Method and device for predicting collision probability, electronic equipment and medium

By constructing a joint envelope of the vehicle and obstacles and using the uncertainty information of trajectory prediction to calculate the collision probability, the problem of ignoring trajectory prediction uncertainty in existing technologies is solved, thereby improving the accuracy and flexibility of autonomous driving decision-making and planning.

CN118953344APending Publication Date: 2024-11-15BEIJING ZHIXINGZHE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310540506.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-13
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

Existing technologies ignore or partially ignore uncertainties in trajectory prediction, leading to inaccurate environmental perception and affecting the accuracy and reliability of autonomous driving decision-making and planning.

Method used

By constructing a joint envelope of the vehicle and obstacles, and utilizing the uncertainty information of trajectory prediction, the problem is transformed into a double integral calculation problem of collision probability, thus making full use of the uncertainty information for environmental cognition.

Benefits of technology

It improves the accuracy and flexibility of autonomous driving decision-making and planning, provides a more reasonable solution space, and enhances the safety of autonomous driving systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118953344A_ABST
    Figure CN118953344A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a method and device for predicting a collision probability, electronic equipment and a medium. The method for predicting the collision probability comprises the steps of obtaining a probability density function of position distribution of an obstacle at a prediction moment; determining an envelope body formed by all possible obstacle positions which have intersections with the target object; and determining an integral result of the probability density function on an integral domain by taking the envelope body as the integral domain, so as to predict the collision probability of the collision between the target object and the obstacle at the prediction moment. According to the method, a more accurate and reasonable solution space can be provided for a downstream decision planning module, and the flexibility of a downstream decision planning algorithm is effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of automatic control technology, and in particular to a method, device, electronic device and medium for predicting collision probability. Background Art

[0002] With the rapid development of artificial intelligence technology in recent years, the automotive industry is transforming towards electrification and intelligence, and the research and development of autonomous driving or assisted driving functions has become a key investment project for automakers and their suppliers.

[0003] Trajectory prediction of dynamic obstacles in the environment is an important part of autonomous vehicles' cognition of the dynamic environment around them. However, trajectory prediction is often subject to many uncertainties due to factors such as the mobility and speed of moving obstacles. How to reasonably utilize the uncertainty in trajectory prediction for environmental cognition is of great research significance.

[0004] In existing technical solutions, the uncertainty of trajectory prediction is usually ignored as a whole or partially and converted into a certain solution space for downstream decision-making planning solutions. This does not fully utilize the prediction uncertainty information and is prone to generating poor planning results or even planning failures. Summary of the invention

[0005] To solve the problems in the related art, the embodiments of the present application provide a method, device, electronic device and medium for predicting collision probability.

[0006] One aspect of an embodiment of the present application provides a method for predicting collision probability, including obtaining a probability density function of the position distribution of obstacles at a prediction time; determining an envelope formed by all possible obstacle positions that intersect with a target object; and using the envelope as an integral domain, determining an integral result of the probability density function on the integral domain, for predicting the collision probability of a target object colliding with an obstacle at the prediction time.

[0007] According to an embodiment of the present application, the target object is a vehicle. Determining the envelope formed by all possible obstacle positions that intersect with the target object includes obtaining a predicted position of the target object at a predicted time; obtaining a target object size and an obstacle size; and determining a closed envelope formed by all possible obstacle positions that intersect with the target object based on the target object size, the obstacle size, and the predicted position.

[0008] According to an embodiment of the present application, the method for predicting collision probability also includes obtaining a lane centerline; taking the projection position of the predicted position on the lane centerline as the origin, and taking the tangent and normal directions of the lane centerline as coordinate axes, establishing a Cartesian coordinate system as a lane coordinate system, wherein the probability density function is a probability density function mapped to the lane coordinate system.

[0009] According to an embodiment of the present application, determining a closed envelope formed by all possible obstacle positions that intersect with the target object includes approximating the target object and the obstacle as circles, respectively, to obtain a target circle with a first radius and an obstacle circle with a second radius; and determining a circular area with the center of the target circle as the center and the sum of the first radius and the second radius as the radius as the envelope.

[0010] According to an embodiment of the present application, determining a closed envelope formed by all possible obstacle positions that intersect with the target object includes approximating the target object and the obstacle as rectangles, respectively, to obtain a target rectangle and an obstacle rectangle; and determining an octagonal envelope based on the target rectangle and the obstacle rectangle.

[0011] According to an embodiment of the present application, the method for predicting collision probability further includes extending four alternate sides of the octagonal envelope to construct a rectangular envelope as an integration domain.

[0012] According to an embodiment of the present application, the four alternating sides in the octagonal envelope are extended to construct a rectangular envelope, including: determining two groups of four alternating sides from the octagonal envelope to construct a first rectangular envelope and a second rectangular envelope; selecting a rectangular envelope from the first rectangular envelope and the second rectangular envelope as an integration domain includes: selecting a larger rectangular envelope among the minimum Mahalanobis distances of the four vertices as the integration domain; or selecting a smaller variance of the Mahalanobis distance distribution of the four vertices as the integration domain.

[0013] According to an embodiment of the present application, an envelope is used as an integral domain, and an integral result of a probability density function on the integral domain is determined to predict the probability of a collision between a target object and an obstacle at a prediction time, including: taking the mean position of the obstacle as the origin and the directions of the two sides of the rectangular envelope as the coordinate axes, a Cartesian coordinate system is established as an obstacle coordinate system; the probability density function is mapped to the obstacle coordinate system by translation and rotation; the probability density function is scaled along the coordinate axis to standardize the covariance matrix of the two-dimensional Gaussian distribution; and the integral of the probability density function on the integral domain is converted into a linear combination of the values ​​of a standard two-dimensional Gaussian distribution function at the four vertices of the rectangular envelope.

[0014] According to the embodiment of the present application, the probability density function conforms to the mixed Gaussian model. Taking the envelope as the integral domain, determining the integral result of the probability density function on the integral domain, which is used to predict the collision probability of the target object and the obstacle at the prediction time, includes taking the envelope as the integral domain, determining multiple integral results of each modal component in the probability density function on the integral domain; weighting the multiple integral results according to the mixing coefficient of the mixed Gaussian model to determine the collision probability of the target object and the obstacle at the prediction time.

