Vehicle risk avoidance planning method, device, vehicle, electronic device and storage medium

Abstract

The present application provides a method, device, vehicle, electronic device and storage medium for risk avoidance planning of a vehicle, wherein the method comprises: obtaining sensor data of the vehicle; performing feature extraction on the sensor data to obtain bird's-eye view features; constructing instance tokens according to the bird's-eye view features; mapping the instance tokens to trajectory space for trajectory prior to obtain potential expected trajectories; obtaining target loss data according to the potential expected trajectory; and performing risk avoidance planning on the trajectory of the vehicle according to the target loss data. The implementation of the present application can improve the safety of the vehicle during driving, improve the accuracy and feasibility of risk avoidance planning, does not rely on pre-defined trajectories and paths, has strong adaptability, can be applied to a variety of complex road environments and driving environments, and provide protection for driving safety and pedestrian safety during vehicle driving.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and more specifically, to a method, device, vehicle, electronic device and storage medium for risk avoidance planning of a vehicle. Background Art

[0002] Automatic risk avoidance technology is one of the key technologies to improve automobile safety performance. By automatically detecting potential dangerous situations and taking risk avoidance measures, it can help avoid or mitigate accidents at critical moments and protect the safety of passengers and pedestrians. With the popularity of electric vehicles, more and more vehicles have led to a more complex traffic environment and more variable driving conditions, and the vehicle's risk avoidance technology has gradually become inapplicable.

[0003] There are many problems with the risk avoidance methods of existing technologies. For example, the interaction modeling is limited. For rule-based and model-based methods, there is a lack of ability to understand and predict the complex interactions between the vehicle and other vehicles in the environment; trajectory planning has poor adaptability. Many existing methods rely on predefined paths or trajectories that cannot adapt well to sudden changes in the environment; there is a lack of realism in the predictions, and it is difficult to incorporate the subtle real behaviors of human drivers and other road users into their predictions; uncertainty processing is insufficient, and the uncertainty in perception, prediction and action execution cannot be fully addressed. Summary of the invention

[0004] The purpose of the present application is to provide a vehicle risk avoidance planning method, device, vehicle, electronic device and storage medium, which can improve the safety of the vehicle during driving, improve the accuracy and feasibility of risk avoidance planning, do not rely on pre-defined trajectories and paths, have strong adaptability, and can be applied to a variety of complex road environments and driving environments, thereby providing protection for driving safety and pedestrian safety during vehicle driving.

[0005] In a first aspect, an embodiment of the present application provides a method for risk avoidance planning of a vehicle, the method comprising:

[0006] Acquiring sensor data of the vehicle;

[0007] Extracting features from the sensor data to obtain bird's-eye view features;

[0008] constructing an instance token according to the bird's-eye view features;

[0009] Mapping the instance tokens to the trajectory space for trajectory prior to obtain potential expected trajectories;

[0010] obtaining target loss data according to the potential desired trajectory;

[0011] The trajectory of the vehicle is planned for risk avoidance according to the target loss data.

[0012] In the above implementation process, the bird's-eye view features are used to construct instance tokens, and then the trajectory prior is performed based on the instance tokens. Finally, the loss data is avoided to achieve reasonable risk avoidance planning, which can improve the safety of the vehicle during driving, improve the accuracy and feasibility of risk avoidance planning, and does not rely on pre-defined trajectories and paths. It has strong adaptability and can be applied to a variety of complex road environments and driving environments, providing protection for driving safety and pedestrian safety during vehicle driving.

[0013] Furthermore, the step of extracting features from the sensor data to obtain bird's-eye view features includes:

[0014] Performing feature extraction on the sensor data to obtain multi-scale image features;

[0015] Converting the multi-scale image features into bird's-eye view spatial data according to a pre-built deformable attention model;

[0016] The bird's-eye view feature spatial data is aligned with the current vehicle coordinates in the sensor data to obtain the bird's-eye view feature.

