Vehicle obstacle avoidance method and device, and computer‑readable storage medium

Abstract

A vehicle obstacle avoidance method and device, and a computer‑readable storage medium. The vehicle obstacle avoidance method comprises: performing collision detection between a predicted motion trajectory of an obstacle in a current driving environment of an ego vehicle and a predicted driving trajectory of the ego vehicle, wherein the obstacle moves towards the predicted driving trajectory of the ego vehicle (S100); when there is a risk of collision between the obstacle and the ego vehicle, creating a safe passage corridor for the obstacle to cross a road on which the ego vehicle is currently driving (S200), wherein the safe passage corridor is planned for the predicted motion trajectory of the obstacle the intention of which is difficult to predict, has a certain degree of inclusiveness, and can solve the problems of uncertainty of the intention of the obstacle and limited sensing capabilities of sensors; and then adding the safe passage corridor to longitudinal planning of the ego vehicle, and on the basis of the predicted driving trajectory, generating a decision trajectory for avoiding the obstacle, so as to control the ego vehicle to perform obstacle‑avoidance driving (S300). Thus, the obstacle avoidance effect is improved, and safe passage of the obstacle and safe driving of the ego vehicle are ensured.

Description

Vehicle obstacle avoidance method, device and computer readable storage medium TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, and in particular to a vehicle obstacle avoidance method, device and computer readable storage medium. BACKGROUND

[0002] During the driving process of a vehicle, the situation of a pedestrian crossing the road often occurs, which puts high requirements on the avoidance ability of the vehicle.

[0003] At present, the avoidance method based on sensors and algorithms is mainly used for the vehicle to avoid the pedestrian crossing the road. These sensors can detect the surrounding environment in real time, including the position, speed, moving direction and other information of the pedestrian. Then, the system uses advanced computer vision and machine learning algorithms to detect and track the perceived pedestrian, and predict its future trajectory and intention.

[0004] However, the behavior of the pedestrian is often affected by various factors, such as personal habits, psychological state, environmental changes, etc., making the behavior of the pedestrian complex and uncertain. These factors make the intention of the pedestrian difficult to predict. In addition, the limitation of current environmental perception is also a difficulty in the avoidance method for the pedestrian crossing the road. Although sensor technology has made significant progress, in some complex environmental conditions, such as insufficient light and bad weather, the perception ability of the sensor will still be limited, resulting in the system being unable to accurately perceive the position and speed of the pedestrian and other information.

[0005] In summary, the current vehicle avoidance technology is not good in some cases due to the above problems.

[0006] SUMMARY

[0007] Therefore, one of the purposes of the present application is to provide a vehicle avoidance method, device and computer readable storage medium to solve at least part of the problems in the related art.

[0008] In a first aspect, the vehicle avoidance method provided by the embodiments of the present application comprises:

[0009] Collision detection is performed on the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle; the obstacle moves towards the direction of the predicted driving trajectory of the ego vehicle;

[0010] In the case that there is a collision risk between the obstacle and the ego vehicle, a safe passing corridor is created for the obstacle to pass through the road currently driven by the ego vehicle;

[0011] Add the safe passage corridor to the longitudinal planning of the ego vehicle, generate a decision trajectory for avoiding the obstacle based on the predicted trajectory, to control the ego vehicle to avoid the obstacle.

[0012] In an optional embodiment, the collision detection on the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted trajectory of the ego vehicle comprises:

[0013] Obtain the predicted motion trajectory of the obstacle generated by the motion data of the obstacle in the current driving environment of the ego vehicle and the predicted trajectory obtained by the driving data of the ego vehicle;

[0014] The collision detection on the predicted motion trajectory of the obstacle and the predicted trajectory of the ego vehicle is performed in space-time, and a collision detection result is obtained.

[0015] Based on the motion data and driving data of continuous multiple frames, the collision detection result is verified.

[0016] After the collision verification passes, a target collision point when the obstacle and the ego vehicle collide is generated.

[0017] In an optional embodiment, the collision detection result includes an initial collision point when the obstacle and the ego vehicle collide,

[0018] The collision detection result is verified based on the motion data and driving data of continuous multiple frames, comprising:

[0019] After the preliminary collision detection, the motion data of the obstacle and the driving data of the ego vehicle of N frames are continuously obtained;

[0020] Based on the motion data of the same frame in the continuous N frames, a new predicted motion trajectory is generated and a new predicted trajectory is generated based on the driving data;

[0021] The new predicted motion trajectory and the predicted trajectory are used for collision detection in space-time, and the collision detection result is verified.

[0022] According to the verification result, the initial collision point is updated.

[0023] In an optional embodiment, the initial collision point is updated according to the verification result, comprising:

[0024] In the case of obtaining the verification result representing the collision between the obstacle and the ego vehicle, the longitudinal position change value between the new collision point in the verification result and the initial collision point is determined, and the longitudinal direction is the driving direction of the ego vehicle.

[0025] In a case where the longitudinal position change value is greater than a set threshold value, updating the initial collision point based on the new collision point;

[0026] The generating of the collision point at which the obstacle collides with the ego vehicle comprises:

[0027] The new collision point is determined as a target collision point at which the obstacle collides with the ego vehicle.

[0028] In an optional embodiment, the updating of the initial collision point according to the verification result comprises:

[0029] In a case where the verification result indicates that the obstacle does not collide with the ego vehicle, the preliminary collision point is deleted.

[0030] In an optional embodiment, the updating of the initial collision point according to the verification result comprises:

[0031] Intention detection is performed on the obstacle;

[0032] In a case where the verification result indicates that the obstacle does not have an intention to cross the road currently traveled by the ego vehicle, the preliminary collision point is deleted.

[0033] In a case where the obstacle has a collision risk with the ego vehicle, a safe passing corridor for the obstacle to pass through the road currently traveled by the ego vehicle is created, comprising:

[0034] If the target collision point exists, it is determined that the obstacle has a collision risk with the ego vehicle;

[0035] In a case where the obstacle has a collision risk with the ego vehicle, a target collision point with a minimum longitudinal distance from the ego vehicle is taken as a base point;

[0036] Based on a preset boundary constraint, the safe passing corridor for a target obstacle corresponding to the target collision point with the minimum longitudinal distance is created.

[0037] In an optional embodiment, the boundary constraint comprises a longitudinal boundary constraint in the same direction as the travel direction of the ego vehicle and a transverse boundary constraint perpendicular to the travel direction of the ego vehicle, and the method further comprises:

[0038] According to the base point and a preset safe walking distance, a longitudinal boundary constraint of the safe passing corridor is obtained;

[0039] According to the width of the road currently traveled by the ego vehicle, a transverse boundary constraint of the safe passing corridor is obtained.

[0040] In an optional embodiment, the safe walking distance is determined according to an error range between the predicted motion trajectory of the obstacle and an actual motion trajectory of the obstacle, and the safe walking distance is in positive correlation with the error range.

[0041] In an optional embodiment, after the safe passing corridor for the target obstacle passing corresponding to the target collision point with the minimum longitudinal distance is created,

[0042] The adding of the safe passing corridor into the longitudinal planning of the ego vehicle generates a decision trajectory for avoiding the obstacle, so as to control the ego vehicle to avoid the obstacle.

[0043] The adding of the safe passing corridor into the longitudinal planning of the ego vehicle generates a decision trajectory for avoiding the target obstacle, so as to control the ego vehicle to avoid the target obstacle.

[0044] In an optional embodiment, the adding of the safe passing corridor into the longitudinal planning of the ego vehicle generates a decision trajectory for avoiding the obstacle, so as to control the ego vehicle to avoid the obstacle, including:

[0045] The safe passing corridor is projected into an ST graph of the longitudinal planning, and a decision trajectory for avoiding the obstacle is planned, so as to control the ego vehicle to avoid the obstacle.

[0046] In an optional embodiment, the control of the ego vehicle to avoid the obstacle includes:

[0047] The predicted motion trajectory of the obstacle is updated based on a preset detection frequency;

[0048] In a case where a change between the updated predicted motion trajectory and the current predicted motion trajectory does not exceed a set range, the decision trajectory is reused to control the ego vehicle to avoid the obstacle based on the decision trajectory.

