Obstacle trajectory prediction method and device

Abstract

The invention provides an obstacle trajectory prediction method and device, and the method comprises the steps: carrying out the feature extraction of vehicle perception data, and obtaining the global feature information under a Cartesian coordinate system; determining a reference line according to the global feature information, converting from a Cartesian coordinate system to a Frenet coordinate system, and then converting to an agent coordinate system to obtain second feature information under the agent coordinate system; according to a pre-constructed trajectory prediction model and the second feature information, obstacle future trajectory prediction is carried out, and future trajectory information is obtained; and converting the future trajectory information from the agent coordinate system to a Cartesian coordinate system to obtain target future trajectory information under the Cartesian coordinate system. According to the method and the device, the curve data can be converted into the straightway through Frenet coordinate system conversion, the difference of maps in a scene is eliminated, then track prediction is performed through the track prediction model, the accuracy is high, meanwhile, the curve prediction accuracy is also improved, and the automatic driving experience is improved.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and more specifically, to a method and device for predicting obstacle trajectories. Background Art

[0002] As an important development direction of future transportation, the core of autonomous driving technology lies in the accurate prediction of the motion trajectories of various objects in the surrounding environment. Autonomous driving motion prediction technology, by integrating the vehicle's historical trajectory, current surrounding environment information and future driving planning, can accurately predict the vehicle's future motion trajectory, which is of great significance to ensuring driving safety and improving traffic efficiency. Current trajectory prediction technologies generally rely on physical models (such as dynamics or kinematic models) to characterize the target's motion pattern, and based on this model combined with the obstacle's immediate state to estimate its future position and dynamics. However, in practice, it is found that the existing methods perform well in straight road scenes, but the accuracy of predicting the future trajectory of obstacles in curved environments is greatly reduced. In addition, it is often difficult to achieve the ideal level of accuracy by relying solely on physical models for prediction. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide an obstacle trajectory prediction method and device, which can convert curve data into straight roads through Frenet coordinate system conversion, eliminate the differences in maps in the scene, and then perform trajectory prediction through a trajectory prediction model with high accuracy. At the same time, it also improves the accuracy of curve prediction and enhances the autonomous driving experience.

[0004] The first aspect of the present application provides an obstacle trajectory prediction method, comprising:

[0005] Collect vehicle perception data of target vehicles on the curve through sensor equipment;

[0006] Extracting features from the vehicle perception data to obtain global feature information in a Cartesian coordinate system;

[0007] Determine a reference line according to the global feature information;

[0008] Acquire at least one obstacle to be inferred that needs to be inferred, and extract obstacle features of the obstacle to be inferred from the global feature information;

[0009] Converting the obstacle feature from the Cartesian coordinate system to the Frenet coordinate system according to the reference line to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system;

[0010] Converting the first feature information from the Frenet coordinate system to the agent coordinate system to obtain second feature information of the obstacle to be inferred in the agent coordinate system;

[0011] Predicting the future trajectory of the obstacle according to the pre-built trajectory prediction model and the second feature information to obtain the future trajectory information of the obstacle to be inferred;

[0012] The future trajectory information is converted from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

[0013] Furthermore, the vehicle perception data includes at least vehicle trajectory information and vehicle surrounding map information;

[0014] The global feature information includes historical trajectory information and surrounding map information of all obstacles within a preset range; wherein the historical trajectory information includes at least position information, speed information and movement direction; and the surrounding map information includes at least lane center point position information and lane type.

[0015] Further, determining the reference line according to the global feature information includes:

[0016] Determine the scene range where the target vehicle is currently located and the target lane where the target vehicle is located according to the global feature information;

[0017] Using a depth-first search algorithm within the scope of the scene to search for other lanes that have a predecessor-successor reachable relationship with the target lane;

[0018] If the other lanes include multiple preceding and succeeding lanes, selecting a preceding and succeeding lane from the other lanes as the preceding and succeeding reachable lane in the scene;

[0019] A reference line is generated according to the target lane and the preceding and succeeding reachable lanes in the scene.

[0020] Furthermore, generating a reference line according to the target lane and the preceding and succeeding reachable lanes in the scene includes:

[0021] The target lane and the preceding and succeeding reachable lanes in the scene are combined in order from the rear of the target vehicle to the front of the target vehicle to obtain a reference line.

[0022] Furthermore, the step of converting the obstacle feature from the Cartesian coordinate system to a Frenet coordinate system according to the reference line to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system includes:

[0023] Obtaining the obstacle history trajectory coordinates and surrounding lane coordinates of the obstacle to be inferred from the obstacle features;

[0024] Based on the reference line, the coordinates of the obstacle historical trajectory are converted from a Cartesian coordinate system to a Frenet coordinate system to obtain Frenet coordinates of the obstacle historical trajectory;

[0025] Based on the reference line, the coordinates of the surrounding lanes are converted from a Cartesian coordinate system to a Frenet coordinate system to obtain Frenet coordinates of the surrounding lanes;

[0026] The obstacle historical trajectory Frenet coordinates and the surrounding lane Frenet coordinates are summarized to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system.

[0027] Furthermore, converting the first feature information from the Frenet coordinate system to the agent coordinate system to obtain the second feature information of the obstacle to be inferred in the agent coordinate system includes:

[0028] Extracting the target trajectory information of the obstacle to be inferred from the Frenet coordinates of the obstacle historical trajectory; wherein the target trajectory information at least includes the current position of the trajectory, the displacement of the trajectory in adjacent frames, the obstacle speed, and the obstacle movement direction;

[0029] Taking the obstacle to be inferred as the center, converting the Frenet coordinates of the obstacle's historical trajectory from the Frenet coordinate system to the agent coordinate system to obtain the agent coordinate system trajectory features in the agent coordinate system;

[0030] Taking the obstacle to be inferred as the center, converting the Frenet coordinates of the surrounding lanes from the Frenet coordinate system to the agent coordinate system to obtain lane features in the agent coordinate system;

[0031] The agent coordinate system trajectory feature and the lane feature are summarized to obtain second feature information of the obstacle to be inferred in the agent coordinate system.

