CN116295495A

CN116295495A - Automatic driving vehicle path planning method, device, equipment and medium

Info

Publication number: CN116295495A
Application number: CN202310286980.5A
Authority: CN
Inventors: 姚志鹏; 李勇强; 吕强; 付一豪
Original assignee: Neolix Technologies Co Ltd
Current assignee: Neolix Technologies Co Ltd
Priority date: 2023-03-22
Filing date: 2023-03-22
Publication date: 2023-06-23

Abstract

The invention discloses a method, a device, equipment and a medium for planning an automatic driving vehicle path. The scheme relates to the unmanned field, and the method comprises the following steps: determining a priori characteristics of at least one obstacle based on the perceived information of the unmanned vehicle; according to the motion state of the vehicle, determining weights matched with at least two candidate areas based on vehicle position division, and determining the sequencing result of the obstacles according to the weights of the candidate areas, the prior characteristics of the obstacles and the positions of the obstacles; and determining a preset number of dangerous obstacles in each obstacle according to the sequencing result, and taking the prior characteristics of each dangerous obstacle as the input of a prediction model to determine the planned path of the automatic driving vehicle. The technical scheme solves the problems of low accuracy, high time cost and the like caused by directly extracting the abstract features of the perception information by the prediction model, and can enhance the interpretability of the prediction model while improving the prediction accuracy of the prediction model and the generation efficiency of the model.

Description

Automatic driving vehicle path planning method, device, equipment and medium

Technical Field

The invention relates to the technical field of unmanned driving, in particular to a method, a device, equipment and a medium for planning an automatic driving vehicle path.

Background

With the continuous progress of technology, the demand for navigation ability of unmanned vehicles under various road conditions is increasingly urgent. The navigation system of the unmanned vehicle needs to acquire sensing information such as radar, vision and the like, and makes reasonable decisions based on the recognition result of the sensing information so as to arrive at a destination timely and accurately on the premise of ensuring driving safety.

At present, a navigation system of an unmanned vehicle mainly depends on predictive models such as Lyft, waymo and the like, the perception information is directly subjected to feature representation based on a deep learning network, and driving decisions are carried out according to abstract features extracted from the perception information.

However, in the prior art, the manner of directly characterizing the perception information through the deep learning network requires a great deal of time to train the deep learning network, and meanwhile, the stability of the training effect is difficult to ensure, and the interpretation of the prediction model is poor.

Disclosure of Invention

The invention provides an automatic driving vehicle path planning method, device, equipment and medium, which are used for solving the problems of low accuracy, high time cost and the like caused by directly extracting abstract features of perception information by a prediction model, and can enhance the interpretation of the prediction model while improving the prediction accuracy of the prediction model and the model generation efficiency.

According to an aspect of the present invention, there is provided a method of automatically driving a vehicle path planning, the method comprising:

determining a priori characteristics of at least one obstacle based on the perception information of the vehicle;

according to the motion state of the vehicle, determining weights matched with at least two candidate areas based on vehicle position division, and determining the sequencing result of the obstacles according to the weights of the candidate areas, the prior characteristics of the obstacles and the positions of the obstacles;

and determining a preset number of dangerous obstacles in each obstacle according to the sequencing result, and taking the prior characteristics of each dangerous obstacle as the input of a prediction model to determine the planned path of the automatic driving vehicle.

According to another aspect of the present invention, there is provided an autonomous vehicle path planning apparatus comprising:

the prior feature determining module is used for determining prior features of at least one obstacle based on the perception information of the vehicle;

the sequencing result determining module is used for determining the matching weight of at least two candidate areas based on the vehicle position division according to the motion state of the vehicle, and determining the sequencing result of the obstacles according to the weight of each candidate area, the prior characteristic of each obstacle and the position of each obstacle;

And the planning path determining module is used for determining a preset number of dangerous obstacles in each obstacle according to the sequencing result, and taking the prior characteristics of each dangerous obstacle as the input of a prediction model to determine the planning path of the automatic driving vehicle.

According to another aspect of the present invention, there is provided an electronic apparatus including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,

the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the autonomous vehicle path planning method according to any of the embodiments of the present invention.

According to another aspect of the present invention, there is provided a computer readable storage medium storing computer instructions for causing a processor to execute the method for autonomous vehicle path planning according to any of the embodiments of the present invention.

According to the technical scheme, the prior characteristic of at least one obstacle is determined through the perception information of the vehicle; then, according to the motion state of the vehicle, determining the weight matched with at least two candidate areas based on the vehicle position division, and determining the sequencing result of the obstacles according to the weight of each candidate area, the prior characteristic of each obstacle and the position of each obstacle; and determining a preset number of dangerous obstacles in each obstacle according to the sequencing result of the obstacles, and taking the prior characteristics of each dangerous obstacle as the input of a prediction model to determine the planned path of the automatic driving vehicle. The method solves the problems of low accuracy, high time cost and the like caused by directly extracting the abstract features of the perception information by the prediction model, and can enhance the interpretability of the prediction model while improving the prediction accuracy of the prediction model and the generation efficiency of the model.

It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the invention or to delineate the scope of the invention. Other features of the present invention will become apparent from the description that follows.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flow chart of a method for automatically driving a vehicle path planning in accordance with a first embodiment of the present invention;

FIG. 2 is a schematic diagram of a prediction model provided according to an embodiment of the present invention;

fig. 3 is a flowchart of an automatic driving vehicle path planning method according to a second embodiment of the present invention;

fig. 4 is a schematic view of candidate region division in a simple driving scenario provided according to an embodiment of the present invention;

fig. 5 is a schematic diagram of candidate region division in a complex driving scenario according to an embodiment of the present invention;

Fig. 6 is a schematic structural view of an automatic driving vehicle path planning apparatus according to a third embodiment of the present invention;

fig. 7 is a schematic structural diagram of an electronic device implementing a method for path planning of an autonomous vehicle according to an embodiment of the present invention.