[0049] In an optional embodiment, after the predicted motion trajectory of the obstacle is updated based on the preset detection frequency, the method further includes:

[0050] In a case where a change between the updated predicted motion trajectory and the current predicted motion trajectory exceeds the set range, a collision detection is performed on a predicted motion trajectory of an obstacle in a current driving environment of the ego vehicle and a predicted driving trajectory of the ego vehicle, so as to update the decision trajectory.

[0051] control the ego vehicle to perform an evasive driving based on the updated decision trajectory.

[0052] In an optional embodiment, the updating the decision trajectory comprises:

[0053] performing collision detection between a predicted motion trajectory of the obstacle in a current driving environment of the ego vehicle and a predicted driving trajectory of the ego vehicle;

[0054] deleting a safe passing corridor for the obstacle to pass a road currently driven by the ego vehicle in a case where there is no collision risk between the obstacle and the ego vehicle;

[0055] updating the decision trajectory.

[0056] In an optional embodiment, before the performing collision detection between a predicted motion trajectory of the obstacle in a current driving environment of the ego vehicle and a predicted driving trajectory of the ego vehicle, the method further comprises:

[0057] detecting an object in the current driving environment of the ego vehicle;

[0058] regarding the object satisfying a crossing condition as the obstacle having an intention of crossing a road currently driven by the ego vehicle;

[0059] wherein the crossing condition comprises all of the following:

[0060] the object is in a non-stationary state;

[0061] an included angle is between a motion direction of the object and a driving direction of the ego vehicle;

[0062] a predicted motion trajectory of the object and a predicted driving trajectory of the ego vehicle have an intersection point.

[0063] In an optional embodiment, the detecting the obstacle in the current driving environment of the ego vehicle comprises:

[0064] detecting the obstacle in the current driving environment of the ego vehicle within a preset screening window range.

[0065] In an optional embodiment, the regarding the object satisfying a crossing condition as the obstacle having an intention of crossing a road currently driven by the ego vehicle comprises:

[0066] determining the object satisfying the crossing condition;

[0067] performing multi-frame data verification on the object satisfying the crossing condition;

[0068] The object satisfying the crossing condition for continuous times is regarded as a crossing obstacle with the intention of crossing the current road of the ego vehicle.

[0069] In a second aspect, the vehicle obstacle avoidance method provided by the embodiments of the present application comprises the following steps:

[0070] In the case that there is a collision risk between the obstacle and the ego vehicle, a set stop point and / or a stop line between the target collision point and the ego vehicle;

[0071] The stop point and / or the stop line are added to the longitudinal planning of the ego vehicle to generate a decision trajectory for avoiding the obstacle, so as to control the ego vehicle to avoid the obstacle.

[0072] In a third aspect, the embodiments of the present application provide a storage medium for storing executable codes, the executable codes being used to execute the vehicle obstacle avoidance method provided by the first aspect of the embodiments of the present application when running.

[0073] In a fourth aspect, the embodiments of the present application provide a storage medium for storing executable codes, the executable codes being used to execute the vehicle obstacle avoidance method provided by the second aspect of the embodiments of the present application when running.

[0074] In a fifth aspect, the embodiments of the present application provide an application program for executing the vehicle obstacle avoidance method provided by the first aspect of the embodiments of the present application when running.

[0075] In a sixth aspect, the embodiments of the present application provide an application program for executing the vehicle obstacle avoidance method provided by the second aspect of the embodiments of the present application when running.

[0076] In a seventh aspect, the embodiments of the present application provide a vehicle obstacle avoidance device, comprising a processor, a memory, a communication interface and a bus.

[0077] The processor, the memory and the communication interface are connected through the bus and complete communication with each other.

[0078] The memory stores executable program codes.

[0079] The processor runs the program corresponding to the executable program codes by reading the executable program codes stored in the memory, so as to execute the vehicle obstacle avoidance method provided by the first aspect of the embodiments of the present application.

[0080] In an eighth aspect, the embodiments of the present application provide a passenger flow information processing device, comprising a processor, a memory, a communication interface and a bus.

[0081] The processor, the memory and the communication interface are connected through the bus and complete communication with each other;

[0082] The memory stores executable program code;

[0083] The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, for executing the passenger flow information processing method provided in the second aspect of the embodiments of the present application. According to another aspect of the embodiments of the present application, an automobile is provided. The automobile comprises the vehicle obstacle avoidance as described in each of the above embodiments.

[0084] The technical solutions provided by the embodiments of the present application can include the following beneficial effects:

[0085] In the first aspect, the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle are subjected to collision detection, and in the case that there is a collision risk between the obstacle and the ego vehicle, a safe passing corridor for the obstacle to pass through the road currently driven by the ego vehicle is created, the safe passing corridor is planned for the predicted motion trajectory of the obstacle whose intention is difficult to predict, and has a certain inclusiveness, which can solve the problems of uncertain intention of the obstacle and limited sensor perception capability.

[0086] In the second aspect, by adding the safe passing corridor to the longitudinal planning of the ego vehicle, the longitudinal planning plans a trajectory that does not cross the safe passing corridor, and therefore a decision trajectory that avoids the obstacle is generated based on the predicted driving trajectory, to control the ego vehicle to avoid driving relative to the obstacle. This process plans a safe and reliable trajectory for avoiding the obstacle for the predicted motion trajectory of the uncertain obstacle, to guarantee the safe passing of the obstacle and the safe driving of the ego vehicle, and the avoidance effect is good.

[0087] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0088] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0089] FIG. 1 is a flow diagram of a vehicle obstacle avoidance method according to an embodiment of the present application.

[0090] FIG. 2 is a schematic diagram of the motion relationship between the ego vehicle and the pedestrian in a vehicle obstacle avoidance method according to an embodiment of the present application.

[0091] FIG. 3 is a schematic diagram of the motion relationship between the ego vehicle and the pedestrian in a vehicle obstacle avoidance method according to an embodiment of the present application.

[0092] FIG. 4 is a schematic diagram of the process of recursively predicting the trajectory of the ego vehicle in a vehicle obstacle avoidance method according to an embodiment of the present application.

[0093] FIG. 5 is a schematic diagram of the process of multi-frame space-time collision detection in a vehicle obstacle avoidance method according to an embodiment of the present application.

[0094] FIG. 6 is a schematic diagram of the process of updating the collision point in a vehicle obstacle avoidance method according to an embodiment of the present application.

[0095] FIG. 7 is another schematic diagram of the process of updating the collision point in a vehicle obstacle avoidance method according to an embodiment of the present application.

[0096] FIG. 8 is a schematic diagram of the process of creating a safe passing corridor in a vehicle obstacle avoidance method according to an embodiment of the present application.

[0097] FIG. 9 is a schematic diagram of the process of longitudinal planning in a vehicle obstacle avoidance method according to an embodiment of the present application.

[0098] FIG. 10 is a block diagram of a vehicle obstacle avoidance device according to an embodiment of the present application.

[0099] FIG. 11 is a schematic block diagram of a vehicle obstacle avoidance device according to an embodiment of the present application. DETAILED DESCRIPTION

[0100] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0101] The exemplary embodiments will be described in detail herein below with reference to the drawings. The following description is only one of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0102] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting. As used in this application and the appended claims, the singular forms "a," "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items. Unless otherwise noted, "front," "rear," "lower," and / or "upper" and like terms are used for convenience only and do not limit the location or spatial orientation of the item described. "Connected" or "coupled" or like terms are not limited to a direct connection or physical coupling, but also include indirect connections or couplings, such as electrical connections or couplings. In this application, "may" can mean "possibility."

[0103] Various embodiments of the application will be described in detail below with reference to the drawings. The features of the embodiments and implementation described below can be combined with each other as long as there is no conflict.

