Driving intention prediction method and device

Abstract

The present disclosure provides an embodiment method performed by a driving intention prediction apparatus for predicting a driving intention of a driver, including: receiving driving environment information including lane information and trajectory information; receiving driving information; outputting a lane feature vector by inputting the lane information into a first encoder; outputting a trajectory feature vector by inputting the trajectory information into a second encoder; and predicting a driving intention by inputting the plurality of feature vectors into a driving intention prediction model.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of Korean Patent Application No. 10-2023-0141272, filed on October 20, 2023, which is hereby incorporated by reference in its entirety. Technical Field

[0003] The present disclosure relates to a driving intention prediction method and device. Background Art

[0004] The description in this section merely provides background information related to embodiments of the present disclosure and does not necessarily constitute known prior art.

[0005] Autonomous vehicles are designed to navigate to the vehicle's destination without driver intervention. Autonomous vehicles below Level 4 require driver intervention and need to predict driving intent in order to maintain safe control while the driver is behind the wheel.

[0006] Traditional methods can only grasp limited driving intentions, such as turning, going straight, changing lanes, and maintaining the vehicle lane based on information such as vehicle speed, steering angle, and yaw rate. It is difficult to judge complex driving intentions, such as sudden lane changes, lane merging, and lane deviation.

[0007] Driving intention has a causal relationship with lane networks (e.g., intersections), traffic signals, and past driving trajectories. However, conventional methods do not take these into account. Therefore, in order to accurately predict driving intention, a driving intention determination method that considers lane networks, traffic signals, and past driving trajectories is required.

[0008] In addition, the traditional method of determining driving intention using artificial intelligence proposes predicting the vehicle's destination coordinates to predict the future route, but the prediction results are unstable due to inputs not considered during artificial intelligence training. Therefore, a method of judging driving intention using a stable model is needed. Summary of the invention

[0009] The present disclosure relates to a method and device for predicting driving intention. Specific embodiments relate to a method and device for predicting driving intention using a grid map.

[0010] The embodiments of the present disclosure provide a reliable driving intention prediction model and a training method thereof, which can accurately predict complex driving intentions.

[0011] The embodiments of the present disclosure are not limited to the above-mentioned embodiments, and other embodiments not mentioned herein will be clearly understood by those skilled in the art from the following description.

[0012] At least one embodiment of the present disclosure provides a method performed by a driving intention prediction device for predicting a driver's driving intention, comprising: receiving driving environment information including lane information and trajectory information and receiving driving information, inputting lane information into a first encoder to output a lane feature vector, inputting trajectory information into a second encoder to output a trajectory feature vector, and predicting driving intention by inputting multiple feature vectors into a driving intention prediction model.

[0013] Another embodiment of the present disclosure provides a driving intention prediction device, including: an input unit, which is configured to receive driving environment information including lane information and trajectory information and receive driving information; a feature extraction unit, which is configured to output a lane feature vector when the lane information is received and output a trajectory feature vector when the trajectory information is received; and a driving intention prediction unit, which is configured to predict driving intention by inputting multiple feature vectors into a driving intention prediction model.

[0014] Implementations of the above-described methods and apparatus may include one or more of the following features.

[0015] Implementations may include training a driving intention prediction model using driving intention and trajectory information.

[0016] The process of outputting a lane feature vector includes normalizing the lane information by removing noise from the lane information and representing the denoised lane information as a graph, and inputting the normalized lane information to a first encoder.

[0017] The process of outputting the trajectory feature vector includes normalizing the trajectory information by removing noise from the trajectory information and representing the denoised trajectory information as a graph, and inputting the normalized trajectory information to a second encoder.

[0018] The first encoder and the second encoder may each include a graph convolutional network (GCN).

[0019] Embodiments may further include generating a grid map using the driving information and the lane information, wherein the driving intention is represented as a destination on the grid map.

[0020] According to an embodiment of the present disclosure, complex driving intentions may be predicted by predicting a destination of a vehicle using a grid map.