[0015] According to an embodiment of the present application, there are multiple obstacles. Taking the envelope as the integral domain, determining the integral result of the probability density function on the integral domain, which is used to predict the collision probability of the target object and the obstacle at the prediction time, includes taking the envelope as the integral domain, respectively determining the integral result of the probability density function corresponding to each obstacle on the integral domain; determining the maximum value of multiple integral results as the collision probability of the target object and the multiple obstacles at the prediction time.

[0016] According to an embodiment of the present application, the target object is a lane. Determining an envelope formed by all possible obstacle positions that intersect with the target object includes obtaining a lane width and an obstacle size; and determining a strip-shaped envelope formed by all possible obstacle positions that intersect with the target object based on the lane width and the obstacle size.

[0017] According to an embodiment of the present application, determining the integral result of the probability density function on the integral domain includes approximating the lane as two parallel lines; mapping the probability density function through an affine transformation so that the coordinate axis is parallel to the boundary of the integral domain; and converting the integral of the probability density function on the integral domain into a linear combination of two standard one-dimensional normal distribution functions in a direction perpendicular to the boundary to calculate the integral result.

[0018] Another aspect of the embodiments of the present application provides a device for predicting collision probability, including an acquisition module, a determination module, and a prediction module. The acquisition module is configured to obtain a probability density function of the position distribution of the obstacle at the prediction time; the determination module is configured to determine an envelope formed by all possible obstacle positions that intersect with the target object; and the prediction module is configured to use the envelope as an integral domain and determine the integral result of the probability density function on the integral domain, so as to predict the collision probability of the target object and the obstacle at the prediction time.

[0019] Another aspect of an embodiment of the present application provides an electronic device, comprising at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the processor implements the method described above.

[0020] Another aspect of an embodiment of the present application provides a computer-readable storage medium having computer-readable instructions stored thereon, which, when executed by a processor, enables the processor to implement the method described above.

[0021] Another aspect of an embodiment of the present application provides an intelligent driving vehicle, comprising the electronic device as described above.

[0022] Another aspect of an embodiment of the present application provides a computer program product, which, when executed by a processor, enables the processor to implement the method described above.

[0023] The technical solution of the embodiment of the present application proposes to use the joint envelope formed by the vehicle and the obstacle to represent the area formed by all possible collision positions. The collision probability is the probability that the obstacle falls within the joint envelope, and the collision probability is converted into a double integral calculation problem with the joint envelope as the integral domain. This method makes full use of the uncertainty information of trajectory prediction for environmental cognition, which can provide a more accurate and reasonable solution space for the downstream decision-making and planning module, and effectively improve the flexibility of the downstream decision-making and planning algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Other features, purposes and advantages of the present application will become more apparent through the following detailed description of non-limiting embodiments in conjunction with the accompanying drawings. In the accompanying drawings:

[0025] Figure 1 A schematic diagram schematically illustrates a system architecture of a method for predicting collision probability using an embodiment of the present application;

[0026] Figure 2 A flowchart schematically illustrates a method for predicting collision probability according to an embodiment of the present application;

[0027] Figure 3 Schematically showing a flow chart of determining an envelope according to an embodiment of the present application;

[0028] Figure 4 A flowchart for determining an envelope according to another embodiment of the present application is schematically shown;

[0029] Figure 5 A schematic diagram schematically showing a determination envelope of an embodiment of the present application;

[0030] Figure 6 A flowchart for determining an envelope according to another embodiment of the present application is schematically shown;

[0031] Figure 7 A schematic diagram schematically showing another embodiment of the present application for determining an envelope;

[0032] Figure 8 A schematic diagram schematically showing a target object and an obstacle in an embodiment of the present application;

[0033] Fig. 9 A schematic diagram schematically illustrates a rectangular envelope constructed in accordance with an embodiment of the present application;

[0034] Fig.10 A schematic diagram schematically shows a rectangular envelope constructed in another embodiment of the present application;

[0035] Fig.11 A flowchart schematically illustrating an embodiment of the present application for determining an integral result;

[0036] Fig.12 A flowchart for determining the collision probability according to an embodiment of the present application is schematically shown;

[0037] Fig.13 A flowchart for determining a collision probability according to another embodiment of the present application is schematically shown;

[0038] Fig.14 A flowchart for determining an envelope according to another embodiment of the present application is schematically shown;

[0039] Fig.15 A schematic diagram schematically showing a target object and an obstacle according to another embodiment of the present application;

[0040] Fig.16 A flowchart schematically illustrating another embodiment of the present application for determining an integral result;

[0041] Fig.17 A block diagram schematically illustrating an apparatus for predicting collision probability according to an embodiment of the present application; and

[0042] Fig.18 The structural diagram of a computer system suitable for implementing the method for predicting collision probability according to an embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0043] Hereinafter, exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings so that those skilled in the art can more easily implement them. In addition, for the sake of clarity, parts not related to the description of the exemplary embodiments are omitted in the accompanying drawings.

[0044] In the present application, it should be understood that terms such as "include" or "have" are intended to indicate the presence of features, numbers, steps, behaviors, components, parts or a combination thereof disclosed in the present specification, and are not intended to exclude the possibility that one or more other features, numbers, steps, behaviors, components, parts or a combination thereof exist or are added.

[0045] It should also be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0046] It should be noted that the acquisition or display of data in this application is authorized, confirmed, or actively selected by the user.

[0047] The embodiment of the present application provides a method for predicting collision probability, including obtaining a probability density function of the position distribution of obstacles at the prediction time; determining an envelope formed by all possible obstacle positions that intersect with the target object; and using the envelope as an integral domain, determining the integral result of the probability density function on the integral domain, which is used to predict the collision probability of the target object and the obstacle at the prediction time. This method makes full use of the uncertainty information of trajectory prediction for environmental cognition, can provide a more accurate and reasonable solution space for the downstream decision planning module, and effectively improve the flexibility of the downstream decision planning algorithm.

[0048] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0049] Figure 1 A schematic diagram of the system architecture of the method for predicting collision probability using an embodiment of the present application is schematically shown.

