CN114104006A

CN114104006A - Method and device for automatically driving vehicle to realize vehicle crossing by mistake

Info

Publication number: CN114104006A
Application number: CN202210103538.XA
Authority: CN
Inventors: 陆一帆; 胡晋; 刘云夫
Original assignee: Alibaba Damo Institute Hangzhou Technology Co Ltd
Current assignee: Alibaba Damo Institute Hangzhou Technology Co Ltd
Priority date: 2022-01-28
Filing date: 2022-01-28
Publication date: 2022-03-01

Abstract

The application discloses a method and a device for realizing vehicle mismeeting by automatically driving a vehicle, which record historical state information of an obstacle vehicle by using a state container based on a key data frame. And further, by reasonably designing the detection logic of the key data frame, the state information of the key moment is selected and recorded, and a large amount of invalid information is eliminated, so that under the limitation of the same physical capacity, more complete state information of the barrier vehicle is recorded, a guarantee is provided for better identifying a vehicle-crossing object, and the safety of automatic driving is also guaranteed. Furthermore, the obstacle vehicle is determined to be the wrong meeting object through the judgment of the three layers, the problems of false detection and missed detection caused by dynamic and static false detection and incomplete historical information are solved, the wrong meeting object is accurately identified, and the safety of automatic driving is better guaranteed.

Description

Method and device for automatically driving vehicle to realize vehicle crossing by mistake

Technical Field

The present application relates to, but not limited to, automatic driving technology, and more particularly, to a method and apparatus for automatically driving a vehicle to meet a vehicle in a wrong way.

Background

As the application field of the autonomous vehicle (own vehicle) is increasingly expanded, driving scenes to be handled are also gradually increased. The narrow road missed car scene is one of the most complex scenes in the L4 level automatic driving decision planning algorithm.

The scene of the vehicle crossing by mistake refers to the scene that the automatic driving vehicle firstly avoids other vehicles running in opposite directions under the scene of a narrow road section, and the vehicle passes through again after the opposite side passes through. However, due to the complex and various environment of the public roads, the number of traffic participants varies greatly in different time periods and road sections, and at the same time, the dynamic behavior of each vehicle is also variable. Therefore, how to accurately and timely identify the wrong meeting is one of the very key challenges for the L4-level automatic driving technology.

Disclosure of Invention

The application provides a method and a device for realizing vehicle mis-meeting by automatically driving a vehicle, which can better identify a vehicle mis-meeting object and ensure the safety of automatic driving.

The embodiment of the invention provides a method for automatically driving a vehicle to realize vehicle crossing by mistake, which comprises the following steps:

recording the state information of the barrier vehicle entering the preset range of the vehicle based on the key data frame;

and identifying whether the obstacle vehicle is a vehicle-crossing object of the vehicle according to the recorded state information of the obstacle vehicle and the road topology structure of the vehicle.

In an exemplary embodiment, before recording the state information of the obstacle vehicle entering the preset range of the host vehicle, the method further includes:

and identifying that the road where the self vehicle is located is a narrow road scene.

In an exemplary embodiment, the recording of the state information of the obstacle vehicle entering the preset range of the own vehicle includes:

acquiring and recording the state information of the barrier vehicle entering the preset range of the self vehicle according to a preset acquisition period;

processing the collected state information to delete the non-key data frames; wherein the first frame data frame of the obstacle vehicle is a key data frame.

In one illustrative example, the processing the collected status information to remove non-key data frames includes:

judging the data frame acquired in the previous acquisition period from the Nth acquisition period of the acquisition period according to the judgment standard of the key data frame, and if the data frame acquired in the previous acquisition period is the key data frame, reserving the data frame acquired in the previous acquisition period; if the data frame acquired in the previous acquisition period is not the key data frame, deleting the data frame acquired in the previous acquisition period; wherein N is an integer greater than or equal to 2.

In one illustrative example, further comprising:

if the number of the key data frames reaches the upper limit threshold of the container, deleting the key data frame with the lowest freshness according to the freshness;

wherein, the freshness refers to the time span between the data generation time and the current time, and the freshness is lower when the span is larger.