[0021] According to an embodiment of the present disclosure, driving intention may be accurately predicted using lane information, trajectory information, and traffic signal information.

[0022] According to an embodiment of the present disclosure, a method for training a reliable driving intention prediction model by predicting driving intention within a limited grid map may be provided.

[0023] The effects of the embodiments of the present disclosure are not limited to the above-mentioned effects, and other unmentioned effects will be clearly understood by those skilled in the art from the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a schematic block diagram of a driving intention prediction device according to at least one embodiment of the present disclosure.

[0025] Figure 2 is a schematic block diagram of a feature extraction unit according to at least one embodiment of the present disclosure.

[0026] Figure 3 is a schematic diagram illustrating lane information and trajectory information according to at least one embodiment of the present disclosure.

[0027] Figure 4A and Figure 4B An exemplary grid map generated by a grid map generation unit according to at least one embodiment of the present disclosure is illustrated.

[0028] Figure 5 is a schematic diagram illustrating a driving intention predicted on a grid map by a driving intention prediction model according to at least one embodiment of the present disclosure.

[0029] Figure 6 is a flowchart of a driving intention prediction method according to at least one embodiment of the present disclosure.

[0030] Figure 7 is a flowchart of a method for training a driving intention prediction model according to at least one embodiment of the present disclosure. DETAILED DESCRIPTION

[0031] Hereinafter, some embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. In the following description, similar figure labels preferably represent similar elements, although these elements are shown in different drawings. In addition, in the following description of some embodiments, for the purpose of clarity and brevity, when it is considered that related known components and functions will obscure the subject matter of the embodiments of the present disclosure, a detailed description thereof will be omitted.

[0032] Various ordinal or letter codes, such as first, second, i), ii), a), b) and other prefixes are only used to distinguish one component from another component, and do not imply or suggest the substance, order or sequence of the components. Throughout the specification, when a part "includes" or "contains" a component, it means that the part also includes other components, rather than excluding other components, unless there is a specific description to the contrary. For example, terms such as "unit", "module" and the like refer to one or more units for processing at least one function or operation, which can be implemented by hardware, software or a combination thereof.

[0033] The following description of the embodiments of the present disclosure in conjunction with the accompanying drawings is intended to describe exemplary embodiments of the present disclosure, rather than merely representing embodiments of the technical concepts that can implement the embodiments of the present disclosure.

[0034] Figure 1 is a schematic block diagram of a driving intention prediction device according to at least one embodiment of the present disclosure. Figure 1 The components shown represent functionally different elements, and in an actual physical environment, at least one component may be implemented to be integrated with other components.

[0035] refer to Figure 1 The driving intention prediction device 10 includes all or part of an input unit 100 , a feature extraction unit 110 , a grid map generation unit 120 , and a driving intention prediction unit 130 .

[0036] According to at least one embodiment of the present disclosure, the driving intention prediction device 10 may use a coordinate system to represent the positions of objects and elements in the environment. For example, the driving intention prediction device 10 may use a coordinate system with the current position of the vehicle as the origin to represent the positions of surrounding objects and elements. The driving intention prediction device 10 may use the following coordinate system to represent the positions of surrounding objects and elements, wherein the current driving direction of the vehicle is used as the x direction and the width direction of the vehicle is used as the y direction.

[0037] The input unit 100 receives the driving information of the vehicle. Here, the driving information of the vehicle includes at least one of the acceleration of the vehicle, the yaw rate of the vehicle, the steering angle of the vehicle, the speed of the vehicle, the input values ​​of the accelerator pedal and the brake pedal, or any combination thereof. In an example, the input unit 100 may receive the driving information from a sensor unit (not shown). In another example, the input unit 100 may receive the driving information of the vehicle from a memory.

[0038] The input unit 100 receives driving environment information of the vehicle. Here, the driving environment information includes at least one of a high-definition map, trajectory information, traffic signal information, the current position of the vehicle, or any combination thereof. The high-definition map may include lane information, crossroad information, etc. The lane information may include information about the center point of the lane, the center line of the lane, and the width of the lane. In an example, the input unit 100 may receive the driving environment information from a server. In another example, the input unit 100 may receive the driving environment information from a memory.

