Trajectory prediction method and device, vehicle and storage medium

Abstract

The application discloses a trajectory prediction method and device, a vehicle and a readable storage medium. The method comprises the following steps: determining a prediction evaluation index result of each target object according to a respective perception trajectory and a prediction trajectory of each target object in a target frame, wherein the prediction trajectory of the target object in the target frame is determined by a prediction model; determining a correction index according to a feedforward compensation gain corresponding to each target object and the prediction evaluation index result; correcting the prediction model according to the correction index to obtain a corrected prediction model; and determining a respective prediction trajectory of each target object in a next frame of the target frame according to the corrected prediction model. In the application, the prediction model is corrected according to the correction index obtained from the target frame, and then the trajectory of the next frame of the target frame is predicted according to the corrected prediction model, so that the accuracy of the trajectory prediction is improved, and the driving risk is reduced.

Description

Technical Field

[0001] This application relates to the field of intelligent driving, and more specifically, to a trajectory prediction method, device, vehicle, and computer-readable storage medium. Background Technology

[0002] With the development of intelligent driving technology, the real-time prediction of one or more driving trajectories for objects around the vehicle is crucial for safe driving. Existing methods use physical models to generate future trajectories of objects based on historical data representing their physical actions.

[0003] However, the above method has low accuracy in predicting the trajectory of objects and poses a high risk to driving safety. Summary of the Invention

[0004] This application proposes a trajectory prediction method, apparatus, vehicle, and computer-readable storage medium to improve upon the aforementioned deficiencies.

[0005] In a first aspect, embodiments of this application provide a trajectory prediction method, the method comprising: determining a prediction evaluation index result for each target object based on the actual trajectory and predicted trajectory of each of multiple target objects in a target frame, wherein the predicted trajectory of the target object in the target frame is determined by a prediction model; determining a correction index based on the feedforward compensation gain corresponding to each target object and the prediction evaluation index result; correcting the prediction model based on the correction index to obtain a corrected prediction model; and determining the predicted trajectory of each of the multiple target objects in the next frame of the target frame based on the corrected prediction model.

[0006] Secondly, embodiments of this application also provide a trajectory prediction device, the device comprising:

[0007] The first determining module is used to determine the prediction evaluation index result of each target object based on the actual trajectory and predicted trajectory of each of the multiple target objects in the target frame. The predicted trajectory of the target object in the target frame is determined by the prediction model.

[0008] The second determination module is used to determine the correction index based on the feedforward compensation gain corresponding to each target object and the predicted evaluation index results.

[0009] The correction module is used to correct the prediction model based on the correction index to obtain the corrected prediction model.

[0010] The output module is used to determine the predicted trajectories of multiple target objects in the next frame after the target frame based on the corrected prediction model.

[0011] Thirdly, embodiments of this application also provide an electronic device, characterized in that the vehicle includes: one or more processors; a memory; one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to perform the methods described above.

[0012] Fourthly, embodiments of this application also provide a computer-readable storage medium storing processor-executable program code, which, when executed by the processor, causes the processor to perform the above-described method.

[0013] This application provides a trajectory prediction method, device, vehicle, and computer-readable storage medium. Based on the actual trajectories and predicted trajectories of multiple target objects in a target frame, the prediction evaluation index results for each target object are determined. The predicted trajectory of the target object in the target frame is determined by a prediction model. Then, based on the feedforward compensation gain and prediction evaluation index results corresponding to each target object, a correction index is determined, establishing a feedback mechanism between model prediction quality and model output. The correction index can accurately indicate the correction trend of the prediction model. Through the correction index, targeted corrections to the prediction model can be achieved, resulting in better prediction performance of the prediction model after correction, improving the accuracy of the prediction after correction, and thus reducing driving risks.

[0014] Meanwhile, compared to the method of exporting vehicle data and then having manual or other platforms obtain corrected indicators based on the vehicle data, obtaining corrected indicators directly from the vehicle improves efficiency.

[0015] Other features and advantages of the embodiments of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the embodiments of this application. The objects and other advantages of the embodiments of this application may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic diagram of a vehicle hardware environment applicable to embodiments of this application is shown.

[0018] Figure 2A flowchart of a trajectory prediction method according to an embodiment of this application is shown.

[0019] Figure 3 A flowchart of a trajectory prediction method according to yet another embodiment of this application is shown.

