Vehicle trajectory prediction method and device, and storage medium

Abstract

The application discloses a vehicle driving track prediction method and device and a storage medium. The method comprises the following steps: acquiring lane information of an area where an obstacle vehicle and a target vehicle are located at a predetermined historical moment, historical state information and a historical priority of the obstacle vehicle at the predetermined historical moment, and the historical priority is used to represent a priority degree of the target vehicle avoiding the obstacle vehicle at the predetermined historical moment; acquiring lane features of the lane information, historical state features of the historical state information, and historical priority features of the historical priority; acquiring first target features according to the lane features, the historical state features and the historical priority features; acquiring a driving state prediction result of the obstacle vehicle according to the first target features; and acquiring a predicted track of the target vehicle according to the driving state prediction result of the obstacle vehicle and an initial planning track of the target vehicle at the current moment. Thus, the technical problem of lacking an accurate self-vehicle track prediction method in the related art is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automatic driving, in particular to a vehicle driving trajectory prediction method and device and storage medium. BACKGROUND

[0002] In the field of automatic driving, the driving trajectory of a vehicle needs to be predicted so that the vehicle can drive according to the trajectory prediction result.

[0003] In the related art, the driving state of the vehicle is adjusted based on the instantaneous state of other vehicles other than the vehicle itself, and the trajectory prediction result of the vehicle is obtained. This method has poor adaptability to environmental changes, resulting in inaccurate trajectory prediction results of the vehicle. That is, in the related art, there is a technical problem of lacking an accurate method for predicting the trajectory of the vehicle.

[0004] At present, there is no effective solution to the above problems. SUMMARY

[0005] The embodiments of the present application provide a vehicle driving trajectory prediction method, device and storage medium to at least solve the technical problem of lacking an accurate method for predicting the trajectory of the vehicle in the related art.

[0006] According to an aspect of an embodiment of the present application, a vehicle driving trajectory prediction method is provided, comprising: obtaining lane information of an area where an obstacle vehicle and a target vehicle are located at a predetermined historical time, and historical state information and a historical priority of the obstacle vehicle at the predetermined historical time, wherein the historical priority is used to represent the priority degree of the target vehicle avoiding the obstacle vehicle at the predetermined historical time; obtaining lane features of the lane information, historical state features of the historical state information, and historical priority features of the historical priority; obtaining first target features according to the lane features, the historical state features and the historical priority features; obtaining a driving state prediction result of the obstacle vehicle according to the first target features; and obtaining a predicted trajectory of the target vehicle according to the driving state prediction result of the obstacle vehicle and an initial planning trajectory of the target vehicle at the present time.

[0007] Optionally, the first target features are obtained according to the lane features, the historical state features and the historical priority features, comprising: splicing the lane features, the historical state features and the historical priority features to obtain the first target features.

[0008] Optionally, further comprising: obtaining historical positions of the obstacle vehicle and the target vehicle at a predetermined historical time, and historical traffic conditions of areas where the obstacle vehicle and the target vehicle are located; obtaining historical position features of the historical positions, and historical traffic condition features of the historical traffic conditions.

[0009] Optionally, the obtaining the first target feature according to the lane feature, the historical state feature and the historical priority feature comprises: obtaining the first target feature according to the lane feature, the historical state feature, the historical priority feature, the historical position feature and the historical traffic condition feature.

[0010] Optionally, the obtaining the driving state prediction result of the obstacle vehicle according to the first target feature comprises: determining a first correspondence relationship between a second target feature of a first vehicle and a driving state prediction result of the first vehicle, wherein the second target feature is obtained according to a lane feature of an area where the first vehicle is located, a historical state feature of the first vehicle and a historical priority feature of the first vehicle; and obtaining the driving state prediction result of the obstacle vehicle according to the first correspondence relationship and the first target feature.

[0011] Optionally, the obtaining the predicted trajectory of the target vehicle according to the driving state prediction result of the obstacle vehicle and the initial planning trajectory of the target vehicle at the current time comprises: obtaining a state prediction feature of the driving state prediction result of the obstacle vehicle and an initial planning trajectory feature of the initial planning trajectory of the target vehicle at the current time; splicing the state prediction feature and the initial planning trajectory feature to obtain a third target feature; and obtaining the predicted trajectory of the target vehicle based on the third target feature.

[0012] Optionally, the obtaining the predicted trajectory of the target vehicle based on the third target feature comprises: determining a second correspondence relationship between a fourth target feature and a predicted trajectory of a second vehicle, wherein the fourth target feature is obtained by splicing a state prediction feature of a third vehicle and an initial planning trajectory feature of the second vehicle, and the third vehicle is an obstacle vehicle of the second vehicle; and obtaining the predicted trajectory of the target vehicle according to the second target feature and the second correspondence relationship.

[0013] Optionally, the obtaining the traffic data feature of the historical traffic data comprises: performing feature extraction on the historical traffic data based on a lightweight network model MobilenetV3 to obtain the traffic data feature.