[0104] During vehicle driving, the situation of crossing pedestrians often occurs, which puts high requirements on the vehicle's obstacle avoidance capability, especially for autonomous vehicles, whose safety performance has become the focus of the public. Therefore, developing an efficient and accurate crossing pedestrian avoidance system is of great significance to improve the safety performance of autonomous vehicles. At present, autonomous vehicles mainly use sensor-based and algorithm-based avoidance methods to avoid crossing pedestrians. These sensors can detect the surrounding environment in real time, including the position, speed, and moving direction of pedestrians. Then, the system uses advanced computer vision and machine learning algorithms to detect and track the perceived pedestrians, predict their future trajectory and intent. However, due to the complexity and uncertainty of pedestrian behavior, existing technologies still cannot accurately predict the intent of pedestrians in some cases, resulting in poor avoidance effect.

[0105] The present application aims to solve the problems of current technology in avoiding crossing pedestrians for autonomous vehicles, and improve the avoidance effect and safety of autonomous vehicles by introducing new planning methods and roughness.

[0106] The vehicle obstacle avoidance method described herein can be applied to all vehicles on the market (such as electric vehicles, hybrid vehicles, autonomous vehicles, etc.), which have the conventional functions of vehicles on the market, such as environmental perception, speed control, etc. The following takes an autonomous vehicle as an example for illustration, and the specific implementation of other vehicles is basically similar, which will not be introduced here.

[0107] The present application provides a first embodiment of a vehicle obstacle avoidance method, as shown in FIG. 1, which is a flowchart of a vehicle obstacle avoidance method according to an embodiment. The method can include the following steps:

[0108] In step 100, collision detection is performed on the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle; the obstacle moves towards the direction of the predicted driving trajectory of the ego vehicle.

[0109] The ego vehicle described herein refers to the automatic driving vehicle itself, which is the main object controlled and managed by the entire automatic driving system. In making various decisions and plans, the current state of the ego vehicle (such as position, speed, driving direction, etc.) is an important reference. The ego vehicle perceives the surrounding environment through various sensors, processes and analyzes information, and makes corresponding driving decisions to achieve safe and efficient driving goals.

[0110] The obstacle described herein refers to an object that moves towards the direction of the predicted driving trajectory of the ego vehicle during the driving process of the ego vehicle, such as a pedestrian, a pedestrian in a passenger vehicle, a moving vehicle, etc. The motion of these objects on the road has uncertainty, and they may suddenly change the driving direction and cross the road currently driven by the ego vehicle, which will affect the driving safety of the ego vehicle and need to be detected, identified and processed by the automatic driving system to take corresponding avoidance measures. The following takes the pedestrian as an example to illustrate the specific embodiments, and other forms are basically the same, which will not be repeated here.

[0111] The crossing described herein refers to the behavior of crossing the road in the driving direction of the ego vehicle at a certain angle, which can be close to a vertical angle or can be inclined to cross, as long as the motion trajectory of the pedestrian or other objects intersects with the driving direction of the ego vehicle, it is considered as an obstacle crossing the road, i.e. a crossing pedestrian. The crossing behavior of the pedestrian and whether it is identified as an obstacle will be described in detail later.

[0112] When the pedestrian is identified as an obstacle with crossing intention, it means that there may be a collision risk between the pedestrian and the ego vehicle, therefore, collision detection needs to be performed on the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle to determine the position of the collision, so as to take certain avoidance measures based on the position subsequently.

[0113] In step 200, in the case where there is a collision risk between the obstacle and the ego vehicle, a safe passage corridor is created for the obstacle to pass through the road currently driven by the ego vehicle.

[0114] After the collision detection between the obstacle and the ego vehicle, if it is determined that there is a collision risk between the obstacle and the ego vehicle, it means that the obstacle hinders the normal driving of the ego vehicle based on the current state. In order to ensure the safe passage of the crossing pedestrian and the safe passage of the ego vehicle, a safe passage corridor for the crossing pedestrian to cross the road where the ego vehicle is located needs to be created. The safe passage corridor can solve the problem of decision error caused by unstable sensor data in harsh environment and unclear prediction intention of the pedestrian. It can be understood that, compared with the traditional trajectory directly predicted by the pedestrian, the safe passage corridor does not excessively rely on the prediction data of the pedestrian, has a certain inclusiveness, and can better cope with actual scenarios such as sudden changes in the intention of the pedestrian.

[0115] In step 300, the safe passage corridor is added to the longitudinal planning of the ego vehicle, and a decision trajectory for avoiding the obstacle is generated based on the predicted driving trajectory to control the ego vehicle to avoid the obstacle.

[0116] The longitudinal planning described herein refers to the planning in the driving direction of an autonomous vehicle, which involves speed control, distance control, and time planning of the ego vehicle in the forward direction. For example, determining the speed change of the vehicle at different time points, the safe distance from the front obstacle, etc. The purpose of longitudinal planning is to ensure that the ego vehicle safely and efficiently advances during driving, while considering various possible situations and constraints.

[0117] Adding the safe passage corridor as a constraint condition to the longitudinal planning of the ego vehicle will plan a decision trajectory that ensures the ego vehicle will not cross the safe passage corridor. The vehicle system controls the ego vehicle to drive based on the decision trajectory to avoid collision with the crossing pedestrian, and realizes efficient and accurate avoidance of the crossing pedestrian.

[0118] In this embodiment, a safe and reliable decision trajectory for avoiding the pedestrian is planned for the prediction motion trajectory of the crossing pedestrian that is unreliable or uncertain, to improve the safety and stability of the autonomous vehicle.

[0119] The present application is mainly applied to avoiding pedestrians in autonomous driving. The vehicle obstacle avoidance method involves determining a crossing pedestrian (i.e. an obstacle), creating / deleting a safe passage corridor for the crossing pedestrian to pass, and adding the safe passage corridor to the longitudinal planning of the ego vehicle. These three parts are closely integrated to form a complete set of efficient and accurate pedestrian avoidance solutions, which is of great significance to improve the safety performance of autonomous vehicles.

[0120] In some embodiments, before the collision detection between the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle, the obstacle in the driving process of the vehicle is determined. The following steps are included:

[0121] Step 401, detecting an object in a current driving environment of the ego vehicle.

[0122] The object with a moving attribute other than the ego vehicle is detected during driving of the ego vehicle, including a pedestrian, another vehicle, and the like in a road environment that can affect driving of the ego vehicle, to determine whether the object has an intention to cross the road, for subsequent avoidance of the object crossing the road on which the ego vehicle is driving.

[0123] Since the vehicle has a relatively clear driving specification or indicated route during driving, its driving trajectory is easier to predict than that of a pedestrian, and its driving intention has a certain certainty, therefore, in the vehicle obstacle avoidance process, the case where the object is a pedestrian is described in detail, and the case where the object is a vehicle is basically the same as the embodiments, which will not be described here.

[0124] In other embodiments, a screening window is provided, and objects in a driving environment within the screening window are detected to adjust the detection range and improve detection efficiency.

[0125] Specifically, the obstacle in the current driving environment of the ego vehicle is detected, including the following steps:

[0126] Step 4011, detecting an obstacle in a preset screening window range in the current driving environment of the ego vehicle.

[0127] The size of the screening window in the driving direction of the ego vehicle described herein can be set according to actual detection requirements, which can be determined based on the distance when the vehicle takes emergency braking, or can be determined according to the theoretical value when the obstacle avoidance test is performed. The size of the screening window in the direction perpendicular to the ego vehicle can be set according to the width of the road in that direction, or can be set according to the width of the road occupied by the ego vehicle during driving, which is not specifically limited here. It should be noted that the range of the screening window covers the area occupied by the predicted driving trajectory of the ego vehicle during subsequent collision detection, thereby ensuring safe driving of the ego vehicle within the screening window range.

[0128] In this embodiment, by setting the range of the screening window, unnecessary consumption of computing resources is reduced, the detection efficiency is improved, and the area that may have a risk or the area that has a higher risk degree from the ego vehicle can be focused on more quickly.

[0129] Step 402, regarding the object satisfying the crossing condition as an obstacle having an intention to cross the road on which the ego vehicle is currently driving.

[0130] Determining whether the object is a crossing pedestrian crossing the road on which the ego vehicle is driving needs to satisfy the following crossing conditions, which include all of the following:

[0131] Condition one: the object is in a non-stationary state, that is, the pedestrian is a dynamic pedestrian.