[0032] Furthermore, the obstacle to be inferred is taken as the center, the Frenet coordinates of the obstacle historical trajectory are converted from the Frenet coordinate system to the agent coordinate system, and the agent coordinate system trajectory features in the agent coordinate system are obtained, including:

[0033] Normalizing the target trajectory information with the obstacle to be inferred as the center to obtain normalized trajectory information;

[0034] Obtaining the obstacle motion heading of the obstacle to be inferred;

[0035] Based on the obstacle movement heading, performing direction alignment processing on the trajectory directions of all obstacles in the normalized trajectory information to obtain trajectory features;

[0036] A mask is added to the trajectory feature to obtain the agent coordinate system trajectory feature in the agent coordinate system.

[0037] Furthermore, the method of converting the surrounding lane Frenet coordinates from the Frenet coordinate system to the agent coordinate system with the obstacle to be inferred as the center to obtain lane features in the agent coordinate system includes:

[0038] Extracting target map information of the obstacle to be inferred within a preset range at the current moment from the surrounding lane Frenet coordinates; wherein the target map information at least includes lane centerline node position, lane node displacement, lane connection relationship, road structure and traffic lights;

[0039] Taking the obstacle to be inferred as the center, normalizing the target map information to obtain normalized map information;

[0040] Direction alignment processing is performed on the lane direction in the normalized map information based on the obstacle movement heading to obtain lane features in the agent coordinate system.

[0041] Furthermore, the converting the future trajectory information from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system includes:

[0042] Converting the future trajectory information from the agent coordinate system to the Frenet coordinate system to obtain the future trajectory information of the obstacle to be inferred in the Frenet coordinate system;

[0043] The future trajectory information is converted from the Frenet coordinate system to the Cartesian coordinate system to obtain target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

[0044] A second aspect of the present application provides an obstacle trajectory prediction device, the obstacle trajectory prediction device comprising:

[0045] A collection unit, used for collecting vehicle perception data of a target vehicle on a curve through a sensor device;

[0046] A feature extraction unit, used to extract features from the vehicle perception data to obtain global feature information in a Cartesian coordinate system;

[0047] A determination unit, configured to determine a reference line according to the global feature information;

[0048] an acquisition unit, configured to acquire at least one obstacle to be inferred and extract obstacle features of the obstacle to be inferred from the global feature information;

[0049] A first coordinate conversion unit, configured to convert the obstacle feature from the Cartesian coordinate system to a Frenet coordinate system according to the reference line, so as to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system;

[0050] a second coordinate conversion unit, configured to convert the first feature information from the Frenet coordinate system to an agent coordinate system, to obtain second feature information of the obstacle to be inferred in the agent coordinate system;

[0051] A trajectory prediction unit, configured to predict the future trajectory of the obstacle according to a pre-built trajectory prediction model and the second feature information, and obtain the future trajectory information of the obstacle to be inferred;

[0052] The third coordinate conversion unit is used to convert the future trajectory information from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

[0053] Furthermore, the vehicle perception data includes at least vehicle trajectory information and vehicle surrounding map information;

[0054] The global feature information includes historical trajectory information and surrounding map information of all obstacles within a preset range; wherein the historical trajectory information includes at least position information, speed information and movement direction; and the surrounding map information includes at least lane center point position information and lane type.

[0055] Furthermore, the determining unit includes:

[0056] A determination subunit, used to determine the scene range where the target vehicle is currently located and the target lane where the target vehicle is located according to the global feature information;

[0057] A search subunit, configured to use a depth-first search algorithm to search for other lanes that have a predecessor-successor reachable relationship with the target lane within the scene range;

[0058] A selection subunit is used to select a predecessor and successor lane from the other lanes as the predecessor and successor reachable lane in the scene when the other lanes include multiple predecessor and successor lanes;

[0059] A generating subunit is used to generate a reference line according to the target lane and the preceding and succeeding reachable lanes in the scene.

[0060] Furthermore, the generating subunit is specifically used to combine the target lane and the preceding and succeeding reachable lanes in the scene in the order from the rear of the target vehicle to the front of the target vehicle to obtain a reference line.

[0061] Furthermore, the first coordinate conversion unit includes:

[0062] An acquisition subunit, used for acquiring the obstacle history trajectory coordinates and surrounding lane coordinates of the obstacle to be inferred from the obstacle features;

[0063] A first conversion subunit, configured to convert the obstacle historical trajectory coordinates from a Cartesian coordinate system to a Frenet coordinate system based on the reference line to obtain the obstacle historical trajectory Frenet coordinates;

[0064] The first conversion subunit is further used to convert the surrounding lane coordinates from a Cartesian coordinate system to a Frenet coordinate system based on the reference line to obtain the surrounding lane Frenet coordinates;

[0065] The first summarizing subunit is used to summarize the Frenet coordinates of the obstacle historical trajectory and the Frenet coordinates of the surrounding lanes to obtain the first feature information of the obstacle to be inferred in the Frenet coordinate system.

[0066] Furthermore, the second coordinate conversion unit includes:

[0067] An extraction subunit, used to extract the target trajectory information of the obstacle to be inferred from the Frenet coordinates of the obstacle historical trajectory; wherein the target trajectory information at least includes the current position of the trajectory, the displacement of the trajectory in adjacent frames, the obstacle speed, and the obstacle movement direction;

[0068] A second conversion subunit is used to convert the Frenet coordinates of the obstacle historical trajectory from the Frenet coordinate system to the agent coordinate system with the obstacle to be inferred as the center, so as to obtain the agent coordinate system trajectory features in the agent coordinate system;

[0069] The second conversion subunit is further used to convert the surrounding lane Frenet coordinates from the Frenet coordinate system to the agent coordinate system with the obstacle to be inferred as the center, so as to obtain the lane features in the agent coordinate system;

[0070] The second summarizing subunit is used to summarize the trajectory features of the agent coordinate system and the lane features to obtain second feature information of the obstacle to be inferred in the agent coordinate system.