[0017] In the above implementation process, the multi-scale image features are converted into bird's-eye view spatial data and then aligned with the current ego-vehicle coordinates, so that the bird's-eye view features can contain more ego-vehicle features and ego-vehicle data, thereby improving the accuracy of risk avoidance planning.

[0018] Furthermore, the step of constructing an instance token according to the bird's-eye view feature comprises:

[0019] Learning the bird's-eye view features according to a pre-built global cross-attention model to obtain a map token set;

[0020] Get his car token collection;

[0021] The instance token is constructed according to the map token set and the other vehicle token set.

[0022] In the above implementation process, a map token set is constructed according to the bird's-eye view features, and then an instance token is constructed according to the other vehicle token set and the map token set. This can expand the effective range of the instance token in the application process and provide data support for subsequent trajectory priors.

[0023] Furthermore, the step of obtaining the other vehicle token set includes:

[0024] Get other car data;

[0025] Obtaining a token for another vehicle according to the other vehicle data;

[0026] The other vehicle tokens are updated according to a pre-built deformable cross-attention model to obtain the other vehicle token set.

[0027] In the above implementation process, updating the other vehicle tokens according to the deformable cross-attention model can improve the rationality and effectiveness of the other vehicle token set, improve data accuracy, and reduce errors generated during the updating process.

[0028] Furthermore, the step of mapping the instance token to the trajectory space to perform trajectory prior to obtain a potential expected trajectory includes:

[0029] obtaining a desired trajectory of the instance token;

[0030] Mapping the expected trajectory to the trajectory space to obtain potential distribution information of the expected trajectory;

[0031] The potential distribution information of the expected trajectory is decoded to obtain the potential expected trajectory.

[0032] In the above implementation process, the expected trajectory is mapped to the trajectory space, and the potential distribution of the expected trajectory is obtained and then decoded, so that the obtained potential expected trajectory is more consistent with the risk avoidance plan of the vehicle, thereby reducing the driving risk of the vehicle.

[0033] Furthermore, the step of obtaining target loss data according to the potential expected trajectory includes:

[0034] Obtaining divergence loss data according to the potential desired trajectory;

[0035] Obtain trajectory loss data, semantic map loss data, and 3D object detection loss data;

[0036] The target loss data is constructed according to the divergence loss data, the trajectory loss data, the semantic map loss data and the 3D object detection loss data.

[0037] In the above implementation process, target loss data is constructed according to divergence loss data, trajectory loss data, semantic map loss data and 3D object detection loss data, which can improve the error rate in the risk avoidance planning process and make the risk avoidance planning more accurate.

[0038] In a second aspect, an embodiment of the present application further provides a risk avoidance planning device for a vehicle, the device comprising:

[0039] An acquisition module, used for acquiring sensor data of the vehicle;

[0040] A feature extraction module, used to extract features from the sensor data to obtain bird's-eye view features;

[0041] A construction module, used for constructing an instance token according to the bird's-eye view features;

[0042] A trajectory prior module, used for mapping the instance token to the trajectory space to perform trajectory prior and obtain a potential expected trajectory;

[0043] A data acquisition module, used for acquiring target loss data according to the potential expected trajectory;

[0044] A risk avoidance planning module is used to perform risk avoidance planning on the trajectory of the vehicle according to the target loss data.

[0045] In the above implementation process, the bird's-eye view features are used to construct instance tokens, and then the trajectory prior is performed based on the instance tokens. Finally, the loss data is avoided to achieve reasonable risk avoidance planning, which can improve the safety of the vehicle during driving, improve the accuracy and feasibility of risk avoidance planning, and does not rely on pre-defined trajectories and paths. It has strong adaptability and can be applied to a variety of complex road environments and driving environments, providing protection for driving safety and pedestrian safety during vehicle driving.

[0046] In a third aspect, an embodiment of the present application provides a vehicle, comprising the risk avoidance planning device for the vehicle described in the second aspect.

[0047] In a fourth aspect, an embodiment of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in any one of the first aspects when executing the computer program.

[0048] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium, on which instructions are stored. When the instructions are executed on a computer, the computer executes the method described in any one of the first aspects.