Detailed Description

In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, apparatus, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus. The data acquisition, storage, use, processing and the like in the technical scheme meet the relevant regulations of national laws and regulations.

Example 1

Fig. 1 is a flowchart of an automatic driving vehicle path planning method according to an embodiment of the present invention, where the embodiment is applicable to a path planning scenario of an automatic driving vehicle, and is particularly applicable to complex traffic scenarios such as congestion, multiple lanes, and the like. The method may be performed by an autonomous vehicle path planning device, which may be implemented in hardware and/or software, which may be configured in an electronic device. As shown in fig. 1, the method includes:

s110, determining prior characteristics of at least one obstacle based on perception information of the vehicle.

The scheme can be executed by a controller of the automatic driving vehicle, and the controller can acquire multidimensional sensing information through sensors such as radars, cameras and the like deployed by the vehicle. The sensing information may include information such as a speed of the vehicle, a speed of the obstacle, a distance between the vehicle and the obstacle, and response time. The controller can identify the types of the obstacles, such as static obstacles and movement obstacles, according to the driving scene images acquired by the camera. Wherein the static object may include a lane line, a green belt, a building, etc., and the dynamic obstacle may include a vehicle, a pedestrian, a bicycle, etc. According to the perception information of the vehicle, the controller can determine the motion characteristics, the navigation map characteristics and the obstacle characteristics of the vehicle. The obstacle features include a priori features of the obstacle, which may include a safety loss of the obstacle, a rate of change of the safety loss, and the like. Specifically, the controller may determine the safety loss of each obstacle according to the information such as the speed of the vehicle, the speed of the obstacle, the distance between the vehicle and the obstacle, and the response time. The controller may determine a rate of change of the safety loss of each obstacle according to the speed of the host vehicle and the distance between the host vehicle and the obstacle.

S120, according to the motion state of the vehicle, determining the weight matched with at least two candidate areas based on the vehicle position division, and determining the sequencing result of the obstacles according to the weight of each candidate area, the prior characteristic of each obstacle and the position of each obstacle.

It can be appreciated that the autonomous vehicle may perform region division on the sensing range according to the current location of the vehicle, to obtain at least two candidate regions. The controller can divide candidate areas of the perception range of the vehicle according to the complexity degree of the driving scene. After determining each candidate region based on the current position of the host vehicle, the controller may determine a weight that each candidate region matches according to the motion state of the host vehicle. The movement state of the autonomous vehicle may include straight, u-turn, left-turn, right-turn, reverse, etc. The controller may set one or more sets of candidate region weights for each motion state, each set of candidate region weights including the weights of the candidate regions. Based on the position of each obstacle, the controller may determine a candidate region to which each obstacle belongs. According to the weights of the candidate areas to which the obstacles belong and the prior characteristics of the obstacles, the controller can calculate the attention degree of the obstacles, order the attention degree and obtain the ordering result of the obstacles. It is readily appreciated that a priori the characteristic safety loss and the rate of change of the safety loss may characterize the risk level of the obstacle. The risk level of the candidate area associated with the motion state of the host vehicle may increase due to the motion of the host vehicle. Therefore, the risk of each obstacle can be accurately estimated in real time according to the prior characteristics of each obstacle and the weight of the candidate region to which each obstacle belongs.

S130, determining a preset number of dangerous obstacles in each obstacle according to the sequencing result, and taking the prior characteristics of each dangerous obstacle as the input of a prediction model to determine the planned path of the automatic driving vehicle.

The controller of the automatic driving vehicle may select a preset number of obstacles having a high risk assessment value as the dangerous obstacle according to the sequencing result. For example, the moving object of the first 64 bits of the risk assessment value in the ranking result is determined as a risk obstacle. The controller may also select, as the dangerous obstacle, an obstacle having a risk evaluation value greater than a preset evaluation threshold according to the sorting result. For example, an obstacle having a risk assessment value of greater than 0.8 in the ranked result is taken as a risk obstacle.

The controller can take the motion characteristics, the navigation map characteristics and the dangerous obstacle characteristics of the vehicle as the input of a prediction model to obtain the prediction information of the automatic driving vehicle. The motion characteristics of the vehicle can comprise the position, the running direction, the motion speed and the like of the vehicle. The navigation map feature may include road features of the host vehicle driving environment, such as road boundary lines, reference lines, etc. The dangerous obstacle features may include a priori features of the dangerous obstacle and may also include features of the location, length, width, direction of movement, and distance from the host vehicle of the dangerous obstacle.

In this solution, the prediction model may be constructed based on a deep learning algorithm, and the prediction information may include information such as a speed, an acceleration, a curvature, and a position of the host vehicle. According to the prediction information, the controller can generate a planned path of the automatic driving vehicle so as to ensure safe driving of the vehicle.

Specifically, the prediction model may be obtained by training sample data in advance, and the training sample data may include multiple sets, so that the prediction model achieves a good training effect. Each set of training sample data may include feature data and tag data. Similar to the application process of the prediction model, the feature data may include a vehicle motion feature, a navigation map feature, and a preset number of dangerous obstacle features. The tag data may be own vehicle motion information matched with the feature data.

