Collision form prediction method, device and storage medium

Abstract

To provide a collision form prediction method, a device and a storage medium.SOLUTION: A collision form prediction method includes the steps of: acquiring ultrasonic sensing data on the basis of a plurality of ultrasonic radars mounted on a current vehicle; and inputting the ultrasonic sensing data respectively into pre-trained attenuation self-adjusting potential functions corresponding to respective collision forms, and determining a form of collision between a peripheral vehicle and the current vehicle. According to technical means of the present invention, detailed classification of collision forms is realized, and the corresponding attenuation self-adjusting potential functions are trained for different collision forms, whereby accuracy and robustness during collision form prediction are significantly improved, so that a subsequent actuation for a corresponding passive safety device can be easily performed, and protective effect and user experience are improved.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular, to a method, apparatus, and storage medium for predicting a collision form.

Background Art

[0002] Currently, for vehicle collision prediction, the probability and intensity of a collision are mainly determined by predicting the trajectories of surrounding vehicles. When the probability and intensity of a collision reach a certain threshold, all passive safety devices in the vehicle are activated to protect the people inside the vehicle.

[0003] However, since it is difficult to determine the collision form of a vehicle by trajectory prediction in the above method, there is a possibility of malfunction in some passive safety devices, which may affect the user experience. In view of this, the present invention is proposed.

Summary of the Invention

[0004] In order to solve the above technical problems, the present invention realizes a detailed classification of collision forms, and trains a corresponding attenuation self-adjusting potential function for different collision forms, thereby greatly improving the accuracy and robustness in predicting collision forms, facilitating the subsequent operation of corresponding passive safety devices, and providing a method, apparatus, and storage medium for predicting a collision form that enhance the protection effect and user experience.

[0005] An embodiment of the present invention provides a method for predicting a collision form. The method includes acquiring ultrasonic sensing data based on a plurality of ultrasonic radars currently installed in a vehicle; inputting the ultrasonic sensing data into pre-trained attenuation self-adjusting potential functions corresponding to respective collision forms to determine a form in which a surrounding vehicle collides with the current vehicle, wherein the attenuation self-adjusting potential functions corresponding to the respective collision forms are trained as follows. For each collision mode, acquire sample ultrasonic data corresponding to the collision mode. Based on the sample ultrasonic data, determine a first range corresponding to a first parameter of the initial self - adjustment potential function, and based on the accuracy and density of the ultrasonic radar, within the first range, determine a second range of the first parameter, and within the second range, determine the numerical value of the first parameter. Train the initial self - adjustment potential function with the sample ultrasonic data to obtain a damping self - adjustment potential function corresponding to the collision mode.

[0006] Embodiments of the present invention provide an electronic device, and the electronic device includes a processor and a memory. The processor is used to execute the steps of the collision mode prediction method described in any embodiment by calling a program or command stored in the memory.

[0007] Embodiments of the present invention provide a computer - readable storage medium. A program or command is stored in the computer - readable storage medium. The program or command causes a computer to execute the steps of the collision mode prediction method described in any embodiment.

Advantages of the Invention

[0008] Based on a plurality of ultrasonic radars currently installed in a vehicle, acquire ultrasonic sensing data, and input the ultrasonic sensing data into pre - trained damping self - adjustment potential functions corresponding to each collision mode respectively to determine the collision mode between the surrounding vehicle and the current vehicle. By using the damping self - adjustment potential function, it is realized to accurately predict in real - time the collision mode between the current vehicle and the surrounding vehicle under various complex collision conditions, solve the problem of malfunction of passive safety devices caused by too low setting of the threshold of collision probability and intensity, greatly improve the accuracy and robustness of the determination of the collision mode, and enhance the protection effect.

Brief Description of the Drawings

[0009] To more clearly explain the specific embodiments of the present invention or the technical means in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. The drawings in the following description are some embodiments of the present invention, and it is obvious that those skilled in the art can also obtain other drawings based on these drawings without creative labor.

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Embodiments for Carrying out the Invention

[0010] In order to more clearly illustrate the object, technical means, and advantages of the present invention, the technical means of the present invention will be described clearly and completely below. It is obvious that the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the present invention.

[0011] The collision form prediction method according to an embodiment of the present invention is mainly suitable for predicting the collision form with surrounding vehicles before the vehicle collides. The collision form prediction method according to an embodiment of the present invention can be executed by an electronic device.

[0012] FIG. 1 is a flowchart of the collision form prediction method according to an embodiment of the present invention. As shown in FIG. 1, the collision form prediction method specifically includes the following steps.

[0013] S110: Based on a plurality of ultrasonic radars mounted on the current vehicle, obtain ultrasonic sensing data.

[0014] Here, the current vehicle is the vehicle for predicting the collision form. The ultrasonic sensing data is the data collected by a plurality of ultrasonic radars, that is, the ultrasonic data obtained when detecting surrounding vehicles.

[0015] Specifically, a plurality of ultrasonic radars are mounted at a plurality of preset positions of the current vehicle, and ultrasonic sensing data, which is the data collected by these ultrasonic radars, is obtained.

[0016] Based on the above embodiment, as follows, ultrasonic sensing data can be obtained based on a plurality of ultrasonic radars mounted on the current vehicle. When the distance between the surrounding vehicle and the current vehicle is less than or equal to a second preset distance and greater than a first preset distance, predict the trajectory of the surrounding vehicle. When the distance between the surrounding vehicle and the current vehicle is less than or equal to a first preset distance, the trajectory prediction of the surrounding vehicle is stopped, and ultrasonic sensing data is acquired based on a plurality of ultrasonic radars mounted on the current vehicle.