[0050] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0051] The terminal devices 101, 102, 103 interact with the server 105 through the network 104 to receive or send messages, etc. The terminal devices 101, 102, 103 can be hardware or software. When the terminal devices 101, 102, 103 are hardware, they can be various special or general electronic devices, including but not limited to vehicle-mounted terminal devices, smart phones, tablet computers, laptop portable computers and desktop computers, etc. When the terminal devices 101, 102, 103 are software, they can be installed in the electronic devices listed above. It can be implemented as multiple software or software modules (for example, multiple software or software modules used to provide distributed services), or it can be implemented as a single software or software module.

[0052] The server 105 may be a server that provides various services, such as a backend server that provides services for client applications installed on the terminal devices 101 , 102 , 103 .

[0053] The server 105 may be hardware or software. When the server 105 is hardware, it may be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When the server 105 is software, it may be implemented as multiple software or software modules (e.g., multiple software or software modules for providing distributed services), or as a single software or software module.

[0054] The method for predicting collision probability provided in the embodiment of the present application can be executed by the server 105, for example, or by the terminal devices 101, 102, 103. Alternatively, the method for predicting collision probability in the embodiment of the present application can be partially executed by the terminal devices 101, 102, 103, and the other part can be executed by the server 105.

[0055] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to the implementation requirements.

[0056] Figure 2 The flowchart of the method for predicting collision probability according to an embodiment of the present application is schematically shown.

[0057] like Figure 2 As shown, the method for predicting collision probability includes operations S210 - S230 .

[0058] In operation S210, a probability density function of the position distribution of obstacles at the prediction time is obtained.

[0059] In operation S220 , an envelope formed by all possible obstacle positions intersecting with the target object is determined.

[0060] In operation S230, the envelope is used as an integration domain, and an integration result of the probability density function on the integration domain is determined to predict a collision probability of the target object colliding with the obstacle at the prediction time.

[0061] According to an embodiment of the present application, when the decision planning samples a certain position, if the position is occupied by a stationary obstacle, the collision probability p=1, otherwise p=0, so the following embodiments of the present invention mainly discuss the case of dynamic obstacles.

[0062] When describing the uncertainty of an object's motion, it can be assumed that the estimated position coordinates conform to a Gaussian distribution. For an object moving on the ground, the direction distribution of the position at a certain moment forms a standard deviation ellipse. The predicted trajectory over a period of time forms multiple elliptical direction distributions. The final predicted trajectory is usually represented by discrete points of a fixed time step, and the mean μ of its central position at time t is t (x t,y t ), towards θ t and the x-direction variance Variance in y direction and position covariance cov(x,y) t The covariance matrix is:

[0063]

[0064] In operation S210, according to the predicted trajectory, the position of the moving obstacle at time t obeys a two-dimensional Gaussian distribution of N(X|μ,∑), and thus its probability density function can be determined.

[0065] According to an embodiment of the present application, the lane centerline can be obtained, and a Cartesian coordinate system is established as a lane coordinate system with the projection position of the predicted position on the lane centerline as the origin and the tangent and normal directions of the lane centerline as coordinate axes, wherein the probability density function is a probability density function mapped to the lane coordinate system.

[0066] For example, a Cartesian coordinate system can be established with the nearest point of the projection of the vehicle's position at any time on the lane centerline as the origin, the tangent direction of the lane centerline as the x-axis, and its normal direction as the y-axis, hereinafter referred to as the lane rectangular system or lane coordinate system.

[0067] In addition, a three-dimensional coordinate system can be established with time t as the vertical axis. Static obstacles in the environment are mapped to the lane rectangular system. Assuming that the obstacle is represented as a rectangle, it will be a cuboid in the three-dimensional coordinate system, and the height of the cuboid is the entire time domain. From a probability perspective, at a certain moment, the collision probability when the decision or plan samples the area where the static obstacle is located and all areas that collide with the cuboid is 1, otherwise it is 0.

[0068] According to the discrete trajectory prediction line of each moving obstacle (i.e., including the mean μ described above t (x t ,y t ), towards θ t and the x-direction variance Variance in y direction and position covariance cov(x,y) t And the covariance matrix ∑ t ), projected to the lane rectangular system. First, the mean absolute coordinates of each predicted point can be transformed into the lane rectangular system to obtain the position mean μ. The corresponding covariance matrix of each point also needs to be transformed by the rotation matrix to obtain a new covariance matrix ∑. Then the position of the moving obstacle at time t obeys Two-dimensional Gaussian distribution. Where X represents the random variable of the position and represents the two-dimensional point coordinates (x, y).

[0069] In the Frenet coordinate system, the probability distribution will be distorted and difficult to calculate. Compared with the Frenet coordinate system, the Cartesian coordinate system can easily convert the predicted probability distribution (usually expressed in the global coordinate system) to this coordinate system for easy calculation.

[0070] According to an embodiment of the present application, in operation S220, at a certain sampling position, if the positions of the target object and the obstacle conflict, there is a collision, otherwise there is no collision. The collision probability of the sampling position is the probability that the obstacle falls within the area of ​​all possible conflicts with the target object, and this area is called an envelope.

[0071] Finally, in operation S230, the envelope is used as the integration domain to determine the integration result of the probability density function in the integration domain, which can generally be used as the collision probability of the target object and the obstacle colliding at the prediction time.

[0072] The technical solution of the embodiment of the present application proposes to use the joint envelope formed by the vehicle and the obstacle to represent the area formed by all possible collision positions. The collision probability is the probability that the obstacle falls within the joint envelope, and the collision probability is converted into a double integral calculation problem with the joint envelope as the integral domain. This method makes full use of the uncertainty information of trajectory prediction for environmental cognition, which can provide a more accurate and reasonable solution space for the downstream decision-making and planning module, and effectively improve the flexibility of the downstream decision-making and planning algorithm.

[0073] According to the embodiment of the present application, the target object may be a vehicle, and the obstacle may be other vehicles or pedestrians, etc. The collision probability of the target object and the obstacle at the predicted time is predicted, that is, the probability of the own vehicle colliding with other vehicles or pedestrians is predicted. This data is of great significance for the autonomous driving function.

[0074] Figure 3 The flowchart of determining an envelope according to an embodiment of the present application is schematically shown.

[0075] like Figure 3 As shown, operation S220 may include operations S310 - S330 .