Fig. 2 is a schematic structural diagram of a prediction model according to an embodiment of the present invention. As shown in fig. 2, the predictive model may include an input branch structure, a multi-headed self-attention structure, a multi-layered perceptron structure, and a long-term memory structure. The input branch structure can be used for extracting characteristics of the motion characteristics of the vehicle, dangerous obstacle characteristics and navigation map characteristics. The multi-head self-attention structure can be used for extracting the relation features among the input features so as to improve the accuracy and the prediction efficiency of the prediction model. The multi-layer perceptron structure may compress the features. The long-short term memory structure can extract time sequence characteristics among multiple frames of input data.

In a preferred scheme, the long-short-term memory structure can take the current frame and the historical four frames as input data, and extract time sequence characteristics among the input data so as to achieve a good prediction effect.

In one possible implementation, the input branch structure includes a first input branch, a second input branch, and a third input branch; the first input branch, the second input branch and the third input branch are respectively used for extracting the motion characteristics, the dangerous obstacle characteristics and the navigation map characteristics of the vehicle; the input branch structure, the multi-head self-attention structure, the multi-layer perceptron structure and the long-period memory structure are sequentially connected.

It is easy to understand that each input branch in the input branch structure is used for extracting the motion characteristics, the dangerous obstacle characteristics and the navigation map characteristics of the vehicle respectively. Each input branch may include at least one full connection layer, and the structures of the input branches may be the same or different. The first input branch, the second input branch and the third input branch can respectively perform feature extraction on the motion feature, the obstacle feature and the navigation map feature of the vehicle to obtain a first feature extraction result, a second feature extraction result and a third feature extraction result.

The controller may input the first feature extraction result, the second feature extraction result, and the third feature extraction result to the multi-head self-attention structure according to a preset input manner, so as to extract a relationship feature between the input features. After obtaining the output characteristics with relation characteristics output by the multi-head self-attention structure, the controller can further extract deep characteristics of the output characteristics by utilizing the multi-layer perceptron mechanism, and input the deep characteristics obtained by extraction into the long-short-period memory structure so as to extract time sequence characteristics among multi-frame deep characteristics.

The prediction model structure of the scheme can realize multi-dimensional feature extraction of the motion features, the dangerous obstacle features and the navigation map features of the vehicle, and is beneficial to improving the robustness and the accuracy of the model.

Based on the above scheme, optionally, the multi-head self-attention structure may include three inputs; wherein the first input and the second input are each determined based on a first feature extraction result of the first input branch, a second feature extraction result of the second input branch, and a third feature extraction result of the third input branch; the third input is determined based on the first feature extraction result of the first input branch.

Specifically, the controller may fuse the first feature extraction result, the second feature extraction result, and the third feature extraction result according to a preset fusion manner, and use the fused feature as the first input or the second input. Wherein the first input and the second input may be the same or different. The controller can directly splice the first feature extraction result, the second feature extraction result and the third feature extraction result according to a certain arrangement sequence to obtain the fusion feature. For example, the first feature extraction result, the second feature extraction result, and the third feature extraction result are represented by A, B and C, respectively, and the fusion feature may be represented as [ a, B, C ].

The controller may also select a part of the feature extraction results in the first feature extraction result, the second feature extraction result, and the third feature extraction result sequentially, and the different feature extraction results are spliced alternately to generate the fusion feature. For example, the fused feature may be represented as [ A1, B1, C1, A2, B2, C2], where A1 represents a first partial feature extraction of the first feature extraction and A2 represents a second partial feature extraction of the first feature extraction, and B1, C1, B2, and C2 are the same.

It will be appreciated that the controller may also take the first feature extraction result as a third input to determine an output feature having a relationship between the host vehicle motion feature and the navigation map feature and the hazard barrier feature based on the first input, the second input, and the third input. The calculation formula of the output characteristic can be expressed as:

wherein K represents a first input, V represents a second input, Q represents a third input, d _k Representing the dimension of the first input, softmax represents the normalized exponential function, and Attention (Q, K, V) represents the output characteristics of the multi-headed attentional mechanism.

The prediction model with the multi-head self-attention structure can improve the prediction efficiency and accuracy.

According to the scheme, the perception information of the vehicle is not directly input into the prediction model, the abstract features of the obstacle are extracted, the driving decision is judged, the prior features of the obstacle are determined according to the perception information, the prior features are used as the basis of the driving decision and are input into the prediction model, and the driving decision of the vehicle is obtained. Therefore, the scheme can improve the decision accuracy and the decision efficiency of the prediction model. According to the technical scheme, the prior characteristic of at least one obstacle is determined through the perception information of the vehicle; then, according to the motion state of the vehicle, determining the weight matched with at least two candidate areas based on the vehicle position division, and determining the sequencing result of the obstacles according to the weight of each candidate area, the prior characteristic of each obstacle and the position of each obstacle; and determining a preset number of dangerous obstacles in each obstacle according to the sequencing result of the obstacles, and taking the prior characteristics of each dangerous obstacle as the input of a prediction model to determine the planned path of the automatic driving vehicle. The method solves the problems of low accuracy, high time cost and the like caused by directly extracting the abstract features of the perception information by the prediction model, and can enhance the interpretability of the prediction model while improving the prediction accuracy of the prediction model and the generation efficiency of the model.

Example two

Fig. 3 is a flowchart of a method for planning a path of an automatic driving vehicle according to a second embodiment of the present invention, where the method is based on the above embodiment, and the calculation of the safety coefficient of a dynamic obstacle and the selection of a dangerous obstacle are refined. As shown in fig. 3, the method includes:

s210, determining the safety coefficient of the vehicle and at least one obstacle according to the perception information of the vehicle; the safety coefficient is the ratio of the current distance between the vehicle and the obstacle to the safety distance.