[0132] Pedestrians are divided into static pedestrians and dynamic pedestrians, where static means a state in which the position is basically unchanged, and dynamic means a pedestrian whose position changes significantly within the detection range. Since static pedestrians do not cross at the current time, they are not considered as obstacles, while dynamic pedestrians are in a situation where they may cross the road at any time. Therefore, it is necessary to determine dynamic pedestrians based on pedestrian information (such as pedestrian motion data) as the basis for subsequent obstacle determination.

[0133] In some embodiments, in order to improve the accuracy of the state determination of the object, the object is detected in multiple consecutive frames to determine its state. It can be understood that in multiple consecutive frames, the position of the pedestrian changes very little or remains basically unchanged, and the pedestrian is a static pedestrian. Similarly, in multiple consecutive frames, the pedestrian can be observed to have a certain displacement, speed, and change in motion direction, indicating that the pedestrian is in a motion state or a non-stationary state, and the pedestrian is a static pedestrian.

[0134] Condition two: the object has an included angle between the motion direction and the driving direction of the ego vehicle.

[0135] Referring to FIG. 2, FIG. 2 is a schematic diagram of the motion relationship between the ego vehicle and the pedestrian in a vehicle obstacle avoidance method according to an embodiment. When the included angle θ between the motion direction of the pedestrian and the driving direction of the ego vehicle is within a certain range (α < θ < β), it is determined that the pedestrian has a crossing motive, which can be crossing perpendicular to the direction of the road on which the ego vehicle is driving, or can be crossing obliquely across the road. It can be understood that when the motion direction of the pedestrian is consistent or opposite to the driving direction of the ego vehicle, or the motion direction of the pedestrian deviates from the road on which the ego vehicle is located, there is no collision risk between the two, and therefore α = 0° and β = 180° can be set, and the pedestrians corresponding to the included angles within this range are determined to have a crossing motive or intention.

[0136] Condition three: the predicted motion trajectory of the object intersects with the predicted driving trajectory of the ego vehicle.

[0137] The motion trajectory of the pedestrian in a preset time is predicted according to the current motion data of the pedestrian, such as the displacement, speed and motion direction of the pedestrian, and so on, to obtain the predicted motion trajectory of the pedestrian. Similarly, the driving trajectory of the ego vehicle in a preset time is predicted according to the current driving data of the ego vehicle, such as the displacement, speed and motion direction of the ego vehicle, and so on, to obtain the predicted driving trajectory of the ego vehicle. The motion trajectory of the ego vehicle can be obtained by recursion, so that the prediction result is quickly obtained, the real-time requirement of the automatic driving system is met, and timely decision and adjustment can be made. Meanwhile, the recursion can be combined with the current speed, acceleration and other dynamic information of the vehicle, so that various changes in the driving process can be better adapted to, and the prediction accuracy is improved.

[0138] It should be noted that, when judging whether the trajectories of the ego vehicle and the pedestrian intersect, the preset time of the ego vehicle can be the same as or different from the preset time of the pedestrian. In the setting, the predicted motion trajectory of the pedestrian and the predicted driving trajectory of the ego vehicle obtained based on the respective preset time are required to exist completely within the screening window range. It can be understood that the end point of the predicted motion trajectory / predicted driving trajectory does not exist within the screening window range, taking the pedestrian / ego vehicle as the starting point. For example, if the predicted motion trajectory of the pedestrian and the predicted driving trajectory of the ego vehicle actually intersect at the 8th second, and the intersection point is within the screening window range, but when the detection is performed, the preset time is set to 6 seconds, the predicted motion trajectory of the pedestrian in the future 6 seconds and the predicted driving trajectory of the ego vehicle in the future 6 seconds are predicted, and the result obtained is that the two trajectories do not intersect, which will result in an incorrect detection result, affecting the subsequent avoidance measures for the crossing pedestrian within the screening window range, and there is a safety hazard. At present, in another embodiment, to solve the problem, the frequency of trajectory prediction of the pedestrian or the ego vehicle can be adjusted, and the frequency is less than the preset time. In the case that the pedestrian does not change the intention, the intersection between the predicted motion trajectory of the pedestrian and the predicted driving trajectory of the ego vehicle can be reflected through the data of the next frame.

[0139] In the present embodiment, whether the predicted motion trajectory of the pedestrian and the predicted driving trajectory of the ego vehicle intersect within the screening window range is detected. If the predicted motion trajectory of the pedestrian and the predicted driving trajectory of the ego vehicle intersect, it is determined that the crossing intention of the pedestrian corresponding to the intersecting predicted motion trajectory may cause the risk of collision with the ego vehicle, and the pedestrian is determined as an obstacle (or a crossing pedestrian).

[0140] It is worth noting that the ego vehicle has a certain physical size, such as the width of the vehicle. Therefore, when performing intersection detection, whether the predicted motion trajectory of the pedestrian and the area occupied by the predicted driving trajectory of the ego vehicle intersect, it can be understood that the extension line of the vehicle body on the side of the ego vehicle close to the pedestrian and the extension line of the vehicle body on the side of the ego vehicle away from the pedestrian form the area occupied by the predicted driving trajectory of the ego vehicle. The pedestrians on both sides of the ego vehicle are detected simultaneously. When the predicted motion trajectory of the pedestrian intersects with any of the two extension lines, it is determined that the pedestrian has the motivation to cross the road and may have a collision risk. The final determination of whether there is a collision risk is described in detail in the subsequent chapters.

[0141] Referring to FIG. 3, FIG. 3 is a motion schematic diagram of the intersection of the trajectories of the ego vehicle and the pedestrian in a vehicle obstacle avoidance method according to an embodiment. According to the detection of each pedestrian in the screening window range, it is determined that the 6th pedestrian and the 3rd pedestrian intersect with the area occupied by the predicted driving trajectory of the ego vehicle, and it is determined that the 6th pedestrian and the 3rd pedestrian are obstacles crossing the road where the ego vehicle is located, that is, crossing pedestrians.

[0142] In this embodiment, the pedestrian who meets the above three conditions at the same time is determined to be a crossing pedestrian. Subsequently, the ego vehicle is longitudinally planned to obtain a decision trajectory for avoiding the crossing pedestrian, and corresponding avoidance measures are taken to ensure the safety of the ego vehicle and the pedestrian.

[0143] In other embodiments, although the above three conditions have met the determination conditions of the crossing pedestrian, the predicted motion trajectory of the pedestrian may appear a jump phenomenon, that is, there is a large and sudden difference between the actual motion trajectory of the pedestrian and the predicted motion trajectory of the predicted pedestrian. This may be due to the sudden behavior change of the pedestrian, the error of the sensor, the environmental disturbance, etc., resulting in discontinuous or unreasonable changes in the predicted motion trajectory of the predicted pedestrian at certain times. On the other hand, the motion of the pedestrian is uncertain, and a single frame of prediction may be inaccurate due to various factors such as sensor noise, instantaneous abnormal behavior disturbance, etc. These situations may lead to misjudgment of whether the pedestrian is truly crossing and whether there is a collision risk. Therefore, in order to improve the prediction accuracy of the motion intention of the pedestrian and the determination accuracy of the crossing pedestrian, based on the above scheme for determining the crossing pedestrian, the trajectory of the pedestrian is further optimized by multi-frame verification.

[0144] Specifically, the object satisfying the crossing condition is taken as an obstacle having the intention to cross the road currently driven by the ego vehicle, including the following steps:

[0145] Step 4021, determining an object satisfying the crossing condition;

[0146] Step 4022, multi-frame data verification is performed on the object satisfying the crossing condition.

[0147] Step 4023, the object satisfying the crossing condition for continuous times is regarded as a crossing obstacle with the intention of crossing the road currently traveled by the ego vehicle.

[0148] The continuous multi-frame data of the pedestrian satisfying the above crossing condition is obtained, the multi-frame data is verified according to the above crossing condition, and if the predicted motion trajectory of the pedestrian exists intersection with the predicted travel trajectory of the ego vehicle for continuous times, the pedestrian is finally determined as a crossing pedestrian with the intention of crossing the road currently traveled by the ego vehicle.

