Information processing method, information processing device, and program

Abstract

[Problem] To automatically recognize connections between lanes. [Solution] An information processing method executed by a computer, the method including generating, on the basis of learning that uses a dataset including sensing data, which is data sensed by a vehicle, and a first topology map, which is of lanes corresponding to the position where the sensing data is acquired, a recognizer that outputs a second topology map to be used as a travel trajectory of the traveling vehicle.

Description

Information processing method, information processing device, and program

[0001] The present disclosure relates to an information processing method, an information processing device, and a program.

[0002] In recent years, technologies for automatically generating various maps related to vehicle travel have been developed. For example, Patent Literature 1 discloses a technology for automatically generating a new map by correcting a map representing pre-stored road information based on a deviation between lane boundaries recognized by a vehicle and the vehicle's travel path.

[0003] Japanese Patent Application Laid-Open No. 2023-149510

[0004] Meanwhile, with the development of autonomous driving technology, navigation technology, etc., there is a demand for automatically generating a topology map of lanes that is used as a trajectory for a vehicle. However, with the technology disclosed in Patent Document 1, it was difficult to generate such a topology map.

[0005] Therefore, the present disclosure proposes a new and improved technology that can automatically recognize connections between lanes.

[0006] According to the present disclosure, there is provided an information processing method executed by a computer, which includes generating a recognizer that outputs a second topology map used as a travel trajectory of a traveling vehicle based on learning using a dataset including sensing data that is data sensed by a vehicle and a first topology map that is a topology map of a lane corresponding to a position where the sensing data was acquired.

[0007] Furthermore, according to the present disclosure, there is provided an information processing device including a generation unit that generates a recognizer that outputs a second topology map used as a travel trajectory of a traveling vehicle based on learning using a dataset that includes sensing data that is data sensed by a vehicle and a first topology map that is a topology map of a lane corresponding to a position where the sensing data was acquired.

[0008] Furthermore, according to the present disclosure, there is provided a program that causes a computer to function as a generator that generates a recognizer that outputs a second topology map used as a travel trajectory of a traveling vehicle based on learning using a dataset that includes sensing data that is data sensed by a vehicle and a first topology map that is a topology map of a lane corresponding to a position where the sensing data was acquired.

[0009] 1 is a block diagram showing the overall configuration of an information processing system according to the present embodiment. FIG. 2 is a block diagram showing the functional configuration of a vehicle 10 according to the present embodiment. FIG. 3 is a diagram for explaining generation of a local lane topology map by a road recognition unit 141 and a self-position estimation unit 142. FIG. 4 is a diagram showing an example of a local topology map 1302. FIG. 5 is a block diagram showing the functional configuration of a server 20 according to the present embodiment. FIG. 6 is a diagram for explaining a method for generating a global topology map by a global topology map generation unit 231. FIG. 7 is a diagram for explaining selection of a representative driving trajectory as a driving line by the global topology map generation unit 231. FIG. 8 is a diagram showing an example of a global topology map 222 generated by the global topology map generation unit 231 by reflecting a driving line. FIG. 9 is a flowchart showing an example operation of the server 20 in generating the global topology map 222. FIG. 10 is a diagram for explaining a method for generating a learning dataset used for training a lane topology recognition AI by a learning data generation unit 232, and a method for generating a lane topology recognition AI by an AI learning unit 233. 1 is a diagram showing each vehicle 10 on the global topology map 222. FIG. 2 is a diagram for explaining lane topology recognition AI learned by the AI ​​learning unit 233. FIG. 3 is a flowchart showing an example of the operation of the server 20 in generating the lane topology recognition AI, and in generating the lane topology recognition AI. FIG. 4 is a diagram for explaining a method for relearning the lane topology recognition AI by the AI ​​learning unit 233. FIG. 5 is a flowchart showing an example of the operation of the server 20 in relearning the lane topology recognition AI by the AI ​​learning unit 233. FIG. 6 is a diagram showing an example of an architecture for realizing processing according to this embodiment. FIG. 7 is a block diagram showing an example of an information processing device 90.

[0010] Preferred embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. In this specification and drawings, components having substantially the same functional configurations are designated by the same reference numerals, and redundant description will be omitted.

[0011] The explanation will be given in the following order: 1. Overview 2. Example of the configuration of the vehicle 10 3. Example of the configuration of the server 20 4. Detailed functions 5. Example of the architecture 6. Example of the hardware configuration 7. Supplementary information

[0012] <<1. Overview>> The present disclosure relates to an information processing system that generates a recognizer that outputs a topology map used as a vehicle travel trajectory. The topology map is a map that shows lane topology, including a high-definition (HD) map showing road information such as lane separators, road boundaries, and crosswalks, and a vehicle travel trajectory (hereinafter also referred to as a "traveling line"). For example, in locations where lanes are set, the center line of the lane may be used as the traveling line. The HD map includes vector data and position information representing each piece of road information. The topology map also includes vector data and position information representing the traveling line.

[0013] Here, it is expected that road information can be automatically recognized from sensing data sensed by a vehicle, a topology map can be generated, and driving lines can be used to perform automatic driving of a vehicle or navigation of the vehicle. However, it has been difficult to derive driving lines in places where lane dividing lines cannot be recognized, such as intersections or alleys where lane dividing lines do not exist, or in places where other vehicles are obstructing the lane dividing lines.

[0014] According to a technique according to an embodiment of the present disclosure, it is possible to generate a recognizer that can output a topology map even in the above-described locations. Hereinafter, such a recognizer will also be referred to as "lane topology recognition AI (Artificial Intelligence)."

[0015] In this embodiment, first, a global topology map, which is a topology map serving as training data, is generated based on the travel trajectories of a plurality of vehicles. The global topology map is a first topology map according to this embodiment. The global topology map includes, as travel lines, trajectories appropriate for the travel of the vehicle 10.

[0016] Next, pairs of sensing data acquired by the vehicle and a topology map of the global topology map at the location where the sensing data was acquired are collected as a training data set. Using this training data set, the lane topology recognition AI is trained to output a topology map to be used as a trajectory of the vehicle traveling at various locations. The topology map output by the lane topology recognition AI is the second topology map according to this embodiment.

[0017] The overall configuration of an information processing system according to an embodiment of the present disclosure will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the overall configuration of an information processing system according to this embodiment.

[0018] 1, the information processing system according to this embodiment includes a plurality of vehicles 10 (10A to 10C) and a server 20. The vehicles 10 and the server 20 are connected for communication via a network 30.

