Vehicle track prediction method and device

Abstract

The invention provides a vehicle trajectory prediction method and device. The method comprises the following steps: acquiring historical sample training data, historical trajectory data of a target vehicle and traffic participants around the target vehicle, and lane information of a road; training the initially constructed trajectory prediction model by using historical sample training data to obtain a target trajectory prediction model; inputting the historical trajectory data and the lane information into a target trajectory prediction model to obtain a longitudinal acceleration prediction sequence and a transverse course angle prediction sequence of the target vehicle; based on the longitudinal acceleration prediction sequence and the transverse course angle prediction sequence, track integration is carried out through a kinematic model, and an initial prediction track is obtained; and performing uncertainty modeling on the initial prediction trajectory to generate a probability trajectory of the target vehicle. Historical trajectory data are predicted through a target trajectory prediction model, and high-precision prediction in a complex traffic scene is realized by combining a kinematic model and uncertainty modeling.

Description

Technical Field

[0001] This invention relates to the field of motion planning technology, and in particular to a method and apparatus for predicting vehicle trajectories. Background Technology

[0002] With the rapid development of autonomous driving technology, vehicle trajectory prediction has become a core component of autonomous driving systems. Its goal is to accurately predict the vehicle's trajectory over a future period using historical vehicle motion information and surrounding environmental data, which is crucial for path planning, decision-making, and safety warnings. However, practical applications face multiple challenges: traffic scenarios are complex and ever-changing, with intricate interactions between vehicles and pedestrians, other vehicles, and the road environment; vehicle motion is constrained by multiple factors such as dynamics, road rules, and traffic signals; and the autonomous driving system needs to make accurate predictions within a short timeframe to ensure safe and efficient driving.

[0003] Existing trajectory prediction technologies fall into two main categories: physics-based and data-driven. Physics-based methods predict trajectories by establishing vehicle dynamics and kinematic models, utilizing initial states and control inputs. Their advantage lies in their ability to accurately describe vehicle physical characteristics; however, they struggle to accurately model inter-vehicle interactions in complex traffic scenarios, resulting in limited prediction accuracy and typically only applicable to short-term predictions. With the development of deep learning, data-driven methods have become mainstream. These methods train neural networks and other models using large amounts of trajectory data for prediction, handling complex scenarios. However, they suffer from three major drawbacks: first, they struggle to fully capture complex interactions among traffic participants (such as multi-vehicle avoidance and following behavior); second, they lack road rule constraints (such as lane restrictions and traffic signals), leading to predicted trajectories that may violate actual driving rules; and third, in long-term predictions, the lack of dynamic constraints significantly reduces accuracy.

[0004] In other words, existing trajectory prediction schemes lack sufficient accuracy and reliability in complex traffic scenarios, making it difficult to meet the decision-making and planning requirements of autonomous driving. Traditional physical models are limited by their interaction modeling capabilities and are only suitable for short-term predictions; while deep learning methods can handle complex scenarios, they suffer from problems such as insufficient capture of interaction relationships, lack of rule constraints, and long-term prediction accuracy degradation. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art, and proposes a vehicle route planning method and apparatus.

[0006] In a first aspect, embodiments of the present invention provide a vehicle trajectory prediction method, comprising:

[0007] Acquire historical sample training data, historical trajectory data of the target vehicle and surrounding traffic participants, and lane information of the road;

[0008] Using the historical sample training data, the initially constructed trajectory prediction model is trained to obtain the target trajectory prediction model;

[0009] The historical trajectory data and the lane information are input into the target trajectory prediction model to obtain the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence of the target vehicle.

[0010] Based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence, the initial predicted trajectory is obtained by trajectory integration through the kinematic model;

[0011] Uncertainty modeling is performed on the initial predicted trajectory to generate the probabilistic trajectory of the target vehicle.

[0012] Optionally, the historical trajectory data and the lane information are input into the target trajectory prediction model to obtain the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence of the target vehicle, including:

[0013] The encoder in the target trajectory prediction model encodes the historical trajectory data and the lane information to obtain a scene feature representation that integrates the interaction relationship between traffic participants and road information.

[0014] The scene feature representation is decoded using the decoder in the target trajectory prediction model to obtain the longitudinal acceleration prediction sequence and the probability distribution of candidate lanes;

[0015] Based on the target candidate lane with the highest probability among the candidate lanes, the lateral heading angle prediction sequence is generated in the Frenet coordinate system using the moving horizon optimization method.