[0014] According to another aspect of the embodiments of the present application, a vehicle trajectory prediction device is provided, characterized by comprising: a first obtaining module, configured to obtain lane information of an area where an obstacle vehicle and a target vehicle are located at a predetermined historical time, and historical state information and a historical priority of the obstacle vehicle at the predetermined historical time, wherein the historical priority is used to represent a priority degree of the target vehicle to avoid the obstacle vehicle at the predetermined historical time; a second obtaining module, configured to obtain lane features of the lane information, historical state features of the historical state information, and historical priority features of the historical priority; a third obtaining module, configured to obtain first target features according to the lane features, the historical state features, and the historical priority features; a fourth obtaining module, configured to obtain a driving state prediction result of the obstacle vehicle according to the first target features; and a fifth obtaining module, configured to obtain a predicted trajectory of the target vehicle according to the driving state prediction result of the obstacle vehicle and an initial planning trajectory of the target vehicle at present.

[0015] According to another aspect of the embodiments of the present application, a nonvolatile storage medium is provided, characterized by comprising a stored program, wherein the program, when executed, controls a device in which the storage medium is located to perform any of the above methods.

[0016] According to another aspect of the embodiments of the present application, a computer device is provided, characterized by comprising a memory and a processor, wherein the memory stores a computer program; and the processor is configured to execute the computer program stored in the memory, and the computer program, when executed, causes the processor to perform any of the above methods.

[0017] In the embodiment of the present application, lane information of an area where the obstacle vehicle and the target vehicle are located at a predetermined historical moment is acquired, as well as historical state information and a historical priority of the obstacle vehicle at the predetermined historical moment, wherein the historical priority is used to represent a priority degree of the target vehicle avoiding the obstacle vehicle at the predetermined historical moment; lane features of the lane information, historical state features of the historical state information, and historical priority features of the historical priority are acquired; a first target feature is acquired according to the lane features, the historical state features, and the historical priority features; a driving state prediction result of the obstacle vehicle is acquired according to the first target feature; and a predicted trajectory of the target vehicle is acquired according to the driving state prediction result of the obstacle vehicle and an initial planning trajectory of the target vehicle at the current moment. In the acquisition of the driving state prediction result of the obstacle vehicle, the lane information is taken into account, so that the driving state prediction result of the obstacle vehicle acquired can adapt to the change of the environment, and the trajectory prediction result of the ego vehicle can also adapt to the change of the environment, thereby overcoming the poor adaptability to the change of the environment in the related art when the trajectory prediction of the ego vehicle is performed, and the trajectory prediction result of the ego vehicle is inaccurate. In addition, in the method of the embodiment, the historical priority of the obstacle vehicle is taken into account, so that the driving state prediction result of the obstacle vehicle acquired has high accuracy, and the trajectory prediction result of the ego vehicle is high. Furthermore, the technical problem of lacking an accurate method for predicting the trajectory of the ego vehicle in the related art is solved. BRIEF DESCRIPTION OF DRAWINGS

[0018] The accompanying drawings, which are included to provide a further understanding of the present application and are incorporated in and constitute a part of this application, illustrate embodiments of the present application and together with the description serve to explain the present application. In the drawings:

[0019] Figure 1 is a flowchart of an optional vehicle driving trajectory prediction method according to an embodiment of the present application;

[0020] Figure 2 is a flowchart of another optional vehicle driving trajectory prediction method according to an embodiment of the present application;

[0021] Figure 3 is a structural block diagram of an optional vehicle driving trajectory prediction device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order for 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 drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0023] It should be noted that the terms "first", "second", and the like in the description and in the claims of the present application and in the above-described drawings are used to distinguish similar objects and are not necessarily used to describe a particular sequential or chronological order. It should be understood that the data thus used can be interchanged, where appropriate, so that the embodiments of the application described herein can be carried out in other than the order shown or described herein. Moreover, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or apparatus that includes a list of steps or units need not be limited to those steps or units that are clearly listed, but can include other steps or units not clearly listed or inherent to such processes, methods, products, or apparatuses.

[0024] Embodiment 1

[0025] According to an embodiment of the present application, a method embodiment of vehicle trajectory prediction is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0026] Figure 1 is a flowchart of a vehicle trajectory prediction method according to an embodiment of the present application, as shown in Figure 1 The method comprises the following steps:

[0027] Step S102, acquiring lane information of a region where an obstacle vehicle and a target vehicle are located at a predetermined historical time, and historical state information and historical priority of the obstacle vehicle at the predetermined historical time, wherein the historical priority is used to indicate the priority degree of the target vehicle avoiding the obstacle vehicle at the predetermined historical time.

[0028] In this optional embodiment, the target vehicle is a self-vehicle that needs to be trajectory predicted.

[0029] In some optional embodiments, there are various methods for acquiring lane information, for example, the lane information can be acquired through a grid map of a region where the obstacle vehicle and the target vehicle are located.

[0030] In some optional embodiments, the predetermined historical time can be set according to user demand, for example, it can be set as 5 time points within 1 second before the current time, wherein the time interval between adjacent time points is 0.2S, or it can be set as multiple time points within other time periods before the current time.