[0019] (Vehicle 10) The vehicle 10 is a vehicle that travels on a road. The type (model) of the vehicle 10 is not particularly limited, and may be a small, medium, or large passenger car, a motorcycle, a moped, a truck, a bus, a towing vehicle, or the like.

[0020] The vehicle 10 senses the surrounding environment while traveling using various sensors mounted on the vehicle 10. The sensing data acquired by the vehicle 10 includes sensing data that can recognize roads, such as images (still images or moving images) captured by an RGB camera, point cloud data acquired by LiDAR (Light Detection and Ranging) or Radar (Radio Detection and Ranging), etc.

[0021] The vehicle 10 recognizes road information based on sensing data of the surrounding environment. The road information includes, for example, information indicating crosswalks, road boundaries, lane dividing lines, etc.

[0022] The vehicle 10 also acquires sensing data related to the traveling of the vehicle 10. The sensing data related to the traveling of the vehicle 10 may include, for example, the rotation speed of the wheels of the vehicle 10 acquired by a wheel speed sensor, the steering angle of the steering wheel of the vehicle 10 acquired by a steering angle sensor, the acceleration and angular velocity of the vehicle 10 acquired by an IMU (Inertial Measurement Unit), and position information of the vehicle 10 acquired by a GNSS (Global Navigation Satellite System) sensor such as a GPS (Global Positioning System).

[0023] The vehicle 10 estimates its own position based on an SD (Standard Distance) map, sensing data of the surrounding environment, and sensing data related to the traveling of the vehicle 10. The SD map is a map that is primarily used for navigation and has lower accuracy than an HD map. The SD map may be pre-stored in a storage unit provided in the vehicle 10, or an SD map of an area as needed may be downloaded from an external database or the like. The vehicle 10 recognizes the trajectory of the estimated own position as the traveling trajectory of the vehicle 10.

[0024] The vehicle 10 may generate a topology map of the locations where the vehicle has traveled based on the recognized road information and travel trajectory. Such a unique topology map generated by each vehicle 10 is hereinafter also referred to as a "local topology map." The local topology map may include the travel trajectory of the vehicle 10 as a travel line.

[0025] The vehicle 10 transmits various types of sensing data and the generated local topology map to the server 20 via the network 30 .

[0026] (Server 20) The server 20 is an information processing device that collects various data from the vehicle 10 via the network 30 and generates a lane topology recognition AI.

[0027] The server 20 executes various processes for generating the lane topology recognition AI. For example, the server 20 generates a global topology map for a location from a local topology map for the same location collected from the vehicle 10.

[0028] The server 20 also collects sensing data from the vehicle 10, pairs the collected sensing data with the global topology map at the location where the sensing data was collected, and generates a training data set for training the lane topology recognition AI. The server 20 then generates the lane topology recognition AI by performing training using the accumulated training data set.

[0029] (Network 30) The network 30 is a communication path between the vehicle 10 and the server 20. The network 30 may take any form, such as the Internet or a public line network.

[0030] <<2. Configuration Example of Vehicle 10>> Next, a configuration example of the vehicle 10 according to this embodiment will be described with reference to Fig. 2. Fig. 2 is a block diagram showing the functional configuration of the vehicle 10 according to this embodiment.

[0031] As shown in FIG. 2 , the vehicle 10 according to this embodiment includes a sensor unit 110 , a communication unit 120 , a storage unit 130 , and a control unit 140 .

[0032] (Sensor Unit 110) The sensor unit 110 is a sensor that acquires sensing data related to the surrounding environment of the vehicle 10 while it is traveling and the traveling of the vehicle 10. The sensor unit 110 may include, for example, an RGB camera (hereinafter also simply referred to as a camera) that acquires image data of the surrounding environment of the vehicle 10, a LiDAR and a Radar that acquire point cloud data of the surrounding environment of the vehicle 10, etc.

[0033] The sensor unit 110 may also include a wheel speed sensor that acquires the rotation speed of the wheels of the vehicle 10, a steering angle sensor that acquires the steering angle of the steering wheel of the vehicle 10, an IMU that acquires the acceleration and angular velocity of the vehicle 10, and a GNSS sensor that acquires the position information of the vehicle 10.

[0034] The sensors included in the sensor unit 110 are not limited to the above examples, and may include other sensors, such as an illuminance sensor, as long as they are capable of acquiring sensing data regarding the surrounding environment of the vehicle 10 or the driving of the vehicle 10.

[0035] (Communication Unit 120) The communication unit 120 communicates information with the server 20 via the network 30. The communication unit 120 transmits, for example, various types of sensing data acquired by the sensor unit 110 to the server 20.

[0036] (Storage Unit 130) The storage unit 130 stores various data used by the vehicle 10. For example, the storage unit 130 may store an SD map.

[0037] (Control Unit 140) The control unit 140 functions as an arithmetic processing unit and a control device, and controls the overall operation within the vehicle 10 in accordance with various programs. Furthermore, the control unit 140 according to this embodiment also functions as a road recognition unit 141 and a self-position estimation unit 142. A local topology map is generated based on the processing by the road recognition unit 141 and the self-position estimation unit 142. The processing by the road recognition unit 141 and the self-position estimation unit 142 will be described below with reference to FIG. 3. FIG. 3 is a diagram for explaining the generation of a local lane topology map by the road recognition unit 141 and the self-position estimation unit 142.

[0038] (Road Recognition Unit 141) The road recognition unit 141 recognizes road information such as crosswalks, road boundaries, and lane dividing lines based on sensing data of the surrounding environment of the vehicle 10 obtained by the sensor unit 110. The road information may be recognized from sensing data at a certain time, or may be recognized by compiling statistics of multiple pieces of sensing data at multiple times.

[0039] 3 , the road recognition unit 141 may recognize road information by inputting sensing data of the surrounding environment of the vehicle 10 acquired by the camera 1101, radar 1102, and LiDAR 1103 to a road recognizer 1411. The road recognizer 1411 is a recognizer that has been trained in advance to output road information using the sensing data as input data. The camera 1101, radar 1102, and LiDAR 1103 are sensors included in the sensor unit 110.