According to the perception information of the vehicle, the controller can calculate the safety coefficient of each obstacle in the vehicle and the perception range, wherein the safety coefficient can be used for evaluating the risk degree of the obstacle to the vehicle. Specifically, the safety coefficient may be represented by a ratio of a current distance between the host vehicle and the obstacle to a safety distance, where the safety distance may be calculated based on a speed of the host vehicle, a speed of the obstacle, and a response time of the host vehicle to a driving decision.

In the scheme, the controller can determine the safety distance between the vehicle and each obstacle according to the speed information of the vehicle, the speed information of the obstacle and the response time; and determining the safety coefficient of the vehicle and each obstacle according to the safety distance and the distance information of the vehicle and the obstacle. The safety factor may include a longitudinal safety factor and a transverse safety factor.

It can be appreciated that, since the driving directions of the dynamic obstacles are different in the driving scene, the controller can analyze the vehicle and each dynamic obstacle from both a transverse angle and a longitudinal angle. According to the information such as the speed of the vehicle, the acceleration of the vehicle, the speed of the dynamic obstacle, the acceleration of the dynamic obstacle, the response time and the like, the controller can calculate the transverse safety distance between the vehicle and each dynamic obstacle. According to the transverse safety distance, the controller can calculate the transverse safety coefficient matched with each dynamic obstacle. According to the information such as the speed of the vehicle, the acceleration of the vehicle, the speed of the dynamic obstacle, the acceleration of the dynamic obstacle, the response time and the like, the controller can also calculate the longitudinal safety distance between the vehicle and each dynamic obstacle. According to the longitudinal safety distance, the controller can calculate the longitudinal safety coefficient matched with each dynamic obstacle.

1. The determining of the longitudinal safety distance may include:

(1) And determining the relative movement direction of the vehicle and the dynamic obstacle according to the speed information of the vehicle and the speed information of the dynamic obstacle.

It is easy to understand that the speed information of the host vehicle may indicate the movement direction of the host vehicle, and the speed information of the dynamic obstacle may indicate the movement direction of the dynamic obstacle. For the longitudinal safety distance, the controller can judge the relative movement direction of the vehicle and the dynamic obstacle according to the running direction of the vehicle and the movement direction of the dynamic obstacle. Based on the transverse and longitudinal directions of the vehicle, the speed of the dynamic obstacle is decomposed, and the relative movement direction of the vehicle and the dynamic obstacle is judged according to the decomposition speed of the dynamic obstacle in the longitudinal direction of the vehicle. The controller can respectively set different longitudinal safety distance calculation modes according to the relative movement direction of the vehicle and the dynamic obstacle.

According to the scheme, the relative movement direction of the vehicle and the dynamic obstacles is determined according to the movement direction indicated by the speed information, and then the longitudinal safety distance between the vehicle and each dynamic obstacle is calculated according to different relative movement directions by utilizing the speed information of the vehicle, the speed information of the dynamic obstacles and the response time. The scheme realizes accurate longitudinal safety distance calculation, and is beneficial to ensuring the accuracy of safety coefficient calculation.

(2) And determining the longitudinal safety distance between the vehicle and each dynamic obstacle according to the relative movement direction, the speed information of the vehicle, the speed information of the dynamic obstacle and the response time.

It will be appreciated that the relative movement direction may include both co-directional and non-directional; the speed information may include speed, lateral speed information, and longitudinal speed information; wherein the longitudinal speed information may include a longitudinal speed, a longitudinal deceleration, and a longitudinal acceleration; the lateral velocity information may include a lateral velocity.

<1> if the relative movement direction of the host vehicle and the dynamic obstacle is the same direction, determining the longitudinal safety distance between the host vehicle and the dynamic obstacle according to the longitudinal speed of the preceding traffic object, the longitudinal speed of the following traffic object, the maximum longitudinal deceleration, the maximum longitudinal acceleration, the minimum longitudinal deceleration and the response time; wherein the maximum longitudinal deceleration, the maximum longitudinal acceleration, and the minimum longitudinal deceleration are determined based on statistics of traffic target attribute parameters.

If the relative movement direction of the vehicle and the dynamic obstacle is the same direction, the controller can calculate the longitudinal safety distance between the vehicle and the dynamic obstacle according to the longitudinal speed of the preceding traffic object, the longitudinal speed of the following traffic object, the maximum longitudinal deceleration, the maximum longitudinal acceleration, the minimum longitudinal deceleration and the response time. Wherein the preceding traffic objective is different from the following traffic objective, which may be the host vehicle or a dynamic obstacle, and similarly, the following traffic objective may be one of the host vehicle or the dynamic obstacle. The maximum longitudinal deceleration, the maximum longitudinal acceleration, and the minimum longitudinal deceleration are determined based on statistics of traffic target attribute parameters. The traffic target attribute parameters may include parameters such as a category, model, size, load range, acceleration range, deceleration range, and the like of the traffic target. The controller can acquire attribute parameters of traffic targets in the traffic environment, and respectively perform parameter statistics on the traffic targets of each class to obtain the maximum longitudinal deceleration, the maximum longitudinal acceleration and the minimum longitudinal deceleration of the traffic targets of each class. The controller may determine the maximum longitudinal deceleration, the maximum longitudinal acceleration, and the minimum longitudinal deceleration of the dynamic obstacle according to the maximum longitudinal deceleration, the maximum longitudinal acceleration, and the minimum longitudinal deceleration corresponding to the traffic target of the category to which the dynamic obstacle belongs.