In an exemplary embodiment, the collecting of the state information of the obstacle vehicle entering the preset range of the host vehicle includes:

in the acquisition period, aiming at the currently received data frames of all the obstacle vehicles, checking whether historical state information exists in the container according to the ID of the obstacle vehicle;

generating a new queue under the condition that the queue of historical state information does not exist in the container, taking the current data frame of the obstacle vehicle as a first frame to be recorded in the queue and marking the first frame as a key data frame;

for the condition that a queue of historical state information of the obstacle vehicle already exists in the container, judging whether the last frame in the queue is a key data frame or not, if the last frame in the queue is not the key data frame, deleting the last frame and then inserting the last frame into the latest data frame collected currently; and if the last frame in the queue is a key data frame, sequentially inserting the latest data frame collected currently into the queue.

In one illustrative example, further comprising:

and after the latest data frame is inserted in sequence, marking whether the latest data frame is a key data frame or not.

In one illustrative example, further comprising:

and periodically checking whether a queue which is not updated within a preset time length exists in the container, and if so, deleting the queue.

In one illustrative example, the identifying whether the obstacle vehicle is a vehicle-crossing object of the own vehicle includes:

acquiring historical state information of each obstacle vehicle in the container; determining that the whole movement process of the obstacle vehicle is approximately opposite to the advancing direction of the vehicle; judging that the short-term motion trail of the obstacle vehicle is opposite to the advancing direction of the vehicle, and judging that the obstacle vehicle is a vehicle-crossing object of the vehicle; alternatively, the first and second electrodes may be,

acquiring historical state information of each obstacle vehicle in the container; determining that the whole movement process of the obstacle vehicle is approximately opposite to the advancing direction of the vehicle; judging that the predicted track of the obstacle vehicle is inconsistent with the advancing direction of the vehicle, and judging that the obstacle vehicle is a vehicle-crossing object of the vehicle; alternatively, the first and second electrodes may be,

acquiring historical state information of each obstacle vehicle in the container; determining that the whole movement process of the obstacle vehicle is approximately opposite to the advancing direction of the vehicle; judging that the short-term motion trail of the obstacle vehicle is not opposite to the advancing direction of the vehicle; judging that the predicted track of the obstacle vehicle is inconsistent with the advancing direction of the vehicle, and judging that the obstacle vehicle is a vehicle-crossing object of the vehicle; alternatively, the first and second electrodes may be,

acquiring historical state information of each obstacle vehicle in the container; determining that the whole movement process of the obstacle vehicle is approximately opposite to the advancing direction of the vehicle; judging that the predicted track of the obstacle vehicle is consistent with the advancing direction of the vehicle; and judging that the short-term motion trail of the obstacle vehicle is opposite to the advancing direction of the vehicle, and judging that the obstacle vehicle is a wrong meeting object of the vehicle.

The embodiment of the application also provides a computer-readable storage medium, which stores computer-executable instructions for executing any one of the above methods for realizing vehicle-crossing by automatically driving the vehicle.

The embodiment of the application further provides a device for realizing vehicle mismeeting by automatically driving the vehicle, which comprises a memory and a processor, wherein the memory stores the following instructions which can be executed by the processor: for performing the steps of the method of any of the above described automated guided vehicle for achieving a missed vehicle crossing.

The embodiment of the present application further provides a device for realizing vehicle mismeeting by automatically driving a vehicle, including: the device comprises a recording module and a first identification module; wherein the content of the first and second substances,

the recording module is used for recording the state information of the barrier vehicle entering the preset range of the vehicle based on the key data frame;

and the first identification module is used for identifying whether the obstacle vehicle is a vehicle-crossing object of the vehicle according to the recorded state information of the obstacle vehicle and the lane topological structure of the vehicle.

In one illustrative example, further comprising: and the second identification module is used for identifying the narrow road scene of the road where the self-vehicle is located.

In one illustrative example, the recording module is to:

acquiring and recording the state information of the barrier vehicle entering the preset range of the self vehicle according to a preset acquisition period; processing the collected state information to delete the non-key data frames; wherein the first frame data frame of the obstacle vehicle is a key data frame.