[0040] (Self-position estimation unit 142) The self-position estimation unit 142 estimates the self-position of the vehicle 10 based on the SD map and sensing data related to the surrounding environment of the vehicle 10 and the traveling of the vehicle 10 obtained by the sensor unit 110. For example, as shown in Fig. 3 , the road recognition unit 141 may estimate the self-position by inputting sensing data of the surrounding environment of the vehicle 10 acquired by the camera 1101, radar 1102, and LiDAR 1103, sensing data related to the traveling of the vehicle 10 acquired by the wheel odometry 1104, the IMU 1105, the steering angle sensor 1106, and the GNSS 1107 which is a GNSS sensor, and the SD map 1301 to a self-position estimator 1421.

[0041] The self-position estimator 1421 is an estimator that has been trained in advance to output its own position using sensing data as input data. The IMU 1105, steering angle sensor 1106, and GNSS 1107 are sensors included in the sensor unit 110. The wheel odometry 1104 outputs the position of the vehicle 10 estimated based on the number of rotations of the wheels acquired by wheel speed sensors included in the sensor unit 110 as sensing data.

[0042] The control unit 140 generates a local topology map 1302 by reflecting the road information recognized by the road recognition unit 141 and the self-position estimated by the self-position estimation unit 142 in an HD map of the vicinity of the self-position. The HD map may be downloaded in advance from an external map server or the like, or may be generated from the road information recognized by the road recognition unit 141.

[0043] FIG. 4 is a diagram showing an example of the local topology map 1302. FIG. 4 shows the local topology map 1302 and a legend 1302n that is a legend for each line included in the local topology map 1302. As shown in FIG. 4, the local topology map 1302 includes road information such as crosswalks, road boundaries, and lane dividing lines recognized by the road recognition unit 141, as well as the travel trajectory of the vehicle 10, which is represented by its own position acquired continuously over time. The travel trajectory represents the trajectory of the vehicle 10 traveling along the lane.

[0044] The communication unit 120 transmits the generated local topology map 1302 to the server 20. In addition to the local topology map 1302, the communication unit 120 also transmits sensing data acquired by the sensor unit 110 at each position on the local topology map 1302.

[0045] Furthermore, the communication unit 120 may transmit information about the vehicle type of the vehicle 10. The vehicle type of the vehicle 10 may be, for example, a small, medium, or large passenger car, a motorcycle, a moped, a truck, a bus, or a towing vehicle.

[0046] <<3. Configuration Example of Server 20>> Next, a configuration example of the server 20 according to this embodiment will be described with reference to Fig. 5. Fig. 5 is a block diagram showing the functional configuration of the server 20 according to this embodiment.

[0047] As shown in FIG. 5, the server 20 includes a communication unit 210 , a storage unit 220 , and a control unit 230 .

[0048] (Communication Unit 210) The communication unit 210 communicates information with the vehicles 10 via the network 30. The communication unit 210 receives, from each vehicle 10, various types of sensing data from the vehicle 10 and a local topology map 1302 specific to the vehicle 10, for example.

[0049] (Storage Unit 220) The storage unit 220 stores various data used by the server 20. As shown in FIG. 5 , the storage unit 220 also stores a travel trajectory DB 221, a global topology map 222, and a learning dataset 223.

[0050] (Traveling Locus DB 221 ) The traveling locus DB 221 is a DB that stores the traveling locus included in the local topology map 1302 unique to each vehicle 10 , which is acquired from the vehicle 10 by the communication unit 210 .

[0051] (Global Topology Map 222) The global topology map 222 is a topology map generated by a global topology map generation unit 231, which will be described later.

[0052] (Learning Data Set 223) The learning data set 223 is a learning data set generated by a learning data generation unit 232 (described later) and used for training the lane topology recognition AI.

[0053] (Control unit 230) The control unit 230 functions as an arithmetic processing unit and a control device, and controls the overall operation within the server 20 in accordance with various programs. The control unit 230 according to this embodiment also functions as a global topology map generation unit 231, a learning data generation unit 232, and an AI learning unit 233. Details of the global topology map generation unit 231, the learning data generation unit 232, and the AI ​​learning unit 233 will be described later.

[0054] <<4. Detailed Functions>> Next, the functions of the server 20 according to this embodiment will be described in detail.

[0055] <4.1. Generation of Global Topology Map> First, a method for generating a global topology map by the global topology map generation unit 231 will be described with reference to Fig. 6. Fig. 6 is a diagram for explaining a method for generating a global topology map by the global topology map generation unit 231.

[0056] In generating the global topology map, first, the communication unit 210 collects a local topology map 1302 including the recognition results of the travel trajectory and road information from each vehicle 10 (S1). Here, the communication unit 210 acquires not only the local topology map 1302 but also sensing data acquired by the vehicle 10 at each position on the local topology map 1302. Here, the communication unit 210 may further acquire information on the vehicle type of the vehicle 10.

[0057] Next, the global topology map generation unit 231 filters and classifies the collected data (S2). The global topology map generation unit 231 refers to the sensing data to recognize the environment in which the sensing data was acquired. If the global topology map generation unit 231 recognizes that the sensing data was acquired at night, in bad weather, or in an environment with poor visibility, it filters the range of the local topology map 1302 corresponding to the sensing data. At night, in bad weather, or in an environment with poor visibility, the vehicle 10 may not be traveling along an appropriate traveling trajectory. Furthermore, at night, in bad weather, or in an environment with poor visibility, the recognition results of the road information transmitted from the vehicle 10 may be less accurate. Therefore, by performing the above filtering process, it is possible to generate the global topology map 222 using only traveling trajectories that indicate more appropriate driving. In other words, traveling trajectories within the filtered range can be excluded from the selection candidates for traveling lines in the global topology map 222.

[0058] Furthermore, the global topology map generating unit 231 calculates the reliability of the sensing data, and then the global topology map generating unit 231 filters the range of the local topology map 1302 corresponding to the sensing data whose reliability is equal to or less than a threshold value.

[0059] The reliability may be calculated to be lower, for example, when there is a contradiction between the contents of sensing data acquired at the same time. For example, when it is determined that different objects are captured in the image and the point cloud, which are the sensing data, the reliability may be calculated to be lower.

[0060] In addition, if the global topology map generation unit 231 determines that road information cannot be recognized due to the presence of an obstacle such as another vehicle on the road, it filters the range of the local topology map 1302 corresponding to the road information.

[0061] The global topology map generation unit 231 may classify each collected data by vehicle type based on the vehicle type information of the vehicle 10 acquired by the communication unit 210. Then, the following processing may be performed on each piece of data classified by vehicle type. This makes it possible to generate a global topology map 222 that includes the optimal driving trajectory for each vehicle type.