[0020] Figure 4 A flowchart of a trajectory prediction method according to yet another embodiment of this application is shown.

[0021] Figure 5 A schematic diagram illustrating the application process of a trajectory prediction method in an embodiment of this application is shown.

[0022] Figure 6 A structural block diagram of a trajectory prediction device according to an embodiment of this application is shown. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, and not all of them. The components of the embodiments of the present application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without inventive effort are within the scope of protection of the present application.

[0024] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0025] Please see Figure 1 , Figure 1 A schematic diagram of a vehicle hardware environment applicable to an embodiment of this application is shown. The vehicle 100 includes a trajectory prediction and evaluation system 110, a processor 113, and a memory 114.

[0026] The trajectory prediction and evaluation system 110 is used to predict the trajectory of a target object through a prediction model, and to evaluate the prediction model through the actual trajectory and the predicted trajectory. The trajectory prediction and evaluation system 110 can be an online prediction and evaluation system or an offline prediction and evaluation system. The trajectory prediction and evaluation system 110 includes a trajectory prediction module 111 and a model evaluation module 112. The trajectory prediction module 111 is used to predict the trajectory of the target object through a prediction model, and the model evaluation module 112 is used to evaluate the prediction model through the actual trajectory and the predicted trajectory.

[0027] The processor 113 may be a microcontroller unit (MCU) with a built-in memory 114 containing a program that can execute the contents of the following embodiments, and the processor 113 can execute the program stored in the memory 114.

[0028] The processor 113 may include one or more processors. The processor 113 connects to various parts of the vehicle 100 using various interfaces and lines, and performs various functions of the vehicle 100 and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 114, and calling data stored in the memory 114.

[0029] The memory 114 may include random access memory (RAM) or read-only memory (ROM). The memory 114 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 114 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as touch functionality, sound playback functionality, image playback functionality, etc.), and instructions for implementing the various method embodiments described below.

[0030] Please see Figure 2 , Figure 2 A flowchart of a trajectory prediction method according to an embodiment of this application is shown, for a vehicle, the method comprising:

[0031] S101. Based on the actual trajectory and predicted trajectory of each target object in the target frame, determine the prediction evaluation index result for each target object.

[0032] The predicted trajectory of the target object in the target frame is determined by a prediction model. The target object can be an object surrounding the vehicle, such as an obstacle vehicle or pedestrian. The length of the target frame can be set according to requirements, such as 5 seconds or 8 seconds.

[0033] The actual trajectory is the real driving trajectory of the target object perceived by the vehicle when it is driving in the target frame. The target frame can include multiple time points. At each time point, the vehicle perceives a trajectory point of the target object. Connecting the multiple trajectory points perceived in the target frame yields the actual trajectory.

[0034] The predicted trajectory is the driving trajectory of the target object predicted by the vehicle's prediction model. The target frame can include multiple time points. At each time point, the vehicle predicts a trajectory point of the target object. Connecting the multiple predicted trajectory points within the target frame yields the predicted trajectory.

[0035] The prediction evaluation index result for each target object is determined based on the gap between the actual trajectory and the predicted trajectory of each target object. The larger the gap, the lower the prediction evaluation index result, indicating a worse prediction effect of the prediction model. Conversely, the smaller the gap, the higher the prediction evaluation index result, indicating a better prediction effect of the prediction model.

[0036] In some implementations, the prediction evaluation index results corresponding to the target object include the average displacement error and the final displacement error between the actual trajectory and the predicted trajectory of the target object; S101 may include: determining the average displacement error and the final displacement error corresponding to each target object based on the predicted trajectory and the actual trajectory corresponding to each target object.

[0037] In some implementations, determining the average displacement error and final displacement error for each target object based on the predicted trajectory and the actual trajectory for each target object may include: for each target object, obtaining the predicted timestamp and predicted coordinates of the trajectory points included in the predicted trajectory of the target object; obtaining the actual timestamp and actual coordinates of the trajectory points included in the actual trajectory of the target object; aligning the actual timestamp and the predicted timestamp to obtain the trajectory points included in the actual trajectory for each of the predicted trajectories; and determining the average displacement error and final displacement error for the target object based on the predicted coordinates of all trajectory points and the actual coordinates of the trajectory points corresponding to the trajectory points.

[0038] The average displacement error (ADE) is the average of the errors between the predicted and actual coordinates of all trajectory points, while the final displacement error (FDE) is the error between the predicted and actual coordinates of the last trajectory point.