[0071] Furthermore, the second conversion subunit includes:

[0072] A processing module, used for normalizing the target trajectory information with the obstacle to be inferred as the center to obtain normalized trajectory information;

[0073] An acquisition module, used for acquiring the obstacle motion heading of the obstacle to be inferred;

[0074] The processing module is further used to perform direction alignment processing on the trajectory directions of all obstacles in the normalized trajectory information based on the obstacle movement heading to obtain trajectory features;

[0075] The adding module is used to add a mask to the trajectory feature to obtain the agent coordinate system trajectory feature in the agent coordinate system.

[0076] Furthermore, the second conversion subunit includes:

[0077] An extraction module is used to extract the target map information of the obstacle to be inferred within a preset range at the current moment from the Frenet coordinates of the surrounding lanes; wherein the target map information at least includes the lane centerline node position, lane node displacement, lane connection relationship, road structure and traffic lights;

[0078] The processing module is further used to perform normalization processing on the target map information with the obstacle to be inferred as the center to obtain normalized map information;

[0079] The processing module is further used to perform direction alignment processing on the lane direction in the normalized map information based on the obstacle movement heading to obtain the lane feature in the agent coordinate system.

[0080] Furthermore, the third coordinate conversion unit is specifically used to convert the future trajectory information from the agent coordinate system to the Frenet coordinate system to obtain the future trajectory information of the obstacle to be inferred in the Frenet coordinate system;

[0081] The third coordinate conversion unit is further configured to convert the future trajectory information from the Frenet coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

[0082] A third aspect of the present application provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the obstacle trajectory prediction method described in any one of the first aspects of the present application.

[0083] A fourth aspect of the present application provides a computer-readable storage medium storing computer program instructions, wherein when the computer program instructions are read and executed by a processor, the obstacle trajectory prediction method described in any one of the first aspects of the present application is executed.

[0084] The beneficial effects of the present application are as follows: the method and device can convert curve data into straight roads through Frenet coordinate system conversion, eliminate the differences in maps in the scene, and then perform trajectory prediction through a trajectory prediction model with high accuracy. At the same time, it also improves the accuracy of curve prediction and enhances the autonomous driving experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0086] Figure 1 A flowchart of an obstacle trajectory prediction method provided in an embodiment of the present application;

[0087] Figure 2 A schematic diagram of another obstacle trajectory prediction method provided in an embodiment of the present application;

[0088] Figure 3 A schematic diagram of the result of predicting the future trajectory of a continuous curve by a trajectory prediction model provided in an embodiment of the present application;

[0089] Figure 4 A schematic diagram of the result of converting the model input features into the Frenet coordinate system and then converting them into the agent coordinate system for trajectory prediction provided by an embodiment of the present application;

[0090] Figure 5 A schematic diagram of converting the prediction results into a Cartesian coordinate system provided in an embodiment of the present application;

[0091] Figure 6 A schematic diagram of the structure of an obstacle trajectory prediction device provided in an embodiment of the present application;

[0092] Figure 7A schematic diagram of the structure of another obstacle trajectory prediction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0093] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0094] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0095] Example 1

[0096] Please see Figure 1 , Figure 1 A flowchart of an obstacle trajectory prediction method provided in this embodiment. The obstacle trajectory prediction method includes:

[0097] S101. Collect vehicle perception data of a target vehicle on a curve through a sensor device.

[0098] S102: extract features from vehicle perception data to obtain global feature information in a Cartesian coordinate system.

[0099] S103: Determine a reference line according to the global feature information.

[0100] S104: Acquire at least one obstacle to be inferred that needs to be inferred, and extract obstacle features of the obstacle to be inferred from the global feature information.

[0101] S105 . Transform the obstacle feature from the Cartesian coordinate system to the Frenet coordinate system according to the reference line to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system.

[0102] S106: Convert the first feature information from the Frenet coordinate system to the agent coordinate system to obtain second feature information of the obstacle to be inferred in the agent coordinate system.

[0103] In this embodiment, each obstacle that needs to be inferred in the scene acts as an agent in turn, that is, the method converts the features in the Frenet coordinate system to the agent coordinate system in turn.

[0104] S107 . Predict the future trajectory of the obstacle according to the pre-built trajectory prediction model and the second feature information to obtain the future trajectory information of the obstacle to be inferred.

[0105] S108, converting the future trajectory information from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

[0106] Generally speaking, the amount of straight data in the public trajectory prediction dataset is much larger than the amount of curve data, and the same is true for the self-collected data. This makes the model's straight prediction generalization ability better than curve prediction (straight prediction effect is better than curve prediction effect). In addition, the distribution of curve map feature data varies greatly, which will limit the generalization ability of the model.

[0107] At present, the commonly used Cartesian coordinate system (Cartesian coordinate system) is difficult to express the relative position relationship between the vehicle and the road, which will lead to a fuzzy relative relationship between the two. The Frenet coordinate system (a rectangular coordinate system based on a reference line) is a coordinate system based on curves and can better describe the position and motion state of objects on curves.

[0108] Specifically, the Frenet coordinate system uses the center line of the road as a reference line (the center line is a series of discrete points output by the map module), and then uses the tangent vector and normal vector of the reference line to establish a coordinate system. The ordinate represents the distance the vehicle travels along the road, and the abscissa represents the distance the vehicle deviates from the center line of the road. This ensures that the abscissa and ordinate at each point on the road are perpendicular. Compared with the Cartesian coordinate system, a curve can be regarded as a straight road in the Frenet coordinate system.