[0017] Here, the surrounding vehicle is another vehicle around the current vehicle. The second preset distance is greater than the first preset distance. The second preset distance is a distance threshold for distinguishing between the case of performing the trajectory prediction of the surrounding vehicle and the case of not performing the trajectory prediction. The first preset distance is a distance threshold for distinguishing between the case of identifying the collision form of the surrounding vehicle and the case of not identifying the collision form.

[0018] Specifically, the distance between the surrounding vehicle around the current vehicle and the current vehicle is monitored. When the distance is less than or equal to the second preset distance and greater than the first preset distance, the trajectory prediction of the surrounding vehicle is performed to indicate that the surrounding vehicle has entered the trajectory prediction range of the current vehicle. When the distance between the surrounding vehicle and the current vehicle is less than or equal to the first preset distance, at this time, since it indicates that there is a possibility of collision between the surrounding vehicle and the current vehicle, the trajectory prediction of the surrounding vehicle is stopped, and it is necessary to identify the collision form of the surrounding vehicle. That is, in order to facilitate the subsequent identification of the collision form, an operation of acquiring ultrasonic sensing data is executed based on a plurality of ultrasonic radars mounted on the current vehicle.

[0019] Based on the above embodiments, it is possible to determine whether the distance between the surrounding vehicle and the current vehicle is less than or equal to the first preset distance as follows. When the ultrasonic signal intensity between the current vehicle and the surrounding vehicle is greater than a preset signal intensity, it is determined that the distance between the surrounding vehicle and the current vehicle is less than or equal to the first preset distance.

[0020] Here, the ultrasonic signal intensity is the intensity of the ultrasonic signal received by the ultrasonic radar and reflected by the surrounding vehicles. The preset signal intensity is the preset intensity of the ultrasonic signal for determining whether the distance between the surrounding vehicle and the current vehicle is less than or equal to a first preset distance, and can be calibrated according to the first preset distance.

[0021] Specifically, the intensity of the ultrasonic signal reflected by the surrounding vehicle is obtained as the ultrasonic signal intensity by each ultrasonic radar mounted on the current vehicle. When the ultrasonic signal intensity is greater than the preset signal intensity, it is determined that the distance between the surrounding vehicle and the current vehicle is less than or equal to the first preset distance.

[0022] Based on the above embodiments, as follows, when the distance between the surrounding vehicle and the current vehicle is less than or equal to a second preset distance and greater than the first preset distance, the trajectory prediction of the surrounding vehicle can be performed. A vehicle coordinate system is established based on the current vehicle. When the distance between the surrounding vehicle and the current vehicle is less than or equal to the second preset distance and greater than the first preset distance, the current number of trajectory predictions is determined, the surrounding vehicle coordinates of the surrounding vehicle are determined in the vehicle coordinate system, and the current number of trajectory predictions and the surrounding vehicle coordinates are input into a pre-trained trajectory prediction model to determine the predicted coordinate values.

[0023] Based on the predicted coordinate values, it is determined whether the coordinate difference between the predicted coordinate values and the origin of the vehicle coordinate system is less than a preset coordinate difference. If so, the operation of determining the surrounding vehicle coordinates of the surrounding vehicle in the vehicle coordinate system is stopped, and the operation of obtaining ultrasonic sensing data is executed based on a plurality of ultrasonic radars mounted on the current vehicle.

[0024] Here, the vehicle coordinate system has the positive direction of the horizontal axis as the direction from the rear to the front of the current vehicle. The horizontal axis is located on the left-right central symmetry plane of the current vehicle, and the origin of the horizontal axis is the point that protrudes the most on the left-right central symmetry plane of the current vehicle. The positive direction of the vertical axis is the direction from the left side to the right side of the current vehicle. The vertical axis is located on a horizontal plane that is perpendicular to the left-right central symmetry plane of the current vehicle and passes through the origin of the horizontal axis. The origin of the vertical axis is the same point as the origin of the horizontal axis, that is, the origin of the vehicle coordinate system. A schematic diagram of the vehicle coordinate system is shown in FIG. 2. The current number of trajectory predictions is defined as the number of times to perform current trajectory predictions for surrounding vehicles. It can be understood that the prediction of the position of a surrounding vehicle at the next prediction time is one trajectory prediction. Also, when the same surrounding vehicle enters the trajectory prediction range multiple times, it can be understood that zero-set processing of the current number of trajectory predictions is performed each time it enters. The surrounding vehicle coordinates are the coordinate values of the surrounding vehicles in the vehicle coordinate system. The trajectory prediction model is pre-trained and used to predict the coordinates of the surrounding vehicles in the vehicle coordinate system at the next prediction time. The predicted coordinate value is the coordinate value of the surrounding vehicle in the vehicle coordinate system at the next prediction time predicted by the trajectory prediction model. The preset coordinate difference is a coordinate difference threshold for determining whether a collision occurs between the surrounding vehicle and the current vehicle, that is, whether it is necessary to end the trajectory prediction and enter the collision form identification. The ultrasonic sensing data is the data collected by each ultrasonic radar.