[0076] In operation S310, a predicted position of the target object at the predicted time is obtained.

[0077] In operation S320 , the target object size and the obstacle size are obtained.

[0078] In operation S330 , a closed envelope formed by all possible obstacle positions intersecting with the target object is determined based on the target object size, the obstacle size, and the predicted position.

[0079] According to an embodiment of the present application, when the envelopes of the ego vehicle and the obstacle intersect at a certain sampling position, there is a collision, otherwise there is no collision. The collision probability of the sampling position is the probability that the obstacle falls within all areas that may intersect with the ego vehicle.

[0080] In addition to using the true shapes of the target and the obstacle, you can also use an approximate shape of a simple geometric figure for at least one of the target and the obstacle. Figure 4-Figure 10 The various embodiments shown are described below.

[0081] Figure 4 The flowchart of determining an envelope according to another embodiment of the present application is schematically shown.

[0082] like Figure 4 As shown, operation S220 may include operations S410 and S420.

[0083] In operation S410, the target object and the obstacle are respectively approximated as circles to obtain a target circle with a first radius and an obstacle circle with a second radius.

[0084] In operation S420, a circular area having a center of the target circle as a center and a radius that is a sum of the first radius and the second radius as a radius is determined as an envelope.

[0085] Figure 5 A schematic diagram of determining an envelope according to an embodiment of the present application is schematically shown.

[0086] like Figure 5 As shown, at time t, the target position is K, and the shape of the target is approximately circular, that is, Figure 5 The small circle with K as the center is the target circle, and its radius is the first radius. At the same time, the obstacle mean position F, the obstacle shape can also be approximated as a circle, that is, Figure 5 The obstacle circle with F as the center has a radius of the second radius. The position of the obstacle circle is changed so that the obstacle circle overlaps the target circle at least partially. The position of F is constrained to Figure 5 Inside the large circle (i.e., the circular envelope) with K as the center, its radius is the sum of the first radius and the second radius.

[0087] According to an embodiment of the present application, in some scenarios where there are few obstacles and the distance is far, a rectangular outer envelope circle can be used to represent it. In this case, the joint envelope is a joint envelope circle composed of the sum of the radii of the two circles, which can be easily converted into a double integral calculation problem in a polar coordinate system.

[0088] Figure 6 The flowchart of determining an envelope according to another embodiment of the present application is schematically shown.

[0089] like Figure 6As shown, operation S220 may include operations S610 and S620.

[0090] In operation S610, the target object and the obstacle are respectively approximated as rectangles to obtain a target rectangle and an obstacle rectangle.

[0091] In operation S620, an octagonal envelope is determined based on the target rectangle and the obstacle rectangle.

[0092] According to the embodiment of the present application, both the target object and the obstacle can be simplified into a rectangle. The target object can be, for example, a vehicle represented as the self in the algorithm, hereinafter referred to as the self-vehicle. The pose sampled by the self-vehicle is [P ego (x ego ,y ego ),θ ego ], where θ ego represents the heading angle of the vehicle head in the global coordinate system xOy. Then, the area formed by all possible positions where obstacles and the vehicle intersect (the center position of obstacles) is the joint envelope of the octagon centered at the center position of the vehicle, as shown in Figure 7 shown.

[0093] Since the obstacle center position P obs (x obs ,y obs ) obeys a two-dimensional Gaussian distribution, so the collision probability of the ego-vehicle at this sampling position is the double integral of the two-dimensional Gaussian probability density function with the octagonal joint envelope as the integral domain. obs is the heading angle of the obstacle vehicle head in the global coordinate system xOy. Figure 8 As shown. Figure 8 In the equation, K represents the target position at time t, and F represents the center position of the obstacle.

[0094] The double integral with an octagonal integral domain is obviously not easy to calculate. In order to reduce the complexity and computational time, the embodiment of the present application can extend the four alternate sides of the octagonal envelope to construct a rectangular envelope as the integral domain, such as Figure 8 The rectangular area GHIJ is shown.

[0095] According to the embodiment of the present application, since the octagon has two sets of alternate sides, two rectangular envelopes can be constructed, such as Fig. 9 and Fig.10 As shown. For the two envelopes, one can be selected, or a better rectangular envelope can be selected according to certain selection criteria for the next step of calculation. That is, two groups of four alternating sides can be determined from the octagonal envelope to construct a first rectangular envelope and a second rectangular envelope, and a rectangular envelope can be selected from the first rectangular envelope and the second rectangular envelope as the integral domain.

[0096] The following is a method for determining the probability of selecting the one closer to the octagonal envelope integration domain from the two rectangular integration domains.

[0097] According to the embodiment of the present application, a larger rectangular envelope among the minimum Mahalanobis distances of the four vertices may be selected as the integration domain.

[0098] The Mahalanobis distance is the modulus of the vector Z′ obtained by affine transformation from the center of the distribution to a certain point into a standard Gaussian distribution. To some extent, it can also be regarded as the z-score in high dimensions. For multi-dimensional Gaussian distribution, the Mahalanobis distance calculation formula is:

[0099]

[0100] According to the embodiment of the present application, the Mahalanobis distance can be used to quickly evaluate the degree of increase in collision probability to a lesser extent. Take the rectangle with the larger average Mahalanobis distance of the four corner points as an approximation, so that the four vertices are farther away from the distribution center. Furthermore, the minimum value of the Mahalanobis distance of the four corner points can be used to replace the spliced ​​Mahalanobis distance, because the smallest corner point is closer to the distribution center, the probability density value of the two-dimensional Gaussian distribution is larger, and the impact on the final integral is greater. Therefore, the larger of the two minimum Mahalanobis distances should be used (the farther the corresponding rectangle is from the distribution center), the smaller the degree of increase in collision probability.

[0101] According to the embodiment of the present application, the four vertices with the smaller variance of the Mahalanobis distance distribution can also be selected as the integration domain.

[0102] The technical solution of the embodiment of the present application provides a judgment method for selecting the integral domain that is closer to the original octagon in the rectangular approximation by calculating the maximum and minimum values ​​of the Mahalanobis distance (Mahalanobisdistance) of the rectangle vertices and comprehensively evaluating the distribution, thereby simplifying the calculation of the double integral.