[0016] Optionally, before obtaining the initial predicted trajectory by performing trajectory integration through a kinematic model based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence, the method further includes:

[0017] The acceleration prediction sequence and the lateral heading angle prediction sequence are corrected.

[0018] Optionally, the acceleration prediction sequence and the lateral heading angle prediction sequence are corrected, including:

[0019] The scene feature representation, the longitudinal acceleration prediction sequence, and the lateral heading angle prediction sequence are input into a pre-trained correction sub-network;

[0020] The longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence are fine-tuned by the correction sub-network to output the corrected longitudinal acceleration prediction sequence and the corrected lateral heading angle prediction sequence.

[0021] Optionally, the kinematic model is:

[0022] ;

[0023] in,( , () represents the current position. , () represents the predicted position for the next moment. The speed at the current moment, For time step, This is the predicted lateral heading angle at the current moment.

[0024] Optionally, the longitudinal acceleration prediction sequence is obtained by decoding the scene feature representation through the decoder in the target trajectory prediction model, including:

[0025] Based on the scene feature representation, predict the discrete label of acceleration at each future time step;

[0026] According to the preset mapping relationship between the label and the acceleration value, the discrete acceleration label is mapped to the corresponding continuous acceleration value to form the longitudinal acceleration prediction sequence.

[0027] Optionally, uncertainty modeling is performed on the initial predicted trajectory to generate the probabilistic trajectory of the target vehicle, including:

[0028] Each trajectory point in the initial predicted trajectory is defined as a two-dimensional Gaussian distribution mean to form a mean trajectory;

[0029] A covariance matrix is ​​assigned to each trajectory point in the mean trajectory to characterize the uncertainty at that location;

[0030] The probability trajectory is constructed based on the mean trajectory and the covariance matrix corresponding to each trajectory point.

[0031] Secondly, embodiments of the present invention provide a vehicle trajectory prediction device, comprising:

[0032] The acquisition module is used to acquire historical sample training data, historical trajectory data of the target vehicle and its surrounding traffic participants, and lane information of the road;

[0033] The model training module is used to train the initially constructed trajectory prediction model using the historical sample training data to obtain the target trajectory prediction model.

[0034] The sequence prediction module is used to input the historical trajectory data and the lane information into the target trajectory prediction model to obtain the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence of the target vehicle.

[0035] The initial trajectory acquisition module is used to obtain the initial predicted trajectory by performing trajectory integration through a kinematic model based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence.

[0036] The probability trajectory acquisition module is used to perform uncertainty modeling on the initial predicted trajectory and generate the probability trajectory of the target vehicle.

[0037] Optionally, the sequence prediction module includes:

[0038] The encoding submodule is used to encode the historical trajectory data and the lane information through the encoder in the target trajectory prediction model to obtain a scene feature representation that integrates the interaction relationship between traffic participants and road information;

[0039] The decoding submodule is used to decode the scene feature representation through the decoder in the target trajectory prediction model to obtain the longitudinal acceleration prediction sequence and the probability distribution of the candidate lanes;

[0040] The heading angle prediction submodule is used to generate the lateral heading angle prediction sequence in the Frenet coordinate system using the moving horizon optimization method based on the target candidate lane with the highest probability among the candidate lanes.

[0041] Optionally, the probability trajectory acquisition module includes:

[0042] The mean trajectory forming submodule is used to define each trajectory point in the initial predicted trajectory as a two-dimensional Gaussian distribution mean to form a mean trajectory;

[0043] The covariance matrix allocation submodule is used to allocate a covariance matrix to each trajectory point in the mean trajectory to characterize the uncertainty of that location;

[0044] The probability trajectory construction submodule is used to construct the probability trajectory based on the mean trajectory and the covariance matrix corresponding to each trajectory point.

[0045] Thirdly, the present invention provides an electronic device including a processor and a memory, the memory storing computer-readable instructions that, when executed by the processor, perform the steps of the method provided in the first aspect above.

[0046] Fourthly, the present invention provides a storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the method provided in the first aspect above.

[0047] Fifthly, the present invention provides a computer program product comprising a computer program that, when executed by a processor, performs the steps of the method provided in the first aspect above.