[0049] In a sixth aspect, an embodiment of the present application provides a computer program product, which, when running on a computer, enables the computer to execute the method as described in any one of the first aspects.

[0050] Other features and advantages of the present disclosure will be set forth in the following description, or some features and advantages may be inferred or unambiguously determined from the description, or may be learned by implementing the above-mentioned technology of the present disclosure.

[0051] And it can be implemented according to the contents of the specification. The following is a detailed description of the preferred embodiments of the present application with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the range values. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 A schematic diagram of a flow chart of a method for risk avoidance planning of a vehicle provided in an embodiment of the present application;

[0054] Figure 2 A schematic diagram of the structure of a risk avoidance planning device for a vehicle provided in an embodiment of the present application;

[0055] Figure 3 A schematic diagram of the structural composition of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0057] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0058] The specific implementation methods of the present application are further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present application, but are not used to limit the scope of the present application.

[0059] Embodiment 1

[0060] Figure 1 is a flow chart of a method for risk avoidance planning of a vehicle provided in an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0061] S1, obtain the sensor data of the vehicle;

[0062] S2, extract features from sensor data to obtain bird’s-eye view features;

[0063] S3, constructs instance tokens based on bird’s-eye view features;

[0064] S4, maps instance tokens to trajectory space for trajectory prior to obtain potential expected trajectories;

[0065] S5, obtaining target loss data based on potential expected trajectory;

[0066] S6, performing risk avoidance planning for the vehicle's trajectory according to the target loss data.

[0067] In the above implementation process, the bird's-eye view features are used to construct instance tokens, and then the trajectory prior is performed based on the instance tokens. Finally, the loss data is avoided to achieve reasonable risk avoidance planning, which can improve the safety of the vehicle during driving, improve the accuracy and feasibility of risk avoidance planning, and does not rely on pre-defined trajectories and paths. It has strong adaptability and can be applied to a variety of complex road environments and driving environments, providing protection for driving safety and pedestrian safety during vehicle driving.

[0068] In S1, this application uses a small electric car to collect data. The car is equipped with 6 cameras and 5 radars to monitor the surrounding environment in all directions. 84 driving records were selected, totaling about 15 hours, covering a distance of 242 kilometers, with an average speed of 16 kilometers per hour. The data includes various scenes in cities, residential areas, suburbs and industrial areas, including daytime, night, sunny, rainy and cloudy weather. Finally, the data are divided into 1,000 segments, each about 20 seconds long. The collection process is divided into an offline stage, using laser point clouds to generate high-definition maps; in the online stage, the odometer and laser data are combined to locate the vehicle through the Monte Carlo positioning algorithm, and the vehicle's position error is controlled within 10 centimeters.

[0069] Furthermore, S2 includes:

[0070] Extract features from sensor data to obtain multi-scale image features;

[0071] Convert multi-scale image features into bird’s-eye-view spatial data based on a pre-built deformable attention model;

[0072] The bird's-eye view feature is obtained by aligning the bird's-eye view feature spatial data with the current vehicle coordinates in the sensor data.

[0073] In the above implementation process, the multi-scale image features are converted into bird's-eye view spatial data and then aligned with the current ego-vehicle coordinates, so that the bird's-eye view features can contain more ego-vehicle features and ego-vehicle data, thereby improving the accuracy of risk avoidance planning.

[0074] This application uses convolutional neural networks and feature pyramid networks to obtain multi-scale image features F from sensor data and transform H×W bird's-eye view (BEV) tokens Initialize and use deformable cross-attention to transfer information from multi-scale image features F to obtain bird's-eye view feature spatial data.

[0075] ;

[0076] Among them, B represents the bird's-eye view spatial data, DA (Q, K, V) represents the deformable attention model, Q, K and V are used as query, key and value respectively, and the embodiment of the present application fills the bird's-eye view token and multi-scale image features in sequence.

[0077] Furthermore, S3 includes:

[0078] The bird’s-eye view features are learned according to the pre-built global cross-attention model to obtain a set of map tokens;

[0079] Get his car token collection;

[0080] Construct an instance token based on the map token set and the other car token set.

