Trajectory prediction method and device, electronic equipment and storage medium

Abstract

The invention discloses a trajectory prediction method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining the road information of a position where a target vehicle is located, and obtaining the state information related to a traffic participant; encoding and decoding the obtained road information and state information through a preset network model to obtain global features, determining a future trajectory of the traffic participant and Gaussian distribution of the future trajectory according to the global features, and determining the traffic participant according to features corresponding to the state information in the global features. And determining an interaction relationship between the traffic participant and the target vehicle and a static starting probability of the traffic participant. Since the model outputs the prediction trajectory and Gaussian distribution of the trajectory, the uncertainty of the trajectory can be represented. And the interaction relationship of the traffic participant relative to the vehicle and the static starting probability of the traffic participant output by the model represent the semantics of the traffic participant relative to the vehicle, so that the vehicle can conveniently plan the driving track of the vehicle according to the semantic information.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and specifically to a trajectory prediction method, device, electronic device and storage medium. Background Art

[0002] In the field of vehicle autonomous driving, autonomous vehicles need to predict the trajectories of surrounding traffic participants (such as cars, pedestrians, bicycles, electric vehicles, etc.) in real time in order to plan their own driving.

[0003] At present, neural network models are usually selected as the main predictor, that is, a genetic algorithm is used to sample semantic points based on some prior information (for example, it is believed that the end point of the trajectory of most vehicles will fall on the lane), and then the key points of the trajectory are generated to ensure that the end point of the trajectory is within the drivable range. The neural network model is then used to generate the future trajectory of the traffic participants based on the key points of the trajectory.

[0004] However, since generating trajectory key points requires more genetic algorithm calculations, such predictors usually find it difficult to balance the relationship between prediction effect and time consumption, and the output future trajectory of traffic participants lacks semantic information relative to the autonomous driving vehicle, which is not conducive to the autonomous driving vehicle planning its own driving. Summary of the invention

[0005] The purpose of the present application is to propose a trajectory prediction method, device, electronic device and storage medium to address the deficiencies of the above-mentioned prior art, and this purpose is achieved through the following technical solutions.

[0006] The first aspect of the present application provides a trajectory prediction method, the method comprising:

[0007] Acquire road information of the target vehicle's location, and acquire status information related to a traffic participant; the traffic participant is one of the surrounding targets perceived by the target vehicle, and the status information is data perceived by the target vehicle around the traffic participant;

[0008] The acquired road information and state information are encoded and decoded through a preset network model to obtain global features, and the future trajectory of the traffic participant and the Gaussian distribution of the future trajectory are determined based on the global features, and the interaction relationship of the traffic participant with respect to the target vehicle and the stationary starting probability of the traffic participant are determined based on the features corresponding to the state information in the global features.

[0009] A second aspect of the present application provides a trajectory prediction device, the device comprising:

[0010] A data acquisition module is used to acquire road information of the target vehicle and to acquire status information related to a traffic participant; the traffic participant is one of the surrounding targets perceived by the target vehicle, and the status information is data perceived by the target vehicle around the traffic participant;

[0011] The prediction module is used to encode and decode the acquired road information and state information through a preset network model to obtain global features, and determine the future trajectory of the traffic participant and the Gaussian distribution of the future trajectory according to the global features, and determine the interaction relationship of the traffic participant with respect to the target vehicle and the stationary starting probability of the traffic participant according to the features corresponding to the state information in the global features.

[0012] The third aspect of the present application proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described in the first aspect above.

[0013] The fourth aspect of the present application proposes a computer-readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement the steps of the method described in the first aspect above.

[0014] Based on the trajectory prediction method and device described in the first and second aspects above, the present application has at least the following beneficial effects or advantages:

[0015] By obtaining the road structure at the location of the vehicle and the historical status related to the traffic participants as the input of the network model, the network model encodes and decodes the input road structure and the historical status of the traffic participants to obtain global features that are conducive to trajectory prediction and semantic prediction, thereby predicting the future trajectory and Gaussian distribution of the traffic participants based on the global features, and predicting the interaction relationship between the traffic participants and the vehicle and the stationary start probability of the traffic participants based on the features corresponding to the historical status in the global features. Since the network model outputs the Gaussian distribution of the trajectory in addition to the predicted trajectory, it can represent the uncertainty of the trajectory. In addition, the interaction relationship between the traffic participants and the vehicle and the stationary start probability of the traffic participants output by the network model represent the semantics of the traffic participants relative to the vehicle, which is conducive to the vehicle planning its own driving trajectory based on the semantic information.