[0103] exist Figure 8 and Fig. 9 In the illustrated embodiment, the coordinates of each vertex can be calculated by the following formula:

[0104]

[0105] The length L of the rectangular area GHIJ rec and width W rec for

[0106] L rec =L+L′cos(|θ ego -θ obs |)+W′sin(|θ ego -θ obs |),

[0107] W rec =W+L′sin(|θ ego -θ obs |)+W′cos(|θ ego -θ obs |). (2)

[0108] Then the coordinates of the four vertices of rectangle GHIJ are

[0109]

[0110] Where R(α) is the rotation matrix

[0111]

[0112] Here we take α=-θ ego .

[0113] exist Fig.10 In the illustrated embodiment, the length L of G'H'I'J' rec and width W rec for:

[0114] L rec =L′+L cos(|θ ego -θ obs |)+W sin(|θ ego -θ obs |)

[0115] W rec =W′+L sin(|θ ego -θ obs |)+W cos(|θ ego -θ obs |) (5)

[0116] The calculation method of the four vertex coordinates of the rectangle G'H'I'J' is similar to formula (3), except that the rotation angle of the rotation matrix is ​​changed to α = -θ obs .

[0117] According to the embodiment of the present application, the obstacle position X(x obs ,y obs ) obeys a two-dimensional Gaussian distribution Its probability density function is

[0118]

[0119] At the sampling position of the vehicle [P ego (x ego ,y ego ),θ ego], the collision probability with the obstacle is the double integral of the above probability density function in the rectangle GHIJ (denoted as S):

[0120]

[0121] According to the technical solution of the embodiment of the present application, when the ego-vehicle and the obstacle are represented by rectangles, the joint envelope becomes an octagon, and the use of a rectangular approximate integral domain can simplify the probability calculation.

[0122] Fig.11 The flowchart of determining the integration result in an embodiment of the present application is schematically shown.

[0123] like Fig.11 As shown, operation S230 may include operations S1110 - S1040 .

[0124] In operation S1110 , a Cartesian coordinate system is established as an obstacle coordinate system with the obstacle mean position as the origin and the directions of the two sides of the rectangular envelope as coordinate axes.

[0125] In operation S1120, the probability density function is mapped to the obstacle coordinate system by means of translation and rotation.

[0126] In operation S1130, the probability density function is scaled along the coordinate axis to normalize the covariance matrix of the two-dimensional Gaussian distribution.

[0127] In operation S1140, the integral of the probability density function over the integral domain is converted into a linear combination of values ​​of a standard two-dimensional Gaussian distribution function at four vertices of the rectangular envelope.

[0128] The following describes the invention in conjunction with specific embodiments.

[0129] Please continue to refer to Figure 8 According to the embodiment of the present application, in order to facilitate the integral solution, according to the vector Translation and angle θ F Rotate to the obstacle mean position μ as the center, and use the directions of the two sides of the integration domain rectangle as the x-axis and y-axis coordinate systems, denoted as coordinate system F, as follows: Figure 8 As shown in x2Fy2. F It is the heading angle of the x-axis of the custom coordinate system F on the global coordinate system x1Oy1. After the coordinate conversion, the variable is recorded as X′=R(θ F )(X-μ), the covariance matrix of the obstacle is recorded as ∑′=R T (θ F )∑R(θ F ), then the probability density function becomes:

[0130]

[0131] The orthogonal property of the rotation matrix is ​​used here, and the determinants of orthogonal matrices are equal, so |∑′|=|∑|, where the covariance matrix is

[0132]

[0133] The correlation coefficient is

[0134]

[0135] Then scale along the coordinate axis X″=SX′, where the scaling matrix

[0136]

[0137] Get the covariance matrix of the standard two-dimensional Gaussian distribution

[0138]

[0139] The standard two-dimensional Gaussian probability density function is

[0140]

[0141] After the above translation, rotation and scaling changes (which can be unified into an affine transformation), the integral domain also changes accordingly, and the four vertices of the newly obtained rectangular integral domain S′ become G”H”I”J”:

[0142]

[0143] Since scaling the coordinate axis during integral calculation is equivalent to integral substitution, the collision probability becomes:

[0144]

[0145] Among them, H and J are two diagonal points, GH and JI in the final coordinate system are parallel to the x-axis of the x2Fy2 coordinate system, and GJ and HI are parallel to its y-axis. The x and y of point J are both the smallest, which is the lower limit of integration, and the x and y of point H are both the largest, which is the upper limit of integration.

[0146] The standard two-dimensional Gaussian distribution function at point B = (b1, b2) is

[0147]

[0148] The collision probability is finally converted into a linear combination of the values ​​of the standard two-dimensional Gaussian distribution function at different points:

[0149]

[0150] The calculation of the standard two-dimensional Gaussian distribution function can be conveniently solved by calling third-party solvers such as mvncdf. At this point, the collision probability of the vehicle with a moving obstacle at a certain sampling position can be approximately calculated.

[0151] According to the technical solution of the embodiment of the present application, the general two-dimensional Gaussian distribution is converted into an integral calculation problem under the standard two-dimensional Gaussian distribution through affine transformation, and the collision probability is decomposed into a linear combination of the values ​​of four standard two-dimensional Gaussian distribution functions at the four rectangular vertex positions, which greatly simplifies the calculation.

[0152] According to the embodiment of the present application, an obstacle may have multiple predicted trajectories, each predicted trajectory has a corresponding probability, and the sum of all trajectory probabilities is 1. This situation is called multimodal trajectory prediction. In the case of multimodal trajectory prediction, the probability density function conforms to the mixed Gaussian model.

[0153] For an obstacle i with a multimodal mixed Gaussian model predicting trajectory distribution, it is necessary to calculate the collision probability p for each mode of Gaussian distribution at a certain sampling point according to formula (17): k , because each mode has a different orientation at the same sampling point, resulting in different integration domains, so the collision probability must be calculated separately for the Gaussian distribution of different modes. Fig.12 The illustrated embodiment is described.

[0154] Fig.12 The flowchart of determining the collision probability according to an embodiment of the present application is schematically shown.

[0155] like Fig.12 As shown, the operation S230 may include operations S1210 and S1220.