[0031] Step S104, acquiring lane features of the lane information, historical state features of the historical state information, and historical priority features of the historical priority.

[0032] In an optional embodiment, the grid map image frames of the areas where the obstacle vehicle and the target vehicle are located are collected at a plurality of predetermined historical moments respectively, and lane information in the grid map image frames at the plurality of historical moments is extracted according to the trained feature extraction model. The trained feature extraction model is obtained by training the feature extraction model through a plurality of sets of grid map image frame samples. The types of the feature extraction model are various, and preferably, a light network model MobilenetV3 is used for lane feature extraction. The lane information is obtained according to the grid map image frames of the areas where the obstacle vehicle and the target vehicle are located, and the lane information thus obtained can comprehensively and accurately reflect the lane conditions of the environment where the vehicle is located, thereby improving the accuracy of the results of subsequent driving state prediction of the obstacle vehicle.

[0033] In step S106, the first target feature is obtained according to the lane feature, the historical state feature and the historical priority feature.

[0034] In some optional embodiments, the first target feature is obtained according to the lane feature, the historical state feature and the historical priority feature, including: splicing the lane feature, the historical state feature and the historical priority feature to obtain the first target feature. By fusing the lane feature, the historical state feature and the historical priority feature, the first target feature obtained contains multi-dimensional feature information, thereby improving the accuracy of the first target feature.

[0035] In some optional embodiments, the obtained information includes lane information of the areas where the obstacle vehicle and the target vehicle are located, and further includes historical positions of the obstacle vehicle and the target vehicle, historical traffic conditions of the areas where the obstacle vehicle and the target vehicle are located; a historical position feature of the historical position and a historical traffic condition feature of the historical traffic condition are obtained. The first target feature is obtained according to the lane feature, the historical state feature and the historical priority feature, including: the first target feature is obtained according to the lane feature, the historical state feature and the historical priority feature, and the historical position feature and the historical traffic condition feature. The first target feature can be obtained by splicing the lane feature, the historical state feature and the historical priority feature, and the historical position feature and the historical traffic condition feature. In the process of obtaining the first target feature, not only the historical state feature and the historical priority feature of the obstacle vehicle are considered, but also the lane feature and the historical traffic condition feature are considered, and the first target feature thus obtained can reflect the environmental features of the environment where the vehicle is located, can adapt to the change of the environment where the vehicle is located, and improves the accuracy of the prediction results.

[0036] In some optional embodiments, the traffic condition can include various information, for example, can include the position of a traffic light in the corresponding area, the road type in the corresponding area, and the like, wherein the road type can include a lane changing, a straight road, an intersection, and the like. In this optional embodiment, the lane information, the historical traffic condition, and the historical position of the obstacle vehicle and the target vehicle jointly constitute the environmental information of the area where the obstacle vehicle and the target vehicle are located at a predetermined historical moment, and the first target feature is obtained by taking into account the environmental information, so that the adaptability of the first target feature and the environment obtained is high.

[0037] In step S108, the driving state prediction result of the obstacle vehicle is obtained according to the first target feature.

[0038] In some optional embodiments, obtaining the driving state prediction result of the obstacle vehicle according to the first target feature includes: determining a first correspondence relationship between the second target feature of the first vehicle and the driving state prediction result of the first vehicle, wherein the second target feature is obtained according to the lane feature of the area where the first vehicle is located, the historical state feature of the first vehicle, and the historical priority feature of the first vehicle; and obtaining the driving state prediction result of the obstacle vehicle according to the first correspondence relationship and the first target feature. According to the first correspondence relationship between the second target feature of the first vehicle and the driving state prediction result of the first vehicle, and the first target feature of the obstacle vehicle, the driving state prediction result of the obstacle vehicle can be quickly and accurately obtained.

[0039] In step S110, the predicted trajectory of the target vehicle is obtained according to the driving state prediction result of the obstacle vehicle and the initial planning trajectory of the target vehicle at the current moment.

[0040] In some optional embodiments, obtaining the predicted trajectory of the target vehicle according to the driving state prediction result of the obstacle vehicle and the initial planning trajectory of the target vehicle at the current moment includes: obtaining a state prediction feature of the driving state prediction result of the obstacle vehicle and an initial planning trajectory feature of the initial planning trajectory of the target vehicle at the current moment; splicing the state prediction feature and the initial planning trajectory feature to obtain a third target feature; and obtaining the predicted trajectory of the target vehicle based on the third target feature.

[0041] In some optional embodiments, based on the third target feature, the predicted trajectory of the target vehicle is obtained, including: determining a second correspondence relationship between a fourth target feature and the predicted trajectory of the second vehicle, wherein the fourth target feature is obtained by splicing the state prediction feature of the third vehicle and the initial planning trajectory feature of the second vehicle, and the third vehicle is an obstacle vehicle of the second vehicle; and obtaining the predicted trajectory of the target vehicle according to the second target feature and the second correspondence relationship. In this way, the predicted trajectory of the target vehicle can be accurately and quickly obtained. In an embodiment, the process of obtaining the predicted trajectory of the target vehicle according to the second target feature and the second correspondence relationship is based on a trained prediction model, and the prediction model can be a multilayer perceptron (MLP).