[0039] The feature extraction unit 110 pre-processes the driving information and driving environment information received by the input unit 100, and encodes the pre-processed information to extract a feature vector.

[0040] References below Figure 2 and Figure 3 The feature extraction unit 110 is described.

[0041] Figure 2 is a schematic block diagram of a feature extraction unit 110 according to at least one embodiment of the present disclosure. Figure 2 The components shown represent functionally different elements, and in an actual physical environment, at least one component may be implemented to be integrated with other components.

[0042] Figure 3 is a schematic diagram illustrating lane information and trajectory information according to at least one embodiment of the present disclosure.

[0043] refer to Figure 2 and Figure 3 The feature extraction unit 110 includes all or part of the lane information unit 200 , the trajectory information unit 210 and the driving information unit 220 .

[0044] The lane information unit 200 receives lane information from the input unit 100. The lane information is information about a lane included in the high-definition map. The lane information may include, for example, information about a lane center point and a lane center line.

[0045] The lane information unit 200 pre-processes the received lane information. The lane information unit 200 may remove noise from the lane information using a moving average filter. The lane information unit 200 may convert the coordinates of each point of the lane information into relative coordinates based on the current position of the vehicle and store them as a graph.

[0046] The lane information unit 200 may convert the coordinates of each point of the denoised lane information into relative coordinates. Here, the relative coordinates may be coordinates in a coordinate system with the current position of the vehicle as the origin. The lane information may be represented by a set of points. For example, if the lane in which the vehicle is currently located is a straight lane, the lane may be represented by a set of coordinates such as (1,0), (2,0), and (3,0). As another example, a left turn lane may be represented by a set of coordinates such as (1,0.2), (2,1), and (2.2,1.5). These coordinates are for illustrative purposes only and are not intended to be important or limiting factors in the present disclosure.

[0047] The lane information unit 200 stores the denoised lane information as a graph. One graph represents one lane. For example, a straight lane may be represented by a graph connecting coordinates (1,0), (2,0) and (3,0). In another example, a left turn lane may be represented by a graph connecting coordinates (1,0.2), (2,1) and (2.2,1.5).

[0048] The lane information unit 200 inputs the preprocessed lane information into a first encoder. The first encoder may include a pretrained artificial neural network. The pretrained artificial neural network may be trained by machine learning or deep learning. The pretrained artificial neural network may be, for example, a graph convolutional network (GCN). The lane information unit 200 extracts a feature vector from the lane information using the first encoder and outputs the feature vector. For example, the lane information unit 200 may extract a one-dimensional feature vector from the lane information and output the one-dimensional feature vector. The feature vector output by the lane information unit 200 is referred to as a lane feature vector.

[0049] The lane information unit 200 may only use lane information within a certain distance from the current position of the vehicle. For example, the lane information unit 200 may only use lane information of three lanes including the current lane of the vehicle to output a one-dimensional feature vector.

[0050] The trajectory information unit 210 receives trajectory information from the input unit 100. The trajectory information is information on a past travel trajectory of the vehicle.

[0051] The trajectory information unit 210 pre-processes the received trajectory information. The trajectory information unit 210 may remove noise from the trajectory information using a moving average filter. The trajectory information unit 210 may convert the coordinates of each point of the trajectory information into relative coordinates based on the current position of the vehicle and store them as a graph.

[0052] The trajectory information unit 210 can convert the coordinates of each point of the denoised trajectory information into relative coordinates. Here, the relative coordinates may be coordinates in a coordinate system with the current position of the vehicle as the origin. The trajectory information may be represented by a set of points. For example, if the vehicle has been going straight, the trajectory may be represented by a set of (-3,0), (-2,0) and (-1,0). If the vehicle turns right to reach the current position, the trajectory may be represented by a set of (-2,-1.5), (-1.5,-0.5) and (-1,-0.2). The aforementioned coordinates are for illustrative purposes only and do not imply any meaning or limit the present disclosure.