In one particular approach, the controller may determine a maximum displacement of the subsequent traffic target within the response time based on the longitudinal speed, the maximum longitudinal acceleration, and the response time of the subsequent traffic target. The controller may determine a maximum braking distance between the preceding traffic target and the following traffic target within the response time based on the longitudinal speed of the following traffic target, the longitudinal speed of the preceding traffic target, the maximum longitudinal acceleration, the minimum longitudinal deceleration, the maximum longitudinal deceleration, and the response time. And adding the maximum displacement and the maximum braking distance of the subsequent traffic targets as the safety distance. Specifically, when the relative motion direction of the vehicle and the dynamic obstacle is in the same direction, the calculation formula of the longitudinal safety distance can be expressed as follows:

wherein v is _r Representing longitudinal speed, v, of a following traffic object _f Representing the longitudinal speed of a preceding traffic target, t representing the response time, a _{lon,max,accel} Indicating maximum longitudinal acceleration, a _{lon,min,brake} Representing the minimum longitudinal deceleration, a _{lon,max,brake} Indicating the maximum longitudinal deceleration.

<2> if the relative movement direction of the host vehicle and the dynamic obstacle is in the opposite direction, determining the longitudinal safety distance between the host vehicle and the dynamic obstacle according to the longitudinal speed of the host vehicle, the longitudinal speed of the dynamic obstacle, the minimum longitudinal deceleration, the maximum longitudinal acceleration, the minimum longitudinal deceleration of the host vehicle and the response time.

If the relative movement direction of the vehicle and the dynamic obstacle is different, the controller can determine the longitudinal safety distance between the vehicle and the dynamic obstacle according to the longitudinal speed of the vehicle, the longitudinal speed of the dynamic obstacle, the minimum longitudinal deceleration, the maximum longitudinal acceleration, the minimum longitudinal deceleration and the response time of the vehicle.

Taking the forward running of the vehicle and the reverse running of the dynamic obstacle as an example, the controller can determine the maximum speed of the vehicle in the response time according to the longitudinal speed, the maximum longitudinal acceleration and the response time of the vehicle. Similarly, the controller may determine a maximum velocity of the dynamic obstacle within the response time based on the longitudinal velocity, the maximum longitudinal acceleration, and the response time of the dynamic obstacle. The controller may calculate the maximum displacement of the host vehicle based on the host vehicle's longitudinal speed, the host vehicle's maximum speed within the response time, the host vehicle's minimum longitudinal deceleration, and the response time. Based on the longitudinal velocity of the dynamic obstacle, the maximum velocity of the dynamic obstacle within the response time, the minimum longitudinal deceleration, and the response time, the controller may calculate the maximum displacement of the dynamic obstacle. And adding the maximum displacement of the vehicle and the maximum displacement of the dynamic obstacle as a safety distance. When the vehicle runs reversely and the dynamic obstacle runs forward in the same way, and the relative movement direction of the vehicle and the dynamic obstacle is different, the calculation formula of the longitudinal safety distance can be expressed as follows:

v _1lon,t ＝v _1lon +ta _{lon,max,accel} ；

v _2lon,t ＝|v _2lon |+ta _{lon,max,accel} ；

Wherein v is _1lon Representing the speed, v, of a traffic target travelling in forward direction _2lon Representing the speed of a traffic target traveling in reverse, t representing the response time, a _{lon,max,accel} Indicating maximum longitudinal acceleration, a _{lon,min,brake} Representing the minimum longitudinal deceleration, a _{lon,min,brake,correct} Representing the minimum longitudinal deceleration of the vehicle.

2. The determination of the lateral safety distance may include:

(1) And determining the lateral safety distance between the vehicle and each dynamic obstacle according to the lateral speed of the vehicle, the lateral speed of the dynamic obstacle, the maximum lateral acceleration, the minimum lateral deceleration and the response time.

It will be appreciated that the lateral velocity information also includes lateral acceleration and lateral deceleration. For the lateral safety distance, the controller can determine the lateral safety distance between the vehicle and each dynamic obstacle according to the relative position between the vehicle and the dynamic obstacle. The controller may bring the two traffic targets close to each other with maximum lateral acceleration and brake with minimum lateral deceleration until lateral displacement of the motion stop as lateral safe distance.

In a specific scheme, the calculation formula of the lateral safety distance can be expressed as follows:

v _1lat,t ＝v _1lat +ta _{lat,max,accel} ；

v _2lat,t ＝v _2lat -ta _{lat,max,accel} ；

wherein μ represents a fault tolerant space, v _1lat Representing lateral speed, v, of traffic targets located on the left _2lat Represents the lateral speed of a traffic target located on the right, t represents the response time, a _{lat,max,accel} Indicating maximum lateral acceleration, a _{lat,min,brake} Representing the minimum lateral deceleration.

According to the scheme, the safety distances are calculated from the transverse angle and the longitudinal angle respectively, so that the accurate calculation of the safety coefficient is realized, and the driving safety of the vehicle is guaranteed to the greatest extent. In some embodiments, mu is 0.2m, and fault tolerance is increased on the basis of calculating the safety distance so as to ensure running safety.

After the transverse safety distance and the longitudinal safety distance between the vehicle and the dynamic obstacle are obtained, the transverse safety coefficient and the longitudinal safety coefficient can be calculated through the following calculation formulas:

wherein lon_dis represents the current longitudinal distance between the vehicle and the dynamic obstacle, d _min Representing the longitudinal safety distance, lon_safe_coeff representing the longitudinal safety factor; lat_dis represents the current lateral distance between the vehicle and the dynamic obstacle, d _min,lat Represents the lateral safety distance, lat_safe_coeff represents the lateral safety factor.