[0048] As can be seen from the above technical solutions, the present invention has the following advantages:

[0049] This invention provides a vehicle trajectory prediction method and apparatus. The method includes: acquiring historical sample training data, historical trajectory data of a target vehicle and surrounding traffic participants, and lane information of the road; training an initially constructed trajectory prediction model using the historical sample training data to obtain a target trajectory prediction model; inputting the historical trajectory data and the lane information into the target trajectory prediction model to obtain a longitudinal acceleration prediction sequence and a lateral heading angle prediction sequence for the target vehicle; performing trajectory integration using a kinematic model based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence to obtain an initial predicted trajectory; and performing uncertainty modeling on the initial predicted trajectory to generate a probabilistic trajectory for the target vehicle. By predicting lane information and historical trajectory data of the target vehicle and surrounding traffic participants using the target trajectory prediction model, the longitudinal acceleration prediction sequence and lateral heading angle prediction sequence for the target vehicle are obtained. Combining the kinematic model and uncertainty modeling, a probabilistic trajectory is obtained, achieving high-precision prediction in complex traffic scenarios. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating a vehicle routing method according to a first embodiment of the present invention.

[0051] Figure 2 This is a flowchart illustrating a second embodiment of the vehicle routing method of the present invention;

[0052] Figure 3 This is a schematic diagram of the encoder data processing flow of a second embodiment of the vehicle routing method of the present invention;

[0053] Figure 4 This is a schematic diagram of the decoder data processing flow of a second embodiment of the vehicle routing method of the present invention.

[0054] Figure 5 This is a schematic diagram of the Frenet coordinate system in a second embodiment of the vehicle routing method of the present invention;

[0055] Figure 6 This is a structural block diagram of an embodiment of the vehicle routing device of the present invention;

[0056] Figure 7 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0057] To enable those skilled in the art to better understand the technical solutions of the present invention, exemplary embodiments of the present invention are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0058] Where there is no conflict, the various embodiments of the present invention and the features thereof may be combined with each other.

[0059] As used herein, the term “and / or” includes any and all combinations of one or more related enumerated entries.

[0060] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used herein, the singular forms “a” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Terms such as “connected” or “linked” are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0061] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having the meaning consistent with their meaning in the context of the relevant art and the invention, and will not be interpreted as having an idealized or overly formal meaning unless expressly so defined herein.

[0062] In the technical solution of this invention, the collection, storage, use, processing, transmission, provision, and disclosure of user personal information all comply with relevant laws and regulations and do not violate public order and good morals. The use of user data in this technical solution follows relevant national laws and regulations (e.g., the "Information Security Technology - Personal Information Security Specification"). For example: appropriate measures are taken for personal information access control; restrictions are imposed on the display of personal information; the purpose of using personal information does not exceed the scope of direct or reasonable association; and explicit identity targeting is eliminated when using personal information to avoid precisely locating a specific individual.

[0063] To address at least one of the technical problems existing in the aforementioned related technologies, the present invention provides a control method for displaying vehicle driving range. Figure 1 This is a flowchart illustrating a vehicle trajectory prediction method provided in an embodiment of the present invention; the method includes:

[0064] Step S101: Obtain historical sample training data, historical trajectory data of the target vehicle and its surrounding traffic participants, and lane information of the road;

[0065] In this embodiment, a large amount of vehicle driving data under different scenarios is collected, including various traffic conditions and road types. This data includes historical trajectory information of the target vehicle and surrounding traffic participants, as well as corresponding lane information. At the same time, using sensors on the vehicle and sensors in the road infrastructure, historical trajectory information of the target vehicle, pedestrians, or obstacles over a period of time is recorded, along with corresponding lane information, including lane geometry (such as lane width, curvature, etc.) and rules (such as speed limits, lane driving direction, etc.).

[0066] Step S102: Using the historical sample training data, train the initially constructed trajectory prediction model to obtain the target trajectory prediction model;

[0067] In the embodiments of this application, the encoder part includes a fully connected network, a nonlinear activation function, a long short-term memory network (LSTM), and a graph attention network; the decoder part combines a neural network and rule-based constraints.