[0053] The trajectory information unit 210 stores the denoised trajectory information as a graph. One graph represents one trajectory. For example, the trajectory of the vehicle when it continues to go straight can be represented by a graph connecting coordinates (-3,0), (-2,0) and (-1,0). In another example, if the vehicle turns right to reach its current position, the trajectory can be represented by a graph connecting coordinates (-2,-1.5), (-1.5,-0.5) and (-1,-0.2). Typically, the vehicle travels along the center line of the lane. Therefore, according to at least one embodiment of the present disclosure, a graph can be a graph with the same shape as the center line of the lane.

[0054] The trajectory information unit 210 inputs the preprocessed trajectory information into a second encoder. Here, the second encoder may include a pretrained artificial neural network. The pretrained artificial neural network may be trained by machine learning or deep learning. The pretrained neural network may be, for example, a graph convolutional network (GCN). The trajectory information unit 210 extracts a feature vector from the trajectory information using the second encoder and outputs the feature vector. For example, the trajectory information unit 210 may extract a one-dimensional feature vector from the trajectory information and output the one-dimensional feature vector. The feature vector output by the trajectory information unit 210 is called a trajectory feature vector.

[0055] The trajectory information unit 210 may only use trajectory information within a certain distance from the current position of the vehicle. For example, the trajectory information unit 210 may only use trajectory information of 1 kilometer in length to output a one-dimensional feature vector.

[0056] The driving information unit 220 receives the driving information of the vehicle from the input unit 100. The driving information unit 220 preprocesses the driving information of the vehicle. Preprocessing is a process of removing noise. Here, the driving information unit 220 may preprocess the driving information of the vehicle using a low pass filter (LPF).

[0057] The driving information unit 220 inputs the preprocessed driving information into a third encoder. The third encoder may include a pretrained artificial neural network. Here, the pretrained artificial neural network may be trained by machine learning or deep learning. The driving information unit 220 extracts a feature vector from the driving information using the third encoder and outputs the feature vector. For example, the driving information unit 220 may extract a one-dimensional feature vector from the driving information and output the one-dimensional feature vector. The feature vector output by the driving information unit 220 is referred to as a driving feature vector.

[0058] The grid map generation unit 120 generates a grid map. Figure 4A and Figure 4B The grid map generating unit 120 is described.

[0059] Figure 4A and Figure 4B An exemplary grid map generated by a grid map generation unit according to at least one embodiment of the present disclosure is illustrated.

[0060] The grid map generation unit 120 generates a grid map of the driving available area based on the driving information and the driving environment information. The grid map generation unit 120 determines the x-direction length of the grid map using the speed and acceleration of the vehicle. The grid map generation unit 120 determines the y-direction length of the grid map using the width of the driving lane and one or more adjacent lanes. The grid map generation unit 120 generates a grid map by dividing the area into a plurality of grids. Here, the grid map generation unit 120 generates a grid map only for the driving direction of the vehicle. In other words, a grid map is generated only for the area in front of the vehicle.

[0061] According to at least one embodiment of the present disclosure, the size of each grid in the grid map may be the same. Figure 4A , all grids in a grid map may have the same length in the x direction and the same length in the y direction.

[0062] According to another embodiment of the present disclosure, the size (area) of each grid in the grid map may be different. Figure 4B , the size of each grid in the lane where the vehicle is currently located may be smaller than the size of each grid in other lanes. For example, the size of each grid in the current lane of the vehicle may be 0.5 times the size of the grid in other lanes. In other words, the y-direction length of each grid in the current lane of the vehicle may be 0.5 times the y-direction length of the grid in other lanes.

[0063] Alternatively, each grid of the grid map may be constructed differently from the above grids.