In this aspect, optionally, the a priori features further include a rate of change of the security loss; the safety loss change rate is determined based on the distance between the vehicle and the dynamic obstacle and the relative speed.

The security loss rate of change may include a lateral security loss rate of change and a longitudinal security loss rate of change. The transverse safety loss change rate can be determined according to the ratio of the current transverse relative speed of the vehicle and the dynamic obstacle to the current transverse distance, and the longitudinal safety loss change rate can be determined according to the ratio of the current longitudinal relative speed of the vehicle and the dynamic obstacle to the current longitudinal distance.

According to the scheme, the safety loss change rate can be calculated and added to the priori characteristics of the dynamic obstacle, so that the risk assessment of the dynamic obstacle from two dimensions of the safety loss and the safety loss change is facilitated.

S220, according to the safety coefficients, based on the inverted Gaussian model, determining the safety loss matched with each obstacle.

It will be appreciated that the controller of the autonomous vehicle may constrain the respective safety coefficients to be within the (0, 1) range based on inverting the gaussian model, resulting in a normalized safety loss for the respective obstacle matches.

Specifically, if the safety coefficient of the obstacle is smaller than the preset coefficient threshold value, determining the safety loss matched with the obstacle based on the inverted Gaussian model. According to the calculation method of the safety factor in S210, the greater the current distance is with respect to the safety distance, the greater the safety factor is, and the lower the risk of the obstacle is. If the safety coefficient is greater than or equal to a preset coefficient threshold value, the fact that the obstacle corresponding to the safety coefficient is not dangerous or has low risk degree is indicated for the vehicle. The controller can directly set the safety loss of the barrier without danger or with low danger degree to 0, and the safety loss calculation is not needed.

If the safety coefficient is smaller than the preset coefficient threshold, the controller can restrict the safety coefficient matched with each obstacle to be between (0 and 1) according to the inverted Gaussian model, so that the convergence speed of the prediction model is increased in the training process, and the driving decision accuracy of the prediction model is improved.

The controller may determine a longitudinal safety loss from the longitudinal safety factor and a lateral safety loss from the lateral safety factor based on the inverse gaussian model; determining the safety loss matched with the obstacle according to the longitudinal safety loss, the transverse safety loss and a predetermined weight coefficient; wherein the weight coefficient may be determined based on a security loss rate of change.

Specifically, the calculation formulas of the longitudinal safety loss and the transverse safety loss can be expressed as:

lon_safe_cost＝-1×[1-gaussian(K1-lon_safe_coeff)]；

lat_safe_cost＝-1×[1-gaussian(K2-lat_safe_coeff)]；

the method comprises the steps of determining a longitudinal safety loss of a vehicle, wherein the lon_safe_cost represents the longitudinal safety loss, the lon_safe_coeff represents the longitudinal safety coefficient, the lat_safe_cost represents the transverse safety loss, the lat_safe_coeff represents the transverse safety coefficient, the gaussian represents the gaussian distribution, K1 and K2 represent preset coefficient thresholds, and K1 and K2 can be the same or different.

The controller can determine the overall safety loss of the operation target according to the transverse safety loss, the longitudinal safety loss and the weight coefficient. Specifically, the calculation formula of the security loss can be expressed as:

safe_cost＝w_lon_safe×lon_safe_cost+w_lat_safe×lat_safe_cost；

Where w_lon_safe represents the weight coefficient of the longitudinal safety loss, and w_lat_safe represents the weight coefficient of the transverse safety loss.

The weight coefficient may be statistically derived from historical security loss data, for example, the weight coefficient of lateral security loss may be set to 0.54, and the weight coefficient of longitudinal security loss may be set to 0.46. The weight coefficient may also be determined based on a security loss rate of change. The controller can determine the change rate of the safety loss according to the current distance between the vehicle and the obstacle and the current speed. The security loss rate of change may include a lateral security loss rate of change and a longitudinal security loss rate of change. The transverse safety loss change rate may be determined according to a ratio of a current transverse speed to a current transverse distance between the vehicle and the obstacle, and the longitudinal safety loss change rate may be determined according to a ratio of a current longitudinal speed to a current longitudinal distance between the vehicle and the obstacle.

The calculation formulas of the longitudinal and lateral loss change rates can be expressed as:

wherein lon_speed represents the relative speed of the vehicle and the dynamic obstacle in the longitudinal direction, and lat_speed represents the relative speed of the vehicle and the dynamic obstacle in the transverse direction.

According to the scheme, the matched weight coefficients can be set for the transverse safety loss and the longitudinal safety loss so as to calculate the overall safety loss of the obstacle, and the accurate evaluation of the safety of the obstacle is facilitated.

S230, determining at least two candidate areas according to the position of the vehicle.

After the prior characteristics of each obstacle are obtained, the controller can divide candidate areas for the perception range of the vehicle according to the vehicle position. The region division may be different according to different driving scenarios. Fig. 4 is a schematic view of candidate region division in a simple driving scenario according to an embodiment of the present invention. As shown in fig. 4, in a narrow one-way driving scenario, the controller may divide 3 candidate areas, namely candidate area (1), candidate area (2) and candidate area (3), according to the current position of the host vehicle. The controller can also divide the perception range of the vehicle according to the current position of the vehicle in a gradient manner according to the distance, for example, the perception range is divided into 5 candidate areas, namely, a candidate area (1), a candidate area (2), a candidate area (3), a candidate area (4) and a candidate area (5).

Fig. 5 is a schematic diagram of candidate region division in a complex driving scenario according to an embodiment of the present invention. In a complex driving scenario such as traffic jam, the vehicle control apparatus may divide 9 candidate areas shown in fig. 5 with the candidate area (5) where the host vehicle is located as a center area, and the vehicle control apparatus may set the area size of each candidate area according to the driving scenario.