[0016] In addition, since the present application does not need to use genetic algorithms to calculate trajectory key points, the network model can be used to directly predict and output the trajectory of traffic participants and the semantic information of relative vehicles. Therefore, the present application can not only improve the prediction effect, but also reduce the prediction time, and can strike a good balance between the prediction effect and time consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 This is a flow chart of an embodiment of a trajectory prediction method according to an exemplary embodiment of the present application;

[0019] Figure 2 This is a schematic diagram of a network model structure according to an exemplary embodiment of the present application;

[0020] Figure 3 This application is based on Figure 2 The illustrated embodiment shows a schematic diagram of the structure of a first encoder;

[0021] Figure 4 This application is based on Figure 2 A schematic diagram of the structure of a second encoder shown in the illustrated embodiment;

[0022] Figure 5 This is a schematic structural diagram of a trajectory prediction device according to an exemplary embodiment of the present application;

[0023] Figure 6 This is a schematic diagram of the hardware structure of an electronic device according to an exemplary embodiment of the present application;

[0024] Figure 7 The figure is a schematic diagram of the structure of a storage medium according to an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0025] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0026] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0027] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0028] As mentioned above, currently it is necessary to use genetic algorithms to generate trajectory predictors for key points of trajectories. This not only makes it difficult to balance the relationship between prediction effect and time consumption, but also only outputs the future trajectory of traffic participants. It lacks semantics relative to the autonomous driving vehicle (that is, the main vehicle), which is not conducive to the autonomous driving vehicle planning its own driving.

[0029] In order to solve the above technical problems, the present application still adopts a neural network model for prediction, but abandons the use of a genetic algorithm to generate trajectory key points. The trajectory prediction method proposed in the present application is to obtain the road structure at the vehicle's location and the historical status related to the traffic participants as the input of the network model, and the network model encodes and decodes the input road structure and the historical status of the traffic participants to obtain global features that are conducive to trajectory prediction and semantic prediction, thereby predicting the future trajectory of the traffic participants and the Gaussian distribution of the future trajectory based on the global features, and predicting the interaction relationship between the traffic participants and the vehicle and the probability of the traffic participants starting from a stationary state based on the features in the global features corresponding to the historical status.

[0030] Based on the above description, since the network model of this application outputs not only the predicted trajectory but also the Gaussian distribution of the trajectory, it can represent the uncertainty of the trajectory. In addition, the interactive relationship between the traffic participants and the vehicle and the stationary start probability of the traffic participants output by the network model represent the semantics of the traffic participants relative to the vehicle, which is beneficial for the vehicle to plan its own driving according to the semantic information.

[0031] In addition, since the present application does not need to use genetic algorithms to calculate trajectory key points, the network model can be used to directly predict and output the trajectory of traffic participants and the semantic information of relative vehicles. Therefore, the present application can not only improve the prediction effect, but also reduce the prediction time, and can strike a good balance between the prediction effect and time consumption.

[0032] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0033] Figure 1This is a flow chart of an embodiment of a trajectory prediction method according to an exemplary embodiment of the present application, comprising the following steps:

[0034] Step 101: Obtain road information of the target vehicle's location, and obtain status information related to traffic participants.

[0035] In this step, the traffic participant is one of the surrounding targets perceived by the target vehicle, such as cars, pedestrians, bicycles, electric vehicles, etc., which may affect the driving of the target vehicle. The acquired state information is the data perceived by the target vehicle around the traffic participant. The acquired road information may include description data of road elements and description data of road elements relative to the target vehicle. Road elements may be lane lines, zebra crossings, etc.

[0036] In one example, the status information related to the traffic participant describes the status of the traffic participant, such as the position, speed, bounding box size, heading angle, object category, etc. of the traffic participant at each historical time stamp, by converting these data into a coordinate system centered on the traffic participant for subsequent prediction; it also describes the status of the traffic participant relative to other traffic participants, such as relative heading direction, relative distance, relative position, relative speed, etc.; and it also describes the status of the target vehicle relative to the traffic participant, and the description of this status is similar to the description of the status of the traffic participant relative to other traffic participants.