[0064] The driving intention prediction unit 130 predicts the driving intention. The driving intention prediction unit 130 uses a driving intention prediction model to predict the driving intention. The driving intention prediction model can be pre-trained by machine learning or deep learning. The driving intention prediction model can be implemented, for example, based on a deep neural network (DNN).

[0065] The driving intention prediction unit 130 receives a feature vector from the feature extraction unit 110. The driving intention prediction unit 130 receives traffic signal information from the input unit 100. The feature vector includes at least one of a driving feature vector, a lane feature vector, a trajectory feature vector, or any combination thereof. The driving intention prediction unit 130 may use a plurality of feature vectors and traffic signal information to predict driving intention and train a driving intention prediction model.

[0066] The driving intention prediction model receives the feature vector and the traffic signal information as input and predicts the driving intention. According to at least one embodiment of the present disclosure, the driving intention predicted by the driving intention prediction model may be represented as a grid area on a grid map.

[0067] Figure 5is a schematic diagram illustrating a grid map depicting a driving intention predicted by a driving intention prediction model according to at least one embodiment of the present disclosure.

[0068] refer to Figure 5 , if the predicted driving intention is a left turn, the driving intention prediction model may predict the 4th grid area of ​​the grid map as the destination, as shown in 500. If the predicted driving intention is a sudden lane change, the driving intention prediction model may predict the 14th grid area of ​​the grid map as the destination, as shown in 510. If the predicted driving intention is a slow lane change, the driving intention prediction model may predict the 15th grid area of ​​the grid map as the destination, as shown in 520. If the predicted driving intention is a directional driving within the lane, the driving intention prediction model may predict the 12th grid area of ​​the grid map as the destination, as shown in 530. If the predicted driving intention is a U-turn, the 1st grid area of ​​the grid map may be predicted as the destination, as shown in 540.

[0069] Although Figure 4A and Figure 4B A grid map with 12 or 16 grids is shown, respectively, but they are only for illustration purposes and are not intended to limit the present disclosure. The number of grids in the grid map may be more than 16 or less than 12.

[0070] The driving intention prediction unit 130 may use the driving intention predicted by the driving intention prediction model to predict the vehicle's driving route. For example, the driving intention prediction unit 130 may use the driving intention and the path equation to predict the vehicle's driving route. That is, the driving intention prediction unit 130 may use the destination and the path equation on the grid map to predict the vehicle's driving route. The driving intention prediction model may be pre-trained using training data generated from driving history data collected from multiple vehicles. The driving intention prediction unit 130 may mark the actual driving route of the vehicle, and use the marked actual driving route as training data to train the driving intention prediction model. The actual driving route of the vehicle may be obtained from the trajectory information. For example, the actual driving route may be the same as the vehicle trajectory contained in the trajectory information.

[0071] Figure 6 is a flowchart of a driving intention prediction method according to at least one embodiment of the present disclosure.

[0072] refer to Figure 6 The input unit 100 receives driving information and driving environment information (S600). The driving environment information includes at least one of lane information, track information, traffic signal information, or any combination thereof.

[0073] The feature extraction unit 110 preprocesses the lane information, the trajectory information, and the driving information, and extracts their respective feature vectors (S610). For example, the feature extraction unit 110 may represent the lane information and the trajectory information from which noise has been removed as a graph. For example, the feature extraction unit 110 may remove noise from the driving information. The feature extraction unit 110 may input the preprocessed lane information into a first encoder to extract a lane feature vector. The feature extraction unit 110 may input the trajectory information into a second encoder to extract a trajectory feature vector. The feature extraction unit 110 may input the preprocessed driving information into a third encoder to extract a driving feature vector.

[0074] The grid map generation unit 120 generates a grid map using the driving information and the driving environment information (S620). The grid map generation unit 120 may determine an area to generate a grid map using lane width, vehicle speed, and vehicle acceleration, and may generate a grid map by dividing the area to generate a grid map.

[0075] The driving intention prediction unit 130 concatenates feature vectors. Here, the feature vector is a feature vector extracted from lane information, track information, and driving information. The driving intention prediction unit 130 inputs the concatenated feature vector into a driving intention prediction model (S630).