[0156] In operation S1210, the envelope is used as an integral domain, and multiple integral results of each modal component in the probability density function on the integral domain are determined. According to an embodiment of the present application, each modal component of the probability density function conforms to a standard Gaussian distribution, and the integral results can be calculated for each modal component using the method described above.

[0157] For a random variable X, the Gaussian mixture model can be expressed as follows:

[0158]

[0159] In operation S1220, the plurality of integration results are weighted according to a mixing coefficient of the Gaussian mixture model to determine a collision probability of the target object colliding with the obstacle at the prediction time.

[0160] According to the embodiment of the present application, as described above, each predicted trajectory has a corresponding probability, and these probabilities are called the mixing coefficient π of the mixed Gaussian distribution. k (mixture coefficient), satisfying The final collision probability can be weighted summed using the mixing coefficient:

[0161]

[0162] According to the embodiment of the present application, there may be multiple obstacles. For example, when a vehicle is driving automatically, there may be more than two other vehicles or pedestrians around it. At this time, the autonomous vehicle should avoid collision with any moving obstacles. Fig.13 The illustrated embodiment is described.

[0163] Fig.13 A flowchart for determining a collision probability according to another embodiment of the present application is schematically shown.

[0164] like Fig.13 As shown, the operation S230 may include operations S1310 and S1320.

[0165] In operation S1310, the envelope is used as an integral domain, and the integral results of the probability density function corresponding to each obstacle on the integral domain are determined respectively. According to the embodiment of the present application, as long as the probability density function of each obstacle can be obtained, multiple integral results can be calculated respectively using the method described above, and each integral result corresponds to an obstacle.

[0166] In operation S1320, a maximum value of the plurality of integration results is determined as a collision probability of predicting that the target object collides with the plurality of obstacles at the prediction time.

[0167] Assume that the target position P at time t ego There are k obstacles nearby, and the collision probability of obstacle i (=1, 2, ..., k) is denoted as p i , we can determine that the maximum value of all probabilities is the total collision probability of the sampling points:

[0168]

[0169] Compared with calculating the probability of no collision with each obstacle, this method simplifies the calculation while maintaining a high reference value.

[0170] According to the embodiment of the present application, the target object may be a lane, and the prediction of the collision probability between the obstacle and the target object at a certain moment is converted into the prediction of the probability of a moving obstacle invading the lane at that moment, for example, a vehicle in the adjacent lane merges into the lane where the vehicle is located, causing an impact on the vehicle in the lane. Figure 14-16 The illustrated embodiment is described.

[0171] Fig.14 The flowchart of determining an envelope according to another embodiment of the present application is schematically shown.

[0172] like Fig.14 As shown, operation S220 may include operations S1410 and S1420.

[0173] In operation S1410 , a lane width and an obstacle size are obtained.

[0174] In operation S1420 , a strip-shaped envelope formed by all possible obstacle positions intersecting with the target object is determined based on the lane width and the obstacle size.

[0175] According to the embodiment of the present application, it can be assumed that the lanes near the mean position of the obstacles are straight lines, and the probability of an obstacle invading a lane means the probability that the obstacle falls on the lane and expands to at least half the width of the obstacle on the left and right sides based on the lane width.

[0176] Fig.15 A schematic diagram schematically illustrates a target object and an obstacle according to another embodiment of the present application.

[0177] like Fig.15 As shown, the global coordinate system is xOy, F is the mean position of the obstacle, θ obs is the heading angle of the obstacle vehicle head in the global coordinate system xOy. O1 is the closest projection point of F on the lane, the lane rectangular system x1O1y1, θ1 is the heading angle of the lane rectangular system x1 axis in the global coordinate system xOy. The distance from the obstacle mean position to the lane projection point is l obs , that is, O1 F. L obs and W obs are the length and width of the obstacle respectively.

[0178] Specifically, the expansion width ΔW on each side is

[0179]

[0180] Assuming that the lane width at the nearest projection point of the mean position of the obstacle on the lane reference line at time t is W, the width of the parallel lines formed by all possible positions of the obstacle invading the lane is W′=W+2ΔW.

[0181] The probability of an obstacle intruding the lane is the double integral of the two-dimensional Gaussian probability density function with the distance from the mean position of the obstacle [b2, b1] as the integral domain (also denoted as S here), which is similar to equation (6), where b1 = l obs +W′ / 2,b2=l obs -W′ / 2.

[0182] By approximating the lane near the obstacle as two parallel lines, the probability of a moving obstacle invading the lane is simplified to a two-dimensional Gaussian distribution with the expanded parallel lines as the integral interval, which is conducive to simplifying the calculation.

[0183] Fig.16 A flowchart for determining an integration result according to another embodiment of the present application is schematically shown.

[0184] like Fig.16 As shown, operation S230 may include operations S1610 - S1630 .

[0185] In operation S1610, the lane is approximated as two parallel lines.

[0186] In operation S1620, the probability density function is mapped through affine transformation so that the coordinate axis is parallel to the boundary of the integration domain.

[0187] In operation S1630, in a direction perpendicular to the boundary, an integral of the probability density function over the integral domain is converted into a linear combination of two standard one-dimensional normal distribution functions to calculate an integral result.

[0188] According to the embodiment of the present application, similar to the above method for calculating the collision probability, the obstacle probability density function is transformed by translation and rotation X′=R(θ obs -θ1)(X-μ) is mapped to a two-dimensional Gaussian distribution with the x-axis parallel to the boundary of the integral domain. The covariance matrix ∑′ is now in the form shown in equation (9) above, and the probability density function f X′ (x′, y′) becomes the form shown in equation (8) above. Since the translation and rotation transformation does not change the relative position relationship between the integral domain and the probability density function, the width of the road and the width of the parallel lines formed by all possible ranges of obstacles invading the lane will not change, and the probability of obstacles invading the lane becomes:

[0189]

[0190] Since the marginal probability of the two-dimensional Gaussian distribution obeys the one-dimensional normal distribution, the marginal probability distribution along the y-axis is Then the integral is reduced to a single integral:

[0191]

[0192] where Φ(·) is the standard one-dimensional normal distribution function.

[0193] At this point, at time t, the probability of a moving obstacle invading a lane is approximately calculated. This calculation method is suitable for multi-lane scenes with small curvature.