[0149] In the embodiment, by multi-frame data verification, observing the intersection of the pedestrian trajectory at continuous time points and the decision trajectory of the ego vehicle, the risk of misjudgment caused by single-frame data error and trajectory jump of the pedestrian can be reduced, and the accuracy of the judgment on the pedestrian crossing behavior is improved.

[0150] Through the above embodiments, the dynamic and static detection of the pedestrian, the pedestrian orientation angle detection, whether the predicted motion trajectory of the pedestrian and the predicted travel trajectory of the ego vehicle exist intersection, and continuous multi-frame data verification are used to determine whether the object is a road crossing obstacle, and the accuracy of the obstacle judgment is improved.

[0151] In the process of creating / deleting a safety passage corridor for the crossing pedestrian, the safety passage corridor corresponding to the pedestrian is mainly created by collision detection of the predicted motion trajectory of the pedestrian and the predicted travel trajectory of the ego vehicle. If there is a time-space intersection point, it means that the pedestrian will collide with the ego vehicle and a safety passage corridor needs to be established for avoidance.

[0152] The process specifically includes the following steps:

[0153] Step one, determining the predicted travel trajectory of the ego vehicle. It can be obtained by recursion. It should be noted that when the predicted travel trajectory is recursively obtained, the collision detection is not performed on the obstacle which has been determined as a crossing pedestrian. Referring to FIG. 4, which is a schematic diagram of recursively obtaining the predicted travel trajectory of the ego vehicle in a vehicle obstacle avoidance method according to an embodiment. In order to prevent collision when the trajectory is recursively obtained, resulting in failure to obtain the predicted travel trajectory.

[0154] Step two, generation / update / deletion of the collision point.

[0155] Specifically, in the step 100 of performing collision detection on the predicted motion trajectory of the obstacle in the current travel environment of the ego vehicle and the predicted travel trajectory of the ego vehicle, the following steps are included:

[0156] In step 101, a predicted motion trajectory of the obstacle is obtained based on motion data of the obstacle in the current driving environment of the ego vehicle, and a predicted driving trajectory of the ego vehicle is obtained based on driving data of the ego vehicle.

[0157] As described above, the driving data of the ego vehicle is obtained, and the predicted driving trajectory of the ego vehicle is obtained in a recursive manner, which is an optimal trajectory for avoiding the crossing pedestrian scenario and is simulated based on the driving data of the ego vehicle. The predicted motion trajectory of the crossing pedestrian is obtained based on the motion data of the crossing pedestrian, for example, after the sensor on the ego vehicle collects data of the surrounding environment, the position and motion trend of the pedestrian are determined through a series of algorithm processing, and the predicted motion trajectory of the pedestrian is generated.

[0158] In step 102, a collision detection result is obtained by performing collision detection in space-time between the predicted motion trajectory of the obstacle and the predicted driving trajectory of the ego vehicle.

[0159] The collision detection in space-time between the predicted driving trajectory of the ego vehicle and the predicted motion trajectory of the crossing pedestrian means that whether the two will collide is determined by considering both time and space dimensions. In the spatial dimension, whether the two trajectories will intersect at different positions is detected, and in the time dimension, the time point of intersection is determined. If the ego vehicle and the pedestrian reach the same position at a certain time point, the two have a risk of collision, and then a collision detection result including the position (i.e., the collision point) is obtained, which is used for creating a safe passing corridor subsequently, which is described in detail in subsequent chapters. Otherwise, if the predicted driving trajectory of the ego vehicle and the predicted motion trajectory of the crossing pedestrian do not have a risk of collision in space-time, it means that the crossing pedestrian will not affect the driving of the ego vehicle, and the crossing pedestrian is excluded in the avoidance planning of the ego vehicle, and a result without collision is obtained without a collision point.

[0160] In some embodiments, in order to improve the accuracy of the collision detection, the collision detection result needs to be verified subsequently, which is described in detail below. The collision point obtained before verification is referred to as an initial collision point, and the collision point after verification is referred to as a target collision point.

[0161] In step 103, the collision detection result is verified based on motion data and driving data of a plurality of continuous frames.

[0162] In step 103, the collision detection result includes an initial collision point at which the crossing pedestrian and the ego vehicle collide in the preliminary collision detection.

[0163] Referring to FIG. 5, FIG. 5 is a schematic diagram of multi-frame space-time collision detection in a vehicle obstacle avoidance method according to an embodiment. Specifically, in step 103 of verifying the collision detection result based on motion data and driving data of a plurality of continuous frames, the following steps are included.

[0164] Step 1031, after the preliminary collision detection, continuously acquire N frames of the motion data of the obstacle and the driving data of the ego vehicle;

[0165] Step 1032, generate a new predicted motion trajectory based on the motion data of the same frame in the continuous N frames and a new predicted driving trajectory based on the driving data;

[0166] Step 1033, perform a collision detection in space-time using the new predicted motion trajectory and the predicted driving trajectory, and verify the collision detection result.

[0167] Step 1034, update the initial collision point according to the verification result.

[0168] After generating the preliminary collision point, the predicted driving trajectory of the crossing pedestrian may change because the crossing pedestrian may change the original behavior, and the initial collision point between the predicted driving trajectory of the crossing pedestrian and the predicted driving trajectory of the ego vehicle will also change, so the collision point needs to be updated.

[0169] In this process, the verification result refers to the result of the second collision detection, and the verification of the continuous N frames includes N verification results. These results are compared with the collision detection result in the preliminary collision detection to determine whether there is a collision point between the predicted driving trajectory of the ego vehicle and the predicted motion trajectory of the crossing pedestrian, and the position information of the collision point.

[0170] Step 104, after the collision verification passes, generate the target collision point when the obstacle and the ego vehicle collide.

[0171] The target collision described herein is the position where the final collision between the current predicted driving trajectory of the ego vehicle and the current predicted motion trajectory of the crossing pedestrian is determined after collision detection.

[0172] Among them, there are several cases:

[0173] Case one, when the verification result indicates that the current predicted driving trajectory of the ego vehicle and the current predicted motion trajectory of the crossing pedestrian still have a collision risk in space-time, and both have a collision point. The collision point in the latest verification result is taken as the target collision point, so as to create a safe passing corridor for the crossing pedestrian to pass in the subsequent process.

[0174] Case two, after the initial collision point exists, the longitudinal position of the initial collision point changes more than a set threshold and changes continuously for multiple frames, then the collision point is updated.

[0175] Referring to FIG. 6, FIG. 6 is a schematic diagram of a collision point updating process in a vehicle obstacle avoidance method according to an embodiment. Specifically, the updating of the initial collision point according to the verification result includes the following processes:

[0176] In a case where the verification result representing that the obstacle collides with the ego vehicle is obtained, a longitudinal position change value between a new collision point in the verification result and the initial collision point is determined, the longitudinal direction being a driving direction of the ego vehicle;

[0177] In a case where the longitudinal position change value is greater than a set threshold value, the initial collision point is updated based on the new collision point;

[0178] The new collision point is determined as a target collision point when the obstacle collides with the ego vehicle.

[0179] Case three, referring to FIG. 7, FIG. 7 is another schematic diagram of a collision point updating process in a vehicle obstacle avoidance method according to an embodiment. Specifically, in a case where the verification result representing that the obstacle does not collide with the ego vehicle is obtained, the initial collision point is deleted.

[0180] The predicted motion track of the crossing pedestrian does not collide with the predicted driving track of the ego vehicle in the continuous N frames, which means that there is no collision risk between the crossing pedestrian and the ego vehicle, and thus the initial collision point is deleted. The automatic driving system of the ego vehicle continues to perform collision detection on other crossing pedestrians to ensure the safe passage of all crossing pedestrians within the screening window.

[0181] Case four, still referring to FIG. 7, when the pedestrian is no longer determined as a crossing pedestrian, the initial collision point is deleted.

[0182] Specifically, the updating of the initial collision point according to the verification result includes the following processes:

[0183] Intention detection is performed on the obstacle;

[0184] In a case where the verification result representing that the obstacle does not have the intention to cross the current road on which the ego vehicle is driving is obtained, the initial collision point is deleted.