[0076] The driving intention prediction model predicts the driving intention (S640). The driving intention prediction model may predict the driving intention by determining a specific grid area on the grid map as a destination.

[0077] According to at least one embodiment of the present disclosure, the driving intention prediction model may be a model trained by a learning method, Figure 7 Give a description.

[0078] Figure 7 is a flowchart of a method for training a driving intention prediction model according to at least one embodiment of the present disclosure.

[0079] refer to Figure 7 The driving intention prediction model predicts the driving intention (S700). The driving intention prediction model may predict the driving intention by determining a specific grid area on the grid map as a destination.

[0080] The driving intention prediction model may generate a driving route to a specific grid area determined as a destination using a path equation (S710). The driving intention prediction model may predict a driving route to a destination using, for example, a cubic path equation.

[0081] The driving intention prediction model may receive the actual driving route. For example, the actual driving route may be obtained from the trajectory information. The driving intention prediction model may annotate the actual driving route or the trajectory information (S720).

[0082] The driving intention prediction model (S730) may be trained using the annotated actual driving route or the annotated trajectory information. For example, the driving intention prediction model may be trained by comparing the annotated actual driving route with the predicted driving route and using the comparison result to respectively change the weights of multiple layers.

[0083] According to the device or method of the embodiment of the present disclosure, the corresponding components can be arranged to be implemented as hardware or software or a combination of hardware and software. In addition, each component can be implemented by software in function, and when implemented, for each component, the microprocessor can implement the function by software.

[0084] Various illustrative embodiments of the systems and methods described herein may be implemented by digital electronic circuits, integrated circuits, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), computer hardware, firmware, software, and / or combinations thereof. These different embodiments may include embodiments implemented in one or more computer programs executable on a programmable system. The programmable system includes at least one programmable processor, at least one input device, and at least one output device, at least one programmable processor coupled to receive data and instructions from a storage system and to send data and instructions to the storage system, wherein the programmable processor may be a dedicated processor or a general-purpose processor. A computer program (also referred to as a program, software, software application, or code) contains instructions for a programmable processor and is stored in a "computer-readable recording medium".

[0085] The computer-readable recording medium includes any type of recording device on which data readable by a computer system can be recorded. Examples of computer-readable recording media include non-volatile or non-temporary media such as ROM, CD-ROM, magnetic tape, floppy disk, memory card, hard disk, optical / magnetic disk, storage device, etc. The computer-readable recording medium may also include temporary media such as data transmission media. In addition, the computer-readable recording medium may be distributed in computer systems connected via a network, where the computer-readable code may be stored and executed in a distributed manner.

[0086] Although the steps in each flowchart / sequence diagram are described as being performed in sequence, they only illustrate the technical concepts of some embodiments of the present disclosure. Therefore, a person skilled in the art can perform these steps by changing the order described in each flowchart / sequence diagram or by performing two or more of these steps in parallel, so the steps in each flowchart / sequence diagram are not limited to the sequence shown.

[0087] Although exemplary embodiments of the present disclosure are described for illustrative purposes, it will be appreciated by those skilled in the art that various modifications, additions, and substitutions may be made without departing from the concept and scope of the claimed invention. Therefore, for the sake of brevity and clarity, exemplary embodiments of the present disclosure are described. The scope of the technical concept of the embodiments of the present disclosure is not limited by the diagrams. Therefore, it will be appreciated by those of ordinary skill in the art that the scope of the claimed invention is not limited by the embodiments explicitly described above, but is defined by the claims and their equivalents.

Claims

1. A method for predicting a driver's driving intention performed by a driving intention prediction device, the method comprising the following steps: Receiving driving environment information including lane information and trajectory information; Receive driving information; Outputting a lane feature vector by inputting the lane information into a first encoder; Outputting a trajectory feature vector by inputting the trajectory information into a second encoder; as well as The driving intention is predicted by inputting a plurality of feature vectors into a driving intention prediction model.