[0037] Step 102: The acquired road information and state information are encoded and decoded through a preset network model to obtain a global feature, and the future trajectory of the traffic participant and the Gaussian distribution of the future trajectory are determined based on the global feature, and the interaction relationship of the traffic participant with respect to the target vehicle and the stationary start probability of the traffic participant are determined based on the feature in the global feature corresponding to the acquired state information.

[0038] In this step, the acquired road information and status information are used as the input of the network model. The network model combines the input road information and status information to perform a series of processing and predictions, and finally outputs the future trajectory of the traffic participant and the Gaussian distribution of the future trajectory, as well as the interaction relationship between the traffic participant and the target vehicle and the stationary start probability of the traffic participant.

[0039] In an embodiment of the present application, the network model can output multiple future trajectories (for example, output three trajectories with the highest probability). These trajectories are possible routes for traffic participants to travel in the future. Each trajectory corresponds to an occurrence probability, and the Gaussian distribution of the trajectory includes the mean and variance of the trajectory points, and the xy covariance to represent the uncertainty of the trajectory.

[0040] Furthermore, the interaction relationship between the traffic participant and the target vehicle can include three categories: avoidance, overtaking, and no interaction, which means whether the traffic participant will avoid the main vehicle or overtake the main vehicle in the future, or whether there is no interaction with the main vehicle. The stationary start probability indicates the possibility that the traffic participant is currently stationary and then starts. The semantic representation of the above traffic participants relative to the vehicle can play a good role in guiding the vehicle to plan its driving.

[0041] In one possible implementation, if Figure 2 The network model structure shown in the figure includes a first encoder, a second encoder, and a decoder for encoding and decoding the input content. In the process of encoding and decoding to obtain global features, the first encoder can be used to parse the road information to obtain the road encoding features, and the second encoder can be used to parse the state information to obtain the state encoding features, and then the decoder can use the self-attention mechanism to decode the road encoding features and the state encoding features to obtain the global features.

[0042] Among them, since road information and the status information of traffic participants belong to two different data sources, their representation forms are also different. Therefore, two encoders are used to parse and encode them separately, and then they are combined and decoded together through a decoder to obtain global features containing complete and comprehensive information.

[0043] Depend on Figure 2 It can be seen that the global features include the Agent feature corresponding to the state of the traffic participant, the Actor feature corresponding to the state of the traffic participant relative to other traffic participants, the AV feature corresponding to the state of the target vehicle relative to the traffic participant, and the Map feature corresponding to the road information.

[0044] In an alternative embodiment, by Figure 2 As shown, in the process of parsing and encoding road information through the first encoder, the road information can be classified according to the road element category, and then for each type of road information, the first encoder corresponding to the type of road information is used to parse the type of road information to obtain the road coding features of the type of road information, thereby merging the road coding features of each type of road information into the final road coding features.

[0045] Among them, different types of road elements have different effects on the target vehicle. For example, the center line of the lane has an attractive effect on the vehicle trajectory, and the vehicle hopes to drive along the center line of the lane, while the lane boundary line (that is, the lane line, including the stop line) has a repulsive effect on the vehicle trajectory, and the vehicle hopes not to drive on the line. The zebra crossing has a reference effect on the pedestrian trajectory, and pedestrians cross the road on the zebra crossing. Road information belonging to different types of road elements is parsed and encoded using different first encoders with the same structure to reduce the coding interference between different types of road elements.

[0046] Optional, by Figure 2 It can be seen that road information can be divided into three categories: lane center line, lane boundary, and zebra crossing.

[0047] In specific implementation, Figure 3 The first encoder structure shown in FIG. 1 includes a plurality of parallel multilayer perceptrons and a feature concatenation layer. In the process of using the corresponding first encoder to parse a certain type of road information, different types of description data of road elements in the road information can be input into a multilayer perceptron respectively, and the input description data can be encoded by each multilayer perceptron respectively, and then the encoded features output by each multilayer perceptron can be concatenated by the feature concatenation layer to obtain the road features of the road information.