It should be noted that, the above candidate region division according to the complexity of the driving scenario is only one of the candidate region division modes, and the controller may also perform the candidate region division according to factors such as the lane where the host vehicle is located, the driving road segment of the host vehicle, and the like. The present embodiment does not limit the division manner of the candidate region. The areas of the candidate regions may be the same or different, and the shapes of the candidate regions may be regular or irregular.

S240, determining the weight of each candidate area according to the motion state of the vehicle and a preset weight distribution principle.

The movement state of the autonomous vehicle may include straight, u-turn, left-turn, right-turn, reverse, etc. The controller may set weights of the candidate regions to which the motion states match according to the degree of influence of each motion state on the candidate regions. Taking the candidate region division manner as shown in fig. 5 as an example, in the straight-running state of the vehicle, the affected candidate regions may include (3), (5), (6) and (9), where the degree of influence of the candidate regions (3), (5), (6) and (9) may be ordered as (5), (6), (3) and (9), and then the weights of the candidate regions (1) - (9) that are matched in the straight-running state may be respectively: 1. 1, 1.5, 1, 2, 1.8, 1 and 1.5. The controller can also gradient the weight of the target candidate region according to the sequence of the candidate regions affected by the motion state of the vehicle. Still taking the candidate region division manner as shown in fig. 5 as an example, when the host vehicle sequentially affects the candidate regions (5), (6), (3), (2) and (1) in the left turn state, the weights of the candidate regions (1) to (9) matched in the left turn state may be respectively: 1.2, 1.4, 1.6, 1, 2, 1.8, 1 and 1.

S250, determining the candidate area of each obstacle according to the position of each obstacle.

The controller may compare the positions of the obstacles with the ranges of the candidate regions to determine the candidate regions to which the obstacles belong.

S260, determining the sorting result of the obstacles according to the weight of each candidate area, the candidate area of each obstacle and the safety loss of each obstacle.

In one possible implementation, the determining the ranking result of the obstacles according to the weight of each candidate area, the candidate area of each obstacle, and the safety loss of each obstacle includes:

determining the weight of the candidate area of each obstacle, and determining the weighted safety loss according to the weight of the candidate area of each obstacle and the safety loss of each obstacle;

and determining the sequencing result of the barriers according to the weighted safety loss of each barrier.

The controller may determine the weight of each obstacle safety loss match according to the weight of each candidate region and the candidate region to which each obstacle belongs. Multiplying the safety loss of each obstacle by the matched weight, the controller can obtain the weighted safety loss of each obstacle. The weighted safety losses of the obstacles are ranked, and the controller can output the ranking result of the obstacles.

The scheme can determine the weighted safety loss of each obstacle, and is beneficial to realizing the reliability of obstacle risk degree evaluation.

S270, selecting a preset number of barriers as dangerous barriers according to the sequence of the weighted safety loss from large to small according to the sorting result.

The controller can determine the dangerous degree of each obstacle according to the weighted safety loss in the sequencing result of the obstacles, and select a preset number of obstacles with relatively high dangerous degree from each obstacle as dangerous obstacles.

According to the scheme, a certain number of barriers are selected from the barriers to serve as dangerous barriers, so that the driving decision timeliness is guaranteed, and meanwhile, the safety and reliability of the driving decision are improved.

S280, taking the prior characteristics of each dangerous obstacle as the input of a prediction model to determine the planned path of the automatic driving vehicle.

Example III

Fig. 6 is a schematic structural diagram of an automatic driving vehicle path planning device according to a third embodiment of the present invention. As shown in fig. 6, the apparatus includes:

a priori feature determining module 310, configured to determine a priori feature of at least one obstacle based on the perception information of the host vehicle;

the ranking result determining module 320 is configured to determine, according to a motion state of the vehicle, a weight that is matched based on at least two candidate areas of the vehicle location division, and determine a ranking result of the obstacles according to the weight of each candidate area, the prior feature of each obstacle, and the position of each obstacle;

the planned path determining module 330 is configured to determine a preset number of dangerous obstacles from among the obstacles according to the sorting result, and use a priori feature of each dangerous obstacle as an input of a prediction model to determine a planned path of the autonomous vehicle.

In this scenario, optionally, the a priori features include a security loss;

the a priori feature determination module 310 includes:

the safety coefficient determining unit is used for determining the safety coefficient of the vehicle and at least one obstacle according to the perception information of the vehicle; the safety coefficient is the ratio of the current distance between the vehicle and the obstacle to the safety distance;

And a security loss determination unit for determining security loss matching each obstacle based on the inverse gaussian model according to each security coefficient.

In one possible implementation, the ranking result determining module 320 includes:

a candidate region determining unit for determining at least two candidate regions according to the vehicle position;

the weight determining unit is used for determining the weight of each candidate area according to the motion state of the vehicle and a preset weight distribution principle.

On the basis of the above solution, optionally, the ranking result determining module 320 includes:

the affiliated area determining unit is used for determining affiliated candidate areas of the barriers according to the positions of the barriers;

and the sequencing result determining unit is used for determining the sequencing result of the obstacles according to the weight of each candidate area, the candidate area of each obstacle and the safety loss of each obstacle.

In this embodiment, optionally, the ranking result determining unit includes:

a weighted loss determination subunit, configured to determine a weight of a candidate area to which each obstacle belongs, and determine a weighted safety loss according to the weight of the candidate area to which each obstacle belongs and the safety loss of each obstacle;

And the sequencing result determining subunit is used for determining the sequencing result of the barriers according to the weighted safety loss of each barrier.