[0025] Specifically, a vehicle coordinate system is established for the current vehicle as follows. The horizontal axis points from the rear to the front of the current vehicle and is located on the left-right central symmetry plane of the current vehicle. The origin of the horizontal axis is the point that protrudes the most on the left-right central symmetry plane of the current vehicle. The vertical axis points from the left side to the right side of the current vehicle and is located on a horizontal plane that is perpendicular to the left-right central symmetry plane of the current vehicle and passes through the origin of the horizontal axis. The origin of the vertical axis is the same point as the origin of the horizontal axis. When the distance between the surrounding vehicle and the current vehicle is less than or equal to a second preset distance and greater than a first preset distance, the number of current trajectory predictions for predicting the trajectory of the current surrounding vehicle is determined, and the coordinates of the surrounding vehicle in the vehicle coordinate system are determined as the surrounding vehicle coordinates. Next, the number of current trajectory predictions and the surrounding vehicle coordinates are input into a pre-trained trajectory prediction model, and the output result of the model is determined as the predicted coordinate value of the surrounding vehicle. Further, the coordinate difference between the predicted coordinate value and the origin of the vehicle coordinate system is calculated, and it is determined whether the coordinate difference is smaller than a preset coordinate difference. If so, to indicate that the surrounding vehicle needs to end the trajectory prediction process and enter the collision form identification process, the operation of determining the surrounding vehicle coordinates of the surrounding vehicle in the vehicle coordinate system is stopped, and the operation of acquiring ultrasonic sensing data based on a plurality of ultrasonic radars mounted on the current vehicle is started.

[0026] Based on the above embodiments, after determining whether the coordinate difference between the predicted coordinate value and the origin of the vehicle coordinate system is smaller than a preset coordinate difference, for the case where the coordinate difference is greater than or equal to the preset coordinate difference, the following can be executed. If so, an actual coordinate value having a time correspondence relationship with the predicted coordinate value is acquired. When the difference between the predicted coordinate value and the actual coordinate value is within a preset error range, it is determined whether the distance between the surrounding vehicle and the current vehicle is greater than a second preset distance.

[0027] If so, the operation of determining the surrounding vehicle coordinates of the surrounding vehicle in the vehicle coordinate system is stopped. Otherwise, update the current trajectory prediction count and the surrounding vehicle coordinates, input the current trajectory prediction count and the surrounding vehicle coordinates into a pre-trained trajectory prediction model, and return to the operation of determining the predicted coordinate values for execution.

[0028] Here, the actual coordinate value is the coordinate value of the surrounding vehicle having a time correspondence relationship with the predicted coordinate value in the vehicle coordinate system. For example, when the predicted coordinate value is the coordinate value in the vehicle coordinate system corresponding to the time point T obtained by prediction, it is the actual coordinate value of the surrounding vehicle in the vehicle coordinate system at the time point T. The preset error range is a preset error range for determining whether the trajectory prediction is accurate.

[0029] Specifically, when the coordinate difference between the predicted coordinate value and the origin of the vehicle coordinate system is greater than or equal to the preset coordinate difference, it is necessary to determine whether the error of the trajectory prediction meets the requirements, that is, obtain the actual coordinate value having a time correspondence relationship with the predicted coordinate value, and determine whether the difference between the predicted coordinate value and the actual coordinate value is within the preset error range. If it is within the preset error range, it indicates that the trajectory prediction is accurate. Furthermore, it is necessary to determine whether the distance between the surrounding vehicle and the current vehicle is greater than a second preset distance, that is, whether the surrounding vehicle is away from the current vehicle and outside the range of the trajectory prediction. When the distance between the surrounding vehicle and the current vehicle is greater than the second preset distance, at this time, since the surrounding vehicle is away from the current vehicle and outside the range of the trajectory prediction, it is necessary to stop the trajectory prediction of the surrounding vehicle, that is, stop the operation of determining the surrounding vehicle coordinates of the surrounding vehicle in the vehicle coordinate system. When the distance between the surrounding vehicle and the current vehicle is not greater than the second preset distance, it indicates that the surrounding vehicle is still within the range of the trajectory prediction of the current vehicle. In order to perform the next trajectory prediction of the surrounding vehicle, it is necessary to update the current trajectory prediction count, that is, add 1 to the current trajectory prediction count, update the surrounding vehicle coordinates as well, input the current trajectory prediction count and the surrounding vehicle coordinates into a pre-trained trajectory prediction model, and return to the operation of determining the predicted coordinate values for execution.

[0030] Based on the above examples, after obtaining the actual coordinate values having the predicted coordinate values and the time correspondence relationship, if the difference between the predicted coordinate values and the actual coordinate values is not within a preset error range, it can be executed as follows. When the difference between the predicted coordinate values and the actual coordinate values is not within a preset error range, it is determined whether the distance between the surrounding vehicle and the current vehicle is less than or equal to a first preset distance.

[0031] If so, the operation of determining the surrounding vehicle coordinates of the surrounding vehicle in the vehicle coordinate system is stopped, and the operation of acquiring ultrasonic sensing data based on a plurality of ultrasonic radars mounted on the current vehicle is executed. If not, based on the predicted coordinate values and the actual coordinate values, the training parameters corresponding to the current trajectory prediction times in the trajectory prediction model are updated, the current trajectory prediction times and the surrounding vehicle coordinates are updated, the current trajectory prediction times and the surrounding vehicle coordinates are input into the pre-trained trajectory prediction model, and the operation returns to determining the predicted coordinate values and is executed.