[0039] The formula for calculating the average displacement error is as follows:

[0040]

[0041] Where ADE is the average displacement error, x t (i) represents the x-coordinate of the i-th trajectory point in the actual coordinates, x p(i) represents the x-coordinate and y-coordinate of the predicted coordinates of the i-th trajectory point. t (i) represents the y-coordinate of the i-th trajectory point in the actual coordinates, y p (i) represents the y-coordinate of the predicted coordinates of the i-th trajectory point, n represents the total number of trajectory points, and i represents the i-th trajectory point.

[0042] The final displacement error is calculated as follows:

[0043]

[0044] Where ADE is the final displacement error, x t (―1) is the x-coordinate of the actual coordinates of the last trajectory point, x p (―1) represents the x-coordinate and y-coordinate of the predicted coordinates of the last trajectory point. t (―1) represents the y-coordinate of the last trajectory point in the actual coordinates. p (―1) is the y-coordinate in the predicted coordinates of the last trajectory point.

[0045] S102. Determine the correction index based on the feedforward compensation gain and prediction evaluation index results for each target object.

[0046] Among them, the feedforward compensation gain is the weight value of the prediction evaluation index. The feedforward compensation gain includes the weight values ​​corresponding to the average displacement error and the final displacement error, which can be set according to business needs.

[0047] The feedforward compensation gain can reflect the relative importance of the average displacement error and the final displacement error, so that the correction index determined by the feedforward compensation gain can accurately indicate the adjustment range of the prediction model, thus making the correction index more accurate.

[0048] S103. Correct the prediction model according to the correction index to obtain the corrected prediction model.

[0049] The prediction model can be any model capable of trajectory prediction. By inputting the actual data, map data, and location data of the previous frame of the target frame into the prediction model, the predicted trajectory of the target frame can be obtained. The prediction model can be a deep learning model.

[0050] In some implementations, correcting the prediction model based on the correction index can be achieved by adjusting the values ​​of the parameters in the prediction model according to the correction index. If the correction index is too large, the parameter values ​​are adjusted negatively; if the correction index is too small, the parameter values ​​are adjusted positively. The positive and negative directions of the correction index are consistent with the positive and negative directions of the parameters.

[0051] For example, if we define the predicted trajectory as being to the right of the actual trajectory as positive, the larger the correction index, the more to the right the predicted trajectory predicted by the prediction model will be. We need to adjust the values ​​of the prediction model parameters so that the predicted trajectory output by the corrected prediction model is more to the left than that of the uncorrected prediction model, that is, adjust the values ​​of the parameters in the negative direction.

[0052] S104. Determine the predicted trajectories of multiple target objects in the next frame of the target frame based on the corrected prediction model.

[0053] After obtaining the corrected prediction model, the actual data, map data, and positioning data of the target frame are input into the corrected prediction model to obtain the predicted trajectories of multiple target objects in the next frame after the target frame.

[0054] In this embodiment, based on the actual and predicted trajectories of multiple target objects in the target frame, the prediction evaluation index result for each target object is determined. The predicted trajectory of the target object in the target frame is determined by the prediction model. Then, based on the feedforward compensation gain and prediction evaluation index result corresponding to each target object, a correction index is determined, establishing a feedback mechanism between model prediction quality and model output. The feedback correction index can accurately indicate the correction trend of the prediction model. Through the correction index, targeted corrections to the prediction model can be achieved, resulting in better prediction performance of the prediction model after correction, improving the accuracy of the corrected prediction model, and thus reducing driving risks. Furthermore, compared to the method of exporting vehicle data and then manually or through other platforms obtaining correction indices from the vehicle data, obtaining correction indices directly from the vehicle improves efficiency.

[0055] Please see Figure 3 , Figure 3 A flowchart of a trajectory prediction method according to another embodiment of this application is shown, for a vehicle, the method comprising:

[0056] S201. Based on the actual trajectory and predicted trajectory of each target object in the target frame, determine the prediction evaluation index result for each target object.

[0057] The other descriptions of S201 are the same as those of S101 above, and will not be repeated here.

[0058] S202. Based on the feedforward compensation gain and prediction evaluation results for each target object, determine the intermediate results for each target object.