[0068] Step S103: Input the historical trajectory data and the lane information into the target trajectory prediction model to obtain the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence of the target vehicle;

[0069] In this embodiment, for the historical trajectory data of the target vehicle and surrounding traffic participants, features are first embedded using a fully connected network and a nonlinear activation function. Then, a long short-term memory network is used to encode the embedded features to capture the temporal dependencies in the time-series data. Lane information is treated as sequential data and is also encoded using a long short-term memory network. A graph attention network is introduced to model the interaction relationships between traffic participants. By calculating the attention weights between the target vehicle and surrounding traffic participants, information from adjacent traffic participants is aggregated, and the interaction relationships between the target vehicle and surrounding lanes are processed to obtain the feature representation of the target vehicle, the interaction information with surrounding traffic participants, and the feature representation of lane information.

[0070] Subsequently, by decoding the feature representation, the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence of the target vehicle are obtained.

[0071] Step S104: Based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence, the trajectory is integrated through the kinematic model to obtain the initial predicted trajectory;

[0072] By using a kinematic model to perform trajectory integration, the position at each future time step by step is calculated to obtain the initial predicted trajectory.

[0073] Step S105: Perform uncertainty modeling on the initial predicted trajectory to generate the probability trajectory of the target vehicle.

[0074] Using a Gaussian distribution to describe the uncertainty of the final position, the initial predicted trajectory is transformed into a probabilistic trajectory by assigning a covariance matrix to each trajectory point of the initial predicted trajectory.

[0075] This invention provides a vehicle trajectory prediction method. It acquires historical sample training data, historical trajectory data of the target vehicle and surrounding traffic participants, and road lane information. Using the historical sample training data, an initially constructed trajectory prediction model is trained to obtain a target trajectory prediction model. The historical trajectory data and lane information are input into the target trajectory prediction model to obtain a longitudinal acceleration prediction sequence and a lateral heading angle prediction sequence for the target vehicle. Based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence, trajectory integration is performed using a kinematic model to obtain an initial predicted trajectory. Uncertainty modeling is applied to the initial predicted trajectory to generate a probabilistic trajectory for the target vehicle. By predicting lane information and historical trajectory data of the target vehicle and surrounding traffic participants using the target trajectory prediction model, the longitudinal acceleration prediction sequence and lateral heading angle prediction sequence for the target vehicle are obtained. Combining the kinematic model and uncertainty modeling, a probabilistic trajectory is obtained, achieving high-precision prediction in complex traffic scenarios.

[0076] This invention also provides a vehicle trajectory prediction method; please refer to [link / reference]. Figure 2 , Figure 2 This is a flowchart illustrating a second embodiment of the vehicle trajectory prediction method of the present invention. The steps include:

[0077] Step S201: Obtain historical sample training data, historical trajectory data of the target vehicle and its surrounding traffic participants, and lane information of the road;

[0078] Step S202: Using the historical sample training data, train the initially constructed trajectory prediction model to obtain the target trajectory prediction model;

[0079] In this embodiment, during training, historical sample training data is input into the initial model, and the model parameters are continuously adjusted using the backpropagation algorithm to minimize the error between the model's prediction and the actual trajectory. Common loss functions, such as mean squared error loss, can be used during training.

[0080] Step S203: The historical trajectory data and lane information are encoded by the encoder in the target trajectory prediction model to obtain a scene feature representation that integrates the interaction relationship between traffic participants and road information;

[0081] In this embodiment, historical trajectory data and lane information are encoded using a fully connected network, a nonlinear activation function, a long short-term memory network, and a graph attention network to obtain feature representations of the target vehicle, interaction information with surrounding traffic participants, and lane information, such as... Figure 3 As shown, this information is fused together to form a scene feature representation.

[0082] Step S204: The scene feature representation is decoded by the decoder in the target trajectory prediction model to obtain the longitudinal acceleration prediction sequence and the probability distribution of the candidate lanes;

[0083] The decoder section combines neural networks and rule-based constraints to predict longitudinal acceleration and lateral trajectory, generating future trajectory predictions that conform to vehicle dynamics and road rules. The decoder structure is as follows: Figure 4 As shown in the diagram. This part not only considers the vehicle's dynamic characteristics but also improves prediction accuracy through neural network correction, providing reliable trajectory prediction results for autonomous driving decisions.

[0084] In this embodiment of the application, based on the scene feature representation, the discrete acceleration label for each future time step is predicted; according to the preset mapping relationship between the label and the acceleration value, the discrete acceleration label is mapped to the corresponding continuous acceleration value to form the longitudinal acceleration prediction sequence.