[0032] Here, the training parameters are the parameters corresponding to each prediction in the trajectory prediction model, and the corresponding training parameters are different each time a trajectory prediction is performed.

[0033] Specifically, when the difference between the predicted coordinate value and the actual coordinate value is not within a preset error range, it indicates that the prediction error corresponding to the current trajectory prediction count is too large, and it is necessary to adjust the trajectory prediction model. However, before adjustment, in order to ensure the safety of the vehicle, it is necessary to determine whether the distance between the surrounding vehicle and the current vehicle is less than or equal to a first preset distance, that is, whether there is a risk of collision, and whether the ultrasonic signal intensity between the current vehicle and the surrounding vehicle is greater than a preset signal intensity. If so, in order to indicate that there is a risk of collision with the current vehicle, the trajectory prediction process is terminated and the collision form identification process is entered. That is, the operation of determining the surrounding vehicle coordinates of the surrounding vehicle in the vehicle coordinate system is stopped, and the operation of acquiring ultrasonic sensing data based on a plurality of ultrasonic radars mounted on the current vehicle is executed. Otherwise, in order to indicate that there is no risk of collision with the current vehicle, the training parameters corresponding to the current trajectory prediction count can be updated, and the next trajectory prediction can be performed. That is, the training parameters corresponding to the current trajectory prediction count in the trajectory prediction model are updated, the current trajectory prediction count and the surrounding vehicle coordinates are updated, and the current trajectory prediction count and the surrounding vehicle coordinates are input into the pre-trained trajectory prediction model, and the operation returns to determining the predicted coordinate value and is executed.

[0034] Preferably, the trajectory prediction model is obtained by pre-training as follows. Based on the sample trajectory data, a sample data set is constructed. For each sample trajectory in the sample data set, the training parameters corresponding to each sample trajectory prediction count in the trajectory prediction model are trained with each sample trajectory prediction count corresponding to the sample trajectory and the sample actual coordinate value corresponding to each sample trajectory prediction count, and the trajectory prediction model is updated. When training the training parameters corresponding to each sample trajectory prediction count in the trajectory prediction model based on each sample trajectory of the sample data set, a pre-trained trajectory prediction model is determined.

[0035] Here, the sample trajectory data is the trajectory data within the trajectory prediction range of the sample surrounding vehicles with respect to the sample current vehicle, includes the sample actual coordinate values corresponding to each sample trajectory prediction count, and the time intervals between adjacent time points corresponding to the sample trajectory prediction counts are the same. The sample dataset includes each sample trajectory prediction count in each sample trajectory and the sample actual coordinate values corresponding to each sample trajectory prediction count.

[0036] Specifically, the sample dataset is constructed by each sample trajectory data obtained by pre - collection. Further, a trajectory prediction model is trained with each sample trajectory data. For each sample trajectory in the sample dataset, the training parameters corresponding to each sample trajectory prediction count in the trajectory prediction model are trained by each sample trajectory prediction count corresponding to the sample trajectory and the sample actual coordinate values corresponding to each sample trajectory prediction count, and the trained model is used as a new trajectory prediction model. When the training of the trajectory prediction model by each sample trajectory data is completed, the final trajectory prediction model is the pre - trained trajectory prediction model.

[0037] Exemplarily, the trajectory prediction model is [x i ,y i =φ(ω i [x i-1 ,y i-1 +b i ), i = 1, 2, 3, …. Here, i is the sample trajectory prediction count, x, y are coordinate values, Φ is an activation function to make the trajectory prediction model meet the non - linear fitting requirement, ω i and b iIt is the training parameter corresponding to the sample trajectory prediction count i. For each preset time interval Δt of each sample trajectory, the sample trajectory prediction count and the sample actual coordinate values are determined. Input the first group of sample actual coordinate values in the sample dataset into the trajectory prediction model, obtain the sample predicted coordinate values corresponding to the second group of sample actual coordinate values, and calculate the error between the second group of sample actual coordinate values and the sample predicted coordinate values. If the error meets the preset error requirement, save the training parameter of the first group as the training parameter corresponding to the sample trajectory prediction count 1. If the error does not meet the preset error requirement, update the training parameter of the first group. The update method is to construct an error function between the sample predicted coordinate values and the sample actual coordinate values, and update the training parameter of this group by the gradient method. By repeating the above process, the training parameter corresponding to each sample trajectory prediction count can be obtained, thereby obtaining a pre-trained trajectory prediction model.

[0038] S120: Input the ultrasonic sensing data into the pre-trained attenuation self-adjusting potential function corresponding to each collision form respectively to determine the form of collision between the surrounding vehicle and the current vehicle.

[0039] Here, the collision forms include, for surface-to-surface collisions, front-side collision, front-oblique side collision, front-front collision, and front-rear collision; for small overlapping surface-to-surface collisions, left front-side collision, right front-side collision, left front-left rear collision, and right front-left front collision; for surface-to-corner collisions, front-corner collision. The attenuation self-adjusting potential function corresponds to each collision form respectively, and the ultrasonic sensing data is used to determine whether it belongs to the corresponding collision form. The collision form is the collision form obtained by prediction that the current vehicle and the surrounding vehicle are about to collide.