[0062] Next, the global topology map generation unit 231 matches each local topology map 1302 based on the self-position and road information of each traveling trajectory included in the local topology map 1302 (S3). The global topology map generation unit 231 identifies a map of the surrounding area of ​​the corresponding self-position from the already generated global topology map 222 based on the self-position on the traveling trajectory and road information included in each local topology map 1302 that has been subjected to filtering processing. Note that if the global topology map 222 does not include a corresponding map, the global topology map generation unit 231 generates a global topology map 222 that reflects the road information included in the local topology map 1302.

[0063] Then, the global topology map generation unit 231 stores the travel trajectories in the travel trajectory DB 221 in association with positions in the global topology map 222. Note that if the data collected from the vehicles 10 is classified by vehicle type, the global topology map generation unit 231 classifies the travel trajectories by vehicle type and stores them in the travel trajectory DB 221.

[0064] Next, the global topology map generation unit 231 checks the collection status of the travel trajectories collected in the travel trajectory DB 221 for each area in the global topology map 222 (S4). Here, the area in the global topology map 222 may be an area where no lanes are set, such as an intersection where no lane dividing lines exist or an alley where no dividing lines exist as road boundaries.

[0065] If the collection status of travel trajectories in the area satisfies a criterion, the global topology map generation unit 231 selects a representative travel trajectory (S5). For example, for an area where a predetermined number or more travel trajectories have been collected, the global topology map generation unit 231 selects a representative travel trajectory as a travel line.

[0066] A typical example of selecting a driving path will be described with reference to FIG. 7. Here, an example will be described in which a driving path that serves as a driving line at an intersection where no lane separator exists is selected. The driving line at an intersection is information that indicates the connection between the lanes that connect to the intersection.

[0067] 7 is a diagram for explaining the selection of a representative travel trajectory as a travel line by the global topology map generation unit 231. As shown in Fig. 7, the global topology map generation unit 231 generates a travel trajectory superimposed map TO by superimposing multiple travel trajectories T (TA to TC, ...) included in local topology maps 1302 (1302A to 1302C, ...) of the same area.

[0068] The global topology map generator 231 then extracts travel trajectories connecting the endpoints of the lanes that connect to the intersection. Fig. 7 shows an example in which a travel trajectory set TOab, which is a set of travel trajectories connecting the endpoint of lane La to the endpoint of lane Lb, is extracted.

[0069] The global topology map generation unit 231 selects the centrally located driving trajectory CLab from the extracted driving trajectory set TOab as the driving line. This allows for the acquisition of an optimal driving trajectory that connects the lanes that connect to the intersection. The global topology map generation unit 231 reflects the selected driving line in the global topology map 222.

[0070] The above describes a method for selecting a driving line in an area where no lanes are set. In this way, by collecting and integrating the driving trajectories of multiple vehicles 10, a general-purpose driving line is selected that suppresses variations in driving trajectories due to the driving habits of drivers, etc.

[0071] In an area where lanes are set and which is partitioned by road boundaries and lane separation lines, the global topology map generating unit 231 reflects the center line of each lane as a driving line in the global topology map 222.

[0072] An example of the global topology map 222 reflecting the driving lines as described above is shown in Fig. 8. Fig. 8 is a diagram showing an example of the global topology map 222 generated by the global topology map generation unit 231. Fig. 8 shows the global topology map 222 and a legend 222n that is a legend for each line included in the global topology map 222.

[0073] 8, the global topology map 222 includes road information including driving lines, which are information indicating the connections between the lanes. In this way, the driving lines included in the global topology map 222 indicate an appropriate trajectory for the vehicle to travel.

[0074] (Example of Operation of Generating Global Topology Map) The above has described in detail how the server 20 generates the global topology map 222. Next, a description will be given of an example of operation of the server 20 in generating the global topology map 222. FIG. 9 is a flowchart showing an example of operation of the server 20 in generating the global topology map 222.

[0075] First, the communication unit 210 receives from each vehicle 10 a local topology map 1302 including a travel trajectory and road recognition results, and sensor data at each position on the local topology map 1302 (S101). Here, the communication unit 210 also receives vehicle model information for each vehicle 10. The global topology map generation unit 231 classifies the received data by selecting a map appropriate for the vehicle model according to the received vehicle model information (S102). The map may be, for example, a global topology map 222 that has already been generated, or an HD map. Below, an example will be described in which a global topology map 222 that has already been generated is selected.

[0076] The global topology map generation unit 231 determines whether the data received by the communication unit 210 is suitable for the global topology map 222 (S103). For example, it may be determined whether various data is suitable for the global topology map 222 based on the recognition result of the environment in which the sensing data is acquired.

[0077] If the data received by the communication unit 210 is not suitable for the global topology map 222 (S103 / NO), the processing ends. If the data received by the communication unit 210 is suitable for the global topology map 222 (S103 / YES), the global topology map generation unit 231 determines whether the received data indicates an area that exists on the global topology map 222 (S104). Note that if some of the data is not suitable for the global topology map 222, the area of ​​the local topology map 1302 that corresponds to the data may be filtered, and then the processing may proceed to step S104.

[0078] If the received data indicates an area that exists on the global topology map 222 (S104 / YES), the global topology map generation unit 231 aligns the global topology map 222 by matching the received data with the global topology map 222 (S105). If the received data does not indicate an area that exists on the global topology map 222 (S104 / NO), the global topology map generation unit 231 registers road information such as road boundaries, lane separators, and crosswalks in the global topology map 222 (S106).

[0079] Then, the global topology map generation unit 231 accumulates the travel trajectories included in the received local topology map 1302 in the travel trajectory DB 221 (S107). The global topology map generation unit 231 determines whether or not a certain number of travel trajectories have been collected in the travel trajectory DB 221 for each area of ​​the global topology map 222 (S108).

[0080] If the driving trajectory DB 221 does not collect a certain number of driving trajectories for each region of the global topology map 222 (NO in S108), the processing ends. On the other hand, if the driving trajectory DB 221 collects a certain number of driving trajectories for each region of the global topology map 222 (YES in S108), the global topology map generation unit 231 selects a representative driving trajectory for the region where the certain number of driving trajectories has been collected (S109). Then, the global topology map generation unit 231 generates a new global topology map 222 by registering the selected driving trajectory in the global topology map 222 as a driving line for that region (S110).