[0059] In some implementations, the prediction evaluation results corresponding to the target object include the average displacement error and the final displacement error between the actual trajectory and the predicted trajectory of the target object; S202 may include: constructing an error vector corresponding to each target object based on the average displacement error and the final displacement error of each target object; calculating the inner product of the error vector corresponding to each target object and the feedforward compensation gain, and determining the intermediate result corresponding to each target object based on the inner product.

[0060] The error vector of the i-th target object in the target frame is (ADE(i), FDE(i)), and the feedforward compensation gain is vector (K1, K2), where K1 is the weight value of ADE and K2 is the weight value of FDE.

[0061] The inner product, also known as the dot product or scalar product, is a binary operation that takes two vectors over a real number R and returns a real scalar. The formula for calculating the inner product of the error vector corresponding to the i-th target object and the feedforward compensation gain is:

[0062] S(i)=(ADE(i),FDE(i))·(K1,K2)=ADE(i)·K1+FDE(i)·K2.

[0063] In some implementations, calculating the inner product of the error vector and the feedforward compensation gain corresponding to each target object, and determining the intermediate result corresponding to each target object based on the inner product, may include: for each target object, calculating the inner product of the error vector and the feedforward compensation gain corresponding to the target object; if the predicted trajectory corresponding to the target object is on the first side of the actual trajectory corresponding to the target object, calculating the product of a first value and the inner product as the intermediate result corresponding to the target object; if the predicted trajectory corresponding to the target object is on the second side of the actual trajectory corresponding to the target object, calculating the product of a second value and the inner product as the intermediate result corresponding to the target object, wherein the first value and the second value are opposites of each other.

[0064] The first side can be either the right or left side, and the second side can be either the left or right side. The first and second sides can be defined according to requirements. The first and second values ​​can be 1 and -1, respectively.

[0065] Taking the first side as the right and the second side as the left as an example, when the predicted trajectory corresponding to the target object is to the right of the actual trajectory corresponding to the target object, the formula for calculating the intermediate result corresponding to the target object is:

[0066] P(i) = a·S(i),

[0067] Where i is the i-th target object, a is the first value, and S(i) is the inner product of the error vector corresponding to the i-th target object and the feedforward compensation gain.

[0068] When the predicted trajectory corresponding to the target object is to the left of the actual trajectory corresponding to the target object, the formula for calculating the intermediate result corresponding to the target object is:

[0069] P(i) = -a·S(i),

[0070] Where i is the i-th target object, -a is the second value, and S(i) is the inner product of the error vector corresponding to the i-th target object and the feedforward compensation gain.

[0071] S203. Accumulate the intermediate results corresponding to multiple target objects to obtain the correction index.

[0072] In some implementations, after obtaining intermediate results corresponding to multiple target objects in the target frame, these intermediate results can be transmitted to the vehicle's accumulator for accumulation to obtain the correction index. The accumulator is a temporary register used to store the intermediate results generated during calculation. The accumulator can store not only the correction index of the target frame but also the correction indices of multiple frames preceding the target frame.

[0073] If there is only one target object, then directly obtain the intermediate result corresponding to the target object as the correction indicator.

[0074] S204. The prediction model is modified according to the correction index to obtain the modified prediction model.

[0075] S205. Determine the predicted trajectories of multiple target objects in the next frame of the target frame based on the corrected prediction model.

[0076] The other descriptions of S204-S205 are the same as those of S103-S104 above, and will not be repeated here.

[0077] In this embodiment, based on the actual trajectories and predicted trajectories of multiple target objects in the target frame, the prediction evaluation index result for each target object is determined. Based on the feedforward compensation gain and prediction evaluation index result for each target object, intermediate results for each target object are determined. The intermediate results for multiple target objects are accumulated to obtain a correction index. The prediction model is then corrected based on the correction index to obtain a corrected prediction model. By combining the prediction evaluation results of multiple target objects in the target frame to obtain the correction index and further correcting the model, the accuracy of the model correction is improved. Finally, trajectory prediction is performed based on the corrected prediction model, thus improving the accuracy of trajectory prediction.

[0078] Please see Figure 4 , Figure 4 A flowchart of a trajectory prediction method according to another embodiment of this application is shown, for a vehicle, the method comprising:

[0079] S301. Based on the actual trajectory and predicted trajectory of each of the multiple target objects in the target frame, determine the prediction evaluation index result of each target object.

[0080] S302. Determine the correction index based on the feedforward compensation gain and prediction evaluation index results for each target object.