[0040] Specifically, ultrasonic sensing data is respectively input into pre-trained attenuation self-adjusting potential functions corresponding to each collision form, and result values output by each attenuation self-adjusting potential function are obtained. If the output result value is greater than zero, it indicates that the ultrasonic sensing data of the group may belong to the collision form corresponding to the attenuation self-adjusting potential function. If the output result value is zero or less, it indicates that the ultrasonic sensing data of the group does not belong to the collision form corresponding to the attenuation self-adjusting potential function. When it is determined that there is only one possible collision form, the collision form is set as the form of collision between the surrounding vehicle and the current vehicle. When it is determined that there are at least two possible collision forms, the collision form corresponding to the maximum value among the output result values corresponding to the at least two collision forms is set as the form of collision between the surrounding vehicle and the current vehicle.

[0041] Here, the attenuation self-adjusting potential functions corresponding to each collision form are obtained by training as follows. For each collision form, sample ultrasonic data corresponding to the collision form is acquired. Based on the sample ultrasonic data, a first range corresponding to the first parameter of the initial self-adjusting potential function is determined. Based on the accuracy and density of the ultrasonic radar, within the first range, a second range of the first parameter is determined, and within the second range, the numerical value of the first parameter is determined; Based on the sample ultrasonic data, the initial self-adjusting potential function is trained to obtain an attenuation self-adjusting potential function corresponding to the collision form.

[0042] Here, the sample ultrasonic data is pre-stored ultrasonic data collected by the sample current vehicle before the collision. The first parameter is an adjustable parameter that controls the attenuation rate of the potential function. The first range is the larger range of the first parameter determined by the sample ultrasonic data. The second range is the smaller range of the first parameter determined by the accuracy and density of the ultrasonic radar, and the second range is smaller than the first range. The initial self-adjusting potential function is an untrained initial potential function.

[0043] Specifically, according to the collision form, each sample ultrasonic data is classified, and the sample ultrasonic data corresponding to each collision form is determined. Since it is necessary to train and obtain the corresponding attenuation self-adjustment potential function for each collision form, an arbitrary collision form will be used as an example for explanation. Based on the sample ultrasonic data, a first range corresponding to the first parameter of the initial self-adjustment potential function is determined. Further, according to the accuracy and density of the ultrasonic radar, a second range smaller than the range of the first range is determined, and the initial first parameter is selected therefrom as needed. The initial attenuation self-adjustment potential function is constructed by the first parameter, trained using the sample ultrasonic data, and the result obtained through training is the attenuation self-adjustment potential function corresponding to the collision form.

[0044] Exemplarily, the basic formula for selecting the potential function is as follows. JPEG2025081198000002.jpg24154

[0045] Here, K is the potential function, X is the input data of the potential function, X k is each sample ultrasonic data, and α is the first parameter for controlling the attenuation rate of the potential function. By adjusting the numerical value of α in combination with the complexity of the classification interface between each collision form, the accuracy of the sensor itself, and the arrangement density of the sensor according to the characteristics of the sample ultrasonic data, the attenuation rate of the potential function is controlled, the efficiency and accuracy of classification are improved, and the acquired data is analyzed in a very short time before the collision between the surrounding vehicles and the current vehicle, realizing accurate classification judgment.

[0046] In the training process, as the basis for judging that the training is accurate, μ 1 is a set consisting of sample ultrasonic data corresponding to a certain collision form, X t is the data of μ 1 n is the data volume of μ 1 that is, X t ∈μ 1 where t = 1, 2, …, n. Attenuation self - adjustment potential function JPEG2025081198000003.jpg16145 As can be seen from JPEG2025081198000004.jpg17170 Otherwise, the obtained attenuation self - adjustment potential function through training will be inaccurate. And, in order to make the attenuation rate of the potential function fast and the classification efficiency high, the absolute value of the first parameter α can be set to a larger value. If the classification is accurate, save the current initial self - adjustment potential function for subsequent classification calculations. When a classification error occurs, update the current initial self - adjustment potential function, and the formula is as follows.

[0047] JPEG2025081198000005.jpg17170

[0048] JPEG2025081198000006.jpg17170

[0049] Here, β is the second parameter of the initial self - adjustment potential function and can be determined by the training process. When updating the initial self - adjustment potential function, by further adjusting the absolute value of the first parameter α, the attenuation rate of the potential function can be changed, and the efficiency and accuracy of classification can be controlled until the classification is accurate.

[0050] Based on the above - mentioned embodiments, for each collision form, before obtaining the sample ultrasonic data corresponding to the collision form, it is also possible to further determine the collision form corresponding to each sample ultrasonic data. Specifically, it may be as follows.

[0051] For the sample ultrasonic data of each group in the collision ultrasonic data set, for each ultrasonic data of the sample ultrasonic data, determine each collision gradient of the connection line between each ultrasonic radar corresponding to each ultrasonic data and the vehicle body reflection point. Based on the preset gradient relationship corresponding to each ultrasonic data, each collision gradient, and each collision form, determine the initial form corresponding to the sample ultrasonic data of the group and the gradient reliability corresponding to the initial form. When the gradient reliability is equal to or greater than the preset reliability, set the initial form as the collision form corresponding to the sample ultrasonic data of the group. When the gradient reliability is less than the preset reliability, determine the collision form corresponding to the sample ultrasonic data of the group based on the voting method.

[0052] Here, the collision ultrasonic data set is a data set composed of pre-acquired sample ultrasonic data before collision. The collision gradient is the gradient of the connecting line between each ultrasonic radar corresponding to the sample ultrasonic data and the vehicle body reflection point. The preset gradient relationship is the preset correspondence between the ultrasonic data, the collision gradient, and the collision form. The initial form is the collision form determined by the preset gradient relationship. The gradient reliability is the reliability of the initial form determined by the preset gradient relationship. The preset reliability is a reliability threshold for determining whether the initial form is reliable.