[0081] <4.2. Generation of training data set and generation of lane topology recognition AI> Next, a description will be given of a method for generating a training data set used for training the lane topology recognition AI by the training data generation unit 232, and a method for generating the lane topology recognition AI by the AI ​​learning unit 233. Fig. 10 is a diagram for explaining a method for generating training data used for training the lane topology recognition AI by the training data generation unit 232, and a method for generating the lane topology recognition AI by the AI ​​learning unit 233. The training data generation unit 232 generates a training data set after the global topology map 222 is completed.

[0082] To generate a training data set used for training the lane topology recognition AI, first, the communication unit 210 collects the vehicle's own position, sensing data, and road recognition results from each vehicle 10 (S6). Figure 11 is a diagram showing each vehicle 10 (10A to 10C) on the global topology map 222. Training data is generated using data collected from each vehicle 10 traveling at a position included in the global topology map 222, as shown in Figure 11.

[0083] If the global topology map 222 is generated for each vehicle type, vehicle type information is also collected from the vehicle 10. Then, the subsequent processing is executed for each vehicle type, and a learning data set is generated for each vehicle type.

[0084] Next, the training data generation unit 232 samples various data acquired from the vehicle 10 (S7). The training data generation unit 232 samples various data based on a plurality of sampling conditions, such as illuminance, weather, season, and the presence or absence of obstacles. Whether each sampling condition is satisfied may be determined, for example, based on sensing data or road recognition results received from the vehicle 10. The training data generation unit 232 extracts data that satisfies the conditions that are lacking in the training data set 223. This makes it possible to acquire a wide variety of data evenly, thereby obtaining a training data set for training a lane topology recognition AI that outputs topology maps in response to various situations.

[0085] The learning data generation unit 232 refers to the global topology map 222 corresponding to the surrounding area of ​​the vehicle's own position acquired from the vehicle 10, and performs matching between the sensing data and the global topology map 222 (S8). As a result, the position of the vehicle 10 in the global topology map 222 is identified.

[0086] The learning data generation unit 232 stores the sensing data obtained by the vehicle 10 at a position corresponding to the matching self-position in the learning data set 223 in pairs with the global topology map 222 around the self-position.

[0087] When a predetermined number or more pairs of sensing data and global topology map 222 that satisfy each sampling condition are stored in the learning dataset 223, the AI ​​learning unit 233 performs learning of the lane topology recognition AI (S9). FIG. 12 is a diagram for explaining the lane topology recognition AI learned by the AI ​​learning unit 233. The learning data generation unit 232 performs learning so as to output an AI recognition topology map 2332 by inputting input data such as that shown in FIG. 12. Specifically, the lane topology recognition AI 2331 receives, as input data, sensing data of the surrounding environment obtained by the camera 1101, radar 1102, LiDAR 1103, etc., possessed by the vehicle 10, as well as position information obtained by the GNSS 1107 for identifying the vehicle's own position and an SD map 1301 of a position corresponding to the position information.

[0088] The AI-recognized topology map 2332 outputs the AI-recognized topology map 2332, which is a topology map of the vicinity of the vehicle's own position, according to the sensing data included in the input data. The AI-recognized topology map 2332 indicates a driving line that is suitable as the driving trajectory of the vehicle 10 for which the sensing data was acquired. By using such an AI-recognized topology map 2332 to control the automatic driving of the vehicle 10 in real time along the driving line, it becomes possible to perform automatic driving of the vehicle 10.

[0089] Furthermore, the lane topology recognition AI 2331 is capable of outputting an AI-recognition topology map 2332 according to the location where the vehicle 10 is traveling, not limited to locations included in the global topology map 222, making it possible to perform automated driving in a variety of locations.

[0090] Furthermore, if the road information shown by the global topology map 222 differs from the actual road conditions due to construction or the like, an AI-recognized topology map 2332 appropriate to the actual road conditions at the location included in the global topology map 222 can be output.

[0091] (Example of operation for generating learning data used for training lane topology recognition AI) The method for generating a learning dataset used for training lane topology recognition AI by the server 20 and the method for generating lane topology recognition AI have been described in detail above. Next, an example of operation of the server 20 for generating a learning dataset used for training lane topology recognition AI and generating lane topology recognition AI will be described. Figure 13 is a flowchart showing an example of operation of the server 20 for generating a learning dataset used for training lane topology recognition AI and generating lane topology recognition AI.

[0092] First, the communication unit 210 receives the vehicle position, sensing data, and road recognition results (S201). Here, the communication unit 210 also receives vehicle model information for each vehicle 10. The learning data generation unit 232 selects a global topology map 222 suitable for the vehicle model based on the received vehicle model information (S202).

[0093] The training data generation unit 232 samples data of the sampling conditions that are missing from the training data set 223 (S203). Then, the training data generation unit 232 determines whether the self-position of the vehicle 10 included in the sampled data indicates an area that exists on the global topology map 222 (S204). If the self-position of the vehicle 10 included in the sampled data does not indicate an area that exists on the global topology map 222 (S204 / NO), the processing ends.

[0094] If the vehicle 10's own position included in the sampled data indicates an area on the global topology map 222 (YES in S204), the learning data generation unit 232 aligns the sensing data by matching it with the global topology map 222 (S205). Then, the sensing data is paired with the global topology map 222 around the vehicle's own position from which the sensing data was acquired, and these are recorded as learning data in the learning dataset 223 (S206). When the number of pairs of sensing data and global topology map 222 that satisfy each of the recorded sampling conditions as described above reaches a predetermined number or more, the AI ​​learning unit 233 learns the lane topology recognition AI 2331.

[0095] <4.3. Re-learning of lane topology recognition AI by AI learning unit 233> Next, a method for relearning the lane topology recognition AI 2331 by the AI ​​learning unit 233 will be described. The AI ​​learning unit 233 collects learning data corresponding to scenes in which the lane topology recognition AI 2331 has difficulty outputting a topology map. Then, the AI ​​learning unit 233 re-learns the lane topology recognition AI 2331 using the collected learning data.

[0096] FIG. 14 is a diagram for explaining a method for relearning the lane topology recognition AI by the AI ​​learning unit 233.

[0097] In the relearning of the lane topology recognition AI 2331 by the AI ​​learning unit 233, first, sensing data is collected by each vehicle 10 (S10). Then, each vehicle 10 estimates its own position from the sensing data (S11).

[0098] The vehicle 10 obtains an AI-recognized topology map 2332 around its own position by inputting sensing data and its own position into the AI-recognized topology map 2332 generated by the server 20 using the AI ​​learning unit 233 (S12).