[0048] Optional, by Figure 3 As shown in the figure, the description data of road elements can include sampling coordinate point sequences of road elements, road element attributes (such as starting and ending point coordinates, indicating line segment length, starting point tangent, minimum distance relative to vehicles, distance from end point coordinates to vehicle minimum distance point, speed limit information, etc.), bound traffic light information, line categories (such as solid line, dashed line, curb), and each type of description data is input into a multilayer perceptron for encoding. Among them, the multilayer perceptron for inputting road element coordinate point sequences is used to encode the road features of the coordinate position; the multilayer perceptron for inputting road element attributes is used to encode the features of the road element attributes; the multilayer perceptron for inputting traffic light information is used to encode the features of the bound traffic lights; the multilayer perceptron for inputting road element line categories is used to encode the features representing the road element categories.

[0049] It should be noted that a multi-layer perceptron (MLP), also known as an artificial neural network, includes an input layer, a hidden layer, and an output layer.

[0050] In yet another alternative embodiment, Figure 4As shown, the second encoder includes a dynamic attribute encoding network, a static attribute encoding network and a feature concatenation layer. With respect to the process of parsing and encoding the state information through the second encoder, the state information belonging to the dynamic attribute in the state information can be parsed through the dynamic attribute encoding network to obtain dynamic attribute encoding features, and the state information belonging to the static attribute in the state information can be parsed through the static attribute encoding network to obtain static attribute encoding features, and then the dynamic attribute encoding features and the static attribute encoding features can be concatenated through the feature concatenation layer to obtain state encoding features.

[0051] Among them, the state information of traffic participants can be divided into two attributes: dynamically changing and statically fixed. Using different coding networks for state information belonging to different attributes can achieve complete and comprehensive extraction of state coding features.

[0052] Depend on Figure 4 It can be seen that the state information belonging to dynamic attributes includes the historical trajectory sequence of the traffic participant (including the coordinates, speed, heading angle, displacement, etc. of each trajectory point), the timestamp of the sequence, the interaction information between the traffic participant and other traffic participants, and the interaction information between the traffic participant and the target vehicle. The interaction information includes relative heading direction, relative distance, relative position, relative speed, etc. The state information belonging to static attributes includes the type of traffic participant (such as bicycles, pedestrians, cars, etc.) and the size of the bounding box (including the length and width of the traffic participant).

[0053] In specific implementation, Figure 4 As shown, the dynamic attribute coding network includes a multi-layer perceptron (MLP) and a multi-layer perceptron with a spatial gating unit (GMLP, Gated Multi-Layer Per-ceptron). In the process of parsing the state information belonging to the dynamic attribute in the state information by the dynamic attribute coding network, the state information belonging to the dynamic attribute can be positionally encoded by the multi-layer perceptron to obtain position features, and the temporal relationship between the position features can be encoded by the multi-layer perceptron with a spatial gating unit to obtain dynamic attribute coding features.

[0054] For example, the historical trajectory sequence of traffic participants is composed of a series of trajectory points. The timestamp of the trajectory point represents the timing information of each trajectory point. The position features are obtained by parsing these trajectory points through MLP, and the timing information between the position features of these trajectory points is parsed through GMLP. Therefore, the dynamic attribute coding features obtained contain the position information and timing information of the traffic participants.

[0055] In another possible implementation, if Figure 2The network model structure shown in the figure predicts the future trajectory based on the global features, including a cross-attention network and a key point generation network. In the process of determining the future trajectory of the traffic participant and the Gaussian distribution of the future trajectory based on the global features, the key point generation network generates a trajectory key point vector based on the feature Agent feature corresponding to the state of the traffic participant in the global features, and then the cross-attention network is used to query the global features based on the trajectory key point vector to obtain the future trajectory of the traffic participant and the Gaussian distribution of the future trajectory.

[0056] Among them, the embodiment of the present application can reduce the time consumption of using genetic algorithm calculation by converting the generation of trajectory key points into a learnable network learning output trajectory key point vector feature.

[0057] Furthermore, if Figure 2 The network model structure shown in the figure predicts the semantics of the traffic participant relative to the target vehicle in the future according to the global features, including a classification network and a start prediction network. In the process of determining the interactive relationship between the traffic participant and the target vehicle and the stationary start probability of the traffic participant according to the features corresponding to the state information in the global features, the classification network determines the category of the interactive relationship between the traffic participant and the target vehicle according to the features Agent feature corresponding to the state of the traffic participant in the global features and the features AV feature corresponding to the state of the target vehicle relative to the traffic participant, and the start prediction network predicts the stationary start probability of the traffic participant according to the features Agent feature corresponding to the state of the traffic participant in the global features.