[0053] Specifically, for the sample ultrasonic data of each group in the collision ultrasonic dataset, the ultrasonic data corresponding to each ultrasonic radar of the sample ultrasonic data of this group is used to determine each collision gradient of the connecting line between each ultrasonic radar and the vehicle body reflection point. According to the preset gradient relationship corresponding to each collision form, each collision gradient of the sample ultrasonic data of this group is analyzed together with each ultrasonic data to determine the initial form and the gradient reliability corresponding to the initial form. Furthermore, it is necessary to determine whether the initial form is reliable and compare the gradient reliability with the preset reliability. When the gradient reliability is equal to or higher than the preset reliability, since it indicates that the initial form is relatively reliable, the initial form can be set as the collision form corresponding to the sample ultrasonic data of this group. When the gradient reliability is less than the preset reliability, since it indicates that the initial form is not reliable, based on the voting method, the sample ultrasonic data of this group is classified to determine the collision form corresponding to the sample ultrasonic data of this group.

[0054] Exemplarily, FIG. 3 is a schematic diagram of the ultrasonic radar of the current vehicle, and 1 to 5 are five ultrasonic radars. FIG. 4 is a diagram showing a frontal and side collision, FIG. 5 is a diagram showing a frontal and diagonal side collision, FIG. 6 is a diagram showing a frontal and frontal collision, and FIG. 7 is a diagram showing a frontal and rear collision. As shown in FIGS. 4 to 7, the gradients of the connecting lines between all the ultrasonic radars and the vehicle body reflection points are the same and are determined as surface-to-surface collisions. However, since the corresponding detection distances of each ultrasonic radar are different, different specific forms in surface-to-surface collisions can be distinguished. FIG. 8 is a diagram showing a left front and side collision, FIG. 9 is a diagram showing a right front and side collision, FIG. 10 is a diagram showing a left front and left rear collision, and FIG. 11 is a diagram showing a right front and left front collision. As shown in FIG. 8, the ultrasonic data obtained by the ultrasonic radars 1 to 3 included in the collision surface is the vertical distance to the side of the surrounding vehicle. The gradients of the connecting lines between all the ultrasonic radars corresponding to the collision surface and the vehicle body reflection points are the same, but the ultrasonic radars 4 and 5 corresponding to the non-collision surface do not match, and both are the distances to the side rear corner of the vehicle body of the surrounding vehicle. Similarly, the preset gradient relationships corresponding to FIGS. 9 to 11 can be analyzed. FIG. 12 is a diagram showing a frontal and corner collision. As shown in FIG. 12, when the left front corner of the surrounding vehicle is within the detection range of the ultrasonic radar, the distances of all the ultrasonic radars are the distances to a certain corner. Due to the irregular shape of the vehicle body, the connecting lines between the ultrasonic radars and the vehicle body reflection points are not parallel, so a preset gradient relationship can be constructed.

[0055] Based on the above embodiments, after determining the collision form between the surrounding vehicle and the current vehicle, the passive safety device can be activated as follows. Determine the safety device response parameters according to the collision form, and adjust the passive safety device of the current vehicle according to the safety device response parameters.

[0056] Here, the safety device response parameter is an activation threshold parameter corresponding to the activation of passive safety devices such as the response threshold of the airbag. The passive safety device is a device that operates passively to protect the driver and passengers during a vehicle collision, such as airbags, side curtain airbags, and protection measures such as emergency braking and emergency avoidance operations.

[0057] Specifically, by constructing in advance the safety device response parameters corresponding to each collision mode according to different collision modes, differential processing of passive safety devices corresponding to different collision modes is realized. According to the collision mode, the safety device response parameters corresponding to the collision mode are determined, and the passive safety device of the current vehicle is adjusted according to the determined safety device response parameters, so that the passive safety device of the current vehicle can adapt to the collision of the collision mode that is about to occur.

[0058] Exemplarily, when entering the collision mode identification, a series of preparatory operations can be performed, such as sounding the collision warning sound of the current vehicle in advance to send a collision warning to the passengers, or reducing the response threshold of the airbag to shorten the response time of the airbag and seat belt pretensioners. After obtaining the collision mode by predicting the collision mode, further adaptive adjustment of the passive safety device can be performed.

[0059] Exemplarily, different safety device response parameter settings can be performed for passive safety devices in different positions according to different collision modes so as to adjust the passive safety device as desired.

[0060] In addition, when the distance between the surrounding vehicle and the current vehicle is less than the second preset distance and greater than the first preset distance, the trajectory prediction of the surrounding vehicle is performed to improve the efficiency and accuracy of the trajectory prediction. When the distance between the surrounding vehicle and the current vehicle is less than or equal to the first preset distance, the trajectory prediction of the surrounding vehicle is stopped, and ultrasonic sensing data is acquired based on a plurality of ultrasonic radars mounted on the current vehicle. The ultrasonic sensing data is respectively input into the pre-trained attenuation self-adjusting potential functions corresponding to each collision form to determine the form of collision between the surrounding vehicle and the current vehicle. By combining the trajectory prediction and the collision form prediction, it is possible to accurately predict the driving trajectory and the collision form of the surrounding vehicle in real time under various complex collision conditions, solve the problem of malfunction caused by setting the collision intensity threshold too low, and greatly improve the accuracy and robustness of the classification of the collision form. Furthermore, according to the form of collision, the safety device response parameter is determined, and the passive safety device of the current vehicle is adjusted by the safety device response parameter to reduce the possibility of early or delayed activation of the passive safety device.