[0099] The server 20 acquires the AI-recognized topology map 2332 from the vehicle 10 via the communication unit 210. Then, the AI ​​learning unit 233 matches the AI-recognized topology map 2332 with the global topology map 222 based on the vehicle 10's own position corresponding to the AI-recognized topology map 2332 (S13). The AI ​​learning unit 233 calculates the difference between the AI-recognized topology map 2332 and the global topology map 222 in the same area (S14). More specifically, the AI ​​learning unit 233 obtains the difference by comparing corresponding vector information included in each map.

[0100] If the difference is greater than or equal to a threshold, the AI ​​learning unit 233 pairs the sensing data, which is input data for the lane topology recognition AI 2331, with the global topology map 222 around the vehicle's own position from which the sensing data was acquired, and records the paired data as learning data for the difficult scene in the learning dataset 223 (S15).

[0101] As described above, the AI ​​learning unit 233 uses the recorded learning data of difficult scenes to re-learn the lane topology recognition AI 2331. This allows for automatically collecting learning data that identifies, as difficult scenes, scenes in which there is a discrepancy between the appropriate driving trajectory of the vehicle 10 and the AI-recognition topology map 2332 obtained by the lane topology recognition AI 2331, and then focusing on learning the difficult scenes. Therefore, it is possible to re-learn the lane topology recognition AI 2331 so as to output the AI-recognition topology map 2332 that includes a driving line that is more suitable for driving by the vehicle 10. By repeating this re-learning of the lane topology recognition AI 2331, a more robust lane topology recognition AI 2331 is completed, thereby increasing the feasibility of autonomous driving of the vehicle 10 using the lane topology recognition AI 2331.

[0102] (Example of operation for generating learning data used to learn lane topology recognition AI) The method for relearning the lane topology recognition AI 2331 by the AI ​​learning unit 233 using the server 20 has been described in detail above. Next, an example of operation of the server 20 when the lane topology recognition AI 2331 is relearned by the AI ​​learning unit 233 will be described. Figure 15 is a flowchart showing an example of operation of the server 20 when the lane topology recognition AI 2331 is relearned by the AI ​​learning unit 233.

[0103] First, the communication unit 210 acquires from the vehicle 10 its own position, sensing data, and an AI-recognized topology map 2332 generated from the own position and sensing data (S301).

[0104] The AI ​​learning unit 233 checks whether the area of ​​the acquired AI-recognition topology map 2332 exists in the global topology map 222 (S302). If the area of ​​the AI-recognition topology map 2332 does not exist in the global topology map 222 (S303 / NO), the processing ends. If the area of ​​the AI-recognition topology map 2332 exists in the global topology map 222 (S303 / YES), the AI ​​learning unit 233 aligns the AI-recognition topology map 2332 with the global topology map 222 (S304).

[0105] Then, the AI ​​learning unit 233 calculates the difference between the maps by comparing the corresponding vector information between the AI-recognized topology map 2332 and the global topology map 222 (S304).

[0106] If the calculated difference is smaller than the threshold value (S306 / NO), the AI ​​learning unit 233 ends the process. If the calculated difference is equal to or larger than the threshold value (S306 / YES), the AI ​​learning unit 233 pairs the sensing data acquired by the communication unit 210 with the global topology map 222 around the subject's own position from which the sensing data was acquired, and records the paired data as learning data for the difficult scene in the learning dataset 223 (S307).

[0107] <<5. Architecture Example>> Next, an example of an architecture for realizing the processing according to this embodiment will be described. Fig. 16 is a diagram showing an example of an architecture for realizing the processing according to this embodiment.

[0108] First, estimation of panoptic segmentation by sensor fusion using the camera 1101 and LiDAR 1103 will be described.

[0109] The RGB image acquired by the camera 1101 is subjected to feature extraction (RGB image feature extraction).

[0110] First, depth conversion is performed on the 3D point cloud acquired by the LiDAR 1103. Depth conversion is a process of converting the 3D point cloud acquired by the LiDAR 1103 into a 2D depth image. Next, feature extraction (depth image feature extraction) is performed from the depth image acquired as described above.

[0111] The RGB image feature extraction and the depth image feature extraction may be performed using a network configuration using a CNN (Convolutional Neural Network), a Deformable CNN, or a Transformer. A pixel decoder may be added after the CNN or the Transformer to aggregate features between multiscales. The network for extracting the depth image feature and the network for extracting the RGB image feature may be prepared separately or may be the same.

[0112] Next, feature fusion is performed to fuse the depth image feature and the RGB image feature, and fusion feature is obtained.

[0113] The fusion feature obtained as described above is input to the segmentation head, which is an output device that performs segmentation on the image region mask. The segmentation head outputs the results of semantic segmentation and instance segmentation.

[0114] Next, estimation of depth, 3D bounding box (3DBBox), velocity, and lane topology by sensor fusion using camera 1101, radar 1102, and LiDAR 1103 will be described.

[0115] First, the extracted fusion feature amount and camera parameters are input to the depth head, and an estimated depth is output.

[0116] The above camera parameters are various parameters for projecting the point cloud of LiDAR 1103 onto an RGB image, and include information such as the attitude (rotation matrix) and position (translation vector) of camera 1101, focal length, and center position of the lens.

[0117] Next, a Bird's Eye View (BEV) transform is performed based on the depth estimated as described above.

[0118] On the other hand, radar feature extraction is performed on the point cloud acquired by the radar 1102, and BEV Transform is performed based on the extracted radar feature.

[0119] Next, BEV feature fusion is performed based on the results of the depth-based BEV Transform and the radar feature-based BEV Transform, and the resulting fusion feature is input to a 3D Detection Head to obtain 3D BBox and velocity estimation results.Furthermore, the resulting fusion feature is input to a Lane Topology Head to obtain lane topology estimation results and their reliability.

[0120] Next, estimation of an HD map by sensor fusion using the camera 1101, radar 1102, LiDAR 1103, and SD map 1301 will be described.

[0121] The SD map 1301 is input to a map encoder and converted into a format that can be input to a map decoder. The converted SD map 1301 and the fusion feature obtained by BEV feature fusion are input to the map decoder, whereby an HD map is obtained.

[0122] 6. Hardware Configuration Example>> Each embodiment of the present disclosure has been described above. The above-described information processing is realized by cooperation between software and hardware. Below, a hardware configuration example that can be applied to the vehicle 10 and the server 20 will be described.