[0085] In the specific implementation, during decoding, based on the vehicle's dynamic constraints, the maximum and minimum accelerations are set, and the acceleration is divided into multiple candidate values ​​according to a certain step size. Each candidate discrete acceleration has a corresponding label. Based on the encoded information output by the encoder, longitudinal discrete latent variables are obtained through a fully connected network, and acceleration labels are then sampled from them. Finally, the corresponding acceleration values ​​are obtained based on the sampled labels, forming a longitudinal acceleration prediction sequence. Subsequently, based on the scene feature representation output by the encoder, lateral latent variables are obtained through a fully connected network, and then the probability of candidate lanes is calculated using a graph attention network to obtain the probability distribution of candidate lanes.

[0086] Step S205: Based on the target candidate lane with the highest probability among the candidate lanes, generate the lateral heading angle prediction sequence in the Frenet coordinate system using the moving horizon optimization method;

[0087] In this embodiment, the target candidate lane with the highest probability is selected from the probability distribution of candidate lanes. In the Frenet coordinate system, the moving horizon optimization method is used to estimate the possible future heading angle, incorporating road information.

[0088] In the specific implementation, based on the feature information output by the encoder, lateral latent variables are obtained through a fully connected network, and then the probability of candidate lanes is calculated using a graph attention network. Next, the moving horizon optimization (MHO) method is used to estimate the possible future heading angles, in conjunction with road information. On structured roads, the vehicle's heading angle is estimated in the Frenet coordinate system based on prior knowledge, such as candidate lane information. The Frenet coordinate system is a local coordinate system referenced to the road centerline, suitable for describing the motion of vehicles on roads, such as... Figure 5 As shown, the curve segment formed by the curve and the arrow is the reference line, and the curve segment formed by the dashed curve and the arrow is the trajectory, with the included angle ∠a of the trajectory being the lateral heading angle. In the Frenet coordinate system, the vehicle's motion is decomposed into longitudinal motion along the lane direction and lateral motion perpendicular to the lane direction. By assuming small lateral distances and small road curvature, the model is simplified, and the velocity is treated as a constant, resulting in a linearized motion model, which is then discretized. To meet the requirements of reference path tracking and path smoothing, the objective function is minimized. The vehicle's heading angle is optimized to obtain the optimal solution. The first value of this optimal solution is taken as the predicted heading angle value at the current moment, and this process is repeated to form a sequence of lateral heading angle predictions. The objective function consists of two parts: the squared error of the predicted lateral distance and a smoothness constraint on the change in heading angle. By solving the optimization problem, the optimal solution for the heading angle is obtained, and the first value of this optimal solution is taken as the predicted heading angle value at the current moment.

[0089] Step S206: Correct the acceleration prediction sequence and the lateral heading angle prediction sequence;

[0090] In this embodiment of the application, the scene feature representation, the longitudinal acceleration prediction sequence, and the lateral heading angle prediction sequence are input into a pre-trained correction sub-network; the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence are fine-tuned by the correction sub-network, and the corrected longitudinal acceleration prediction sequence and the corrected lateral heading angle prediction sequence are output.

[0091] In the specific implementation, a neural network is used for minor corrections to obtain corrected longitudinal acceleration prediction sequences and corrected lateral heading angle prediction sequences. The formulas are expressed as follows:

[0092] ;

[0093] in , These are the corrected longitudinal acceleration prediction sequences and the corrected lateral heading angle prediction sequences. It is a fully connected layer. It is a non-linear activation function. As a representation of scene features, This is a sequence of predicted lateral heading angles. This is a longitudinal acceleration prediction sequence.

[0094] Step S207: Based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence, the trajectory is integrated using a kinematic model to obtain the initial predicted trajectory;

[0095] In this embodiment of the application, the kinematic model is:

[0096] ;

[0097] in,( , () represents the current position. , () represents the predicted position for the next moment. The speed at the current moment, For time step, This is the predicted lateral heading angle at the current moment.

[0098] In the specific implementation, the scene feature representation, longitudinal acceleration prediction sequence, and lateral heading angle prediction sequence are input into a pre-trained correction sub-network. The correction sub-network fine-tunes the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence, and outputs the corrected longitudinal acceleration prediction sequence and the corrected lateral heading angle prediction sequence.