[0042] In the optional embodiment, the lane information of the area where the obstacle vehicle and the target vehicle are located at a predetermined historical moment, the historical state information of the obstacle vehicle at the predetermined historical moment, and the historical priority of the target vehicle at the predetermined historical moment are obtained, wherein the historical priority is used to represent the priority of the target vehicle avoiding the obstacle vehicle at the predetermined historical moment; lane features of the lane information, historical state features of the historical state information, and historical priority features of the historical priority are obtained; the first target feature is obtained according to the lane features, the historical state features, and the historical priority features; the driving state prediction result of the obstacle vehicle is obtained according to the first target feature; and the predicted trajectory of the target vehicle is obtained according to the driving state prediction result of the obstacle vehicle and the initial planning trajectory of the target vehicle at the current moment. In the optional embodiment, the driving state prediction result of the obstacle vehicle is obtained based on the lane information, that is, when the driving trajectory of the target vehicle is predicted, the lane information of the area where the obstacle vehicle and the target vehicle are located is considered, so that the driving state prediction result obtained can adapt to the environmental characteristics of the area where the obstacle vehicle and the target vehicle are located, thereby solving the problem that the adaptability of the trajectory prediction method to environmental changes is poor in the related art, resulting in inaccurate trajectory prediction result of the ego vehicle, and improving the accuracy of the trajectory prediction result of the vehicle. In addition, in the optional embodiment, when the driving trajectory of the target vehicle is predicted, the priority of the obstacle vehicle is considered, and the driving trajectory of the target vehicle is predicted based on the priority of the obstacle vehicle, thereby improving the accuracy of the trajectory prediction result, improving the driving safety of the target vehicle, and improving the user experience.

[0043] Based on the above embodiments and optional embodiments, the present application provides an optional implementation of a vehicle driving trajectory prediction method.

[0044] With the development of artificial intelligence, autonomous driving has become the development front line of future transportation. Autonomous driving not only has the functions of acceleration, deceleration, turning, etc. of a regular car, but also integrates the functions of environment perception, behavior decision, path planning, vehicle control, etc. In a high-level autonomous driving system (such as Level 4), scenarios such as other vehicles cutting in, vehicle conflicts at intersections, pedestrians crossing the road, etc. need to be dealt with. In related technologies, the instantaneous state is used to make decisions about the driving of the ego vehicle. This method has poor adaptability to the development and changes of the environment, and when dealing with changing scenarios, it is easy to produce delays, thus leading to collision risks. Specifically, in related technologies, a behavior rule library is established by combining obstacle behavior and trajectory according to formal rules, traffic regulations, driving common sense, etc. Vehicle states are divided according to different scenarios, and obstacle behavior and trajectory are predicted according to rule logic. This method has the following problems: due to the state cutting division condition, the vehicle behavior is not coherent, the behavior rule library trigger condition is easy to overlap, thus causing system failure, and the rule-based behavior and trajectory prediction has the disadvantage of being unable to cover all unexpected situation scenarios. That is, in related technologies, there is a problem of being unable to accurately predict the vehicle trajectory.

[0045] In view of this, the present optional embodiment provides an optional embodiment for solving the problem of being unable to accurately predict the vehicle trajectory in related technologies. The following is a specific description.

[0046] Figure 2 is a flowchart of another optional vehicle driving trajectory prediction method according to an embodiment of the present application. As shown in Figure 2 The method comprises the following steps:

[0047] Obtain the state data of the obstacle vehicle in the driving process and the environmental data within a certain range of priority.

[0048] Specifically, the historical state information set T = {T1, T2,..., Tt,..., Tn} of the obstacle vehicle is obtained through the historical trajectory of the obstacle vehicle, wherein T1, T2, Tt, Tn represent the state information of the obstacle vehicle at the 1st historical time, the 2nd historical time, the tth historical time, and the nth historical time, respectively, wherein the state information of each historical time includes the speed, acceleration, heading, etc. of the obstacle vehicle.

[0049] Specifically, the data set of the historical priority of the obstacle vehicle is P = {P1, P2,..., Pt,..., Pn}, where P1, P2, Pt, and Pn represent the priority of the obstacle vehicle at the 1st historical time, the 2nd historical time, the tth historical time, and the nth historical time, respectively. The historical priority is obtained according to historical speed, historical acceleration, and historical position of the obstacle vehicle, and in an embodiment, a corresponding relationship between the priority and the speed, acceleration, and position of the vehicle is obtained, and the historical priority of the obstacle vehicle is determined according to the corresponding relationship and the historical speed, historical acceleration, and historical position of the obstacle vehicle at the corresponding time. The priority is the degree of possible interaction between the obstacle and the ego vehicle and the traffic danger of the obstacle itself.