[0123] Fig. 17 is a block diagram showing an example of an information processing device 90. Note that the hardware configuration example of the information processing device 90 described below is merely an example of the hardware configuration of the vehicle 10 and the server 20. Therefore, the vehicle 10 and the server 20 do not necessarily have to have the entire hardware configuration shown in Fig. 17 . Furthermore, the vehicle 10 and the server 20 may not necessarily have some of the hardware configuration shown in Fig. 17 .

[0124] 17 , the information processing device 90 includes a CPU 901, a ROM (Read Only Memory) 903, and a RAM 905. The information processing device 90 may also include a host bus 907, a bridge 909, an external bus 911, an interface 913, an input device 915, an output device 917, a storage device 919, a drive 921, a connection port 923, and a communication device 925. Instead of or in addition to the CPU 901, the information processing device 90 may include a processing circuit such as a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), or an ASIC (Application Specific Integrated Circuit).

[0125] The CPU 901 functions as an arithmetic processing unit and control unit, and controls all or part of the operations within the information processing device 90 in accordance with various programs recorded in the ROM 903, RAM 905, storage device 919, or removable recording medium 927. The ROM 903 stores programs and calculation parameters used by the CPU 901. The RAM 905 temporarily stores programs used in the execution of the CPU 901 and / or parameters that change as appropriate during the execution. The CPU 901, ROM 903, and RAM 905 are interconnected by a host bus 907, which is composed of an internal bus such as a CPU bus. Furthermore, the host bus 907 is connected to an external bus 911, such as a PCI (Peripheral Component Interconnect / Interface) bus, via a bridge 909.

[0126] The CPU 901, in cooperation with the ROM 903, the RAM 905, and software, can realize the functions of the control unit 140 and the control unit 230, for example.

[0127] The input device 915 is a device operated by a user, such as a button. The input device 915 may include a mouse, a keyboard, a touch panel, a switch, a lever, or the like. The input device 915 may also include a microphone that captures the user's voice. The input device 915 may be, for example, a remote control device that uses infrared or other radio waves, or an externally connected device 929 such as a mobile phone that supports operation of the information processing device 90. The input device 915 includes an input control circuit that generates an input signal based on information input by the user and outputs the signal to the CPU 901. The user operates the input device 915 to input various data and instruct processing operations to the information processing device 90.

[0128] The input device 915 may also include an imaging device and a sensor. The imaging device is a device that captures real space and generates a captured image using an imaging element such as a CCD (Charge Coupled Device) image sensor or a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor, and various components such as a lens for controlling the formation of a subject image on the imaging element. The imaging device may capture still images or may capture moving images.

[0129] The sensors include various types of sensors, such as a distance measurement sensor, an acceleration sensor, a gyro sensor, a geomagnetic sensor, a vibration sensor, a light sensor, and a sound sensor. The sensors acquire information about the state of the information processing device 90 itself, such as the attitude of the housing of the information processing device 90, or information about the surrounding environment of the information processing device 90, such as the brightness or noise around the information processing device 90. The sensors may also include a GNSS sensor that receives GNSS signals and measures the latitude, longitude, and altitude of the device.

[0130] The output device 917 is configured with a device capable of visually or audibly notifying the user of acquired information. The output device 917 may be, for example, a display device such as an LCD (Liquid Crystal Display) or an organic EL (Electro-Luminescence) display, or an audio output device such as a speaker or headphones. The output device 917 may also include a PDP (Plasma Display Panel), a projector, a hologram, a printer, or the like. The output device 917 outputs the results obtained by the processing of the information processing device 90 as video such as text or images, or as sound such as voice or audio. The output device 917 may also include a lighting device that brightens the surrounding area.

[0131] The storage device 919 is a data storage device configured as an example of a storage unit of the information processing device 90. The storage device 919 is configured, for example, by a magnetic storage device such as a hard disk drive (HDD), a semiconductor storage device, an optical storage device, or a magneto-optical storage device. This storage device 919 stores programs or various data executed by the CPU 901, as well as various data acquired from the outside.

[0132] The drive 921 is a reader / writer for a removable recording medium 927 such as a magnetic disk, optical disk, magneto-optical disk, or semiconductor memory, and is built into or externally attached to the information processing device 90. The drive 921 reads information recorded on the attached removable recording medium 927 and outputs the information to the RAM 905. The drive 921 also writes information to the attached removable recording medium 927.

[0133] The connection port 923 is a port for directly connecting a device to the information processing device 90. The connection port 923 may be, for example, a USB (Universal Serial Bus) port, an IEEE 1394 port, a SCSI (Small Computer System Interface) port, or the like. The connection port 923 may also be an RS-232C port, an optical audio terminal, an HDMI (registered trademark) (High-Definition Multimedia Interface) port, or the like. By connecting an external device 929 to the connection port 923, various types of data can be exchanged between the information processing device 90 and the external device 929.

[0134] The communication device 925 is a communication interface configured with a communication device for connecting to a network 931, such as a local network or a communication network with a wireless communication base station. The communication device 925 may be, for example, a wired or wireless LAN, Bluetooth, Wi-Fi, or a communication card for WUSB (Wireless USB). The communication device 925 may also be a router for optical communication, a router for ADSL (Asymmetric Digital Subscriber Line), or a modem for various communications. The communication device 925 transmits and receives signals, for example, between the Internet and other communication devices using a predetermined protocol such as TCP / IP. The local network or communication network with the base station connected to the communication device 925 is a wired or wireless network, such as the Internet, a home LAN, infrared communication, radio wave communication, or satellite communication.

[0135] <<7. Supplementary Information>> Although preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings, the technical scope of the present disclosure is not limited to such examples. It is clear that a person skilled in the art of the present disclosure can conceive of various modified or altered examples within the scope of the technical idea described in the claims, and it is understood that these also naturally fall within the technical scope of the present disclosure.

[0136] For example, although the above describes an example in which the lane topology recognition AI 2331 is trained using, as training data, a pair of sensing data from the vehicle 10 and the global topology map 222 of the position where the sensing data was obtained, the training data set 223 may include other training data. For example, based on the road structure obtained from the global topology map 222, data corresponding to the sensing data and in line with the road structure may be generated by CG, generation AI, or the like. Then, a pair of the generated data and the global topology map 222 corresponding to the data may be used as training data.