In one illustrative example, the recording module is further configured to:

when the number of the key frames reaches the upper limit threshold of the container, deleting the key data frame with the lowest freshness according to the freshness; wherein, the freshness refers to the time span between the data generation time and the current time, and the larger the span is, the lower the freshness is.

In one illustrative example, the first identification module is configured to:

According to the method for realizing the vehicle wrong meeting by the automatic driving vehicle, the historical state information of the obstacle vehicle is recorded by using the state container based on the key data frame. And further, by reasonably designing the detection logic of the key data frame, the state information of the key moment is selected and recorded, and a large amount of invalid information is eliminated, so that under the limitation of the same physical capacity, more complete state information of the barrier vehicle is recorded, a guarantee is provided for better identifying a vehicle-crossing object, and the safety of automatic driving is also guaranteed.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Drawings

The accompanying drawings are included to provide a further understanding of the claimed subject matter and are incorporated in and constitute a part of this specification, illustrate embodiments of the subject matter and together with the description serve to explain the principles of the subject matter and not to limit the subject matter.

FIG. 1 is a schematic flow chart illustrating a method for automatically driving a vehicle to implement a vehicle meeting by mistake in an embodiment of the present application;

FIG. 2 is a schematic flow chart illustrating a process of collecting status information of a barrier vehicle according to an embodiment of the present application;

FIG. 3 is a schematic flow chart illustrating the identification of a wrong carriage return object in the embodiment of the present application;

FIG. 4 is a key data frame diagram of an obstacle vehicle according to an embodiment of the present application;

fig. 5 is a schematic structural diagram of a device for automatically driving a vehicle to realize a vehicle meeting by mistake in the embodiment of the application.

Detailed Description

To make the objects, technical solutions and advantages of the present application more apparent, embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be noted that the embodiments and features of the embodiments in the present application may be arbitrarily combined with each other without conflict.

In one exemplary configuration of the present application, a computing device includes one or more processors (CPUs), input/output interfaces, a network interface, and memory.

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include non-transitory computer readable media (transient media), such as modulated data signals and carrier waves.

The steps illustrated in the flow charts of the figures may be performed in a computer system such as a set of computer-executable instructions. Also, while a logical order is shown in the flow diagrams, in some cases, the steps shown or described may be performed in an order different than here.

Fig. 1 is a schematic flowchart of a method for automatically driving a vehicle to implement a vehicle meeting by mistake in an embodiment of the present application, as shown in fig. 1, including:

step 101: and recording the state information of the barrier vehicle entering the preset range of the vehicle based on the key data frame.

In an exemplary embodiment, the step may be preceded by:

presetting a preset range taking the self-vehicle as a center, such as a circle taking the self-vehicle as a circle center and a preset value as a radius;

the vehicle that has entered the preset range is regarded as an obstacle vehicle.

Among others, vehicles may include, but are not limited to, such as: automotive vehicles, non-automotive vehicles, and the like.

In one illustrative example, historical status information of the obstacle vehicle may be recorded based on a status container of the key data frame. In the embodiment of the present application, the information of each frame of data of the obstacle vehicle is recorded in the container, but the non-key data frames are periodically cleaned (i.e., deleted).

In one embodiment, the decision criteria for a key data frame may include, but are not limited to, such as: a first frame data frame in which an obstacle vehicle is present; data frames with a movement distance greater than a preset distance threshold, such as 2.0 meters, from the previous key data frame; data frames with an angle change greater than a preset angle threshold, such as 7 degrees, from the previous key data frame; comparing the data frames with the changed dynamic and static states of the previous key data frame; the latest data frame, etc.

In one illustrative example, step 101 may comprise:

acquiring and recording the state information of the barrier vehicle entering the preset range of the self vehicle according to a preset acquisition period; the collected status information is processed to delete non-critical data frames. In one embodiment, the data frames acquired in the previous acquisition period, i.e., (N-1), may be processed to delete non-key data frames starting from the nth acquisition period, where the first frame data frame of the obstacle vehicle is a key data frame, so N may be an integer greater than or equal to 2.