[0081] In the above implementation process, a map token set is constructed according to the bird's-eye view features, and then an instance token is constructed according to the other vehicle token set and the map token set. This can expand the effective range of the instance token in the application process and provide data support for subsequent trajectory priors.

[0082] Since semantic map elements are usually sparse in the bird's-eye view space, this application uses map tokens M to represent the semantic map. M can be decoded into a set of points in the bird's-eye view space by a map decoder, representing the categories of map elements and their corresponding positions. This application considers three types of map elements (i.e., lane dividers, road boundaries, and crosswalks).

[0083] ;

[0084] in, To initialize the map token, B represents the bird's-eye view special spatial data, CA (Q, K, V) is the global cross-attention model, and Q, K and V are used as query, key and value respectively. The embodiment of the present application will initialize the map token and the bird's-eye view special spatial data in the bird's-eye view feature in sequence.

[0085] Furthermore, the step of obtaining the other vehicle token set includes:

[0086] Get other car data;

[0087] Obtain other car tokens based on other car data;

[0088] The other car tokens are updated according to the pre-built deformable cross attention model to obtain the other car token set.

[0089] In the above implementation process, updating the other vehicle tokens according to the deformable cross-attention model can improve the rationality and effectiveness of the other vehicle token set, improve data accuracy, and reduce errors generated during the updating process.

[0090] This application uses a set of other vehicle tokens A to represent the 3D position of each surrounding instance. The embodiment of this application uses a deformable cross attention model to obtain updated other vehicle tokens from the bird's-eye view feature spatial data in the bird's-eye view feature. , as shown below:

[0091] ;

[0092] in, To initialize his car token.

[0093] Since prediction and planning are mainly focused on instances of other vehicles and the vehicle itself, in order to comprehensively and effectively represent the autonomous driving scenario, the embodiment of the present application adds a vehicle token e to the vehicle token A, and map semantic information is added to the vehicle token e, that is, the vehicle token e is constructed based on the map token set and contains map semantic information.

[0094] Construct an instance token I = concat (e, A). The embodiment of the present application realizes high-order interaction between the self-vehicle and other vehicles through self-attention operation.

[0095] ;

[0096] Among them, SA (Q, K, V) is a self-attention block composed of self-attention layers, using Q, K and V as query, key and value respectively. Through the learning of the attention module, the instance token contains high-order other-vehicle-self-vehicle interaction and semantic mapping. These features contain the necessary map and instance information required to perform trajectory planning.

[0097] Furthermore, S4 includes:

[0098] Get the expected trajectory of the instance token;

[0099] Map the expected trajectory to the trajectory space to obtain the potential distribution information of the expected trajectory;

[0100] The potential distribution information of the expected trajectory is decoded to obtain the potential expected trajectory.

[0101] In the above implementation process, the expected trajectory is mapped to the trajectory space, and the potential distribution of the expected trajectory is obtained and then decoded, so that the obtained potential expected trajectory is more consistent with the risk avoidance plan of the vehicle, thereby reducing the driving risk of the vehicle.

[0102] The goal of the method proposed in this application can be formalized as inferring the expected trajectory T of a given map-aware instance token. This application models it as a trajectory generation problem T~p(T|I). Taking into account its uncertainty, a variational autoencoder (VAE) architecture is used to learn. A ground truth trajectory encoder is used to model p(z|T(T, f)), which maps the expected trajectory T(T, f) to a diagonal Gaussian distribution in the trajectory space z. The encoder outputs two vectors, representing the mean and variance of the Gaussian distribution. The learned distribution p(z|T(T, f)) contains prior information about the ground truth trajectory, which can be used to improve the authenticity of motion prediction and planning of other vehicles and self-vehicles, where p is the sign of the probability distribution and f is the mapping function.