[0109] Therefore, this application uses the Frenet coordinate system to convert curves into straight roads, reduce the differences in the distribution of curve map features, and use the strong generalization ability of straight road prediction to improve the curve prediction effect. Among them, by converting the curve data into a straight road through the Frenet coordinate system conversion, the differences in the maps in the scene can be eliminated, making the input more standardized, and then performing model reasoning, which can also reduce the requirements for the model. After the reasoning is completed, the prediction line is transformed back to the Cartesian coordinate system, thereby improving the accuracy of curve prediction, effectively reducing problems such as braking caused by predicting straight lines in curves as cut-ins or cut-ins being incorrectly predicted as straight lines, and improving the autonomous driving experience.

[0110] For example, the method can first use sensor equipment to collect the trajectory information of the vehicle on the curve and the surrounding map information; then, extract the global feature information required for trajectory prediction from the collected data; then, select a reference line and convert the historical trajectory coordinates and surrounding lane coordinates in the global features from the Cartesian coordinate system to the Frenet coordinate system based on the reference line, and other information not involving coordinates remains unchanged; then, convert the features in the Frenet coordinate system to the agent coordinate system, and then send them to the trajectory prediction model for trajectory prediction; finally, convert the prediction results from the Frenet coordinate system to the Cartesian coordinate system.

[0111] In this embodiment, the execution subject of the method may be a computing device such as a computer or a server, and no limitation is made in this embodiment.

[0112] In this embodiment, the execution subject of the method may also be a smart device such as a smart phone, a tablet computer, etc., which is not limited in this embodiment.

[0113] It can be seen that the obstacle trajectory prediction method described in this embodiment can convert the curve data into a straight road through the Frenet coordinate system conversion, eliminate the difference in the map in the scene, and then perform trajectory prediction through the trajectory prediction model with high accuracy. It also improves the accuracy of curve prediction and enhances the autonomous driving experience.

[0114] Example 2

[0115] Please see Figure 2 , Figure 2 A flowchart of an obstacle trajectory prediction method provided in this embodiment. The obstacle trajectory prediction method includes:

[0116] S201. Collect vehicle perception data of a target vehicle on a curve through a sensor device.

[0117] In this embodiment, the vehicle perception data includes at least vehicle trajectory information and vehicle surrounding map information.

[0118] In this embodiment, the method can use sensor equipment (such as lidar, camera, millimeter wave radar, etc.) to perceive and locate the surrounding environment, and collect trajectory information of surrounding traffic participants and positioning map information.

[0119] S202: extract features from the vehicle perception data to obtain global feature information in a Cartesian coordinate system.

[0120] In this embodiment, the global feature information includes historical trajectory information and surrounding map information of all obstacles within a preset range; wherein the historical trajectory information includes at least position information, speed information and movement direction; and the surrounding map information includes at least lane center point position information and lane type.

[0121] In this embodiment, the method can extract historical trajectory information and surrounding map information of all obstacles within a specified range from the perception data. The historical trajectory information includes but is not limited to location information, speed information, and movement direction. The surrounding map information includes but is not limited to lane center location information and lane type.

[0122] S203: Determine the scene range where the target vehicle is currently located and the target lane where the target vehicle is located according to the global feature information.

[0123] S204: Use a depth-first search algorithm to search for other lanes that have a predecessor-successor reachable relationship with the target lane within the scene range.

[0124] S205: When the other lanes include multiple preceding and succeeding lanes, select a preceding and succeeding lane from the other lanes as the preceding and succeeding reachable lane in the scene.

[0125] In this embodiment, if the lane is long enough, the front lane and the following lane are not needed, and only the target lane is sufficient. It can be seen that the existence of this step is to make the reference line cover the scene range as much as possible to facilitate Frenet projection.

[0126] S206. Combine the target lane and the preceding and succeeding reachable lanes in the scene in order from the rear of the target vehicle to the front of the target vehicle to obtain a reference line.

[0127] In this embodiment, since the Frenet coordinate conversion requires the selection of a reference line, the coverage of the reference line needs to cover the history and trajectory of the obstacle to be inferred, the future trajectory, and the lane feature range as much as possible.

[0128] In this embodiment, the method uses the lane where the main vehicle in the scene is located and the predecessor and successor lanes within the scene range as reference lines. The reason is: when the vehicle to be predicted is a vehicle near the main vehicle, the lane where the main vehicle is located and the predecessor and successor are used as reference lines, so that the obstacles to be inferred and the nearby map can be within the reference line range as much as possible. Among them, if there are multiple predecessors and successors, only one is selected. In addition, a depth-first search is used to determine the predecessor of the predecessor, the successor of the successor, and so on, until all lanes in the scene range that have a predecessor and successor reachable relationship with the lane where the main vehicle is located are selected, and multiple lanes are combined in a fixed order from the rear of the main vehicle to the front of the main vehicle to form a reference line.

[0129] S207: Acquire at least one obstacle to be inferred that needs to be inferred, and extract obstacle features of the obstacle to be inferred from the global feature information.

[0130] S208. Transform the obstacle feature from the Cartesian coordinate system to the Frenet coordinate system according to the reference line to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system.

[0131] As an optional implementation, the obstacle feature is converted from the Cartesian coordinate system to the Frenet coordinate system according to the reference line to obtain the first feature information of the obstacle to be inferred in the Frenet coordinate system, including:

[0132] Obtain the obstacle history trajectory coordinates and surrounding lane coordinates of the obstacle to be inferred from the obstacle features;

[0133] Based on the reference line, the coordinates of the obstacle history trajectory are converted from the Cartesian coordinate system to the Frenet coordinate system to obtain the Frenet coordinates of the obstacle history trajectory;

[0134] Based on the reference line, the coordinates of the surrounding lanes are converted from the Cartesian coordinate system to the Frenet coordinate system to obtain the Frenet coordinates of the surrounding lanes;

[0135] The Frenet coordinates of the obstacle's historical trajectory and the Frenet coordinates of the surrounding lanes are summarized to obtain the first feature information of the obstacle to be inferred in the Frenet coordinate system.