[0099] Step S208: Define each trajectory point in the initial predicted trajectory as a two-dimensional Gaussian distribution mean to form a mean trajectory;

[0100] In this embodiment of the application, each trajectory point in the initial predicted trajectory is taken as the mean of a two-dimensional Gaussian distribution to form a mean trajectory.

[0101] Step S209: Assign a covariance matrix to each trajectory point in the mean trajectory to characterize the uncertainty of that location;

[0102] In this embodiment, a covariance matrix is ​​assigned to each trajectory point in the mean trajectory to characterize the uncertainty at that location, denoted as: ,in, For simplicity, the predicted locations share a common covariance matrix. .

[0103] Step S210: Based on the mean trajectory and the covariance matrix corresponding to each trajectory point, construct the probability trajectory.

[0104] In the embodiments of this application, each trajectory point in the mean trajectory and its corresponding covariance matrix are represented as a two-dimensional Gaussian distribution, thereby forming a probability trajectory.

[0105] This invention provides a vehicle trajectory prediction method, comprising: acquiring historical sample training data, historical trajectory data of a target vehicle and surrounding traffic participants, and lane information of the road; training an initially constructed trajectory prediction model using the historical sample training data to obtain a target trajectory prediction model; inputting the historical trajectory data and the lane information into the target trajectory prediction model to obtain a longitudinal acceleration prediction sequence and a lateral heading angle prediction sequence of the target vehicle; performing trajectory integration based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence using a kinematic model to obtain an initial predicted trajectory; and performing uncertainty modeling on the initial predicted trajectory to generate a probabilistic trajectory of the target vehicle. By introducing a graph attention network to fully capture the complex interaction relationships between traffic participants, the target trajectory prediction model is used to predict the longitudinal acceleration prediction sequence of the target vehicle and the probability distribution of candidate lanes. Modeling is performed based on the probability distribution of candidate lanes in the Frenet coordinate system, and the heading angle is estimated using a moving horizon optimization method. This ensures that the predicted trajectory conforms to actual driving rules such as lane restrictions and traffic signals, improving the reliability and practicality of the prediction results and achieving high-precision prediction in complex traffic scenarios.

[0106] This invention also provides a vehicle trajectory prediction device; please refer to [link / reference]. Figure 6 , Figure 6 This is a structural block diagram of an embodiment of a vehicle trajectory prediction device according to the present invention. The device includes:

[0107] The acquisition module 301 is used to acquire historical sample training data, historical trajectory data of the target vehicle and its surrounding traffic participants, and lane information of the road;

[0108] The model training module 302 is used to train the initially constructed trajectory prediction model using the historical sample training data to obtain the target trajectory prediction model.

[0109] The sequence prediction module 303 is used to input the historical trajectory data and the lane information into the target trajectory prediction model to obtain the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence of the target vehicle.

[0110] The initial trajectory acquisition module 304 is used to obtain the initial predicted trajectory by performing trajectory integration through a kinematic model based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence.

[0111] The probability trajectory acquisition module 305 is used to perform uncertainty modeling on the initial predicted trajectory and generate the probability trajectory of the target vehicle.

[0112] In an optional embodiment, the sequence prediction module 303 includes:

[0113] The encoding submodule is used to encode the historical trajectory data and the lane information through the encoder in the target trajectory prediction model to obtain a scene feature representation that integrates the interaction relationship between traffic participants and road information;

[0114] The decoding submodule is used to decode the scene feature representation through the decoder in the target trajectory prediction model to obtain the longitudinal acceleration prediction sequence and the probability distribution of the candidate lanes;

[0115] The heading angle prediction submodule is used to generate the lateral heading angle prediction sequence in the Frenet coordinate system using the moving horizon optimization method based on the target candidate lane with the highest probability among the candidate lanes.

[0116] In an optional embodiment, it further includes:

[0117] The correction module is used to correct the acceleration prediction sequence and the lateral heading angle prediction sequence.

[0118] In an optional embodiment, the correction module includes:

[0119] The input submodule is used to input the scene feature representation, the longitudinal acceleration prediction sequence, and the lateral heading angle prediction sequence into a pre-trained correction subnetwork;

[0120] The output submodule is used to fine-tune the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence through the correction subnetwork, and output the corrected longitudinal acceleration prediction sequence and the corrected lateral heading angle prediction sequence.