[0061] This embodiment has the following technical effects. Ultrasonic sensing data is acquired based on a plurality of ultrasonic radars mounted on the current vehicle, and the ultrasonic sensing data is respectively input into the pre-trained attenuation self-adjusting potential functions corresponding to each collision form to determine the form of collision between the surrounding vehicle and the current vehicle. By using the attenuation self-adjusting potential function, it is possible to accurately predict the collision form between the current vehicle and the surrounding vehicle in real time under various complex collision conditions, solve the problem of malfunction of the passive safety device caused by setting the threshold of the collision probability and intensity too low, greatly improve the accuracy and robustness of the determination of the collision form, and enhance the protection effect.

[0062] FIG. 13 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. As shown in FIG. 13, the electronic device 400 includes one or more processors 401 and a memory 402.

[0063] The processor 401 may be a central processing unit (CPU) or other form of processing unit having data processing capabilities and / or command execution capabilities, and can control other components within the electronic device 400 to execute desired functions.

[0064] The memory 402 may include one or more computer program products, which may include various forms of computer-readable storage media such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program commands can be stored in the computer-readable storage medium. The processor 401 can execute the program commands to implement the collision form prediction method of any embodiment of the present invention described above and / or other desired functions. Various contents such as initial external parameters and thresholds can be further stored in the computer-readable storage medium.

[0065] In one example, the electronic device 400 may further include an input device 403 and an output device 404. These components are interconnected by a bus system and / or other form of connection mechanism (not shown). The input device 403 may include, for example, a keyboard, a mouse, etc. The output device 404 can output various information such as alarm prompt information and braking force to the outside. The output device 404 may include, for example, a display, a speaker, a printer, a communication network, and remote output devices connected thereto.

[0066] Of course, for simplicity, only some of the components related to the present invention in the electronic device 400 are shown in FIG. 13, and components such as buses and input / output interfaces are omitted. Furthermore, according to specific uses, the electronic device 400 may include any other appropriate components.

[0067] In addition to the methods and apparatuses described above, embodiments of the present invention may further be a computer program product including computer program commands that, when executed by a processor, cause the processor to execute the steps of the collision form prediction method according to any embodiment of the present invention.

[0068] The computer program product may create program code for executing the operations of the embodiments of the present invention by any combination of one or more programming languages. The programming languages include object-oriented programming languages such as Java and C++, and further include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed completely on a user's computer, partially executed on a user's computer, executed as a stand-alone software package, partially executed on a user's computer and partially executed on a remote computer, or executed completely on a remote computer or server.

[0069] Also, embodiments of the present invention may further be a computer-readable storage medium storing computer program commands that, when executed by a processor, cause the processor to execute the steps of the collision form prediction method according to any embodiment of the present invention.

[0070] The computer-readable storage medium can use any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (non-exhaustive listing) of the readable storage medium include an electrical connector having one or more conductors, a portable disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0071] It should be noted that the terms used in the present invention are used to describe specific embodiments and do not limit the scope of the present application. As shown in the specification of the present invention, unless the context clearly indicates otherwise, terms such as "one", "a", "a kind", and / or "the" not only refer to the singular form but can also include the plural form. The terms "comprising", "including", or any variation thereof are intended to have a non-exclusive inclusion, and a process, method, or device including a series of elements includes not only those elements but also other elements not expressly listed, or elements specific to those processes, methods, or devices. Without further limitation, the elements defined by the phrase "including one..." do not exclude the presence of other identical elements in the process, method, or device including the said elements.

[0072] In addition, the orientation or positional relationship indicated by terms such as "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for facilitating and simplifying the description of the present invention, and does not indicate or imply that the mentioned device or component must have a specific orientation and be configured and operate in a specific orientation. Therefore, it should not be understood as limiting the present invention. Unless there are clear regulations and limitations, the terms "mounting", "connecting", "coupling" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection or an integral connection, a mechanical connection or an electrical connection, a direct connection or an indirect connection via an intermediate medium, or a communication inside two components. A person skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific situation.

[0073] Finally, it should be noted that each of the above embodiments is only used to illustrate the technical means of the present invention and does not limit the present invention. Although the present invention has been described in detail with reference to each of the above embodiments, a person skilled in the art can modify the technical means described in each of the above embodiments or perform equivalent substitutions for some or all of the technical features. It should be understood that by these modifications and substitutions, the essence of the corresponding technical means does not deviate from the technical means of each embodiment of the present invention.

