method of claim 1, wherein the motion intent is obtained by clustering and labeling the motion trace samples in the motion trace sample set that do not contain motion intent.

8. The method according to claim 1, wherein the data format of the motion trail information is that the motion trail information is represented by at least one track point recorded sequentially, and each track point records a binary set of a speed of the dynamic obstacle moving along a lane line direction and a speed of the dynamic obstacle moving perpendicular to the lane line direction.

9. A vehicle control apparatus characterized by comprising:

10. An electronic device, comprising:

at least one processor; and

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the vehicle control method of any of claims 1-8.

11. A non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the vehicle control method according to any one of claims 1 to 8.