[0081] The other descriptions of S301-S302 are the same as those of S101-S102 above, and will not be repeated here.

[0082] S303. Determine the target threshold range in which the correction index is located.

[0083] In some implementations, target threshold intervals are divided by setting multiple target thresholds. If the correction index is greater than or equal to the first threshold P0 and less than or equal to the second threshold P1, then the target threshold interval is the first threshold interval. If the correction index is greater than the third threshold - P0 and less than or equal to the first threshold P0, then the target threshold interval is the second threshold interval. If the correction index is greater than the fourth threshold - P1 and less than or equal to the third threshold - P0, then the target threshold interval is the third threshold interval. If the correction index is less than the fourth threshold - P1 or greater than the second threshold P1, then the target threshold interval is the fourth threshold interval.

[0084] The values ​​of P0 and P1 can be set according to requirements, for example, P0 = 0.5 and P1 = 10.

[0085] S304. Obtain the target configuration file corresponding to the target threshold range.

[0086] Different target threshold ranges correspond to different configuration files, which include a first configuration file, a second configuration file, and a third configuration file. The first configuration file is an over-predicted configuration file, the second configuration file is the original configuration file, and the third configuration file is an under-predicted configuration file.

[0087] In some implementations, such as Figure 5 As shown, S304 may include: if the target threshold interval is a first threshold interval, obtaining a first configuration file as the target configuration file, the first configuration file being used to increase the prediction magnitude of the prediction model on the second side; if the target threshold interval is a second threshold interval, obtaining a second configuration file as the target configuration file, the second configuration file being used to maintain the values ​​of the configuration parameters of the prediction model; if the target threshold interval is a third threshold interval, obtaining a third configuration file as the target configuration file, the third configuration file being used to increase the prediction magnitude of the prediction model on the first side.

[0088] If the target threshold interval is the first threshold interval, it means that the predicted trajectory of the target frame predicted by the prediction model is located on the first side of the actual trajectory. Before starting the trajectory prediction of the next frame of the target frame, it is necessary to adjust the configuration parameters of the prediction model to the second side to increase the prediction amplitude of the prediction model on the second side, so that the predicted trajectory predicted by the corrected prediction model is shifted to the second side. Adjusting the configuration parameters of the prediction model to the second side only requires calling the prediction configuration file as the configuration file of the prediction model.

[0089] If the target threshold interval is the second threshold interval, it means that the predicted trajectory of the target frame predicted by the prediction model overlaps with the actual trajectory within the error allowable range. The current prediction model configuration parameters can still be used, and the original configuration file can be called as the prediction model configuration file.

[0090] If the target threshold interval is the third threshold interval, it means that the predicted trajectory of the target frame predicted by the prediction model is located on the second side of the actual trajectory. Before starting trajectory prediction for the next frame after the target frame, it is necessary to adjust the configuration parameters of the prediction model to the first side to increase the prediction amplitude of the prediction model on the first side, so that the predicted trajectory predicted by the corrected prediction model is shifted to the first side. Adjusting the configuration parameters of the prediction model to the first side only requires calling the underprediction configuration file as the configuration file of the prediction model.

[0091] The configuration file is in the format of .config. Different configuration files contain the same configuration parameters, but the values ​​of the configuration parameters are different. The configuration parameters cover all aspects of the model, such as the prediction trajectory time length used to constrain the output duration of the predicted trajectory, the prediction trajectory time resolution used to standardize the time interval of the output of the predicted trajectory points, the lateral movement rate threshold dl_threshold used as one of the judgment conditions to determine whether the obstacle vehicle has the intention to change lanes, and the number of historical frames hist numin lateral fluctuation used to check whether the actual output results fluctuate, etc.

[0092] Taking the original configuration file as an example, the specific values ​​of each parameter can be, for example, prediction trajectory timelength = 5.0, prediction trajectory time resolution = 0.2, dl_threshold = 0.45, hist numin lateral fluctuation = 3, etc.

[0093] In some implementations, if the target threshold interval is the fourth interval, the historical correction index of the historical frame corresponding to the target frame is obtained; the historical threshold interval in which the historical correction index is located is determined, and the target configuration file corresponding to the historical threshold interval is obtained. Here, the historical frame can be the latest historical frame, or it can be the latest frame containing the target object contained in the target frame.