[0058] Among them, the types of interaction relationships include avoidance, overtaking, and no interaction, indicating whether the traffic participant will avoid the main vehicle or overtake the main vehicle in the future, or whether there will be no interaction with the main vehicle. The stationary start probability indicates the possibility that the traffic participant is currently stationary and then starts. These semantic representations of the future of the traffic participant relative to the vehicle can play a good role in guiding the vehicle's driving planning.

[0059] At this point, the above Figure 1The trajectory prediction process shown in the figure obtains the road structure at the location of the vehicle and the historical status related to the traffic participants as the input of the network model. The network model encodes and decodes the input road structure and the historical status of the traffic participants to obtain global features that are conducive to trajectory prediction and semantic prediction, thereby predicting the future trajectory and Gaussian distribution of the traffic participants based on the global features, and predicting the interaction relationship between the traffic participants and the vehicle and the stationary start probability of the traffic participants based on the features corresponding to the historical status in the global features. Since the network model outputs the Gaussian distribution of the trajectory in addition to the predicted trajectory, it can represent the uncertainty of the trajectory. In addition, the interaction relationship between the traffic participants and the vehicle and the stationary start probability of the traffic participants output by the network model represent the semantics of the traffic participants relative to the vehicle, which is conducive to the vehicle planning its own driving trajectory based on the semantic information.

[0060] In addition, since the present application does not need to use genetic algorithms to calculate trajectory key points, the network model can be used to directly predict and output the trajectory of traffic participants and the semantic information of relative vehicles. Therefore, the present application can not only improve the prediction effect, but also reduce the prediction time, and can strike a good balance between the prediction effect and time consumption.

[0061] Corresponding to the above-mentioned embodiment of the trajectory prediction method, the present application also provides an embodiment of a trajectory prediction device.

[0062] Figure 5 FIG. 1 is a schematic diagram of a trajectory prediction device according to an exemplary embodiment of the present application. The device is used to execute the trajectory prediction method provided in any of the above embodiments, such as Figure 5 As shown, the trajectory prediction device includes:

[0063] The data acquisition module 510 is used to acquire the road information of the target vehicle and the status information related to the traffic participant; the traffic participant is one of the surrounding targets perceived by the target vehicle, and the status information is the data perceived by the target vehicle around the traffic participant;

[0064] The prediction module 520 is used to encode and decode the acquired road information and state information through a preset network model to obtain global features, and determine the future trajectory of the traffic participant and the Gaussian distribution of the future trajectory based on the global features, and determine the interaction relationship of the traffic participant with respect to the target vehicle and the probability of the traffic participant starting from a stationary state based on the features corresponding to the state information in the global features.

[0065] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0066] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0067] The embodiment of the present application also provides an electronic device corresponding to the trajectory prediction method provided in the above embodiment, so as to execute the above trajectory prediction method.

[0068] Figure 6 This is a hardware structure diagram of an electronic device according to an exemplary embodiment of the present application, and the electronic device includes: a communication interface 601, a processor 602, a memory 603 and a bus 604; wherein the communication interface 601, the processor 602 and the memory 603 communicate with each other through the bus 604. The processor 602 can execute the trajectory prediction method described above by reading and executing the machine executable instructions corresponding to the control logic of the trajectory prediction method in the memory 603. The specific content of the method is referred to the above embodiment, and will not be repeated here.

[0069] The memory 603 mentioned in this application can be any electronic, magnetic, optical or other physical storage device, and can contain storage information, such as executable instructions, data, etc. Specifically, the memory 603 can be RAM (Random Access Memory), flash memory, storage drive (such as hard disk drive), any type of storage disk (such as optical disk, DVD, etc.), or similar storage medium, or a combination thereof. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 601 (which can be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. can be used.

[0070] The bus 604 may be an ISA bus, a PCI bus or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 603 is used to store programs, and the processor 602 executes the programs after receiving execution instructions.

[0071] The processor 602 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 602. The above processor 602 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor to be executed, or the hardware and software modules in the decoding processor can be combined and executed.

[0072] The electronic device provided in the embodiment of the present application and the trajectory prediction method provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, operated or implemented therein.

[0073] The present application also provides a computer-readable storage medium corresponding to the trajectory prediction method provided in the above embodiment. Figure 7 As shown, the computer-readable storage medium shown is a CD 30 on which a computer program (ie, a program product) is stored. When the computer program is executed by a processor, the trajectory prediction method provided by any of the aforementioned embodiments will be executed.