[0050] Specifically, the data set of the environment data is X = {X1, X2,..., Xt,..., Xn}, where X1, X2, Xt, and Xn represent the environment data of the obstacle vehicle at the 1st historical time, the 2nd historical time, the tth historical time, and the nth historical time, respectively. In an embodiment, the environment data at each historical time includes an environment sampling picture (i.e., a map image in the environment sampling picture) of the region where the obstacle vehicle and the ego vehicle are located at the corresponding historical time, where the data in the environment sampling picture includes the coordinates of the ego vehicle and the obstacle vehicle, the lane line situation of the road segment where the ego vehicle and the obstacle vehicle are located, the position and the scene of the traffic light in front of the ego vehicle and the obstacle vehicle, the semantic information of the region where the ego vehicle and the obstacle vehicle are located, and the like. The scene includes road conditions, and the road conditions include one of the following: roundabout, intersection, straight road, and the like. Figure 2

[0051] The environment data set is obtained based on pictures collected at multiple historical times. When the pictures are sampled, the sampling frequency of the picture sampling is the same as the sampling frequency of the historical priority and the historical state information.

[0052] The historical state feature, the historical priority feature, and the environment data feature are obtained according to the historical state information, the historical priority information, and the environment data information, and after the historical state feature, the historical priority feature, and the environment data feature are obtained, they are processed into feature vectors in the range of [0, 1], and subsequent processing is performed based on the processed feature vectors.

[0053] Referring to Figure 2 ​In the optional embodiment, the environment data features are extracted by a lightweight network model MobilenetV3, and then a 1280-dimensional feature vector is obtained. The 1280-dimensional feature vector is processed by an MLP model to obtain a 256-dimensional environment data feature vector. The environment data features include lane features, historical position features, and historical traffic condition features in the foregoing embodiments. The historical state information is extracted by the MLP model to obtain a 64-dimensional historical state feature vector.

[0054] The historical priority is extracted by the MLP model to obtain a 64-dimensional historical priority feature vector.

[0055] The environment data feature vector, the historical state feature vector, and the historical priority feature vector are spliced to obtain a spliced feature. Specifically, the feature_map obtained after the feature is extracted by the lightweight network model mobilenetV3 is expanded into a column after global average pooling, and is fully connected with the historical state feature vector and the historical priority feature vector to obtain the spliced feature.

[0056] The lightweight network model mobilenetV3 used in the optional embodiment can be divided into the following three parts: the starting part includes one convolutional layer, which extracts features by 3x3 convolution; the middle part includes multiple convolutional layers, and the number of layers and parameters are different in different versions; and the last part replaces the full connection by two 1x1 convolutional layers and outputs the category. In the convolutional neural network, each convolutional layer is composed of a plurality of convolutional units, and the parameters of each convolutional unit are obtained by the back propagation algorithm. The purpose of convolution operation is to extract different features of the input. The first convolutional layer can only extract some low-level features such as edges, lines, and angles. More layers of network can iteratively extract more complex features from low-level features.

[0057] The spliced features are input into a state prediction model for state prediction, to obtain a state prediction result of the obstacle vehicle. In an embodiment, the state prediction model is trained from an MLP model. The state prediction result of the obstacle vehicle includes driving trajectories, confidence, driving intention and priority of the obstacle vehicle in a plurality of predetermined driving modes within a predetermined time period. The predetermined driving modes include one of the following states of the vehicle: left turn, right turn, left lane change, right lane change, straight driving, etc. It should be understood that the state prediction process of the obstacle vehicle can be understood as a mapping process. Specifically, according to the spliced features of the environmental data, the historical state information and the historical priority, a prediction result corresponding to the spliced features is obtained. In an embodiment, the rasterization information of 5 frames of pictures within one second before the current time of the obstacle vehicle is obtained, and the picture information is obtained by sampling the area where the obstacle vehicle is located, wherein the time interval between adjacent frames is 0.2S. The dimension of the rasterization information is W1=300x300x(5+(5+1)x2. The historical state information includes data corresponding to 5 predetermined historical time points, and the dimension W2=3x(5+1). The dimension of the historical priority is W3=1x6.

[0058] The output of the state prediction model includes two parts, one of which is the predicted intention, trajectory, probability and priority of the obstacle. In an embodiment, the predicted trajectory is composed of coordinates at a plurality of predetermined time points within 8s, and the time difference between adjacent time points in the plurality of predetermined time points is 0.2s. The output parameter dimension of the state prediction model is: 5x(40x2+1+1)+1.

[0059] In this optional embodiment, a mobilenetV3 neural network model is built based on the python language, combined with the openCV library and using the Tensorflow framework. When training the model, the sample set includes map pictures of the area where the vehicle is located at a plurality of consecutive time points. When training the model, the map pictures are normalized, for example, processed into pictures with a size of 300x300. The mean square error of the predicted trajectory of the obstacle within the prediction period relative to the real driving trajectory is calculated, the probability of the real trajectory in the mixed probability density is expected to be maximized, and the maximum likelihood function of the joint probability distribution is used as the loss function. The exponential decay learning rate is set to 0.001, the gradient descent algorithm is used for training, and the training is stopped when the mean square error no longer decreases.