[0103] After obtaining the potential distribution of the desired trajectory as a prior, it is necessary to decode it from the trajectory space. This application does not decode the entire trajectory at one time, but uses a simple MLP-based decoder to decode a waypoint from the trajectory space. Then, a gated recurrent unit (GRU) is used as the desired trajectory generator to model the time evolution of the waypoint. The GRU takes the current potential representation as input and converts it to the next state. The waypoint decoder can be used to decode the waypoint at the (t + 1)th timestamp. The generated trajectory is more realistic and lifelike after considering the prior knowledge in the learned structured latent space.

[0104] Furthermore, S5 includes:

[0105] Obtain divergence loss data based on potential expected trajectories;

[0106] Obtain trajectory loss data, semantic map loss data, and 3D object detection loss data;

[0107] Target loss data is constructed based on divergence loss data, trajectory loss data, semantic map loss data, and 3D object detection loss data.

[0108] In the above implementation process, target loss data is constructed according to divergence loss data, trajectory loss data, semantic map loss data and 3D object detection loss data, which can improve the error rate in the risk avoidance planning process and make the risk avoidance planning more accurate.

[0109] This application uses two trainings, map segmentation and 3D object detection, using a map decoder on the map token and a target decoder on the other car token to obtain the predicted map and 3D object detection results, and applies semantic map loss and 3D object detection loss to train the network respectively, so as to construct the target loss data as follows:

[0110] ;

[0111] in, , and is the balancing factor, is the divergence loss data, is the trajectory loss data, Loss data for semantic maps, Loss data for 3D object detection.

[0112] This application abandons the expected trajectory encoder and samples a latent state from the instance distribution as the input of the trajectory generator and waypoint decoder, models end-to-end autonomous driving as a generation problem, and performs risk avoidance prediction and planning in a structured latent space, taking into account the prior knowledge of the real trajectory to produce high-quality trajectory prediction and planning.

[0113] Embodiment 2

[0114] In order to execute the method corresponding to the above embodiment 1 to achieve the corresponding functions and technical effects, a vehicle risk avoidance planning device is provided below, such as Figure 2 As shown, the device comprises:

[0115] Acquisition module 1, used to acquire sensor data of the vehicle;

[0116] Feature extraction module 2, used to extract features from sensor data to obtain bird's-eye view features;

[0117] Building module 3, for building instance tokens according to bird's-eye view features;

[0118] Trajectory prior module 4, used to map instance tokens to trajectory space for trajectory prior and obtain potential expected trajectory;

[0119] A data acquisition module 5, for obtaining target loss data according to a potential expected trajectory;

[0120] The risk avoidance planning module 6 is used to perform risk avoidance planning on the trajectory of the vehicle according to the target loss data.

[0121] In the above implementation process, the bird's-eye view features are used to construct instance tokens, and then the trajectory prior is performed based on the instance tokens. Finally, the loss data is avoided to achieve reasonable risk avoidance planning, which can improve the safety of the vehicle during driving, improve the accuracy and feasibility of risk avoidance planning, and does not rely on pre-defined trajectories and paths. It has strong adaptability and can be applied to a variety of complex road environments and driving environments, providing protection for driving safety and pedestrian safety during vehicle driving.

[0122] Furthermore, the feature extraction module 2 is also used for:

[0123] Extract features from sensor data to obtain multi-scale image features;

[0124] Convert multi-scale image features into bird’s-eye-view spatial data based on a pre-built deformable attention model;

[0125] The bird's-eye view feature is obtained by aligning the bird's-eye view feature spatial data with the current vehicle coordinates in the sensor data.

[0126] In the above implementation process, the multi-scale image features are converted into bird's-eye view spatial data and then aligned with the current ego-vehicle coordinates, so that the bird's-eye view features can contain more ego-vehicle features and ego-vehicle data, thereby improving the accuracy of risk avoidance planning.

[0127] Furthermore, building block 3 is also used for:

[0128] The bird’s-eye view features are learned according to the pre-built global cross-attention model to obtain a set of map tokens;

[0129] Get his car token collection;

[0130] Construct an instance token based on the map token set and the other car token set.