[0194] According to the technical solution of the embodiment of the present application, through translation and rotation changes and the definition of marginal probability density function, it is further converted into an interval probability calculation problem of standard one-dimensional normal distribution, thereby greatly simplifying the calculation of intrusion probability.

[0195] According to the embodiment of the present application, some rules and criteria can be formulated to further combine the probability with the direction of the mean position of the obstacle and the direction of multiple frames to determine semantic information such as whether the obstacle cuts in / out, merges / diverges or crosses the lane.

[0196] The joint envelope of the embodiment of the present application takes into account the position and shape of the ego-vehicle and the obstacle, so this collision probability calculation method is theoretically complete.

[0197] Based on the same inventive concept, the present application also provides a device for predicting collision probability, as shown below. Fig.17 The device for predicting collision probability according to an embodiment of the present application is described.

[0198] Fig.17 The block diagram of the device 1700 for predicting collision probability according to an embodiment of the present application is schematically shown. The device 1700 can be implemented as part or all of an electronic device through software, hardware, or a combination of both.

[0199] like Fig.17 As shown, the device 1700 for predicting collision probability includes an obtaining module 1710, a determining module 1720 and a predicting module 1730. The device 1700 for predicting collision probability can execute the various methods described above.

[0200] The obtaining module 1710 is configured to obtain a probability density function of the position distribution of obstacles at the prediction time.

[0201] The determination module 1720 is configured to determine an envelope formed by all possible obstacle positions that intersect with the target object.

[0202] The prediction module 1730 is configured to use the envelope as an integration domain, determine an integration result of the probability density function on the integration domain, and predict the collision probability of the target object colliding with the obstacle at the prediction time.

[0203] Fig.18The structural diagram of a computer system suitable for implementing the method for predicting collision probability according to an embodiment of the present application is schematically shown.

[0204] like Fig.18 As shown, the computer system 1800 includes a processing unit 1801, which can perform various processes in the above-mentioned embodiments according to a program stored in a read-only memory (ROM) 1802 or a program loaded from a storage part 1808 into a random access memory (RAM) 1803. In the RAM 1803, various programs and data required for the operation of the system 1800 are also stored. The processing unit 1801, the ROM 1802, and the RAM 1803 are connected to each other via a bus 1804. An input / output (I / O) interface 1805 is also connected to the bus 1804.

[0205] The following components are connected to the I / O interface 1805: an input section 1806 including a keyboard, a mouse, etc.; an output section 1807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage section 1808 including a hard disk, etc.; and a communication section 1809 including a network interface card such as a LAN card, a modem, etc. The communication section 1809 performs communication processing via a network such as the Internet. A drive 1810 is also connected to the I / O interface 1805 as needed. A removable medium 1811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1810 as needed, so that a computer program product read therefrom is installed into the storage section 1808 as needed. Among them, the processing unit 1801 can be implemented as a processing unit such as a CPU, a GPU, a TPU, an FPGA, an NPU, etc.

[0206] In particular, according to an embodiment of the present application, the method described above can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program tangibly included on a machine-readable medium, and the computer program includes a program code for executing the above method. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 1809, and / or installed from a removable medium 1811.

[0207] Another aspect of the present application also provides an intelligent driving vehicle, comprising the electronic device as described above.

[0208] The "intelligent driving vehicles" referred to in this application include but are not limited to vehicles with L0-L5 autonomous driving technology levels as established by the Society of Automotive Engineers International (SAE International) or the Chinese national standard "Automotive Driving Automation Classification".

[0209] In some embodiments, the intelligent driving vehicle may be a vehicle device or a robot device having the following functions:

[0210] (1) Passenger-carrying function, such as family cars and buses;

[0211] (2) Cargo-carrying functions, such as ordinary trucks, van trucks, trailer trucks, closed trucks, tank trucks, flatbed trucks, container trucks, dump trucks, trucks with special structures, etc.;

[0212] (3) Tool functions, such as logistics delivery vehicles, automated guided vehicles (AGVs), patrol cars, cranes, hoists, excavators, bulldozers, forklifts, road rollers, loaders, off-road engineering vehicles, armored engineering vehicles, sewage treatment vehicles, sanitation vehicles, vacuum trucks, floor scrubbers, water sprinklers, sweeping robots, food delivery robots, shopping guide robots, lawn mowers, golf carts, etc.;

[0213] (4) Entertainment functions, such as entertainment vehicles, amusement park self-driving devices, balance vehicles, etc.;

[0214] (5) Special rescue functions, such as fire trucks, ambulances, power repair vehicles, engineering rescue vehicles, etc.

[0215] With the rapid development of Advanced Driving Assistance System (ADAS) and unmanned driving technology, various sensors are installed on cars to sense the surrounding environment. For example, the vehicle's recognition of the external environment is completed by the perception module, which receives data from various sensors, processes the data through algorithms, and outputs the recognized surrounding vehicles, pedestrians, trees, and buildings. The perception module has various data sources, including cameras, lidar, millimeter-wave radar, and satellite navigation.

[0216] For example, in daily driving, cameras are used as sensors to capture two-dimensional images within a specific viewing angle. Then, autonomous driving is implemented / assisted based on image recognition technology. Among them, object detection is an important part of autonomous driving technology, which requires the ability to accurately detect all objects on the road to ensure driving safety.

[0217] For example, Laser Detecting and Ranging (Lidar) determines the distance by emitting and receiving laser beams, measuring the time difference and phase difference of laser signals, measuring the angle by horizontal rotation scanning, and establishing a two-dimensional polar coordinate system based on these two parameters, and then obtaining three-dimensional height information by obtaining signals at different pitch angles. High-frequency lasers can obtain a large amount of (e.g., 1.5 million) position point information (called point clouds) within one second, and perform three-dimensional modeling based on this information. Lidar has the advantages of high resolution, strong anti-active interference capability, good low-altitude detection performance, and light weight and portability.

[0218] When a device (intelligent driving vehicle with storage computing media) (mobile robot or vehicle) autonomously navigates in an indoor or outdoor environment, it recognizes the environment through various types of sensors. For example, the relevant technology can capture images through a camera, obtain point cloud data through a radar, and identify obstacles through image recognition technology and / or point cloud recognition technology. Then, the point cloud data, image data, and recognition results are projected onto the main layer as a map layer for path planning.