[0094] S305. Based on the target configuration file, the prediction model is modified to obtain the modified prediction model.

[0095] In some implementations, the prediction model is modified according to the target configuration file, which is manifested in the prediction model calling the first configuration file, the second configuration file, or the third configuration file as the configuration file for the prediction model.

[0096] S306. Based on the corrected prediction model, determine the predicted trajectories of multiple target objects in the next frame after the target frame.

[0097] The other descriptions of S306 are the same as those in S104 above, and will not be repeated here.

[0098] In this embodiment, based on the actual trajectory and predicted trajectory of each target object in the target frame, the prediction evaluation index result of each target object is determined. Based on the feedforward compensation gain and prediction evaluation index result of each target object, the correction index is determined. Then, the target threshold range corresponding to the correction index and the target configuration file corresponding to each target threshold range are obtained. By directly calling the target configuration file, the model is corrected, which improves the efficiency and accuracy of model correction. Finally, trajectory prediction is performed based on the corrected prediction model, which improves the accuracy of trajectory prediction.

[0099] See appendix Figure 6 , Figure 6 This illustration shows a structural block diagram of a trajectory prediction device according to one embodiment of this application. For use in a vehicle, the device 400 includes:

[0100] The first determining module 401 is used to determine the prediction evaluation index result of each target object based on the actual trajectory and predicted trajectory of each of the multiple target objects in the target frame. The predicted trajectory of the target object in the target frame is determined by the prediction model.

[0101] The second determining module 402 is used to determine the correction index based on the feedforward compensation gain corresponding to each target object and the predicted evaluation index results.

[0102] The correction module 403 is used to correct the prediction model according to the correction index to obtain the corrected prediction model.

[0103] Output module 404 is used to determine the predicted trajectory of each of the multiple target objects in the next frame of the target frame based on the corrected prediction model.

[0104] Optionally, the prediction evaluation index results corresponding to the target object include the average displacement error and the final displacement error between the actual trajectory and the predicted trajectory corresponding to the target object. The first determining module 401 is also used to determine the average displacement error and the final displacement error corresponding to each target object based on the predicted trajectory and the actual trajectory corresponding to each target object.

[0105] Optionally, the second determining module 402 is further configured to determine the intermediate result corresponding to each target object based on the feedforward compensation gain and the predicted evaluation index result corresponding to each target object; and to accumulate the intermediate results corresponding to multiple target objects to obtain the correction index.

[0106] Optionally, the prediction evaluation index results corresponding to the target object include the average displacement error and the final displacement error between the actual trajectory and the predicted trajectory of the target object. The second determining module 402 is also used to construct the error vector corresponding to each target object based on the average displacement error and the final displacement error of each target object; calculate the inner product of the error vector corresponding to each target object and the feedforward compensation gain; and determine the intermediate result corresponding to each target object based on the inner product.

[0107] Optionally, the second determining module 402 is further configured to calculate, for each target object, the inner product of the error vector corresponding to the target object and the feedforward compensation gain; if the predicted trajectory corresponding to the target object is on the first side of the actual trajectory corresponding to the target object, calculate the product of the first value and the inner product as the intermediate result corresponding to the target object; if the predicted trajectory corresponding to the target object is on the second side of the actual trajectory corresponding to the target object, calculate the product of the second value and the inner product as the intermediate result corresponding to the target object, wherein the first value and the second value are opposites of each other.

[0108] Optionally, the correction module 403 is also used to determine the target threshold range in which the correction index is located; obtain the target configuration file corresponding to the target threshold range, with different target threshold ranges corresponding to different configuration files; and correct the prediction model according to the target configuration file to obtain the corrected prediction model.

[0109] Optionally, the correction module 403 is further configured to: if the target threshold interval is a first threshold interval, obtain a first configuration file as the target configuration file, the first configuration file being used to increase the prediction magnitude of the prediction model on the second side; if the target threshold interval is a second threshold interval, obtain a second configuration file as the target configuration file, the second configuration file being used to maintain the values ​​of the configuration parameters of the prediction model; if the target threshold interval is a third threshold interval, obtain a third configuration file as the target configuration file, the third configuration file being used to increase the prediction magnitude of the prediction model on the first side.

[0110] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described device and module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0111] Furthermore, the functions in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or as a software functional module.

[0112] Furthermore, the functions in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or as a software functional module.