[0136] In this embodiment, the method can convert the historical trajectory coordinates and surrounding lane coordinate points (x, y) in the global features from the Cartesian coordinate system to the Frenet coordinate system based on the reference line in C, and each point is changed from (x, y) to (s, l). In the curved road scene, after the Frenet coordinate system conversion, the curved road map coordinate information will become a straight road, and the curved road historical trajectory will become a straight road. Other global features that do not involve position, such as speed information, direction information, and lane type information, remain unchanged.

[0137] S209: Convert the first feature information from the Frenet coordinate system to the agent coordinate system to obtain second feature information of the obstacle to be inferred in the agent coordinate system.

[0138] As an optional implementation, the first feature information is converted from the Frenet coordinate system to the agent coordinate system to obtain the second feature information of the obstacle to be inferred in the agent coordinate system, including:

[0139] Extract the target trajectory information of the obstacle to be inferred from the Frenet coordinates of the obstacle's historical trajectory; wherein the target trajectory information at least includes the current position of the trajectory, the displacement of the trajectory in adjacent frames, the obstacle speed, and the obstacle movement direction;

[0140] Taking the obstacle to be inferred as the center, the Frenet coordinates of the obstacle's historical trajectory are converted from the Frenet coordinate system to the agent coordinate system to obtain the trajectory features of the agent coordinate system in the agent coordinate system;

[0141] Taking the obstacle to be inferred as the center, the Frenet coordinates of the surrounding lanes are converted from the Frenet coordinate system to the agent coordinate system to obtain the lane features in the agent coordinate system;

[0142] The trajectory features and lane features of the agent coordinate system are summarized to obtain the second feature information of the obstacle to be inferred in the agent coordinate system.

[0143] As a further optional implementation, with the obstacle to be inferred as the center, the Frenet coordinates of the obstacle's historical trajectory are converted from the Frenet coordinate system to the agent coordinate system to obtain the agent coordinate system trajectory features in the agent coordinate system, including:

[0144] The target trajectory information is normalized with the obstacle to be inferred as the center to obtain the normalized trajectory information;

[0145] Obtain the obstacle motion heading of the obstacle to be inferred;

[0146] Based on the obstacle movement heading, the trajectory directions of all obstacles in the normalized trajectory information are aligned to obtain trajectory features;

[0147] Add a mask to the trajectory feature to obtain the agent coordinate system trajectory feature in the agent coordinate system.

[0148] As a further optional implementation, with the obstacle to be inferred as the center, the Frenet coordinates of the surrounding lanes are converted from the Frenet coordinate system to the agent coordinate system to obtain the lane features in the agent coordinate system, including:

[0149] Extract the target map information of the obstacle to be inferred within the preset range at the current moment from the Frenet coordinates of the surrounding lanes; wherein the target map information at least includes the lane centerline node position, lane node displacement, lane connection relationship, road structure and traffic lights;

[0150] Taking the obstacle to be inferred as the center, the target map information is normalized to obtain the normalized map information;

[0151] The lane direction in the normalized map information is aligned based on the obstacle motion heading to obtain the lane features in the agent coordinate system.

[0152] In this embodiment, after the Frenet coordinate system conversion, the method can use the converted coordinate origin as the starting point of the reference line, and select the agent. The perception information and map information within the specified range around the agent constitute a scene. At this time, the scene is represented in a structured manner, and the trajectory information of all obstacles (including the current position of the trajectory, the displacement of the trajectory in adjacent frames, the obstacle speed, and the obstacle movement direction) is extracted from the perception data. The model reasoning needs to convert the features to the agent coordinate system. After the agent is selected, each scene coordinate feature is normalized to be centered on the agent (the obstacle to be inferred) and the trajectory direction and lane direction of all actors (all obstacles in the scene) are aligned with the agent heading. The trajectory length is T, and 0 is filled if it is less than T. In addition, the method can also add a binary 1×T mask to the trajectory feature to indicate whether there are missing values. Extract the map information within a certain range of the obstacle at the current moment from the map data (including the lane centerline node position, lane node displacement, lane connection relationship, road structure, traffic lights, etc., which constitute map features), and convert the map position information in the map scene to be centered on the agent and aligned with the agent's heading. Among them, these features can provide a basis for subsequent motion prediction.

[0153] S210 , predicting the future trajectory of the obstacle according to the pre-built trajectory prediction model and the second feature information, and obtaining the future trajectory information of the obstacle to be inferred.

[0154] In this embodiment, the method can integrate the features in the above-mentioned agent coordinate system into the trajectory prediction model to perform future trajectory prediction.

[0155] S211, converting the future trajectory information from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

[0156] As an optional implementation, the future trajectory information is converted from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system, including:

[0157] Convert the future trajectory information from the agent coordinate system to the Frenet coordinate system to obtain the future trajectory information of the obstacle to be inferred in the Frenet coordinate system;

[0158] The future trajectory information is converted from the Frenet coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

[0159] In this embodiment, the method can convert the future trajectory information obtained by model inference from the agent coordinate system back to the Frenet coordinate system, and then inversely transform from the Frenet coordinate system back to the Cartesian coordinate system, thereby obtaining the future trajectory information in the Cartesian global coordinate system.