[0137] It is also possible to create a computer program for causing hardware such as a CPU, ROM, and RAM built into the vehicle 10 or the server 20 to perform the functions of the vehicle 10 or the server 20. A computer-readable storage medium storing the computer program is also provided.

[0138] Furthermore, the effects described herein are merely descriptive or exemplary and are not limiting. In other words, the technology according to the present disclosure may achieve other effects that will be apparent to those skilled in the art from the description of this specification, in addition to or in place of the above-described effects.

[0139] Note that the following configurations also fall within the technical scope of the present disclosure. (1) An information processing method executed by a computer, comprising: generating a recognizer that outputs a second topology map used as a travel trajectory of a traveling vehicle based on learning using a dataset including sensing data that is data sensed by a vehicle and a first topology map that is a topology map of lanes corresponding to positions where the sensing data was acquired. (2) The information processing method described in (1), in which the recognizer takes sensing data sensed by the traveling vehicle as input data and outputs the second topology map. (3) The information processing method described in (2), in which the first topology map and the second topology map include information indicating connections between the lanes. (4) The information processing method described in (3), in which the information indicating connections between the lanes is information representing a travel trajectory in an area where the lanes are not set. (5) The information processing method according to (3) or (4), further comprising generating the first topology map based on travel trajectories of each vehicle collected from a plurality of vehicles. (6) The information processing method according to (5), wherein, in generating the first topology map in the area where lanes are not set, a travel trajectory that serves as information indicating the connection between the lanes is selected from a plurality of travel trajectories that connect end points of center lines of the lane dividing lines. (7) The information processing method according to (6), wherein the selected information indicating the connection between the lanes is a representative travel trajectory from the plurality of travel trajectories that connect end points of center lines of the lane dividing lines. (8) The information processing method according to any of (5) to (7), wherein, in generating the first topology map, the first topology map is generated for each vehicle type of the plurality of vehicles according to the vehicle types of the plurality of vehicles. (9) The information processing method according to any one of (5) to (8), wherein in generating the first topology map, filtering is performed on each of the plurality of traveling trajectories based on sensing data from each of the plurality of vehicles.(10) The information processing method according to any one of (1) to (9), wherein, in generating the recognizer, a dataset used for the learning is sampled based on whether the sensing data satisfies a predetermined sampling condition. (11) The information processing method according to any one of (3) to (9), includes: after generating the recognizer, determining whether a difference between the first topology map and the second topology map in the same area is equal to or greater than a threshold, collecting input data for generating the second topology map in which the difference is equal to or greater than the threshold, and re-learning the recognizer using a dataset including the collected input data and the first topology map. (12) The information processing method according to any one of (2) to (9), wherein the second topology map is used as a traveling trajectory of the vehicle traveling by autonomous driving. (13) An information processing device including: a generation unit that generates a recognizer that outputs a second topology map used as a traveling trajectory of a traveling vehicle based on learning using a dataset including sensing data that is data sensed by a vehicle and a first topology map that is a topology map of lanes corresponding to positions where the sensing data was acquired. (14) A program that causes a computer to function as: a generation unit that generates a recognizer that outputs a second topology map used as a traveling trajectory of a traveling vehicle based on learning using a dataset including sensing data that is data sensed by a vehicle and a first topology map that is a topology map of lanes corresponding to positions where the sensing data was acquired.

[0140] REFERENCE SIGNS LIST 1 Information processing system 10 Vehicle 110 Sensor unit 120 Communication unit 130 Memory unit 140 Control unit 141 Road recognition unit 142 Self-position estimation unit 20 Server 210 Communication unit 220 Memory unit 221 Travel trajectory DB 222 Global topology map 223 Learning data set 230 Control unit 231 Global topology map generation unit 232 Learning data generation unit 233 AI learning unit 2331 Lane topology recognition AI 2332 AI recognition topology map

Claims

1. An information processing method executed by a computer, comprising: generating a recognizer that outputs a second topology map used as a travel trajectory of a traveling vehicle based on learning using a dataset including sensing data that is data sensed by a vehicle and a first topology map that is a topology map of lanes corresponding to the positions where the sensing data was acquired.

2. The information processing method according to claim 1, wherein the recognizer receives sensing data sensed by the vehicle while it is moving as input data and outputs the second topology map.

3. The information processing method according to claim 2, wherein the first topology map and the second topology map include information indicating connections between the lanes.

4. The information processing method according to claim 3, wherein the information indicating the connection between the lanes is information representing a travel path in an area where the lanes are not set.

5. The information processing method according to claim 3, further comprising generating the first topology map based on the travel trajectories of each vehicle collected from a plurality of vehicles.

6. An information processing method as described in claim 5, wherein, in generating the first topology map in an area where no lanes are set, a driving trajectory that provides information indicating the connection between the lanes is selected from among the plurality of driving trajectories that connect the endpoints of the center lines of the lane dividing lines.

7. An information processing method according to claim 6, wherein the information indicating the connection between the selected lanes is a representative driving trajectory among the plurality of driving trajectories that connect the end points of the center lines of the lane dividing lines.

8. The information processing method according to claim 5, wherein in generating the first topology map, the first topology map is generated for each vehicle type according to the vehicle types of the plurality of vehicles.

9. The information processing method according to claim 5, wherein in generating the first topology map, filtering is performed on each of the plurality of travel trajectories based on sensing data from each of the plurality of vehicles.

10. The information processing method according to claim 1, wherein in generating the recognizer, the data set used for the learning is sampled based on whether the sensing data satisfies a predetermined sampling condition.

11. The information processing method according to claim 3, further comprising: determining whether, after the recognizer is generated, a difference between the first topology map and the second topology map in the same region is equal to or greater than a threshold; collecting input data for generating the second topology map in which the difference is equal to or greater than the threshold; and re-training the recognizer using a data set including the collected input data and the first topology map.

12. The information processing method according to claim 2, wherein the second topology map is used as a travel trajectory of the vehicle traveling by automatic driving.

13. An information processing device comprising: a generation unit that generates a recognizer that outputs a second topology map used as a travel trajectory of a traveling vehicle based on learning using a dataset including sensing data that is data sensed by a vehicle and a first topology map that is a topology map of lanes corresponding to the positions where the sensing data was acquired.

14. A program that causes a computer to function as a generation unit that generates a recognizer that outputs a second topology map used as a travel trajectory of a moving vehicle based on learning using a dataset that includes sensing data that is data sensed by a vehicle and a first topology map that is a topology map of the lane corresponding to the position where the sensing data was acquired.

Images