[0121] In an optional embodiment, the kinematic model is:

[0122] ;

[0123] in,( , () represents the current position. , () represents the predicted position for the next moment. The speed at the current moment, For time step, This is the predicted lateral heading angle at the current moment.

[0124] In an optional embodiment, the decoding submodule includes:

[0125] An acceleration prediction unit is used to predict discrete acceleration labels for each future time step based on the scene feature representation.

[0126] An acceleration prediction sequence generation unit is used to map the discrete acceleration labels to corresponding continuous acceleration values ​​according to a preset mapping relationship between labels and acceleration values, thereby forming the longitudinal acceleration prediction sequence.

[0127] In an optional embodiment, the probability trajectory acquisition module 305 includes:

[0128] The mean trajectory forming submodule is used to define each trajectory point in the initial predicted trajectory as a two-dimensional Gaussian distribution mean to form a mean trajectory;

[0129] The covariance matrix allocation submodule is used to allocate a covariance matrix to each trajectory point in the mean trajectory to characterize the uncertainty of that location;

[0130] The probability trajectory construction submodule is used to construct the probability trajectory based on the mean trajectory and the covariance matrix corresponding to each trajectory point.

[0131] Based on the same inventive concept, embodiments of the present invention also provide an electronic device. Figure 7 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Figure 7 As shown, an embodiment of the present invention provides an electronic device including: one or more processors 101, a memory 102, and one or more I / O interfaces 103. The memory 102 stores one or more programs, which, when executed by the one or more processors, enable the one or more processors to implement any of the vehicle trajectory prediction methods described in the above embodiments; the one or more I / O interfaces 103 are connected between the processors and the memory, configured to enable information interaction between the processors and the memory.

[0132] The processor 101 is a device with data processing capabilities, including but not limited to a central processing unit (CPU); the memory 102 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), and flash memory (FLASH); the I / O interface (read / write interface) 103 is connected between the processor 101 and the memory 102, and can realize information interaction between the processor 101 and the memory 102, including but not limited to a data bus (Bus).

[0133] In some embodiments, the processor 101, memory 102, and I / O interface 103 are interconnected via bus 104, and thus connected to other components of the computing device.

[0134] In some embodiments, the one or more processors 101 include a field-programmable gate array.

[0135] This invention also provides a computer-readable medium. The computer-readable medium stores a computer program, which, when executed by a processor, implements the steps of any of the vehicle trajectory prediction methods described in the above embodiments. The computer-readable storage medium may be volatile or non-volatile.

[0136] This invention also provides a computer program product, including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code. When the computer-readable code is run in the processor of an electronic device, the processor in the electronic device executes the above-described vehicle trajectory prediction method.

[0137] Those skilled in the art will understand that all or some of the steps, systems, and apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software can be distributed on a computer-readable storage medium, which may include computer storage media (or non-transitory media) and communication media (or transient media).

[0138] As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable program instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technologies, portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, it is known to those skilled in the art that communication media typically contain computer-readable program instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0139] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0140] The computer program instructions used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions. This electronic circuitry can execute the computer-readable program instructions to implement various aspects of the invention.

[0141] The computer program product described herein can be implemented specifically through hardware, software, or a combination thereof. In one alternative embodiment, the computer program product is specifically embodied in a computer storage medium; in another alternative embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0142] Various aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0143] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0144] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0145] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0146] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for illustrative purposes only and should be construed as such, and is not intended to be limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of the invention as set forth in the appended claims.

Claims

1. A vehicle trajectory prediction method, characterized by, The method comprises the following steps: obtaining historical sample training data, historical trajectory data of a target vehicle and surrounding traffic participants, and lane information of a road; training an initially constructed trajectory prediction model by using the historical sample training data to obtain a target trajectory prediction model; inputting the historical trajectory data and the lane information into the target trajectory prediction model to obtain a longitudinal acceleration prediction sequence and a lateral heading angle prediction sequence of the target vehicle; performing trajectory integration by using a kinematics model based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence to obtain an initial prediction trajectory; performing uncertainty modeling on the initial prediction trajectory to generate a probability trajectory of the target vehicle.