[0060] According to the current planning trajectory of the ego vehicle, planning trajectory features of the planning trajectory are obtained.

[0061] The trajectory planning feature and the state prediction feature of the state prediction result of the obstacle vehicle are input into a trajectory prediction model to obtain a trajectory prediction result of the ego vehicle. In an embodiment, the trajectory prediction model is obtained by training an MLP model. In an embodiment, the predicted trajectory, intention, probability, and priority of the obstacle are spliced with the input planning information, and adjusted planning information including 8s of planning trajectory point coordinates and corresponding intention information is obtained by decoding through a multi-layer perception machine.

[0062] In the above model training, different network models can be trained according to different scenarios. Different scenarios can include left turn, right turn, left lane change, right lane change, straight driving, and the like. In using the trained model, the scenario in which the vehicle is located at the corresponding time is first determined, and the trained model is selected according to the scenario to predict the corresponding result, thereby improving the accuracy of the prediction result.

[0063] In the optional embodiment, the environment data, corresponding behavior state, priority of the vehicle, and the current planning trajectory of the ego vehicle are obtained, wherein the environment data includes the coordinates of the ego vehicle, the lane line situation of the driving section, the coordinates of the obstacles within a certain range around the vehicle, the situation of the traffic light in front, and the set of the scene. The state information includes speed, acceleration, and yaw rate. The data in a historical time period (such as 1s before the current time) is used as input, and the data in the future 8s is used as output, and MobileNetV3 is used for supervised training. In predicting the driving state of the obstacle vehicle, the output of the prediction model includes the predicted trajectory point, predicted intention, and priority of the obstacle vehicle in the future 8s. In the optional embodiment, each traffic participant has a corresponding priority, and the trajectory planning information of the ego vehicle is adjusted according to different priorities and trajectories. The planning trajectory of the ego vehicle is used as input, the prediction of the obstacle is coupled with the planning of the ego vehicle, the planning of the ego vehicle is adjusted according to the priority, and the interaction with the obstacle is added in the prediction, thereby improving the accuracy of the prediction result, improving the safety of the driving of the autonomous vehicle, and improving the user experience.

[0064] Embodiment 2

[0065] Figure 3 is a structural block diagram of another optional vehicle driving trajectory prediction device according to an embodiment of the application. Referring to FIG. 2, Figure 3 As shown in FIG. 2, the device includes a first obtaining module 302, a second obtaining module 304, a third obtaining module 306, a fourth obtaining module 308, and a fifth obtaining module 310, which will be described in detail below.

[0066] The first obtaining module 302 is configured to obtain lane information of an area where an obstacle vehicle and a target vehicle are located at a predetermined historical time, and historical state information and a historical priority of the obstacle vehicle at the predetermined historical time, wherein the historical priority is used to represent a priority degree of the target vehicle avoiding the obstacle vehicle at the predetermined historical time.

[0067] It should be noted that the first obtaining module 302, the second obtaining module 304, the third obtaining module 306, the fourth obtaining module 308 and the fifth obtaining module 310 correspond to steps S102 to S110 in Embodiment 1, and the several modules and the instances and application scenarios realized by the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment.

[0068] Embodiment 3

[0069] The embodiments of the present application can provide a computer device. Optionally, in the present embodiment, the computer device can be located in at least one network device of a plurality of network devices of a computer network. The computer device comprises a memory and a processor.

[0070] The memory can be used to store software programs and modules, such as program instructions / modules corresponding to the data processing method and device in the embodiments of the present application. The processor executes various functions and data processing by running the software programs and modules stored in the memory, that is, implements the above-mentioned data processing method. The memory can include a high-speed random access memory, and can also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some examples, the memory can further include a memory remotely arranged with respect to the processor, and the remote memory can be connected to the computer terminal through a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0071] The processor can call information and application programs stored in the memory through the transmission device to execute the following steps: obtaining lane information of an area where the obstacle vehicle and the target vehicle are located at a predetermined historical time, and historical state information and a historical priority of the obstacle vehicle at the predetermined historical time, wherein the historical priority is used to represent a priority degree of the target vehicle avoiding the obstacle vehicle at the predetermined historical time; obtaining lane features of the lane information, historical state features of the historical state information, and historical priority features of the historical priority; obtaining first target features according to the lane features, the historical state features and the historical priority features; obtaining a driving state prediction result of the obstacle vehicle according to the first target features; and obtaining a predicted trajectory of the target vehicle according to the driving state prediction result of the obstacle vehicle and an initial planning trajectory of the target vehicle at the current time.

[0072] Optionally, the processor can further execute program codes of the following steps: splicing the lane features, the historical state features and the historical priority features to obtain the first target features.

[0073] Optionally, the processor can further execute program codes of the following steps: obtaining historical positions of the obstacle vehicle and the target vehicle at the predetermined historical time, and historical traffic conditions of the area where the obstacle vehicle and the target vehicle are located; obtaining historical position features of the historical positions, and historical traffic condition features of the historical traffic conditions.