[0113] On the other hand, this application also provides a computer-readable storage medium storing program code that can be called by a processor to execute the methods described in the above method embodiments.

[0114] Computer-readable storage media can be electronic storage devices such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or a cluster of ROMs. Optionally, computer-readable storage media include non-transitory computer-readable storage media. The computer-readable storage medium has storage space for program code that performs any of the method steps described above. This program code can be read from or written to one or more computer program products. The program code can be compressed, for example, in a suitable form.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A trajectory prediction method, characterized in that, The method includes: Based on the actual trajectory and predicted trajectory of multiple target objects in the target frame, the prediction evaluation index result of each target object is determined. The predicted trajectory of the target object in the target frame is determined by the prediction model. The prediction evaluation index result corresponding to the target object includes the average displacement error and the final displacement error between the actual trajectory and the predicted trajectory of the target object. Based on the average displacement error and the final displacement error corresponding to each target object, an error vector is constructed for each target object. Calculate the inner product of the error vector and the feedforward compensation gain for each target object, and determine the intermediate result for each target object based on the inner product; By summing the intermediate results corresponding to the multiple target objects, the correction index is obtained; The prediction model is corrected according to the correction index to obtain the corrected prediction model; The predicted trajectories of the multiple target objects in the next frame of the target frame are determined based on the modified prediction model.

2. The method according to claim 1, characterized in that, The step of determining the prediction evaluation index result for each target object based on the actual trajectory and predicted trajectory of each of the multiple target objects in the target frame includes: Based on the predicted trajectory and the actual trajectory corresponding to each target object, the average displacement error and the final displacement error corresponding to each target object are determined.

3. The method according to claim 1, characterized in that, The step of calculating the inner product of the error vector and the feedforward compensation gain for each target object, and determining the intermediate result for each target object based on the inner product, includes: For each target object, calculate the inner product of the error vector corresponding to the target object and the feedforward compensation gain; If the predicted trajectory corresponding to the target object is on the first side of the actual trajectory corresponding to the target object, calculate the product of the first value and the inner product as the intermediate result corresponding to the target object; If the predicted trajectory corresponding to the target object is on the second side of the actual trajectory corresponding to the target object, the product of the second value and the inner product is calculated as the intermediate result corresponding to the target object, wherein the first value and the second value are opposites of each other.

4. The method according to claim 1, characterized in that, The step of correcting the prediction model according to the correction index to obtain the corrected prediction model includes: Determine the target threshold range in which the correction index is located; Obtain the target configuration file corresponding to the target threshold range, where different target threshold ranges correspond to different configuration files; The prediction model is modified according to the target configuration file to obtain the modified prediction model.

5. The method according to claim 4, characterized in that, The step of obtaining the target configuration file corresponding to the target threshold range includes: If the target threshold interval is the first threshold interval, the first configuration file is obtained as the target configuration file, and the first configuration file is used to increase the prediction magnitude of the prediction model on the second side. If the target threshold range is the second threshold range, obtain the second configuration file as the target configuration file. The second configuration file is used to maintain the values ​​of the configuration parameters of the prediction model. If the target threshold interval is the third threshold interval, the third configuration file is obtained as the target configuration file, and the third configuration file is used to increase the prediction magnitude of the prediction model on the first side.

6. A trajectory prediction device, characterized in that, The device includes: The first determining module is used to determine the prediction evaluation index result of each target object based on the actual trajectory and predicted trajectory of each of the multiple target objects in the target frame. The predicted trajectory of the target object in the target frame is determined by a prediction model. The prediction evaluation index result corresponding to the target object includes the average displacement error and the final displacement error between the actual trajectory and the predicted trajectory of the target object. The second determining module is used to construct an error vector corresponding to each target object based on the average displacement error and the final displacement error corresponding to each target object; calculate the inner product of the error vector corresponding to each target object and the feedforward compensation gain; determine the intermediate result corresponding to each target object based on the inner product; and accumulate the intermediate results corresponding to the multiple target objects to obtain the correction index. The correction module is used to correct the prediction model according to the correction index to obtain the corrected prediction model; The output module is used to determine the predicted trajectory of each of the multiple target objects in the next frame of the target frame based on the corrected prediction model.

7. A vehicle, characterized in that, include: One or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications being configured to perform the method as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores processor-executable program code, which, when executed by the processor, causes the processor to perform the method according to any one of claims 1-5.

Images