Claims

1. A collision mode prediction method, comprising: Acquiring ultrasonic sensing data based on a plurality of ultrasonic radars currently mounted on the vehicle; The ultrasonic sensing data is inputted into a pre-trained damping self-adjusting potential function corresponding to each collision type, respectively, to determine a collision type between the surrounding vehicle and the current vehicle; The damping self-adjusting potential function corresponding to each of the collision types is acquiring, for each crash type, sample ultrasound data corresponding to said crash type; determining a first range corresponding to a first parameter of an initial autoregulatory potential function according to the sample ultrasound data, determining a second range of the first parameter within the first range according to the accuracy and density of an ultrasound radar, and determining a value of the first parameter within the second range; training the initial self-adjusting potential function with the sample ultrasound data to obtain a damped self-adjusting potential function corresponding to the impact mode; This is obtained by practicing For each crash type, before performing the step of acquiring sample ultrasonic data corresponding to said crash type, further comprising: For each group of sample ultrasonic data of the collision ultrasonic data set, each collision gradient of a connecting line between each ultrasonic radar corresponding to each ultrasonic data and a vehicle body reflection point is determined by each ultrasonic data of the sample ultrasonic data; determining an initial form corresponding to the sample ultrasound data of the group and a gradient confidence corresponding to the initial form according to a preset gradient relationship corresponding to each ultrasound data, each impingement gradient, and each impingement form; If the gradient reliability is equal to or greater than a preset reliability, the initial configuration is set as a collision configuration corresponding to the sample ultrasound data of the group; and if the gradient confidence is less than the preset confidence, determining a collision type corresponding to the sample ultrasound data of the group based on a voting method.

2. The step of acquiring ultrasonic sensing data based on a plurality of ultrasonic radars currently mounted on the vehicle includes: A step of predicting a trajectory of the surrounding vehicle when the distance between the surrounding vehicle and the current vehicle is equal to or less than a second preset distance and is greater than a first preset distance, the second preset distance being greater than the first preset distance; The method according to claim 1, further comprising: when a distance between the surrounding vehicles and the current vehicle is equal to or less than the first preset distance, stopping trajectory prediction of the surrounding vehicles and acquiring ultrasonic sensing data based on a plurality of ultrasonic radars mounted on the current vehicle.

3. When the distance between the surrounding vehicle and the current vehicle is equal to or less than a second preset distance and is greater than a first preset distance, the step of predicting a trajectory of the surrounding vehicle includes: establishing a vehicle coordinate system based on a current vehicle; When the distance between the surrounding vehicle and the current vehicle is equal to or less than a second preset distance and greater than a first preset distance, determining a current trajectory prediction number, determining surrounding vehicle coordinates of the surrounding vehicle in the vehicle coordinate system, inputting the current trajectory prediction number and the surrounding vehicle coordinates into a pre-trained trajectory prediction model to determine predicted coordinate values; determining whether a coordinate difference between the predicted coordinate value and an origin of the vehicle coordinate system is smaller than a preset coordinate difference based on the predicted coordinate value; If so, the method of claim 2 further comprises: stopping the operation of determining the surrounding vehicle coordinates of the surrounding vehicles in the vehicle coordinate system, and performing the operation of acquiring ultrasonic sensing data based on a plurality of ultrasonic radars currently mounted on the vehicle.

4. After the step of determining whether or not a coordinate difference between the predicted coordinate value and the origin of the vehicle coordinate system is smaller than a preset coordinate difference, otherwise, obtaining an actual coordinate value having a time correspondence with the predicted coordinate value; determining whether a distance between the nearby vehicle and the current vehicle is greater than the second preset distance when a difference between the predicted coordinate value and the actual coordinate value is within a preset error range; if so, stopping the operation of determining peripheral vehicle coordinates of peripheral vehicles in the vehicle coordinate system; If not, updating the current trajectory prediction count and the surrounding vehicle coordinates, inputting the current trajectory prediction count and the surrounding vehicle coordinates into a pre-trained trajectory prediction model, and returning to and executing the operation of determining predicted coordinate values.

5. After the step of obtaining the actual coordinate values ​​having a time correspondence relationship with the predicted coordinate values, further determining whether a distance between the nearby vehicle and the current vehicle is equal to or less than the first preset distance when a difference between the predicted coordinate value and the actual coordinate value is not within a preset error range; If so, stopping the operation of determining the surrounding vehicle coordinates of the surrounding vehicles in the vehicle coordinate system and performing the operation of acquiring ultrasonic sensing data based on a plurality of ultrasonic radars mounted on the current vehicle; otherwise, updating a training parameter corresponding to the current trajectory prediction number in the trajectory prediction model based on the predicted coordinate value and the actual coordinate value, updating the current trajectory prediction number and the surrounding vehicle coordinates, inputting the current trajectory prediction number and the surrounding vehicle coordinates into a pre-trained trajectory prediction model, and returning to and executing the operation of determining a predicted coordinate value.

6. The trajectory prediction model is constructing a sample data set based on the sample trajectory data, the sample data set including each sample trajectory prediction number in each sample trajectory and a sample actual coordinate value corresponding to each sample trajectory prediction number; For each sample trajectory in the sample dataset, training a training parameter corresponding to each sample trajectory prediction number in a trajectory prediction model according to each sample trajectory prediction number corresponding to the sample trajectory and a sample actual coordinate value corresponding to each sample trajectory prediction number, and updating the trajectory prediction model; and determining a pre-trained trajectory prediction model when training training parameters corresponding to each sample trajectory prediction number in the trajectory prediction model based on each sample trajectory of the sample dataset.

7. After the step of determining a collision mode between the surrounding vehicle and the current vehicle, 2. The method of claim 1, further comprising the steps of: determining a safety device response parameter according to the crash mode; and adjusting passive safety devices of the current vehicle according to the safety device response parameter.

8. An electronic device comprising a processor and a memory, 8. An electronic device, wherein the processor executes the steps of the collision mode prediction method according to claim 1 by calling up a program or a command stored in the memory.

9. A computer-readable storage medium storing a program or commands for causing a computer to execute the steps of the collision mode prediction method according to any one of claims 1 to 7.

Images