[0185] It can be understood that when the pedestrian changes the motion intention and does not have the motivation to cross the lane in which the ego vehicle is located, the pedestrian is no longer determined as a crossing pedestrian, which means that there is no collision risk between the pedestrian and the ego vehicle, and thus the situation of the pedestrian can not be considered when the ego vehicle currently takes avoidance measures, and the avoidance analysis on the crossing pedestrian is performed again when the pedestrian is detected as a crossing pedestrian again. Therefore, in a case where the pedestrian is no longer a crossing pedestrian, the initial collision point obtained based on the collision detection between the predicted motion track of the pedestrian and the predicted driving track of the ego vehicle is deleted.

[0186] In the case where it is determined that there is a collision risk between the crossing pedestrian and the ego vehicle and the target collision point between each crossing pedestrian and the ego vehicle is determined, a safe passage corridor for the crossing pedestrian to pass through the road currently traveled by the ego vehicle is created.

[0187] Referring to FIG. 8, FIG. 8 is a schematic diagram illustrating the creation of a safe passage corridor in a vehicle obstacle avoidance method according to an embodiment.

[0188] In the case where there is a collision risk between the obstacle and the ego vehicle, in step 200 of creating a safe passage corridor for the obstacle to pass through the road currently traveled by the ego vehicle, the following steps are included:

[0189] Step 201, if the target collision point exists, it is determined that there is a collision risk between the obstacle and the ego vehicle.

[0190] Step 202, in the case where there is a collision risk between the obstacle and the ego vehicle, the target collision point with the smallest longitudinal distance from the ego vehicle is taken as the base point.

[0191] In the autonomous driving scenario, time and distance are crucial for avoiding collisions. The collision point closest to the ego vehicle means that it is the most imminent potential danger point, and prioritizing this point can allow the autonomous driving system to react at the fastest speed and take appropriate measures to avoid collisions. Therefore, when multiple target collision points are determined, the safe passage corridor for the crossing pedestrian to pass through is created based on the target collision point with the smallest longitudinal distance from the ego vehicle, focusing on this base point can make the system more efficient in decision-making and planning, and concentrate resources to deal with the most urgent situation. Referring to FIG. 8, in the direction of the ego vehicle, the target collision point of pedestrian No. 6 is closest to the ego vehicle and is the first to have a collision risk, therefore, the safe passage corridor for the crossing pedestrian to pass through is created based on the target collision point detected by pedestrian No. 6 and the ego vehicle, ensuring the safety of pedestrian No. 6 and providing a clear avoidance area for the autonomous vehicle.

[0192] After avoiding the crossing pedestrian closest to the ego vehicle, new target collision points are detected again, and new safe passage corridors are created based on the new target collision points to avoid the next crossing pedestrian. Continuing the above example, referring to FIG. 8, after the ego vehicle avoids pedestrian No. 6, it avoids pedestrian No. 3.

[0193] Step 203, based on the preset boundary constraints, the safe passage corridor for the target obstacle corresponding to the target collision point with the smallest longitudinal distance is created.

[0194] The safe passing corridor described herein is a passing range for a crossing pedestrian to pass through the lane where the ego vehicle is located, which can be regarded as a virtual safe area. The ego vehicle needs to avoid entering this area during driving to prevent collision with the crossing pedestrian. By generating this corridor, the system can more clearly understand which areas need special attention, thereby better planning the path and making decisions.

[0195] The safe passing corridor has a certain inclusiveness and can be compatible with the case where the crossing pedestrian suddenly changes the motion intention. Even if the predicted motion trajectory of the pedestrian is unreliable, it can still ensure that the trajectory planned for the ego vehicle to drive based on the safe passing corridor is safe and reliable. The range is usually determined according to the possible motion range of the crossing pedestrian and the safety requirements of the ego vehicle.

[0196] In some embodiments, as shown in FIG. 8, the boundary constraints of the safe passing corridor include longitudinal boundary constraints in the same direction as the driving direction of the ego vehicle and transverse boundary constraints perpendicular to the driving direction of the ego vehicle.

[0197] Specifically, according to the one or more target collision points detected by the collision, the target collision point with the minimum longitudinal distance from the ego vehicle is set as the base point, the safe walking distance in the longitudinal direction is set, and the longitudinal boundary constraint of the safe passing corridor is obtained.

[0198] The width of the safe passing corridor (i.e., the safe walking distance) can be customized. The wider the width is set, the more conservative the behavior of the ego vehicle is. A wider safe passing corridor means that more safety space is provided for the crossing pedestrian, and the ego vehicle also has more buffer distance to deal with potential dangers.

[0199] In some examples, the safe walking distance is determined according to an error range of the predicted motion trajectory of the obstacle and the actual motion trajectory of the obstacle, and the safe walking distance is in a positive correlation with the error range. It can be understood that the error caused by the uncertainty of the pedestrian intention prediction and the instability of the sensor perception data in the longitudinal direction is solved by the safe walking distance, and the safety and reliability of the avoidance measures taken based thereon are improved.

[0200] On the transverse constraint of the safe passing corridor, the transverse boundary constraint of the safe passing corridor is obtained according to the width of the road currently driven by the ego vehicle, to ensure that the crossing pedestrian passes through the road currently driven by the ego vehicle safely.

[0201] After creating the safe passing corridor in the above manner, the autonomous driving system takes corresponding measures such as deceleration and braking according to the information to avoid collision. When the target collision point does not exist, the corresponding safe passing corridor is deleted.

[0202] Referring to FIG. 9, FIG. 9 is a schematic diagram of a longitudinal planning process in a vehicle obstacle avoidance method according to an embodiment.

[0203] Specifically, in the step 300 of adding the safe passage corridor to the longitudinal planning of the ego vehicle to generate a decision trajectory for avoiding the obstacle to control the ego vehicle to avoid the obstacle, the following process is included:

[0204] The safe passage corridor is projected into the ST graph of the longitudinal planning to plan a decision trajectory for avoiding the obstacle to control the ego vehicle to avoid the obstacle.

[0205] The ST graph (Space-Time graph) described herein is a tool for describing the motion state of a vehicle in the spatial and temporal dimensions. Projecting the safe corridor onto the ST graph of the longitudinal planning is to explicitly mark the area occupied by the safe corridor in this two-dimensional space-time graph. In this way, the planning algorithm can intuitively see the distribution of the safe corridor in space-time when performing longitudinal planning.

[0206] The safe passage corridor is projected as a static obstacle into the ST graph of the longitudinal planning of the ego vehicle, and the longitudinal planning does not plan a trajectory that crosses the safe passage corridor. As shown in FIG. 9, the rectangular box where the 6th pedestrian is located is the projected safe passage corridor, and the st curve below the rectangular box does not cross the rectangular box within the planning time.

[0207] According to the foregoing, the closest collision point to the ego vehicle means that it is the most imminent potential danger point, so creating a safe passage corridor based on the target collision point with the minimum longitudinal distance from the ego vehicle, the decision trajectory planned based on this is a measure to avoid the target obstacle corresponding to the target collision point with the minimum longitudinal distance from the ego vehicle.

[0208] Therefore, in another example, the step of adding the safe passage corridor to the longitudinal planning of the ego vehicle to generate a decision trajectory for avoiding the obstacle to control the ego vehicle to avoid the obstacle includes:

[0209] The safe passage corridor is added to the longitudinal planning of the ego vehicle to generate a decision trajectory for avoiding the target obstacle to control the ego vehicle to avoid the target obstacle. Referring to FIG. 8, the 6th pedestrian closest to the ego vehicle is the target obstacle.

[0210] In some embodiments, in the process of controlling the ego vehicle to avoid the crossing pedestrian based on the decision trajectory, the motion data of the pedestrian is monitored in real time to ensure that the planned decision trajectory continues to be applicable to the current scenario.

[0211] Specifically, the process of controlling the ego vehicle to avoid the obstacle includes the following steps:

[0212] Based on the preset detection frequency, the predicted motion trajectory of the obstacle is updated. That is, based on the detection frequency, the latest trajectory data of the crossing pedestrian is obtained.

[0213] There are two cases between the updated predicted motion trajectory of the pedestrian and the current predicted motion trajectory.