In one embodiment, the status information may include, but is not limited to, such as: current position information of the obstacle vehicle, speed, acceleration, driving state (such as static or dynamic), angle information with respect to the own vehicle, and the like. It should be noted that the state information is obtained by an existing sensing module, and the specific implementation of the sensing module, i.e. how to obtain the state information, is not used to limit the scope of the present application, and the present application emphasizes the use of the state information.

In one embodiment, processing the collected status information to remove non-key data frames may include:

from the third acquisition cycle, judging the data frame acquired in the previous acquisition cycle according to the judgment standard of the key data frame, and if the data frame acquired in the previous acquisition cycle is the key data frame, keeping the data frame acquired in the previous acquisition cycle; and if the data frame acquired in the previous acquisition period is not the key data frame, deleting the data frame acquired in the previous acquisition period.

In the embodiment of the application, in a first acquisition period, the acquired state information is recorded, and the acquired first frame data frame of the obstacle vehicle is a key data frame; in a second acquisition period, recording the acquired state information; and starting from the third acquisition period, recording the state information acquired in the acquisition period in each acquisition period, judging the state information acquired in the previous acquisition period, if the state information acquired in the previous acquisition period is a key data frame, keeping the state information acquired in the previous acquisition period, and if the state information acquired in the previous acquisition period is not a key data frame, deleting the state information acquired in the previous acquisition period. Therefore, the stored data frames are all key information which can be used for judging whether the obstacle vehicle is the wrong meeting object or not, irrelevant information is eliminated, and therefore under the limit of the same physical capacity, a more complete obstacle vehicle state sequence is recorded, and the automatic driving vehicle has a longer view when the wrong meeting object is identified.

In an exemplary embodiment, step 101 may further include:

if the number of key data frames cumulatively reaches the upper threshold of the container, then the key data frame with the lowest freshness is deleted according to the freshness. The freshness refers to a time span between the data generation time and the current time, and the larger the span is, the lower the freshness is.

In one embodiment, collecting the state information of the obstacle vehicle may include:

in the acquisition period, aiming at the currently received data frames of all obstacle vehicles, whether historical state information exists in a container or not is checked according to the ID of the obstacle vehicle;

for the condition that a queue of historical state information of the obstacle vehicle already exists in the container, judging whether the last frame in the queue is a key data frame or not, if the last frame in the queue is not the key data frame, deleting the last frame and then inserting the last frame into the latest data frame collected currently; and if the last frame in the queue is a key data frame, sequentially inserting the latest data frame collected currently into the queue. Furthermore, after the latest data frames are inserted in sequence, whether the latest data frames are key data frames or not can be marked, so that whether the latest data frames are key data frames or not can be judged by directly utilizing the marks when the next acquisition period is judged. It should be noted that, instead of marking, the determination may be performed in the next acquisition cycle.

The method can also comprise the following steps: checking whether the number of the data frames stored in the queue reaches the upper limit threshold value of the container, if so, deleting the frame data frame at the front of the queue (namely, the frame data frame farthest from the current moment), and if not, ending the operation.

In an exemplary embodiment, the method may further include:

and periodically checking whether the queue which is not updated within a preset time length exists in the container, and if so, deleting the whole queue. In this case, the state information of the obstacle vehicle that is not already within the preset range of the own vehicle is deleted to leave a space for recording useful state information of the obstacle vehicle.

Fig. 2 is a schematic flow chart of collecting status information of an obstacle vehicle in the embodiment of the present application, and in an embodiment, as shown in fig. 2, the method may include:

step 200: in the acquisition cycle, the latest state information of a certain obstacle vehicle is acquired, stored in the latest data frame, and waits for the container of the obstacle vehicle to be inserted.

Step 201: judging whether a queue comprising data frames is stored in a container of the barrier vehicle, and if so, entering a step 202; otherwise, generating a new queue and inserting the newest data frame into the queue, and ending.

Step 202: judging whether the last frame data frame is a key data frame, if so, entering step 203; otherwise, step 204 is entered.

Step 203: deleting the last frame of data.