2. The method according to claim 1, further comprising: The driving intention and the trajectory information are used to train the driving intention prediction model.

3. The method according to claim 2, wherein: The steps to train the driving intention prediction model include: marking the trajectory information; generating a predicted driving route using the driving intention; Comparing the annotated trajectory information with the predicted travel path of the route; and The weight of the driving intention prediction model is changed based on a result of the comparison.

4. The method according to claim 1, further comprising: The driving information is input into a third encoder to output a driving feature vector.

5. The method according to claim 4, wherein: The step of outputting the driving feature vector comprises: preprocessing the driving information by removing noise from the driving information; and The pre-processed driving information is input into the third encoder.

6. The method according to claim 1, further comprising: A grid map is generated using the lane information and the driving information, wherein the driving intention is represented as a destination on the grid map.

7. The method according to claim 6, wherein: The steps to generate a grid map include: Determining the size of an area to be generated for a grid map using the driving information and the lane information; and A grid map is generated by dividing the area into a plurality of grids.

8. The method according to claim 7, wherein: The step of generating a grid map by dividing the area into a plurality of grids includes dividing the area so that a size of each grid in the plurality of grids within a driving lane of the vehicle is different from other grids in the plurality of grids on other lanes.

9. The method according to claim 1, wherein: The step of predicting the driving intention includes combining the multiple feature vectors and inputting them into the driving intention prediction model.

10. The method according to claim 1, wherein: The step of predicting the driving intention includes predicting the driving intention by inputting the plurality of feature vectors and traffic signal information included in the driving environment information into the driving intention prediction model.

11. The method according to claim 1, wherein: The step of outputting the lane feature vector comprises: normalizing the lane information by removing noise from the lane information, and representing the normalized lane information as a first graph; and inputting the normalized lane information into the first encoder; and The step of outputting the trajectory feature vector comprises: normalizing the trajectory information by removing noise from the trajectory information, and representing the normalized trajectory information as a second graph; and The normalized trajectory information is input into the second encoder.

12. The method according to claim 1, wherein: The first encoder comprises a graph convolutional network.

13. A driving intention prediction device, comprising: one or more processors; as well as A storage device storing a program to be executed by the one or more processors, the program including instructions for: Receiving driving environment information including lane information and trajectory information; Receive driving information; When the lane information is received, outputting a lane feature vector; When the trajectory information is received, outputting a trajectory feature vector; as well as Driving intention is predicted by inputting multiple feature vectors into a driving intention prediction model.

14. The device according to claim 13, wherein: The program also includes instructions for the following operations: instructions for training the driving intention prediction model using the driving intention and the trajectory information.

15. The device according to claim 14, wherein: The program also includes instructions for: generating a predicted driving route using the driving intention; marking the trajectory information; Comparing the annotated trajectory information with the driving route predicted by the route; as well as The weight of the driving intention prediction model is changed based on a result of the comparison.

16. The apparatus according to claim 13, wherein: The program further includes instructions for outputting a driving feature vector from a third encoder when the driving information is received.

17. The apparatus according to claim 13, wherein: The program also includes instructions for: generating a grid map using the lane information and the driving information; Determine the size of the area to be generated of the grid map by using the driving information and the lane information; generating a grid map by dividing the area into a plurality of grids; as well as The driving intention is represented as a destination on a grid map.

18. The apparatus of claim 13, wherein: The program also includes instructions for: generating normalized lane information and normalized trajectory information by representing the lane information and the trajectory information as graphs, respectively; When the normalized lane information is received, outputting the lane information from the first encoder; as well as Upon receiving the normalized trajectory information, the trajectory information is output from the second encoder.

19. The apparatus according to claim 18, wherein: The first encoder and the second encoder each include a graph convolutional network.

20. The apparatus of claim 13, wherein: The program further includes instructions for predicting the driving intention by inputting the traffic signal information including the plurality of feature vectors and the traveling environment information into the driving intention prediction model.

Images