In a preferred embodiment, the planned path determining module 330 includes:

a dangerous obstacle determining unit, configured to determine a preset number of obstacles as dangerous obstacles in each obstacle according to the sorting result;

and the planning path determining unit is used for taking the prior characteristics of each dangerous obstacle as the input of a prediction model so as to determine the planning path of the automatic driving vehicle.

On the basis of the scheme, optionally, the prior characteristic further comprises a safety loss change rate; wherein the safety loss change rate is determined based on the distance between the host vehicle and the obstacle and the relative speed.

The automatic driving vehicle path planning device provided by the embodiment of the invention can execute the automatic driving vehicle path planning method provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

Example IV

Fig. 7 shows a schematic diagram of an electronic device 410 that may be used to implement an embodiment of the invention. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. Electronic equipment may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed herein.

As shown in fig. 7, the electronic device 410 includes at least one processor 411, and a memory, such as a Read Only Memory (ROM) 412, a Random Access Memory (RAM) 413, etc., communicatively connected to the at least one processor 411, wherein the memory stores a computer program executable by the at least one processor, and the processor 411 may perform various suitable actions and processes according to the computer program stored in the Read Only Memory (ROM) 412 or the computer program loaded from the storage unit 418 into the Random Access Memory (RAM) 413. In the RAM 413, various programs and data required for the operation of the electronic device 410 may also be stored. The processor 411, the ROM 412, and the RAM 413 are connected to each other through a bus 414. An input/output (I/O) interface 415 is also connected to bus 414.

Various components in the electronic device 410 are connected to the I/O interface 415, including: an input unit 416 such as a keyboard, a mouse, etc.; an output unit 417 such as various types of displays, speakers, and the like; a storage unit 418, such as a magnetic disk, optical disk, or the like; and a communication unit 419 such as a network card, modem, wireless communication transceiver, etc. The communication unit 419 allows the electronic device 410 to exchange information/data with other devices through a computer network such as the internet and/or various telecommunication networks.

The processor 411 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of processor 411 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various processors running machine learning model algorithms, digital Signal Processors (DSPs), and any suitable processor, controller, microcontroller, etc. The processor 411 performs the various methods and processes described above, such as an autonomous vehicle path planning method.

In some embodiments, the autonomous vehicle path planning method may be implemented as a computer program tangibly embodied on a computer-readable storage medium, such as storage unit 418. In some embodiments, some or all of the computer program may be loaded and/or installed onto the electronic device 410 via the ROM 412 and/or the communication unit 419. When the computer program is loaded into RAM 413 and executed by processor 411, one or more steps of the autonomous vehicle path planning method described above may be performed. Alternatively, in other embodiments, the processor 411 may be configured to perform the autonomous vehicle path planning method in any other suitable manner (e.g., by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.

A computer program for carrying out methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions/acts specified in the flowchart and/or block diagram block or blocks to be implemented. The computer program may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

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

To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) through which a user can provide input to the electronic device. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), blockchain networks, and the internet.

The computing system may include clients and servers. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical hosts and VPS service are overcome.

It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps described in the present invention may be performed in parallel, sequentially, or in a different order, so long as the desired results of the technical solution of the present invention are achieved, and the present invention is not limited herein.

The above embodiments do not limit the scope of the present invention. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of the present invention.

Claims

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

2. The method of claim 1, wherein the a priori features comprise a security loss;

the determining the prior feature of the at least one obstacle based on the perception information of the vehicle comprises:

according to the perception information of the vehicle, determining the safety coefficient of the vehicle and at least one obstacle; the safety coefficient is the ratio of the current distance between the vehicle and the obstacle to the safety distance;

based on the inverse gaussian model, a security loss matching each obstacle is determined according to each security coefficient.

3. The method of claim 1, wherein determining weights for matching at least two candidate regions based on the vehicle location split based on the motion state of the vehicle comprises:

determining at least two candidate areas according to the position of the vehicle;

and determining the weight of each candidate area according to the motion state of the vehicle and a preset weight distribution principle.

4. The method of claim 2, wherein determining the ranking of the obstacles based on the weights of the candidate regions, the a priori characteristics of the obstacles, and the positions of the obstacles comprises:

Determining the candidate area of each obstacle according to the position of each obstacle;

and determining the sequencing result of the obstacles according to the weight of each candidate region, the candidate region of each obstacle and the safety loss of each obstacle.

5. The method of claim 4, wherein determining the ranking result of the obstacles based on the weight of each candidate region, the candidate region to which each obstacle belongs, and the safety loss of each obstacle comprises:

6. The method of claim 1, wherein determining a preset number of dangerous obstacles from the ranking results, and taking a priori characteristics of each dangerous obstacle as an input to a predictive model to determine a planned path of the autonomous vehicle, comprises:

according to the sorting result, selecting a preset number of barriers as dangerous barriers according to the order of the weighted safety loss from large to small;

The prior characteristics of each dangerous obstacle are used as the input of a prediction model to determine the planned path of the automatic driving vehicle.

7. The method of claims 2 and 6, wherein the a priori features further comprise a rate of change of the security loss; wherein the safety loss change rate is determined based on the distance between the host vehicle and the obstacle and the relative speed.

8. An autonomous vehicle path planning apparatus, the apparatus comprising:

9. An electronic device, the electronic device comprising:

At least one processor; and

the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the autonomous vehicle path planning method of any of claims 1-7.

10. A computer readable storage medium storing computer instructions for causing a processor to perform the autonomous vehicle path planning method of any of claims 1-7.