Step 204: the latest data frame is inserted into the queue of the container of the obstacle vehicle and whether the latest data frame is a key frame is marked.

In the embodiment of the present application, through the processing of step 101, the history state information of the obstacle vehicle is recorded using the state container based on the key data frame. And further, by reasonably designing the detection logic of the key data frame, the state information of the key moment is selected and recorded, and a large amount of invalid information is eliminated, so that under the limit of the same physical capacity, a more complete obstacle vehicle state sequence is recorded, and the automatic driving vehicle has a longer view field when identifying the vehicle object meeting by mistake.

In an exemplary embodiment, before step 101, the method may further include:

step 100: identifying whether the road where the self vehicle is located is a narrow road scene, and if the road is the narrow road scene, continuing to execute the step 101; otherwise, the vehicle crossing by mistake method of the application can not be adopted.

In one embodiment, a collision-free corridor-like space is generated from static obstacles in a scene in a Frenet coordinate system along the advancing direction of a road with an automatic driving vehicle (own vehicle) as a starting point; and judging whether the road where the vehicle is located is a narrow road scene or not by calculating the narrow degree of the space. Here, the Frenet coordinate system describes the position of the vehicle relative to the road, in which s represents the distance along the road, called the ordinate; d represents the displacement from the longitudinal line, called the abscissa.

In one embodiment, identifying whether the road on which the vehicle is located is a narrow road scene may include:

firstly, determining the direction (i.e. winding from the left side or from the right side) of the future obstacle of the self-driving vehicle (self-driving vehicle) according to the relative position relationship between each static obstacle in the surrounding environment of the road where the self-driving vehicle is located and the planned track of the self-driving vehicle;

dividing static obstacles with large occupied area (such as the occupied area is larger than a preset value) into a plurality of small static obstacles at equal intervals along the direction of the center line of the lane;

combining the left and right boundary positions of each static obstacle in a Frenet coordinate system and the determined detour direction to construct a collision-free boundary, so that when the self-vehicle runs in the collision-free boundary, the self-vehicle cannot collide with other static obstacles;

and determining whether the road where the self-vehicle is located is a narrow road scene according to the size of the space left in the middle of the collision-free boundary, for example, if the size of the space left in the middle of the collision-free boundary is smaller than a preset space size threshold value, determining that the road where the self-vehicle is located is the narrow road scene, otherwise, determining that the road where the self-vehicle is located is a non-narrow road scene.

Step 102: and identifying whether the obstacle vehicle is a vehicle-crossing object of the vehicle according to the recorded state information of the obstacle vehicle and the road topology structure of the vehicle.

In step 102, after obtaining the historical state information of the obstacle vehicle, it is possible to determine whether there is a wrong vehicle-crossing object in the obstacle vehicle and which obstacle vehicle is the wrong vehicle-crossing object by performing multilevel matching on the state sequence. In one embodiment, the first layer determination may include: the track distribution of the barrier vehicle in the past whole motion process is matched with a topological structure of a lane where the advancing direction of the automatic driving vehicle is located, whether the motion route of the barrier vehicle is approximately opposite to that of the automatic driving vehicle or not is judged, and the barrier vehicle which is consistent with the motion route of the barrier vehicle, namely the same as the motion route of the barrier vehicle can be excluded through judgment of a first layer; the second layer judgment may include: matching the motion track distribution of the barrier vehicle in a short period (such as a preset time length) with a topological structure of a lane where the advancing direction of the automatic driving vehicle is located, judging whether the motion route of the barrier vehicle is opposite to that of the automatic driving vehicle, and determining whether a wrong meeting object exists in the automatic driving vehicle and which barrier vehicle is a destroyed vehicle object through judgment of a second layer; the third layer of determination may include: and matching the predicted track of the obstacle vehicle with a topological structure of a lane where the advancing direction of the automatic driving vehicle is located, judging whether the future movement route of the obstacle vehicle is opposite to the movement route of the automatic driving vehicle, and finding out the obstacle vehicle which is about to become the vehicle of the self and is the wrong meeting object through the judgment of the third layer. The order of execution of the second layer judgment and the third layer judgment is not strict, and only the first layer judgment and the second layer judgment may be performed, or only the first layer judgment and the third layer judgment may be performed.