[0131] In the above implementation process, a map token set is constructed according to the bird's-eye view features, and then an instance token is constructed according to the other vehicle token set and the map token set. This can expand the effective range of the instance token in the application process and provide data support for subsequent trajectory priors.

[0132] Furthermore, building block 3 is also used for:

[0133] Get other car data;

[0134] Obtain other car tokens based on other car data;

[0135] The other car tokens are updated according to the pre-built deformable cross attention model to obtain the other car token set.

[0136] In the above implementation process, updating the other vehicle tokens according to the deformable cross-attention model can improve the rationality and effectiveness of the other vehicle token set, improve data accuracy, and reduce errors generated during the updating process.

[0137] Furthermore, the trajectory priori module 4 is also used for:

[0138] Get the expected trajectory of the instance token;

[0139] Map the expected trajectory to the trajectory space to obtain the potential distribution information of the expected trajectory;

[0140] The potential distribution information of the expected trajectory is decoded to obtain the potential expected trajectory.

[0141] In the above implementation process, the expected trajectory is mapped to the trajectory space, and the potential distribution of the expected trajectory is obtained and then decoded, so that the obtained potential expected trajectory is more consistent with the risk avoidance plan of the vehicle, thereby reducing the driving risk of the vehicle.

[0142] Furthermore, the data acquisition module 5 is also used for:

[0143] Obtain divergence loss data based on potential expected trajectories;

[0144] Obtain trajectory loss data, semantic map loss data, and 3D object detection loss data;

[0145] Target loss data is constructed based on divergence loss data, trajectory loss data, semantic map loss data, and 3D object detection loss data.

[0146] In the above implementation process, target loss data is constructed according to divergence loss data, trajectory loss data, semantic map loss data and 3D object detection loss data, which can improve the error rate in the risk avoidance planning process and make the risk avoidance planning more accurate.

[0147] The above-mentioned vehicle risk avoidance planning device can implement the method of the above-mentioned embodiment 1. The options in the above-mentioned embodiment 1 are also applicable to this embodiment and will not be described in detail here.

[0148] The rest of the contents of the embodiments of the present application can refer to the contents of the above-mentioned embodiment 1, and will not be repeated in this embodiment.

[0149] Embodiment 3

[0150] An embodiment of the present application provides a vehicle, including the risk avoidance planning device for the vehicle of embodiment 2.

[0151] Embodiment 4

[0152] An embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the vehicle risk avoidance planning method of embodiment one.

[0153] Optionally, the above-mentioned electronic device may be a server.

[0154] See also Figure 3 , Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include a processor 31, a communication interface 32, a memory 33 and at least one communication bus 34. The communication bus 34 is used to realize direct connection and communication between these components. The communication interface 32 of the device in the embodiment of the present application is used to communicate signaling or data with other node devices. The processor 31 may be an integrated circuit chip with signal processing capabilities.

[0155] The processor 31 mentioned above 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. It can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor can be a microprocessor or the processor 31 can also be any conventional processor, etc.

[0156] The memory 33 may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electric erasable programmable read-only memory (EEPROM), etc. The memory 33 stores computer readable instructions. When the computer readable instructions are executed by the processor 31, the device can execute the above-mentioned Figure 1 The method embodiment involves various steps.

[0157] Optionally, the electronic device may further include a storage controller and an input / output unit. The memory 33, the storage controller, the processor 31, the peripheral interface, and the input / output unit are electrically connected to each other directly or indirectly to achieve data transmission or interaction. For example, these components may be electrically connected to each other via a communication bus 34. The processor 31 is used to execute executable modules stored in the memory 33, such as software function modules or computer programs included in the device.

[0158] The input and output unit is used to provide users with the task creation and to create an optional time period or preset execution time for the task to enable interaction between the user and the server. The input and output unit can be, but is not limited to, a mouse and a keyboard.

[0159] Understandably, Figure 3 The structure shown is for illustration only. The electronic device may also include Figure 3 More or fewer components as shown, or with Figure 3 Different configurations are shown. Figure 3Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0160] In addition, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the vehicle risk avoidance planning method of embodiment one.