[0214] In the first case, if the change between the updated predicted motion trajectory and the current predicted motion trajectory does not exceed the set range, the decision trajectory is reused to control the ego vehicle to avoid the obstacle based on the decision trajectory.

[0215] That is, the crossing pedestrian does not suddenly change the motion intention within the detection period, and the motion range is within the aforementioned created safe passage corridor. At this time, the process of repeatedly detecting the collision between the pedestrian's predicted motion data and the ego vehicle's predicted driving trajectory, creating a safe passage corridor, and planning a new decision trajectory is not required, which improves the overall response speed and operation efficiency of the system. At the same time, it avoids introducing new uncertainties and error data when repeatedly calculating the decision trajectory, thereby improving the safety of driving.

[0216] In the second case, if the change between the updated predicted motion trajectory and the current predicted motion trajectory exceeds the set range, the collision detection between the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle is performed to update the decision trajectory.

[0217] The ego vehicle is controlled to avoid the obstacle based on the updated decision trajectory.

[0218] That is, at the current time, the motion state of the crossing pedestrian has changed greatly, which may have a greater impact on the driving safety of the ego vehicle. In this case, the decision trajectory planned by the previous safe passage corridor may not be able to effectively avoid the collision risk, and therefore the predicted motion trajectory generated based on the latest motion data of the pedestrian is needed to perform longitudinal planning on the ego vehicle, and a decision trajectory that avoids the new state of the crossing pedestrian is re-planned. The new decision trajectory can better adapt to the motion state of the crossing pedestrian and take more reasonable avoidance measures, and the ego vehicle is controlled to avoid the obstacle based on the updated decision trajectory, thereby minimizing the collision risk and ensuring the safety of autonomous driving.

[0219] In this case, if it is detected that there is no collision risk or no collision point between the pedestrian and the ego vehicle, the safe passage corridor is deleted and a decision trajectory is re-planned.

[0220] Specifically, the updating the decision trajectory includes the following processes:

[0221] performing collision detection on the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle;

[0222] in the case that there is no collision risk between the obstacle and the ego vehicle, deleting the safe passing corridor for the obstacle to pass through the road currently driven by the ego vehicle;

[0223] updating the decision trajectory.

[0224] That is, when the ego vehicle executes the planned decision trajectory, the motion state of the crossing pedestrian has changed greatly and the change means that there is no collision risk between the ego vehicle and the pedestrian, as shown in FIG. 7, at the t+n frame, there is no collision point between the trajectory of the pedestrian and the trajectory of the ego vehicle, so the current safe passing corridor is deleted, and the ego vehicle is guaranteed to drive normally. Or avoid other crossing pedestrians in the screening window range, as shown in FIG. 8, after the No. 6 pedestrian avoids, it is detected whether the No. 3 pedestrian has a collision point with the ego vehicle, if yes, a safe passing corridor for the No. 3 pedestrian is created, and a decision trajectory for avoiding the No. 3 pedestrian is planned. Further, the avoidance of all obstacles in the driving environment of the ego vehicle is realized.

[0225] Based on the above first embodiment, the second embodiment of the vehicle obstacle avoidance method provided by the present application sets a stop point and / or a stop line, so that the ego vehicle stops at the stop point and / or the stop line, without creating a safe passing corridor.

[0226] Specifically, after the step 100 of performing collision detection on the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle, the method further includes the following steps:

[0227] Step 400, in the case that there is a collision risk between the obstacle and the ego vehicle, setting a stop point and / or a stop line between the target collision point and the ego vehicle.

[0228] The stop point and / or the stop line described herein is a virtual point and / or line marked on the road, which is used as a position for controlling the vehicle to stop when the ego vehicle is planned. The position is determined based on the target collision point between the predicted motion trajectory of the crossing pedestrian and the predicted driving trajectory of the ego vehicle, and is a position at a certain distance from the target collision point in the driving direction of the ego vehicle.

[0229] Step 500, adding the stop point and / or the stop line to the longitudinal planning of the ego vehicle, generating a decision trajectory for avoiding the obstacle, to control the ego vehicle to avoid the obstacle and drive.

[0230] By adding the stop point and / or the stop line to the longitudinal planning, the system can explicitly determine when and where the ego vehicle needs to start deceleration or stop to avoid collision with the obstacle. Based on the longitudinal planning with the added stop point and / or stop line, a decision trajectory for avoiding the obstacle is generated. The decision trajectory takes into account the current state of the ego vehicle, the position and motion state of the obstacle, and the road environment, etc., to determine the driving strategy that the ego vehicle should take, such as deceleration, braking, steering, etc., to ensure that the ego vehicle can safely avoid the obstacle.

[0231] Finally, according to the generated decision trajectory, the ego vehicle is controlled to avoid the obstacle. The autonomous driving system adjusts the speed, direction, and other parameters of the ego vehicle to make the ego vehicle travel according to the decision trajectory, thereby effectively avoiding collision with the obstacle and ensuring safe driving.

[0232] The technical solutions provided by the embodiments of the present application can include the following beneficial effects:

[0233] In a first aspect, the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle are subjected to collision detection, and in the case where there is a collision risk between the obstacle and the ego vehicle, a safe passing corridor for the obstacle to pass through the road currently driven by the ego vehicle is created. The safe passing corridor is planned for the predicted motion trajectory of the obstacle whose intention is difficult to predict, and has a certain inclusiveness, which can solve the problems of uncertain intention of the obstacle and limited sensor perception capability.

[0234] In a second aspect, by adding the safe passing corridor to the longitudinal planning of the ego vehicle, the longitudinal planning plans a trajectory that does not cross the safe passing corridor, and therefore a decision trajectory for avoiding the obstacle is generated based on the predicted driving trajectory, to control the ego vehicle to avoid the obstacle. This process plans a safe and reliable trajectory for avoiding the obstacle for the predicted motion trajectory of the uncertain obstacle, to ensure the safe passing of the obstacle and the safe driving of the ego vehicle, and the avoidance effect is good.

[0235] Corresponding to the foregoing application function implementation method embodiments, the present application also provides application function implementation device embodiments.

[0236] Referring to FIG. 10, FIG. 10 is a block diagram of a vehicle obstacle avoidance device according to an exemplary embodiment, which includes:

[0237] A collision detection module is configured to perform collision detection on the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle. The obstacle moves in the direction of the predicted driving trajectory of the ego vehicle.

[0238] The creating a safe passing corridor module is configured to create a safe passing corridor for the obstacle to pass through a road currently traveled by the ego vehicle in a case where there is a collision risk between the obstacle and the ego vehicle.

[0239] The longitudinal planning module is configured to add the safe passing corridor to a longitudinal plan of the ego vehicle, generate a decision trajectory that avoids the obstacle based on the predicted travel trajectory, and control the ego vehicle to travel to avoid the obstacle.

[0240] For the device embodiment, since it basically corresponds to the method embodiment, the relevant part is described in the method embodiment. The device embodiment described above is only illustrative, and the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., they can be located in one place or distributed on multiple network units. Some or all modules can be selected to achieve the purpose of the embodiment according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0241] FIG. 11 shows a schematic diagram of an entity structure of a vehicle obstacle avoidance device. As shown in FIG. 11, the vehicle obstacle avoidance device can include a processor 810, a communications interface 820, a memory 830, and a communications bus 840, wherein the processor 810, the communications interface 820, and the memory 830 communicate with each other through the communications bus 840. The processor 810 can invoke the logic instructions in the memory 830 to execute the parking path planning method.

[0242] In addition, the logic instructions in the memory 830 described above can be implemented in the form of a software functional unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the current technology or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0243] Correspondingly, the present application also provides a computer readable storage medium, the storage medium stores a computer program, the computer program is used for executing the vehicle obstacle avoidance method of any one of the first aspect.

[0244] Correspondingly, the present application also provides a computer readable storage medium, the storage medium stores a computer program, the computer program is used for executing the vehicle obstacle avoidance method of any one of the second aspect.