In one embodiment, the method for determining whether the barrier vehicle is the wrong vehicle-meeting object through the three-level judgment eliminates the problems of false detection and missed detection caused by dynamic and static false detection and incomplete historical information, identifies the wrong vehicle-meeting object more accurately, and ensures the safety of automatic driving better.

Fig. 3 is a schematic flowchart of a process of identifying a vehicle mis-meeting object in an embodiment of the present application, and in an embodiment, as shown in fig. 3, the process may include:

step 300: historical state information for each obstacle vehicle in the container is obtained.

Step 301: judging whether the whole movement process of the obstacle vehicle is approximately opposite to the advancing direction of the vehicle, if so, entering a step 302; otherwise, step 305 is entered.

In an embodiment, the step may specifically include:

respectively discretizing the historical motion track of the obstacle vehicle and a topological graph of a future driving lane of the automatic driving vehicle into points, wherein the discrete points respectively form two line segments;

matching points on the two line segments by a nearest neighbor matching rule, and if the motion directions of a pair of matching points are approximately opposite, remembering a positive example vote; otherwise, recording a negative example vote;

and weighting and summing the voting results, and if the summation result is greater than a preset first matching threshold value, determining that the whole motion process of the obstacle vehicle is approximately opposite to the advancing direction of the vehicle, wherein the weighting of the voting results is greater closer to the current moment.

Step 302: judging whether the short-term motion trail of the obstacle vehicle is opposite to the advancing direction of the vehicle, if so, entering a step 304; otherwise, step 303 is entered.

The specific implementation of this step may be the same as step 301, for example:

respectively discretizing the short-term motion track of the obstacle vehicle and the future driving lane topological graph of the automatic driving vehicle into points, wherein the discrete points respectively form two line segments;

and weighting and summing the voting results, and if the summation result is greater than a preset second matching threshold value, judging that the short-term motion track of the obstacle vehicle is opposite to the advancing direction of the vehicle, wherein the closer to the current moment, the greater the weight of the voting results is.

Step 303: judging whether the predicted track of the obstacle vehicle is consistent with the advancing direction of the vehicle, if not, entering step 304; otherwise, step 305 is entered.

respectively discretizing the predicted track of the obstacle vehicle and a topological graph of a future driving lane of the automatic driving vehicle into points, wherein the discrete points respectively form two line segments;

and weighting and summing the voting results, and if the summation result is greater than a preset third matching threshold value, judging that the predicted track of the obstacle vehicle is opposite to the advancing direction of the vehicle, wherein the closer to the current moment, the greater the weight of the voting result is.

Step 304: and judging that the obstacle vehicle is a vehicle mismeeting object of the vehicle, and ending.

Step 305: it is determined that the obstacle vehicle is not a wrong-meeting object of the own vehicle.

In the embodiment shown in fig. 3, on the premise that the entire movement process of the obstacle vehicle is substantially opposite to the future travel route of the autonomous vehicle, and when the movement locus of the obstacle vehicle in a short period of time does not coincide with the direction of the future travel route of the autonomous vehicle, the predicted locus of the obstacle vehicle and the direction of the future travel route of the autonomous vehicle are further determined, and the result of whether the obstacle vehicle is a wrong-meeting object is finally obtained.

In another embodiment, the execution sequence of step 302 and step 303 may be exchanged, that is, the step 303 is performed first, and in the case that it is determined that the predicted trajectory of the obstacle vehicle is not consistent with the forward direction of the vehicle, the obstacle vehicle is determined to be the vehicle-crossing-by-mistake object; otherwise, go to step 302; under the condition that the short-term motion trail of the obstacle vehicle is opposite to the advancing direction of the vehicle, determining that the obstacle vehicle is a vehicle-crossing object; when the short-term motion trajectory of the obstacle vehicle is determined to coincide with the direction of forward movement of the vehicle, it is determined that the obstacle is not a wrong-meeting object.