[0161] The embodiment of the present application also provides a computer program product, which, when running on a computer, enables the computer to execute the method described in the method embodiment.

[0162] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of boxes in the block diagram and / or the flowchart, can be implemented with a dedicated hardware-based device that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0163] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0164] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks, or optical disks.

[0165] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application. It should be noted that similar numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0166] The above is only a specific implementation of the present application, but the protection range of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical range disclosed in the present application, which should be included in the protection range of the present application. Therefore, the protection range of the present application should be based on the protection range of the claims.

[0167] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

Claims

1. A vehicle risk avoidance planning method, characterized in that: The method comprises: Acquiring sensor data of the vehicle; Extracting features from the sensor data to obtain bird's-eye view features; constructing an instance token according to the bird's-eye view features; Mapping the instance tokens to the trajectory space for trajectory prior to obtain potential expected trajectories; obtaining target loss data according to the potential desired trajectory; Performing risk avoidance planning for the trajectory of the vehicle according to the target loss data; The step of mapping the instance token to the trajectory space to perform trajectory prior to obtain a potential expected trajectory comprises: obtaining a desired trajectory of the instance token; Mapping the expected trajectory to the trajectory space to obtain potential distribution information of the expected trajectory; The potential distribution information of the expected trajectory is decoded to obtain the potential expected trajectory.

2. The method for avoiding danger of a vehicle according to claim 1, characterized in that: The step of extracting features from the sensor data to obtain bird's-eye view features comprises: Performing feature extraction on the sensor data to obtain multi-scale image features; Converting the multi-scale image features into bird's-eye view spatial data according to a pre-built deformable attention model; The bird's-eye view feature spatial data is aligned with the current vehicle coordinates in the sensor data to obtain the bird's-eye view feature.

3. The method for avoiding danger planning of a vehicle according to claim 1, characterized in that: The step of constructing an instance token according to the bird's-eye view feature comprises: Learning the bird's-eye view features according to a pre-built global cross-attention model to obtain a map token set; Get his car token collection; The instance token is constructed according to the map token set and the other vehicle token set.

4. The method for avoiding danger planning of a vehicle according to claim 3, characterized in that: The step of obtaining the other vehicle token set includes: Get other car data; Obtaining a token for another vehicle according to the other vehicle data; The other vehicle tokens are updated according to a pre-built deformable cross-attention model to obtain the other vehicle token set.

5. The vehicle risk avoidance planning method according to claim 1, characterized in that: The step of obtaining target loss data according to the potential expected trajectory comprises: Obtaining divergence loss data according to the potential desired trajectory; Obtain trajectory loss data, semantic map loss data, and 3D object detection loss data; The target loss data is constructed according to the divergence loss data, the trajectory loss data, the semantic map loss data and the 3D object detection loss data.

6. A vehicle risk avoidance planning device, characterized in that: The device comprises: An acquisition module, used for acquiring sensor data of the vehicle; A feature extraction module, used to extract features from the sensor data to obtain bird's-eye view features; A construction module, used for constructing an instance token according to the bird's-eye view features; A trajectory prior module, used for mapping the instance token to the trajectory space to perform trajectory prior and obtain a potential expected trajectory; A data acquisition module, used for acquiring target loss data according to the potential expected trajectory; A risk avoidance planning module, used for performing risk avoidance planning on the trajectory of the vehicle according to the target loss data; The trajectory prior module is also used for; obtaining a desired trajectory of the instance token; Mapping the expected trajectory to the trajectory space to obtain potential distribution information of the expected trajectory; The potential distribution information of the expected trajectory is decoded to obtain the potential expected trajectory.

7. A vehicle, characterized in that: The vehicle includes the vehicle hazard avoidance planning device according to claim 6.

8. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the risk avoidance planning method for a vehicle according to any one of claims 1 to 5.

9. A storage medium, characterized in that: It stores a computer program, which, when executed by a processor, implements the risk avoidance planning method for a vehicle as described in any one of claims 1 to 5.

Images