[0074] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.

[0075] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the trajectory prediction method provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the method adopted, run or implemented by the application program stored therein.

[0076] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0077] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0078] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A trajectory prediction method, characterized in that: The method comprises: Acquire road information of the target vehicle's location, and acquire status information related to a traffic participant; the traffic participant is one of the surrounding targets perceived by the target vehicle, and the status information is data perceived by the target vehicle around the traffic participant; The acquired road information and state information are encoded and decoded through a preset network model to obtain global features, and the future trajectory of the traffic participant and the Gaussian distribution of the future trajectory are determined based on the global features, and the interaction relationship of the traffic participant with respect to the target vehicle and the stationary starting probability of the traffic participant are determined based on the features corresponding to the state information in the global features.

2. The method according to claim 1, characterized in that The preset network model includes a first encoder, a second encoder, and a decoder; The obtained road information and status information are encoded and decoded through the preset network model to obtain global features, including: Parsing the road information by a first encoder to obtain a road coding feature; Parsing the state information by a second encoder to obtain a state coding feature; The road encoding feature and the state encoding feature are decoded by a decoder using a self-attention mechanism to obtain a global feature.

3. The method according to claim 2, characterized in that The road information is parsed by a first encoder to obtain a road coding feature, including: Classifying the road information according to road element categories; For each type of road information, use a first encoder corresponding to the type of road information to parse the type of road information to obtain a road coding feature of the type of road information; The road coding features of various types of road information are merged into the final road coding features.

4. The method according to claim 3, characterized in that The first encoder includes a plurality of parallel multi-layer perceptrons and feature concatenation layers; The using a first encoder corresponding to the type of road information to parse the type of road information to obtain the road coding features of the type of road information includes: Different types of description data of road elements in the road information are input into a multi-layer perceptron respectively, and the input description data is encoded respectively by each multi-layer perceptron; The encoding features output by each multi-layer perceptron are concatenated through the feature concatenation layer to obtain the road features of this type of road information.

5. The method according to claim 2, characterized in that: The second encoder includes a dynamic attribute encoding network, a static attribute encoding network and a feature concatenation layer; The state information is parsed by a second encoder to obtain a state coding feature, including: Parsing the state information belonging to the dynamic attribute in the state information through the dynamic attribute coding network to obtain dynamic attribute coding features; Parsing the state information belonging to the static attributes in the state information through the static attribute encoding network to obtain static attribute encoding features; The dynamic attribute coding feature and the static attribute coding feature are concatenated through the feature concatenation layer to obtain a state coding feature.

6. The method according to claim 5, characterized in that The dynamic attribute encoding network includes a multilayer perceptron and a multilayer perceptron with a spatial gating unit; The step of parsing the state information belonging to dynamic attributes in the state information through the dynamic attribute coding network to obtain dynamic attribute coding features includes: Position encoding of state information belonging to dynamic attributes is performed by the multi-layer perceptron to obtain position features; The temporal relationship between the position features is encoded by a multi-layer perceptron with a spatial gating unit to obtain a dynamic attribute encoding feature.

7. The method according to claim 1, characterized in that The preset network model also includes a cross attention network and a key point generation network; The determining the future trajectory of the traffic participant and the Gaussian distribution of the future trajectory according to the global feature includes: Generate a trajectory key point vector according to the features corresponding to the state of the traffic participant in the global features through the key point generation network; The global feature is queried according to the trajectory key point vector through the cross attention network to obtain the future trajectory of the traffic participant and the Gaussian distribution of the future trajectory.

8. The method according to claim 1, characterized in that The preset network model also includes a classification network and a start prediction network; Determining the interaction relationship of the traffic participant with respect to the target vehicle and the stationary start probability of the traffic participant according to the feature corresponding to the state information in the global feature includes: Determine, by a classification network, the category of the interaction relationship between the traffic participant and the target vehicle according to the feature corresponding to the state of the traffic participant in the global feature and the feature corresponding to the state of the target vehicle relative to the traffic participant, wherein the category includes one of avoidance, overtaking and no interaction; The stationary start probability of the traffic participant is predicted by a start prediction network according to the features corresponding to the state of the traffic participant in the global features.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes the program to implement the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the steps of the method according to any one of claims 1 to 8.

Images