[0219] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0220] The units or modules involved in the embodiments described in this application may be implemented by software or by programmable hardware. The units or modules described may also be set in a processor, and the names of these units or modules do not constitute limitations on the units or modules themselves in some cases.

[0221] As another aspect, the present application further provides a computer-readable storage medium, which may be a computer-readable storage medium included in the electronic device or computer system in the above-mentioned embodiment; or a computer-readable storage medium that exists independently and is not assembled into the device. The computer-readable storage medium stores one or more programs, and the programs are used by one or more processors to execute the method of the embodiment of the present application.

[0222] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the present application.

Claims

1. A method for predicting collision probability, characterized in that: include: Obtain the probability density function of the obstacle's position distribution at the prediction time; Determine the envelope formed by all possible obstacle positions that intersect with the target object; as well as Taking the envelope as an integral domain, an integral result of the probability density function on the integral domain is determined to predict the collision probability of the target object colliding with the obstacle at the prediction time.

2. The method according to claim 1, characterized in that The target object is a vehicle, and the step of determining an envelope formed by all possible obstacle positions intersecting with the target object includes: Obtaining a predicted position of the target object at the predicted time; Get the target size and obstacle size; Based on the target object size, the obstacle size and the predicted position, a closed envelope formed by all possible obstacle positions that intersect with the target object is determined.

3. The method according to claim 2, characterized in that Also includes: Get the lane centerline; A Cartesian coordinate system is established as a lane coordinate system, with the projection position of the predicted position on the lane centerline as the origin and the tangent and normal directions of the lane centerline as coordinate axes. The probability density function is a probability density function mapped to the lane coordinate system.

4. The method according to claim 1, characterized in that: The step of determining a closed envelope formed by all possible obstacle positions intersecting with the target object includes: Approximating the target object and the obstacle to circles respectively, to obtain a target circle with a first radius and an obstacle circle with a second radius; A circular area with the center of the target circle as the center and the sum of the first radius and the second radius as the radius is determined as the envelope.

5. The method according to any one of claim 1, characterized in that: The step of determining a closed envelope formed by all possible obstacle positions intersecting with the target object includes: Approximating the target object and the obstacle into rectangles respectively, to obtain a target rectangle and an obstacle rectangle; An octagonal envelope is determined based on the target rectangle and the obstacle rectangle.

6. The method according to claim 5, characterized in that Also includes: The four alternate sides of the octagonal envelope are extended to construct a rectangular envelope as an integration domain.

7. The method according to claim 6, characterized in that The step of extending four alternate sides of the octagonal envelope to construct a rectangular envelope comprises: Determine two groups of four alternate sides from the octagonal envelope to construct a first rectangular envelope and a second rectangular envelope; Selecting a rectangular envelope from the first rectangular envelope and the second rectangular envelope as an integration domain includes: Select the larger rectangular envelope with the minimum Mahalanobis distance between the four vertices as the integration domain; or The four vertices with the smaller variance of the Mahalanobis distance distribution are selected as the integration domain.

8. The method according to claim 5 or 6, characterized in that: The step of taking the envelope as an integral domain and determining an integral result of the probability density function on the integral domain to predict the collision probability of the target object colliding with the obstacle at the prediction time includes: With the mean position of the obstacle as the origin and the directions of the two sides of the rectangular envelope as the coordinate axes, a Cartesian coordinate system is established as the obstacle coordinate system; Mapping the probability density function to the obstacle coordinate system by means of translation and rotation; Scale the probability density function along the coordinate axis to standardize the covariance matrix of the two-dimensional Gaussian distribution; The integral of the probability density function over the integral domain is converted into a linear combination of the values ​​of a standard two-dimensional Gaussian distribution function at the four vertices of the rectangular envelope.

9. The method according to any one of claims 1 to 7, characterized in that: The probability density function conforms to a mixed Gaussian model, and the envelope is used as an integral domain, and an integral result of the probability density function on the integral domain is determined to predict the collision probability of the target object colliding with the obstacle at the prediction time, including: Taking the envelope as an integral domain, determining a plurality of integral results of each modal component in the probability density function on the integral domain; The multiple integration results are weighted according to the mixing coefficient of the mixed Gaussian model to determine the collision probability of the target object and the obstacle colliding at the prediction time.

10. The method according to any one of claims 1 to 7, characterized in that: There are multiple obstacles, and the envelope is used as an integral domain to determine the integral result of the probability density function on the integral domain, so as to predict the collision probability between the target object and the obstacle at the prediction time, including: Taking the envelope as the integral domain, respectively determining the integral result of the probability density function corresponding to each obstacle on the integral domain; The maximum value of the plurality of integration results is determined as a prediction of the collision probability of the target object colliding with the plurality of obstacles at the prediction time.

11. The method according to claim 1, characterized in that: The target object is a lane, and the determining of an envelope formed by all possible obstacle positions intersecting with the target object includes: Get lane width and obstacle size; Based on the lane width and the obstacle size, a strip-shaped envelope formed by all possible obstacle positions intersecting with the target object is determined.

12. The method according to claim 11, characterized in that The step of determining an integral result of the probability density function over the integral domain comprises: approximating the lane as two parallel lines; Mapping the probability density function through affine transformation so that the coordinate axis is parallel to the boundary of the integral domain; In a direction perpendicular to the boundary, the integral of the probability density function over the integral domain is converted into a linear combination of two standard one-dimensional normal distribution functions to calculate an integral result.

13. A device for predicting collision probability, characterized in that: include: An acquisition module is configured to obtain a probability density function of the position distribution of obstacles at the prediction time; A determination module is configured to determine an envelope formed by all possible obstacle positions that intersect with the target object; as well as The prediction module is configured to use the envelope as an integration domain, determine an integration result of the probability density function on the integration domain, and predict the collision probability of the target object colliding with the obstacle at the prediction time.

14. An electronic device comprising: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 12.

15. A computer-readable storage medium having computer-readable instructions stored thereon, wherein when the instructions are executed by a processor, the processor is enabled to execute the method according to any one of claims 1 to 12.

16. An intelligent driving vehicle, comprising the electronic device as claimed in claim 14.