[0160] Please see Figure 3 , Figure 4 and Figure 5 .in, Figure 3 The figure shows a schematic diagram of the result of a trajectory prediction model predicting the future trajectory of a continuous curve, including the center point of the lane, the historical trajectory of obstacles in the scene, the historical obstacle trajectory of the agent to be predicted, and the future trajectory predicted by the model; the intersection point of the historical obstacle trajectory of the agent to be predicted and the future trajectory predicted by the model is the current coordinate position. One thing to note is that Figure 3 The predicted future trajectory is wrong.

[0161] Figure 4 The result of converting the model input features into the Frenet coordinate system and then into the agent coordinate system for trajectory prediction is that the future trajectory prediction is correct.

[0162] Figure 5 for Figure 4 The prediction results in are converted into a schematic diagram of the results in the Cartesian coordinate system. The future trajectory prediction is correct.

[0163] Depend on Figures 3 to 5 It can be seen that this method can correctly predict the future trajectory of obstacles in many scenes where the original curves were not well predicted.

[0164] In this embodiment, the execution subject of the method may be a computing device such as a computer or a server, and no limitation is made in this embodiment.

[0165] In this embodiment, the execution subject of the method may also be a smart device such as a smart phone, a tablet computer, etc., which is not limited in this embodiment.

[0166] It can be seen that the obstacle trajectory prediction method described in this embodiment can convert the curve data into a straight road through the Frenet coordinate system conversion, eliminate the difference in the map in the scene, and then perform trajectory prediction through the trajectory prediction model with high accuracy. It also improves the accuracy of curve prediction and enhances the autonomous driving experience.

[0167] Example 3

[0168] Please see Figure 3 , Figure 3 FIG. 1 is a schematic diagram of the structure of an obstacle trajectory prediction device provided in this embodiment. Figure 3 As shown, the obstacle trajectory prediction device includes:

[0169] A collection unit 310 is used to collect vehicle perception data of a target vehicle on a curve through a sensor device;

[0170] A feature extraction unit 320 is used to extract features from the vehicle perception data to obtain global feature information in a Cartesian coordinate system;

[0171] A determination unit 330, configured to determine a reference line according to the global feature information;

[0172] An acquisition unit 340 is used to acquire at least one obstacle to be inferred and extract obstacle features of the obstacle to be inferred from the global feature information;

[0173] A first coordinate conversion unit 350 is used to convert the obstacle feature from the Cartesian coordinate system to the Frenet coordinate system according to the reference line, so as to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system;

[0174] A second coordinate conversion unit 360, used to convert the first feature information from the Frenet coordinate system to the agent coordinate system to obtain second feature information of the obstacle to be inferred in the agent coordinate system;

[0175] A trajectory prediction unit 370 is used to predict the future trajectory of the obstacle according to the pre-built trajectory prediction model and the second feature information to obtain the future trajectory information of the obstacle to be inferred;

[0176] The third coordinate conversion unit 380 is used to convert the future trajectory information from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

[0177] In this embodiment, the explanation of the obstacle trajectory prediction device can refer to the description in Example 1 or Example 2, and will not be further elaborated in this embodiment.

[0178] It can be seen that the obstacle trajectory prediction device described in this embodiment can convert the curve data into a straight road through the Frenet coordinate system conversion, eliminate the difference in the map in the scene, and then perform trajectory prediction through the trajectory prediction model with high accuracy. It also improves the accuracy of curve prediction and enhances the autonomous driving experience.

[0179] Example 4

[0180] Please see Figure 7 , Figure 7 FIG. 1 is a schematic diagram of the structure of an obstacle trajectory prediction device provided in this embodiment. Figure 7 As shown, the obstacle trajectory prediction device includes:

[0181] A collection unit 310 is used to collect vehicle perception data of a target vehicle on a curve through a sensor device;

[0182] A feature extraction unit 320 is used to extract features from the vehicle perception data to obtain global feature information in a Cartesian coordinate system;

[0183] A determination unit 330, configured to determine a reference line according to the global feature information;

[0184] An acquisition unit 340 is used to acquire at least one obstacle to be inferred and extract obstacle features of the obstacle to be inferred from the global feature information;

[0185] A first coordinate conversion unit 350 is used to convert the obstacle feature from the Cartesian coordinate system to the Frenet coordinate system according to the reference line, so as to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system;

[0186] A second coordinate conversion unit 360, used to convert the first feature information from the Frenet coordinate system to the agent coordinate system to obtain second feature information of the obstacle to be inferred in the agent coordinate system;

[0187] A trajectory prediction unit 370 is used to predict the future trajectory of the obstacle according to the pre-built trajectory prediction model and the second feature information to obtain the future trajectory information of the obstacle to be inferred;

[0188] The third coordinate conversion unit 380 is used to convert the future trajectory information from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

[0189] In this embodiment, the vehicle perception data includes at least vehicle trajectory information and vehicle surrounding map information;

[0190] The global feature information includes the historical trajectory information and surrounding map information of all obstacles within a preset range; wherein the historical trajectory information includes at least the position information, speed information and movement direction; and the surrounding map information includes at least the lane center point position information and lane type.

[0191] As an optional implementation manner, the determining unit 330 includes:

[0192] A determination subunit 331 is used to determine the scene range where the target vehicle is currently located and the target lane where the target vehicle is located according to the global feature information;

[0193] A search subunit 332 is used to use a depth-first search algorithm to search for other lanes that have a predecessor and successor reachable relationship with the target lane within the scene range;

[0194] The selection subunit 333 is used to select a preceding and succeeding lane from the other lanes as the preceding and succeeding reachable lane in the scene when the other lanes include multiple preceding and succeeding lanes;

[0195] The generating subunit 334 is used to generate a reference line according to the target lane and the preceding and succeeding reachable lanes in the scene.

[0196] As an optional implementation, the generating subunit 334 is specifically used to combine the target lane and the preceding and succeeding reachable lanes in the scene in order from the rear of the target vehicle to the front of the target vehicle to obtain a reference line.