In still another embodiment, it may be determined whether the obstacle vehicle is a vehicle-crossing object only according to whether the short-term movement locus or predicted locus of the obstacle vehicle is opposite to the future travel route of the autonomous vehicle on the premise that the entire movement process of the obstacle vehicle is substantially opposite to the future travel route of the autonomous vehicle.

In the embodiment of the application, a multi-level state sequence matching method is used, state information of multiple time periods such as history, current and prediction is fully considered to identify the wrong vehicle-meeting object, the problems of false detection and missed detection caused by dynamic and static false detection and incomplete history information are solved, the wrong vehicle-meeting object is identified more accurately, and the safety of automatic driving is better guaranteed.

The present application further provides a computer-readable storage medium storing computer-executable instructions for performing the method of implementing a missed meeting for an autonomous vehicle according to any of fig. 1.

The application further provides a device for realizing wrong meeting of an automatic driving vehicle, which comprises a memory and a processor, wherein the memory stores the following instructions which can be executed by the processor: for performing the steps of the method of any of the automated driving vehicles of fig. 1 for achieving a missed vehicle crossing.

Fig. 4 is a schematic diagram of a key data frame of a certain obstacle vehicle in the embodiment of the present application, and taking a road segment shown in fig. 4 as an example, according to the method for recording state information of an obstacle vehicle entering a preset range of the vehicle in the embodiment of the present application based on the key data frame, in the embodiment shown in fig. 4, only 6 frames of state information need to be recorded, so that the whole motion process of the obstacle vehicle from entering a view (i.e., entering the preset range of the vehicle) to the current time can be recorded relatively completely. Through the 6-frame state information in the present embodiment, it can be analyzed to obtain: 1. the barrier vehicle enters the visual field at a higher speed and then gradually stops; 2. the overall motion trajectory radian of the obstacle vehicle is small, wherein the front section trajectory is very stable (the radian is small), and the rear section trajectory angle orientation is greatly changed (the radian is large).

The historical state information of the obstacle vehicle is stored by using the key frame container, and the more complete obstacle vehicle motion state is recorded with less capacity;

and judging whether the vehicle is a reverse wrong meeting object according to the track distribution of a plurality of layers in the past, the present and the future, and identifying the wrong meeting object earlier and more accurately.

Fig. 5 is a schematic structural diagram of a device for automatically driving a vehicle to implement a vehicle meeting by mistake in the embodiment of the present application, as shown in fig. 5, the device at least includes: the device comprises a recording module and a first identification module; wherein the content of the first and second substances,

In an exemplary embodiment, the method may further include: and the second identification module is used for identifying the narrow road scene of the road where the self-vehicle is located.

In an exemplary embodiment, the recording module may be specifically configured to:

acquiring and recording the state information of the barrier vehicle entering the preset range of the self vehicle according to a preset acquisition period; the collected status information is processed to delete non-critical data frames. Wherein the first frame data frame of the obstacle vehicle is a key data frame.

In an exemplary instance, the recording module can be further configured to:

if the number of key frames cumulatively reaches the upper threshold of the container, then the key data frame with the lowest freshness is deleted according to the freshness. The freshness refers to a time span between the data generation time and the current time, and the larger the span is, the lower the freshness is.

In one illustrative example, the collecting of the state information of the obstacle vehicle in the recording module may include:

In an exemplary example, the first identification module may be specifically configured to:

acquiring historical state information of each obstacle vehicle in the container; determining that the whole movement process of the obstacle vehicle is approximately opposite to the advancing direction of the vehicle; and judging that the short-term motion trail of the obstacle vehicle is opposite to the advancing direction of the vehicle, and judging that the obstacle vehicle is a wrong meeting object of the vehicle.

acquiring historical state information of each obstacle vehicle in the container; determining that the whole movement process of the obstacle vehicle is approximately opposite to the advancing direction of the vehicle; and judging that the predicted track of the obstacle vehicle is inconsistent with the advancing direction of the vehicle, and judging that the obstacle vehicle is a wrong meeting object of the vehicle.

acquiring historical state information of each obstacle vehicle in the container; determining that the whole movement process of the obstacle vehicle is approximately opposite to the advancing direction of the vehicle; judging that the short-term motion trail of the obstacle vehicle is not opposite to the advancing direction of the vehicle; and judging that the predicted track of the obstacle vehicle is inconsistent with the advancing direction of the vehicle, and judging that the obstacle vehicle is a wrong meeting object of the vehicle.