2. The vehicle trajectory prediction method of claim 1, wherein, The method comprises the following steps: encoding the historical trajectory data and the lane information by using an encoder in the target trajectory prediction model to obtain a scene feature representation which integrates the interaction relationship of the traffic participants and the road information; decoding the scene feature representation by using a decoder in the target trajectory prediction model to obtain the longitudinal acceleration prediction sequence and a probability distribution of candidate lanes; generating the lateral heading angle prediction sequence by using a moving horizon optimization method in a Frenet coordinate system based on a target candidate lane with the highest probability in the candidate lanes.

3. The vehicle trajectory prediction method of claim 2, wherein, Before performing trajectory integration by using a kinematics model based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence to obtain an initial prediction trajectory, the method further comprises the following step: correcting the acceleration prediction sequence and the lateral heading angle prediction sequence.

4. The vehicle trajectory prediction method of claim 3, wherein, The method comprises the following steps: inputting the scene feature representation, the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence into a pre-trained correction subnetwork; fine-tuning the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence by using the correction subnetwork to output a corrected longitudinal acceleration prediction sequence and a corrected lateral heading angle prediction sequence.

5. The vehicle trajectory prediction method according to claim 1 or 2, characterized in that, The kinematics model is as follows: ； in,( , () represents the current position. , () represents the predicted position for the next moment. The speed at the current moment, For time step, This is the predicted lateral heading angle at the current moment.

6. The vehicle trajectory prediction method of claim 2, wherein, The method comprises the following steps: predicting an acceleration discrete label at each time step in the future based on the scene feature representation; mapping the acceleration discrete label to a corresponding continuous acceleration value according to a preset mapping relationship between the label and the acceleration value to form the longitudinal acceleration prediction sequence.

7. The vehicle trajectory prediction method of claim 1, wherein, The method comprises the following steps: defining each trajectory point in the initial prediction trajectory as a mean value of a two-dimensional Gaussian distribution to form a mean value trajectory; assigning a covariance matrix to each trajectory point in the mean value trajectory to represent the uncertainty at the position; constructing the probability trajectory based on the mean value trajectory and the covariance matrix corresponding to each trajectory point.

8. A vehicle trajectory prediction device characterized by comprising: The method comprises the following steps: An acquisition module is configured to acquire historical sample training data, historical trajectory data of a target vehicle and surrounding traffic participants, and lane information of a road; A model training module is configured to train an initially constructed trajectory prediction model by using the historical sample training data, to obtain a target trajectory prediction model; A sequence prediction module is configured to input the historical trajectory data and the lane information into the target trajectory prediction model, to obtain a longitudinal acceleration prediction sequence and a lateral heading angle prediction sequence of the target vehicle; An initial trajectory acquisition module is configured to perform trajectory integration by using a kinematics model based on the longitudinal acceleration prediction sequence and the lateral heading angle prediction sequence, to obtain an initial prediction trajectory; A probability trajectory acquisition module is configured to model uncertainty of the initial prediction trajectory, to generate a probability trajectory of the target vehicle.

9. The vehicle trajectory prediction apparatus according to claim 8, characterized by, The sequence prediction module includes: An encoding submodule is configured to encode the historical trajectory data and the lane information by using an encoder in the target trajectory prediction model, to obtain a scene feature representation that fuses traffic participant interaction relationship and road information; A decoding submodule is configured to decode the scene feature representation by using a decoder in the target trajectory prediction model, to obtain the longitudinal acceleration prediction sequence and a probability distribution of a candidate lane; A heading angle prediction submodule is configured to generate the lateral heading angle prediction sequence by using a moving horizon optimization method in a Frenet coordinate system based on a target candidate lane with the highest probability in the candidate lane.

10. The vehicle trajectory prediction apparatus according to claim 8, characterized by, The probability trajectory acquisition module includes: A mean trajectory formation submodule is configured to define each trajectory point in the initial prediction trajectory as a mean value of a two-dimensional Gaussian distribution, to form a mean trajectory; A covariance matrix allocation submodule is configured to allocate a covariance matrix to each trajectory point in the mean trajectory, to represent uncertainty at the position; A probability trajectory construction submodule is configured to construct the probability trajectory based on the mean trajectory and the covariance matrix corresponding to each trajectory point.

11. An electronic device, comprising: The computer program is executed by the processor to implement the steps in the method of any one of claims 1 to 7. The computer program is executed by the processor to implement the steps in the method of any one of claims 1 to 7. ​ ​ 12. A computer readable medium having stored thereon a computer program, characterized in that ​

Images