[0197] As an optional implementation manner, the first coordinate conversion unit 350 includes:

[0198] An acquisition subunit 351 is used to acquire the obstacle history trajectory coordinates and surrounding lane coordinates of the obstacle to be inferred from the obstacle features;

[0199] A first conversion subunit 352 is used to convert the coordinates of the obstacle history trajectory from the Cartesian coordinate system to the Frenet coordinate system based on the reference line to obtain the Frenet coordinates of the obstacle history trajectory;

[0200] The first conversion subunit 352 is further used to convert the surrounding lane coordinates from the Cartesian coordinate system to the Frenet coordinate system based on the reference line to obtain the surrounding lane Frenet coordinates;

[0201] The first summarizing subunit 353 is used to summarize the historical trajectory Frenet coordinates of the obstacle and the surrounding lane Frenet coordinates to obtain the first feature information of the obstacle to be inferred in the Frenet coordinate system.

[0202] As an optional implementation, the second coordinate conversion unit 360 includes:

[0203] The extraction subunit 361 is used to extract the target trajectory information of the obstacle to be inferred from the Frenet coordinates of the obstacle historical trajectory; wherein the target trajectory information at least includes the current position of the trajectory, the displacement of the trajectory in adjacent frames, the obstacle speed, and the obstacle movement direction;

[0204] The second conversion subunit 362 is used to convert the obstacle historical trajectory Frenet coordinates from the Frenet coordinate system to the agent coordinate system with the obstacle to be inferred as the center, and obtain the agent coordinate system trajectory features in the agent coordinate system;

[0205] The second conversion subunit 362 is further used to convert the surrounding lane Frenet coordinates from the Frenet coordinate system to the agent coordinate system with the obstacle to be inferred as the center, so as to obtain the lane features in the agent coordinate system;

[0206] The second summarizing subunit 363 is used to summarize the trajectory features and lane features of the agent coordinate system to obtain second feature information of the obstacle to be inferred in the agent coordinate system.

[0207] As an optional implementation, the second conversion subunit 362 includes:

[0208] A processing module, used for normalizing the target trajectory information with the obstacle to be inferred as the center to obtain normalized trajectory information;

[0209] An acquisition module, used to acquire the obstacle motion heading of the obstacle to be inferred;

[0210] The processing module is further used to perform direction alignment processing on the trajectory directions of all obstacles in the normalized trajectory information based on the obstacle movement heading to obtain trajectory features;

[0211] Add a module to add a mask to the trajectory feature and obtain the agent coordinate system trajectory feature in the agent coordinate system.

[0212] As an optional implementation, the second conversion subunit 362 includes:

[0213] An extraction module is used to extract the target map information of the obstacle to be inferred within the preset range at the current moment from the Frenet coordinates of the surrounding lanes; wherein the target map information at least includes the lane centerline node position, lane node displacement, lane connection relationship, road structure and traffic lights;

[0214] The processing module is also used to normalize the target map information with the obstacle to be inferred as the center to obtain normalized map information;

[0215] The processing module is also used to align the lane direction in the normalized map information based on the obstacle movement heading to obtain the lane feature in the agent coordinate system.

[0216] As an optional implementation, the third coordinate conversion unit 380 is specifically used to convert the future trajectory information from the agent coordinate system to the Frenet coordinate system to obtain the future trajectory information of the obstacle to be inferred in the Frenet coordinate system;

[0217] The third coordinate conversion unit 380 is further configured to convert the future trajectory information from the Frenet coordinate system to the Cartesian coordinate system, so as to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

[0218] In this embodiment, the explanation of the obstacle trajectory prediction device can refer to the description in Example 1 or Example 2, and will not be further elaborated in this embodiment.

[0219] It can be seen that the obstacle trajectory prediction device described in this embodiment can convert the curve data into a straight road through the Frenet coordinate system conversion, eliminate the difference in the map in the scene, and then perform trajectory prediction through the trajectory prediction model with high accuracy. It also improves the accuracy of curve prediction and enhances the autonomous driving experience.

[0220] An embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the obstacle trajectory prediction method in Embodiment 1 or Embodiment 2 of the present application.

[0221] An embodiment of the present application provides a computer-readable storage medium storing computer program instructions. When the computer program instructions are read and executed by a processor, the obstacle trajectory prediction method in Embodiment 1 or Embodiment 2 of the present application is executed.

[0222] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0223] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0224] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0225] The above description is only an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0226] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0227] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

Claims

1. A method for predicting obstacle trajectories, characterized in that: include: Collect vehicle perception data of target vehicles on the curve through sensor equipment; Extracting features from the vehicle perception data to obtain global feature information in a Cartesian coordinate system; Determine a reference line according to the global feature information; Acquire at least one obstacle to be inferred that needs to be inferred, and extract obstacle features of the obstacle to be inferred from the global feature information; Converting the obstacle feature from the Cartesian coordinate system to the Frenet coordinate system according to the reference line to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system; Converting the first feature information from the Frenet coordinate system to the agent coordinate system to obtain second feature information of the obstacle to be inferred in the agent coordinate system; Predicting the future trajectory of the obstacle according to the pre-built trajectory prediction model and the second feature information to obtain the future trajectory information of the obstacle to be inferred; The future trajectory information is converted from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

2. The obstacle trajectory prediction method according to claim 1, characterized in that: The vehicle perception data includes at least vehicle trajectory information and vehicle surrounding map information; The global feature information includes historical trajectory information and surrounding map information of all obstacles within a preset range; wherein the historical trajectory information includes at least position information, speed information and movement direction; and the surrounding map information includes at least lane center point position information and lane type.