[0074] Optionally, the processor can further execute program codes of the following steps: obtaining the first target features according to the lane features, the historical state features and the historical priority features, and the historical position features and the historical traffic condition features.

[0075] Optionally, the processor can further execute program codes of the following steps: determining a first correspondence relationship between second target features of the first vehicle and the driving state prediction result of the first vehicle, wherein the second target features are obtained according to lane features of an area where the first vehicle is located, historical state features of the first vehicle, and historical priority features of the first vehicle; and obtaining the driving state prediction result of the obstacle vehicle according to the first correspondence relationship and the first target features.

[0076] Optionally, the processor can further execute program codes of the following steps: obtaining state prediction features of the driving state prediction result of the obstacle vehicle, and initial planning trajectory features of the initial planning trajectory of the target vehicle at the current time; splicing the state prediction features and the initial planning trajectory features to obtain third target features; and obtaining the predicted trajectory of the target vehicle based on the third target features.

[0077] Optionally, the processor can further execute program codes of the following steps: determining a second correspondence relationship between a fourth target feature and a predicted trajectory of the second vehicle, wherein the fourth target feature is obtained by splicing a state prediction feature of the third vehicle and an initial planning trajectory feature of the second vehicle, and the third vehicle is an obstacle vehicle of the second vehicle; and obtaining the predicted trajectory of the target vehicle according to the second target feature and the second correspondence relationship.

[0078] Optionally, the processor can further execute program codes of the following steps: performing feature extraction on the historical traffic data based on a lightweight network model MobilenetV3 to obtain traffic data features.

[0079] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer readable storage medium, which can include a flash disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0080] Embodiments of the present application also provide a non-volatile storage medium. Optionally, in the present embodiment, the non-volatile storage medium can be used to save the program codes executed by the vehicle driving trajectory prediction method provided in the above-mentioned embodiment 1.

[0081] Optionally, in the present embodiment, the non-volatile storage medium can be located in any one of the computer terminals in the computer terminal group in the computer network, or in any one of the mobile terminals in the mobile terminal group.

[0082] Optionally, in the present embodiment, the non-volatile storage medium is configured to store program codes for performing the following steps: obtaining lane information of an area where the obstacle vehicle and the target vehicle are located at a predetermined historical time, and historical state information and a historical priority of the obstacle vehicle at the predetermined historical time, wherein the historical priority is used to indicate a priority degree of the target vehicle avoiding the obstacle vehicle at the predetermined historical time; obtaining lane features of the lane information, historical state features of the historical state information, and historical priority features of the historical priority; obtaining a first target feature according to the lane features, the historical state features and the historical priority features; obtaining a driving state prediction result of the obstacle vehicle according to the first target feature; and obtaining a predicted trajectory of the target vehicle according to the driving state prediction result of the obstacle vehicle and an initial planning trajectory of the target vehicle at the present time.

[0083] Optionally, in the embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: splicing the lane feature, the historical state feature and the historical priority feature to obtain the first target feature.

[0084] Optionally, in the embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: obtaining the historical position of the obstacle vehicle and the target vehicle, and the historical traffic condition of the area where the obstacle vehicle and the target vehicle are located at a predetermined historical time; obtaining the historical position feature of the historical position and the historical traffic condition feature of the historical traffic condition.

[0085] Optionally, in the embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: obtaining the first target feature according to the lane feature, the historical state feature and the historical priority feature, and the historical position feature and the historical traffic condition feature.

[0086] Optionally, in the embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: determining a first correspondence relationship between the second target feature of the first vehicle and the first vehicle driving state prediction result, wherein the second target feature is obtained according to the lane feature of the area where the first vehicle is located, the historical state feature of the first vehicle, and the historical priority feature of the first vehicle; obtaining the driving state prediction result of the obstacle vehicle according to the first correspondence relationship and the first target feature.

[0087] Optionally, in the embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: obtaining the state prediction feature of the driving state prediction result of the obstacle vehicle and the initial planning track feature of the target vehicle in the current initial planning track; splicing the state prediction feature and the initial planning track feature to obtain the third target feature; and obtaining the predicted track of the target vehicle based on the third target feature.

[0088] Optionally, in the embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: determining a second correspondence relationship between the fourth target feature and the predicted track of the second vehicle, wherein the fourth target feature is obtained by splicing the state prediction feature of the third vehicle and the initial planning track feature of the second vehicle, and the third vehicle is the obstacle vehicle of the second vehicle; obtaining the predicted track of the target vehicle according to the second target feature and the second correspondence relationship.

[0089] Optionally, in the embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: extracting features of the historical traffic data based on the network model MobilenetV3 to obtain traffic data features.

[0090] The above-mentioned embodiment numbers of the present application are only for description, and do not represent the advantages or disadvantages of the embodiments.

[0091] In the above-mentioned embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0092] In several embodiments provided in the present application, it should be understood that the disclosed technical contents can be implemented by other ways. Among them, the above-mentioned device embodiments are only schematic, for example, the division of the units can be a logical function division, and actual implementation can have another division way, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or modules shown or discussed can be indirect coupling or communication connection through some interfaces, units or modules, which can be electrical or other forms.

[0093] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed to multiple units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0094] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above integrated unit can be realized in the form of hardware or in the form of software functional unit.

[0095] The integrated unit, if realized in the form of software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the whole or part of the technical solutions which essentially contribute to the prior art can be embodied in the form of software product, which is stored in a storage medium and includes a plurality of instructions for making a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and various program code storage media.

[0096] The above merely describes the preferred embodiments of the present application, and it should be pointed out that, for those skilled in the art, several improvements and refinements can be made without departing from the principles of the present application, and these improvements and refinements should also be considered as falling within the protection scope of the present application.

Claims

1. A vehicle travel trajectory prediction method characterized by comprising: The method comprises the following steps: obtaining lane information of a region where an obstacle vehicle and a target vehicle are located at a predetermined historical moment, and historical state information and a historical priority of the obstacle vehicle at the predetermined historical moment, wherein the historical priority is used to represent a priority degree of the target vehicle avoiding the obstacle vehicle at the predetermined historical moment; obtaining lane features of the lane information, historical state features of the historical state information, and historical priority features of the historical priority; obtaining first target features according to the lane features, the historical state features and the historical priority features; obtaining a driving state prediction result of the obstacle vehicle according to the first target features; obtaining a predicted trajectory of the target vehicle according to the driving state prediction result of the obstacle vehicle and an initial planning trajectory of the target vehicle at the current moment.

2. The method of claim 1, wherein, The step of obtaining the first target features according to the lane features, the historical state features and the historical priority features comprises the following step: splicing the lane features, the historical state features and the historical priority features to obtain the first target features.

3. The method of claim 1, wherein, The method further comprises the following steps: obtaining historical positions of the obstacle vehicle and the target vehicle and historical traffic conditions of a region where the obstacle vehicle and the target vehicle are located at a predetermined historical moment; obtaining historical position features of the historical positions and historical traffic condition features of the historical traffic conditions.

4. The method of claim 3, wherein, The step of obtaining the first target features according to the lane features, the historical state features and the historical priority features comprises the following step: obtaining the first target features according to the lane features, the historical state features, the historical priority features, the historical position features and the historical traffic condition features.

5. The method of claim 1, wherein, The step of obtaining the driving state prediction result of the obstacle vehicle according to the first target features comprises the following steps: determining a first correspondence relationship between second target features of a first vehicle and the driving state prediction result of the first vehicle, wherein the second target features are obtained according to lane features of a region where the first vehicle is located, historical state features of the first vehicle and historical priority features of the first vehicle; obtaining the driving state prediction result of the obstacle vehicle according to the first correspondence relationship and the first target features.

6. The method of claim 1, wherein, The step of obtaining the predicted trajectory of the target vehicle according to the driving state prediction result of the obstacle vehicle and the initial planning trajectory of the target vehicle at the current moment comprises the following steps: obtaining state prediction features of the driving state prediction result of the obstacle vehicle and initial planning trajectory features of the initial planning trajectory of the target vehicle at the current moment; splicing the state prediction features and the initial planning trajectory features to obtain third target features; obtaining the predicted trajectory of the target vehicle based on the third target features.

7. The method of claim 6, wherein, The step of obtaining the predicted trajectory of the target vehicle based on the third target features comprises the following steps: determine a second correspondence relationship between a fourth target feature and a predicted trajectory of the second vehicle, wherein the fourth target feature is obtained by splicing a state prediction feature of a third vehicle and an initial planning trajectory feature of the second vehicle, and the third vehicle is an obstacle vehicle of the second vehicle; obtain the predicted trajectory of the target vehicle according to the third target feature and the second correspondence relationship.

8. A vehicle travel trajectory prediction device characterized by comprising: Comprise: a first obtaining module, configured to obtain lane information of an area where an obstacle vehicle and a target vehicle are located at a predetermined historical moment, and historical state information and a historical priority of the obstacle vehicle at the predetermined historical moment, wherein the historical priority is used to represent a priority degree of the target vehicle in avoiding the obstacle vehicle at the predetermined historical moment; a second obtaining module, configured to obtain a lane feature of the lane information, a historical state feature of the historical state information, and a historical priority feature of the historical priority; a third obtaining module, configured to obtain a first target feature according to the lane feature, the historical state feature, and the historical priority feature; a fourth obtaining module, configured to obtain a driving state prediction result of the obstacle vehicle according to the first target feature; a fifth obtaining module, configured to obtain a predicted trajectory of the target vehicle according to the driving state prediction result of the obstacle vehicle and an initial planning trajectory of the target vehicle at the current moment.

9. A non-volatile storage medium, comprising: The storage medium comprises a stored program, wherein the program controls a device where the storage medium is located to execute the method in any one of claims 1 to 7 when the program is running.

10. A computer device, comprising: Comprise: a memory and a processor, the memory stores a computer program; the processor is configured to execute the computer program stored in the memory, and the computer program makes the processor execute the method in any one of claims 1 to 7 when running.

Images