[0245] It should be noted that, in this paper, the relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. The terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0246] The above describes the method and device provided by the embodiments of the present application in detail, and the principles and implementation modes of the present application are described by applying specific examples. The above description of the embodiments is only used to help understand the method of the present application and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range will be changed; in view of the above, the content of the specification should not be understood as the limitation of the present application.

Claims

1. A vehicle obstacle avoidance method characterized by, The method includes: Collision detection is performed on the predicted motion trajectory of obstacles in the current driving environment of the vehicle and the predicted driving trajectory of the vehicle; the obstacles move in the direction of the predicted driving trajectory of the vehicle; In the event of a collision risk between the obstacle and the vehicle, a safe passage corridor is created for the obstacle to pass through the road currently being traveled by the vehicle; The safe passage corridor is added to the longitudinal planning of the vehicle, and a decision trajectory to avoid the obstacle is generated based on the predicted driving trajectory, so as to control the vehicle to avoid the obstacle.

2. The method of claim 1, wherein, The collision detection of the predicted motion trajectory of obstacles in the current driving environment of the vehicle and the predicted driving trajectory of the vehicle includes: The predicted motion trajectory of the obstacle is generated by acquiring the motion data of the obstacle in the current driving environment of the vehicle, and the predicted driving trajectory is obtained from the driving data of the vehicle. The predicted motion trajectory of the obstacle and the predicted driving trajectory of the vehicle are subjected to spatiotemporal collision detection to obtain the collision detection result; The collision detection results are verified based on motion and driving data from multiple consecutive frames. After the collision verification is passed, the target collision point when the obstacle collides with the vehicle is generated.

3. The method of claim 2, wherein, The collision detection results include the initial collision point when the obstacle collides with the vehicle. The verification of the collision detection results based on continuous multi-frame motion data and driving data includes: After the initial collision detection, N frames of motion data of the obstacle and driving data of the vehicle are continuously acquired. A new predicted motion trajectory is generated based on the motion data of the same frame in N consecutive frames, and a new predicted driving trajectory is generated based on the driving data. The new predicted motion trajectory and the predicted driving trajectory are used to perform spatiotemporal collision detection, and the collision detection results are verified. The initial collision point is updated based on the verification results.

4. The method of claim 3, wherein, The step of updating the initial collision point based on the verification result includes: Given a verification result indicating a collision between the obstacle and the vehicle, determine the longitudinal position change between the new collision point and the initial collision point in the verification result, where the longitudinal direction is the driving direction of the vehicle. If the longitudinal position change value is greater than a set threshold, the initial collision point is updated based on the new collision point; The generation of the target collision point when the obstacle collides with the vehicle includes: The new collision point is determined as the target collision point when the obstacle collides with the vehicle.

5. The method of claim 3, wherein, The step of updating the initial collision point based on the verification result includes: If a verification result indicating that the obstacle and the vehicle do not collide is obtained, the preliminary collision point is deleted.

6. The method of claim 3, wherein, The step of updating the initial collision point based on the verification result includes: Intent detection is performed on the obstacle; If a verification result indicates that the obstacle does not have the intention to cross the road currently being traveled by the vehicle, the initial collision point is deleted.

7. The method of claim 2, wherein, The method further includes: if the target collision point exists, determining that there is a collision risk between the obstacle and the ego vehicle; in the case that there is a collision risk between the obstacle and the ego vehicle, taking the target collision point with the minimum longitudinal distance between the ego vehicle as a base point; based on a preset boundary constraint, creating the safe passing corridor for the target obstacle corresponding to the target collision point with the minimum longitudinal distance to pass through the road currently traveled by the ego vehicle.

8. The method of claim 7, wherein, The boundary constraint includes a longitudinal boundary constraint in the same direction of travel as the ego vehicle and a transverse boundary constraint perpendicular to the direction of travel of the ego vehicle, The method further includes: obtaining the longitudinal boundary constraint of the safe passing corridor according to the base point and a preset safe walking distance; obtaining the transverse boundary constraint of the safe passing corridor according to the width of the road currently traveled by the ego vehicle.

9. The method of claim 8, wherein, The safe walking distance is determined according to an error range of the predicted motion trajectory of the obstacle and the actual motion trajectory of the obstacle, and the safe walking distance is in a positive correlation with the error range.

10. The method of claim 7, wherein, After the safe passing corridor for the target obstacle corresponding to the target collision point with the minimum longitudinal distance is created, The method further includes: adding the safe passing corridor to the longitudinal planning of the ego vehicle to generate a decision trajectory for avoiding the obstacle, so as to control the ego vehicle to avoid traveling for the obstacle.

11. The method of claim 1, wherein, The method further includes: adding the safe passing corridor to the longitudinal planning of the ego vehicle to generate a decision trajectory for avoiding the target obstacle, so as to control the ego vehicle to avoid traveling for the target obstacle.

12. The method of claim 1, wherein, The method further includes: projecting the safe passing corridor into an ST graph of the longitudinal planning to plan a decision trajectory for avoiding the obstacle, so as to control the ego vehicle to avoid traveling for the obstacle. The method further includes:

13. The method of claim 12, wherein, updating the predicted motion trajectory of the obstacle based on a preset detection frequency; in the case that the change between the updated predicted motion trajectory and the current predicted motion trajectory does not exceed a set range, reusing the decision trajectory to control the ego vehicle to avoid traveling for the obstacle based on the decision trajectory. After the predicted motion trajectory of the obstacle is updated based on the preset detection frequency, the method further includes:

14. The method of claim 13, wherein, in the case that the change between the updated predicted motion trajectory and the current predicted motion trajectory exceeds the set range, performing collision detection on the predicted motion trajectory of the obstacle in the current travel environment of the ego vehicle and the predicted travel trajectory of the ego vehicle to update the decision trajectory; controlling the ego vehicle to avoid traveling for the obstacle based on the updated decision trajectory. The method further includes: performing collision detection on a predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and a predicted driving trajectory of the ego vehicle; deleting a safe passing corridor for the obstacle to pass through a road currently driven by the ego vehicle in a case where there is no collision risk between the obstacle and the ego vehicle; updating the decision trajectory.

15. The method of claim 2, wherein, After the collision detection on the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle, the method further comprises: a set stop point and / or stop line between the target collision point and the ego vehicle in a case where there is a collision risk between the obstacle and the ego vehicle; adding the stop point and / or stop line to longitudinal planning of the ego vehicle to generate a decision trajectory for avoiding the obstacle to control the ego vehicle to perform avoidance driving for the obstacle.

16. The method of claim 1, wherein, Before the collision detection on the predicted motion trajectory of the obstacle in the current driving environment of the ego vehicle and the predicted driving trajectory of the ego vehicle, the method further comprises: detecting an object in the current driving environment of the ego vehicle; regarding the object meeting a crossing condition as an obstacle having an intention of crossing a road currently driven by the ego vehicle; wherein the crossing condition comprises all of the following: the object is in a non-stationary state; an included angle is between a motion direction of the object and a driving direction of the ego vehicle; a predicted motion trajectory of the object and a predicted driving trajectory of the ego vehicle have an intersection point.

17. The method of claim 16, wherein, The detecting of the obstacle in the current driving environment of the ego vehicle comprises: detecting the obstacle in the current driving environment of the ego vehicle within a preset screening window range.

18. The method of claim 16, wherein, The regarding of the object meeting the crossing condition as the obstacle having the intention of crossing the road currently driven by the ego vehicle comprises: determining the object meeting the crossing condition; performing multi-frame data verification on the object meeting the crossing condition; regarding the object meeting the crossing condition for continuous multiple times as a crossing obstacle having the intention of crossing the road currently driven by the ego vehicle.

19. A vehicle obstacle avoidance apparatus characterized by comprising: A vehicle obstacle avoidance device comprises a memory, a processor and a vehicle obstacle avoidance program stored on the memory and executable on the processor, and the processor implements steps of the vehicle obstacle avoidance method according to any one of claims 1-18 when executing the vehicle obstacle avoidance program.

20. A computer-readable storage medium, characterized in that, A computer readable storage medium stores a vehicle obstacle avoidance program, and the vehicle obstacle avoidance program implements steps of the vehicle obstacle avoidance method according to any one of claims 1-18 when executed.

Images