3. The obstacle trajectory prediction method according to claim 1, characterized in that: The determining of the reference line according to the global feature information includes: Determine the scene range where the target vehicle is currently located and the target lane where the target vehicle is located according to the global feature information; Using a depth-first search algorithm within the scope of the scene to search for other lanes that have a predecessor-successor reachable relationship with the target lane; If the other lanes include multiple preceding and succeeding lanes, selecting a preceding and succeeding lane from the other lanes as the preceding and succeeding reachable lane in the scene; A reference line is generated according to the target lane and the preceding and succeeding reachable lanes in the scene.

4. The obstacle trajectory prediction method according to claim 3, characterized in that: The generating of a reference line according to the target lane and the preceding and succeeding reachable lanes in the scene includes: The target lane and the preceding and succeeding reachable lanes in the scene are combined in order from the rear of the target vehicle to the front of the target vehicle to obtain a reference line.

5. The obstacle trajectory prediction method according to claim 1, characterized in that: The step of converting the obstacle feature from the Cartesian coordinate system to the Frenet coordinate system according to the reference line to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system includes: Obtaining the obstacle history trajectory coordinates and surrounding lane coordinates of the obstacle to be inferred from the obstacle features; Based on the reference line, the coordinates of the obstacle historical trajectory are converted from a Cartesian coordinate system to a Frenet coordinate system to obtain Frenet coordinates of the obstacle historical trajectory; Based on the reference line, the coordinates of the surrounding lanes are converted from a Cartesian coordinate system to a Frenet coordinate system to obtain Frenet coordinates of the surrounding lanes; The obstacle historical trajectory Frenet coordinates and the surrounding lane Frenet coordinates are summarized to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system.

6. The obstacle trajectory prediction method according to claim 5, characterized in that: The converting the first feature information from the Frenet coordinate system to the agent coordinate system to obtain the second feature information of the obstacle to be inferred in the agent coordinate system includes: Extracting the target trajectory information of the obstacle to be inferred from the Frenet coordinates of the obstacle historical trajectory; wherein the target trajectory information at least includes the current position of the trajectory, the displacement of the trajectory in adjacent frames, the obstacle speed, and the obstacle movement direction; Taking the obstacle to be inferred as the center, converting the Frenet coordinates of the obstacle's historical trajectory from the Frenet coordinate system to the agent coordinate system to obtain the agent coordinate system trajectory features in the agent coordinate system; Taking the obstacle to be inferred as the center, converting the Frenet coordinates of the surrounding lanes from the Frenet coordinate system to the agent coordinate system to obtain lane features in the agent coordinate system; The agent coordinate system trajectory feature and the lane feature are summarized to obtain second feature information of the obstacle to be inferred in the agent coordinate system.

7. The obstacle trajectory prediction method according to claim 6, characterized in that: The process of converting the obstacle historical trajectory Frenet coordinates from the Frenet coordinate system to the agent coordinate system with the obstacle to be inferred as the center to obtain the agent coordinate system trajectory features in the agent coordinate system includes: Normalizing the target trajectory information with the obstacle to be inferred as the center to obtain normalized trajectory information; Obtaining the obstacle motion heading of the obstacle to be inferred; Based on the obstacle movement heading, performing direction alignment processing on the trajectory directions of all obstacles in the normalized trajectory information to obtain trajectory features; A mask is added to the trajectory feature to obtain the agent coordinate system trajectory feature in the agent coordinate system.

8. The obstacle trajectory prediction method according to claim 7, characterized in that: The method of converting the surrounding lane Frenet coordinates from the Frenet coordinate system to the agent coordinate system with the obstacle to be inferred as the center to obtain lane features in the agent coordinate system includes: Extracting target map information of the obstacle to be inferred within a preset range at the current moment from the surrounding lane Frenet coordinates; wherein the target map information at least includes lane centerline node position, lane node displacement, lane connection relationship, road structure and traffic lights; Taking the obstacle to be inferred as the center, normalizing the target map information to obtain normalized map information; Direction alignment processing is performed on the lane direction in the normalized map information based on the obstacle movement heading to obtain lane features in the agent coordinate system.

9. The obstacle trajectory prediction method according to claim 1, characterized in that: The converting the future trajectory information from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system includes: Converting the future trajectory information from the agent coordinate system to the Frenet coordinate system to obtain the future trajectory information of the obstacle to be inferred in the Frenet coordinate system; The future trajectory information is converted from the Frenet coordinate system to the Cartesian coordinate system to obtain target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

10. An obstacle trajectory prediction device, characterized in that: The obstacle trajectory prediction device comprises: A collection unit, used for collecting vehicle perception data of a target vehicle on a curve through a sensor device; A feature extraction unit, used to extract features from the vehicle perception data to obtain global feature information in a Cartesian coordinate system; A determination unit, configured to determine a reference line according to the global feature information; an acquisition unit, configured to acquire at least one obstacle to be inferred and extract obstacle features of the obstacle to be inferred from the global feature information; A first coordinate conversion unit, configured to convert the obstacle feature from the Cartesian coordinate system to a Frenet coordinate system according to the reference line, so as to obtain first feature information of the obstacle to be inferred in the Frenet coordinate system; a second coordinate conversion unit, configured to convert the first feature information from the Frenet coordinate system to an agent coordinate system, to obtain second feature information of the obstacle to be inferred in the agent coordinate system; A trajectory prediction unit, configured to predict the future trajectory of the obstacle according to a pre-built trajectory prediction model and the second feature information, and obtain the future trajectory information of the obstacle to be inferred; The third coordinate conversion unit is used to convert the future trajectory information from the agent coordinate system to the Cartesian coordinate system to obtain the target future trajectory information of the obstacle to be inferred in the Cartesian coordinate system.

11. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the obstacle trajectory prediction method according to any one of claims 1 to 9.

12. A readable storage medium, characterized in that: The readable storage medium stores computer program instructions, and when the computer program instructions are read and executed by a processor, the obstacle trajectory prediction method according to any one of claims 1 to 9 is executed.

Images