In an exemplary instance, the second identification module may be specifically configured to:

determining a detour direction according to the position relation between the position of each static obstacle in the surrounding environment of the road where the automatic driving vehicle (the self vehicle) is located and the planned track of the self vehicle; cutting a large static obstacle perpendicular to the center line of the lane to divide the large static obstacle into a plurality of small static obstacles; combining the left boundary position of each static obstacle in a Frenet coordinate system and the determined detour direction to construct a collision-free boundary, and when the self-vehicle runs in the collision-free boundary, the self-vehicle cannot collide with other static obstacles; and determining whether the road where the self-vehicle is located is a narrow road scene according to the size of the space left in the middle of the collision-free boundary, for example, if the size of the space left in the middle of the collision-free boundary is smaller than a preset space size threshold value, determining that the road where the self-vehicle is located is the narrow road scene, otherwise, determining that the road where the self-vehicle is located is a non-narrow road scene.

According to the device for realizing the vehicle crossing by the automatic driving vehicle, the historical state information of the obstacle vehicle is recorded by using the state container based on the key data frame. And further, by reasonably designing the detection logic of the key data frame, the state information of the key moment is selected and recorded, and a large amount of invalid information is eliminated, so that under the limitation of the same physical capacity, a more complete obstacle vehicle state sequence is recorded, so that the automatic driving vehicle has a longer view when identifying the vehicle object meeting by mistake, the guarantee is provided for better identifying the vehicle object meeting by mistake, and the safety of automatic driving is also guaranteed.

In an embodiment, the device for automatically driving the vehicle to realize the vehicle meeting by mistake determines whether the barrier vehicle is the vehicle meeting by mistake object or not through the three-level judgment, so that the problems of false detection and missed detection caused by dynamic and static false detection and incomplete historical information are solved, the vehicle meeting by mistake object is accurately identified, and the safety of automatic driving is better guaranteed.

Although the embodiments disclosed in the present application are described above, the descriptions are only for the convenience of understanding the present application, and are not intended to limit the present application. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the disclosure as defined by the appended claims.

Claims

1. A method for automatically driving a vehicle to realize a vehicle meeting by mistake, comprising the following steps:

2. The method of claim 1, before recording the state information of the obstacle vehicle entering the preset range of the host vehicle, further comprising:

3. The method according to claim 1 or 2, wherein the recording of the state information of the obstacle vehicle entering within a preset range of the own vehicle comprises:

4. The method of claim 3, wherein the processing the collected state information to remove non-key data frames comprises:

5. The method of claim 3, further comprising:

6. The method of claim 3, wherein the collecting of the state information of the obstacle vehicle entering the preset range of the own vehicle comprises:

7. The method of claim 6, further comprising:

8. The method of claim 6, further comprising:

9. The method of claim 1 or 2, wherein the identifying whether the obstacle vehicle is a missed vehicle crossing object of the own vehicle comprises:

10. A computer-readable storage medium storing computer-executable instructions for performing the method of implementing a missed meeting for an autonomous vehicle as claimed in any one of claims 1 to 9.

11. An apparatus for automatically driving a vehicle to effect a missed vehicle crossing, comprising a memory and a processor, wherein the memory has stored therein the following instructions executable by the processor: the steps for performing the method of any one of claims 1 to 9 for achieving a missed vehicle crossing.

12. An apparatus for automatically driving a vehicle to make a vehicle cross meeting, comprising: the device comprises a recording module and a first identification module; wherein the content of the first and second substances,

13. The apparatus of claim 12, further comprising: and the second identification module is used for identifying the narrow road scene of the road where the self-vehicle is located.

14. The apparatus of claim 12 or 13, wherein the recording module is to:

15. The apparatus of claim 14, the recording module further to:

16. The apparatus of claim 12 or 13, wherein the first identifying means is configured to: